2013abd Eestab

Departamento de Ingeniería Eléctrica, Electrónica, de
Computadores y de Sistemas
Doctor of Philosophy
Technical and Economical Analysis of Isolated

Microgrids
Ahmed Mohamed Abd el Motaleb

ACKNOWLEDGEMENTS
First of all, the author would like to express his sincere appreciation and
gratitude to his supervisor, Dr. Guzmán Díaz González, for his invaluable
guidance, patience, encouragement and support throughout the course of this
research work and in the preparation of this thesis.
The author would like to express his sincere gratitude to Dr. Cristina
González Morán for her guidance, insightful comments and suggestions during
the research work.
The author would like to thank all colleagues at Escuela Politécnica de
Ingeniería de Gijón for their support and help.
The author would like to thank Ministerio de Ciencia e Innovación for the
nancial support by the scholarship FPI-ene2010-14899.
Finally, the author would like to take this opportunity to acknowledge the
constant encouragement and support from his family and ex-director. Hazem
Makram. The author presents this thesis as a gift to them.

Contents i
Contents
Introduction 1
1 State of the Art 5

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Operational Concepts of Microgrids . . . . . . . . . . . . . . . . 10
1.3 Impacts of Microgrids . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Power Flow in Microgrids . . . . . . . . . . . . . . . . . . . . . 14
1.5 The Interface Control of Microgrids . . . . . . . . . . . . . . . . 16
1.5.1 Power Electronic Interface . . . . . . . . . . . . . . . . . 16
1.5.2 Droop Control in Microgrids . . . . . . . . . . . . . . . . 19
1.5.3 Microgrids Supplied by Non-Dispatchable Generators . . 20
1.6 Capacity Factor Analysis of Droop-Regulat-ed Microgrids . . . . 26
1.7 Benet Evaluation of Microgrids Using Capacity Factor and
Economic Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.8 Microgrids Stability Analysis . . . . . . . . . . . . . . . . . . . . 32
1.9 Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . . 34
2 Capacity Factor Analysis of Droop-Regulated Microgrids 41

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.1.1 Chapter Outlines . . . . . . . . . . . . . . . . . . . . . . 46
2.2 Latin Hypercube Sampling . . . . . . . . . . . . . . . . . . . . . 49
2.2.1 Distribution Functions of Parameters and State Va-riables 49
2.2.2 Calculating the number of samples . . . . . . . . . . . . 50
2.2.3 Dividing the range of each of the parameters into equi-
probable intervals . . . . . . . . . . . . . . . . . . . . . . 50
2.2.4 Creating the LHS table . . . . . . . . . . . . . . . . . . . 51
2.3 Correlation among Dierent Wind Turbines . . . . . . . . . . . 53

ii Contents
2.3.1 Correlation Steps . . . . . . . . . . . . . . . . . . . . . . 56
2.4 Curve Fitting of the Wind Turbine Enercon-E40 . . . . . . . . . 58
2.5 Cumulative Distribution of Wind Power . . . . . . . . . . . . . 59
2.6 Microgrid Layout . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.7 Nonlinear Optimization . . . . . . . . . . . . . . . . . . . . . . . 64
2.7.1 Global Optimization Search . . . . . . . . . . . . . . . . 65
2.7.2 Applied Global Search Optimization Methodology . . . . 66
2.8 Processing Simualtions and Conclusions . . . . . . . . . . . . . . 68
2.9 Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.9.1 The likelihood measure . . . . . . . . . . . . . . . . . . . 71
2.9.2 Akaike Information Criterion . . . . . . . . . . . . . . . 72
2.9.3 Bayesian Information Criterion . . . . . . . . . . . . . . 72
2.10 Microgrid Capacity Factor . . . . . . . . . . . . . . . . . . . . . 75
2.11 Microgrid Bifurcation Analysis . . . . . . . . . . . . . . . . . . . 78
2.12 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3 Benet Evaluation of Microgrids Using Capacity Factor and

Economic Analysis 83
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.2 Sources of Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.2.1 Microgrid Layout . . . . . . . . . . . . . . . . . . . . . . 87
3.2.2 Droop Regulation by Fischer-Burmeister . . . . . . . . . 88
3.2.3 Daily Operation of Batteries . . . . . . . . . . . . . . . . 90
3.2.4 Wind Turbines Characteristics . . . . . . . . . . . . . . . 92
3.2.5 Economic Evaluation Models . . . . . . . . . . . . . . . 94
3.3 Economic Evaluation for a Scaled Down Wi-nd Turbine . . . . . 98
3.3.1 The First Step of Optimization . . . . . . . . . . . . . . 102
3.3.2 The Second Step of Optimization . . . . . . . . . . . . . 103
3.3.3 Optimization Results . . . . . . . . . . . . . . . . . . . . 104
3.4 Economic Evaluation for a Microgrid Supplied by Wind Tur-
bines of Dierent Power Ratings . . . . . . . . . . . . . . . . . . 109
3.5 Economic Evaluation for Microgrids Supplie-d by Various Wind
Turbines and Batteries . . . . . . . . . . . . . . . . . . . . . . . 116

Contents iii
3.6 The Aggregation Eect of Wind Turbines on the Economic Per-
formance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
4 Microgrid Stability Analysis 137

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
4.2 Chapter Outlines . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.3 Microgrid Layout . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.4 Microgrid State Space Modelling . . . . . . . . . . . . . . . . . 140
4.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 140
4.4.2 Voltage-Frequency Inverter State Space Modelling . . . . 145
4.4.3 Doubly-Fed Induction Generator State Space Modelling . 151
4.4.4 Distribution Line State Space Modelling . . . . . . . . . 160
4.4.5 Static Load State Space Modelling . . . . . . . . . . . . 161
4.5 Conducting Participation Factor and Time Response . . . . . . 163
4.6 Wind Variation Impact on the Microgrid Stability . . . . . . . . 165
4.6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 165
4.6.2 Processing Simulations . . . . . . . . . . . . . . . . . . . 166
4.7 Droop Constants variation Impact on the Microgrid Stability . . 181
4.7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 181
4.8 Power Demand Increment Impact on the Microgrid Stability . . 191
4.8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 191
5 Conclusions, Published Articles

and Future Work 211
5.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
5.2 Published Articles . . . . . . . . . . . . . . . . . . . . . . . . . . 214
5.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
5 Conclusiones, Artículos Publicados

y Trabajos Futuros 217
5.1 Conclusiones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
5.2 Artículos Publicados . . . . . . . . . . . . . . . . . . . . . . . . 219
5.3 Trabajos Futuros . . . . . . . . . . . . . . . . . . . . . . . . . . 220

iv Contents
List of gures v
List of Figures
1 EU electricity generation by dierent energy resources . . . . . . 3
1.1 Centralized vs distributed generation . . . . . . . . . . . . . . . 12
1.2 Interconnected systems power exchange . . . . . . . . . . . . . . 23
1.3 Mechanical analogue of swings in a multi-machine system . . . . 24
1.4 An example of wind measurements at dierent sampling rates
in Germany, Ernst: (a) Wind speed. (b) Correlation coecients
vs distance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.1 Categories of system reliability . . . . . . . . . . . . . . . . . . . 42
2.2 World wind energy generation . . . . . . . . . . . . . . . . . . . 43
2.3 Monte Carlo transformation method . . . . . . . . . . . . . . . 46
2.4 Stratied wind speed samples by LHS . . . . . . . . . . . . . . . 51
2.5 Wind speed sampling at scale parameter (3) processed by: (a)
LHS. (b) MCS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.6 Wind speed sampling at scale parameter (5) processed by: (a)
LHS. (b) MCS. . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.7 Cross-correlation vs distance. . . . . . . . . . . . . . . . . . . . 54
2.8 Microgrid layout . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.9 Wind speed time series processed by scale parameter (3): (a)
Correlation included. (b) Correlation excluded. . . . . . . . . . 57
2.10 Curve tting of Enercon-E40 600 kW, manufacturer data indi-
cated by (asterisks). . . . . . . . . . . . . . . . . . . . . . . . . 60
2.11 Cumulative distribution of Enercon-600kW: (a) Processed at
wind speed scale parameter (3). (b) Processed at wind speed
scale parameter (8). . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.13 Active power vs frequency droop characteristics . . . . . . . . . 63

vi List of gures
2.14 Reactive power vs voltage droop characteristics . . . . . . . . . 63
2.15 Flow chart of the Global Search . . . . . . . . . . . . . . . . . . 67
2.16 Histograms of wind turbines and microgrid power samples. Indi-
vidual powers indicated by (thin lines) and aggregated injected
power indicated by (blue, thick line). . . . . . . . . . . . . . . . 69
2.17 Cumulative distribution functions of the microgrid injected power
processed at wind speed scale and shape parameters (8) and
(1.4) respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.18 Errors between the empirical distribution and a set of tted dis-
tirbution functions of the microgrid injected power processed at
wind speed scale and shape parameters (8) and (1.4) respectively. 75
2.19 Microgrid loadability by CPF. Solid line: MSUs without power
limits; dashed line: one MSU with power limitation; dotted line:
two MSUs with power limitation. Source: [1] . . . . . . . . . . . 79
2.20 Box plot of several scenarios. . . . . . . . . . . . . . . . . . . . . 81
3.2 Daily battery charging-discharging process . . . . . . . . . . . . 92
3.3 Daily operation of wind turbines . . . . . . . . . . . . . . . . . . 93
3.4 Wind turbine output power according to several rating ratio
values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.5 Wind turbine capital cost vs rating ratio . . . . . . . . . . . . . 100
3.6 Wind turbines curve tting: (a) Enercon-600 kW. (b) Enercon-
900 kW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.7 Daily power generation percentage: (a) Enercon-600 kW. (b)
Enercon-900 kW. . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.8 Layout of a microgrid supplied by wind turbines of dierent
power ratings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
3.9 Curve tting of contributing wind turbines: (a) Enercon-330
kW. (b) Enercon-900 kW. (c) Enercon-2000 kW. . . . . . . . . . 110
3.10 Capacity factor vs wind speed . . . . . . . . . . . . . . . . . . . 111
3.11 Wind turbines levelized cost of energy vs wind speed . . . . . . 112
3.12 Simple payback vs wind speed . . . . . . . . . . . . . . . . . . . 112
3.13 Net present value of Enercon-900 kW turbine measured at dif-
ferent periods of time . . . . . . . . . . . . . . . . . . . . . . . . 113

List of gures vii
3.16 Daily power production at average wind speed 2 m/sec . . . . . 115
3.17 Daily power production at average wind speed 5 m/sec . . . . . 115
3.18 Total levelized cost of energy at dierent wind speed levels . . . 117
3.19 Capital cost of batteries . . . . . . . . . . . . . . . . . . . . . . 119
3.20 Daily power generation percentage: (a) Lithium battery case.
(b) Vanadium battery case. . . . . . . . . . . . . . . . . . . . . 122
3.21 Batteries levelized cost of energy . . . . . . . . . . . . . . . . . . 123
3.22 Batteries annual energy production . . . . . . . . . . . . . . . . 123
3.23 Curve tting of contributing wind turbines: (a) Enercon-600
kW. (b) Vestas-1800 kW. (c) Enercon-2000 kW. . . . . . . . . . 125
3.25 Wind turbines capacity factor . . . . . . . . . . . . . . . . . . . 127
3.26 Wind turbines annual energy production . . . . . . . . . . . . . 128
3.27 Wind turbines levelized cost of energy . . . . . . . . . . . . . . 128
3.28 Batteries annual energy production . . . . . . . . . . . . . . . . 129
3.29 Batteries levelized cost of energy . . . . . . . . . . . . . . . . . . 129
3.30 Total levelized cost of energy . . . . . . . . . . . . . . . . . . . . 130
3.31 Net present value, Enercon-600 kW . . . . . . . . . . . . . . . . 130
3.32 Net present value, Enercon-2000 kW . . . . . . . . . . . . . . . 131
3.33 Net present value, Vestas-1800 kW . . . . . . . . . . . . . . . . 131
3.34 Daily power percentage, Enercon-600 kW: (a) mean wind speed
(4 m/s). (b) mean wind speed (5 m/s). . . . . . . . . . . . . . . 132
3.35 Daily power generation percentage, Vestas-1800 kW: (a) mean
wind speed (4 m/s). (b) mean wind speed (5 m/s). . . . . . . . 133
3.36 Daily power percentage, Enercon-2000 kW: (a) mean wind speed
(4 m/s). (b) mean wind speed (5 m/s). . . . . . . . . . . . . . . 134
4.1 Isolated micorgrid layout . . . . . . . . . . . . . . . . . . . . . . 140
4.2 Layout of the microgrid connected to the innite bus . . . . . . 141
4.3 RLC lter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.4 Voltage-frequency inverter . . . . . . . . . . . . . . . . . . . . . 145

viii List of gures
4.5 LCL lter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.6 Inverter power controller . . . . . . . . . . . . . . . . . . . . . . 148
4.7 Inverter voltage-current controller . . . . . . . . . . . . . . . . . 149
4.8 Reference frame transformation . . . . . . . . . . . . . . . . . . 150
4.9 DFIG power ow . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.10 Induction machine equivalent circuit . . . . . . . . . . . . . . . 153
4.11 Control scheme of the rotor-side converter: (a) Reactive power
controller. (b) Active power controller. . . . . . . . . . . . . . . 155
4.12 DC-link circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.13 DFIG reference frame: (a) DFIG reference frame transforma-
tion. (b) DFIG phase-locked loop circuit. . . . . . . . . . . . . . 158
4.14 Control scheme of the grid-side converter: (a) Control scheme
in the coordinate-d. (b) Control scheme in the coordinate-q . . . 159
4.15 Microgrid distribution line . . . . . . . . . . . . . . . . . . . . . 161
4.16 Microgrid static Load . . . . . . . . . . . . . . . . . . . . . . . . 162
4.17 Loci of eigenvalues of the isolated microgrid during wind variation-
case 1: (a) Magnetization current mode of the 1st DFIG imdq1 .

(b) DC-link mode of the 1st DFIG vdc1 andζdc1 . (c) Inverters
angle of rotation mode θpll . (d) DC-link mode of the 2nd, 3rd
and 4th DFIGs vdc2,3 andζdc2,3,4 . . . . . . . . . . . . . . . . . . . 169
4.18 Loci of eigenvalues of the isolated microgrid during wind variation-
case 2: (a) Magnetization current mode of the 1st DFIG imdq1 .

(b) Mechanical speed mode of the 2nd, 3rd and 4th DFIGs
wr2,3,4 . (c) DC-link mode of the 2nd and 4th DFIGs vdc2,4 . (d)
DC-link mode of the 1st DFIG ζdc1 , vdc1 andζdc1 . . . . . . . . . . 173
4.19 DC-link tension at low dc-link control gains versus the wind speed174
4.20 Loci of eigenvalues of the grid-connected mode during wind
variation-case 1: (a) DC-link mode of 1st DFIG vdc1 andζdc1 .

(b) DC-link mode of 2nd, 3rd and 4th DFIGs vdc2,3,4 . (c) Me-
chanical speed mode of 1st DFIG wr1 . (d) Mechanical speed
mode of 2nd, 3rd and 4th DFIGs ωr2,3,4 . . . . . . . . . . . . . . . 176
4.21 Loci of eigenvalues of the grid-connected mode during wind
variation-case 2: (a) DC-link mode of 1st DFIG ζdc1 andvdc1 .

(b) Magnetization current mode of 2nd and 3rd DFIGs imdq2,3 .
(c) DC-link mode of 2nd, 3rd and 4th DFIGs vdc2,4 . . . . . . . . 179
List of gures ix
4.22 Loci of eigenvalues of the isolated microgrid during inverters
droop increment-case 1: (a) Magnetization current mode of
DFIGs imdq1,2,3,4 . (b) Inverters angle of rotation modes δ2,3 . (c)
Integral term of inverters voltage controller modes Φvd1,2,3,4 . (d)
DC-link mode of DFIGs ζdc1,2 . . . . . . . . . . . . . . . . . . . . 185
4.23 Loci of eigenvalues of the isolated microgrid during inverters
droop increment-case 2: (a) Magnetization current mode of
DFIGs imdq1,2,3,4 . (b) Inverters angle of rotation modes δ2,3 . (c)
Inverters angle of rotation modes δ2 . (d) DC-link mode of the
3rd DFIG vdc3 , ζdc3 . . . . . . . . . . . . . . . . . . . . . . . . . . 189
4.24 Eect of inverters droop increment: (a) Inverter angle of rota-
tion. (b) Inverter power. (c) Frequency. . . . . . . . . . . . . . . 190
4.25 Loci of eigenvalues of the isolated microgrid during demand in-
crement at mean wind speed 7 m/s: (a) Magnetization current
mode of DFIGs imdq1,2,3,4 . (b) Magnetization current mode of
DFIGs imdq1,2,4 . (c) Inverters angle of rotation modes δ2,3 . (d)
Inverters angle of rotation modes δ2 . . . . . . . . . . . . . . . . 195
4.26 The rest of critical eigenvalues trajectories of the isolated mi-
crogrid during demand increment at mean wind speed 7 m/s:
(a) DC-link mode vdc2,4 andζdc2,4 . (b) DFIG mechanical speed
mode ωr2,3,4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
4.27 Loci of eigenvalues of the grid-connected mode during demand
increment at mean wind speed 7 m/s: (a)Magnetization current
mode of DFIGs imdq1,2,3,4 . (b) Magnetization current mode of
DFIGs imdq1,4 . (c) DFIGs integral terms of phased-locked loop
ζpll1,2,3,4 . (d) DC-link mode vdc1,2,3,4 . . . . . . . . . . . . . . . . . 199
4.28 The rest of critical eigenvalues trajectories of the grid-connected
mode during demand increment at mean wind speed 7 m/s: (a)
DC-link mode vdc1,2 . (b) DFIG mechanical speed mode ωr1,4 . . . 200
4.29 Loci of eigenvalues of the isolated microgrid during demand in-
crement at mean wind speed 12 m/s: (a) Magnetization current
mode of DFIGs imdq1,2,3,4 . (b) Inverters angle of rotation modes
δ2,3 . (c) Inverters angle of rotation mode δ2 . (d) DC-link mode
ζdc3,4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
x List of gures
4.30 The rest of critical eigenvalues trajectories of the isolated mi-
crogrid during demand increment at mean wind speed 12 m/s:
(a) DC-link mode ζdc1,4 . (b) DFIG mechanical speed mode ωr1,4 . 204
4.31 Loci of eigenvalues of the grid-connected mode during demand
increment at mean wind speed 12 m/s: (a) Magnetization cur-
rent mode of DFIGs imdq1,2,3,4 . (b) DFIGs magnetization current
mode of imdq1,4 . (c) DFIGs integral terms of phased-locked loop
mode ζpll1,2,3,4 . (d) DC-link mode vdc1,2,3,4 . . . . . . . . . . . . . . 207
4.32 The rest of critical eigenvalues trajectories of the grid-connected
mode during demand increment at mean wind speed 12 m/s: (a)
DC-link mode ζdc1,2,3,4 . (b) DFIG mechanical speed mode ωr1,2,3,4 . 208
4.33 Frequency vs demand increment . . . . . . . . . . . . . . . . . . 208
List of tables xi
List of Tables
2.1 Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.2 Microgrid Capacity Factor . . . . . . . . . . . . . . . . . . . . . 77
3.1 Wind Turbines Characteristics . . . . . . . . . . . . . . . . . . . 93
3.2 Items of Wind Turbine Capital Cost . . . . . . . . . . . . . . . 101
3.3 Enercon-600 kW Versus Enercon-900 kW at Mean Wind Speed:
4 m/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.4 Enercon-600 kW Versus Enercon-900 kW at Mean Wind Speed:
5 m/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.5 Characteristics of the Simulated Batteries . . . . . . . . . . . . 120
3.6 Economic Evaluation at Mean Wind Speed: 4m/s-Aggregation
Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.7 Economic Evaluation at Mean Wind Speed: 5m/s-Aggregation
Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.1 Dominant Modes of the Isolated MG During Wind Speed Increment-
Case 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
4.2 Dominant Modes of the Isolated MG During Wind Speed Increment-
Case 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.3 Dominant Modes of the Grid-Connected Mode During Wind
Speed Increment-Case 1 . . . . . . . . . . . . . . . . . . . . . . 175
4.4 Dominant Modes of the Grid-Connected Mode During Wind
Speed Increment-Case 2 . . . . . . . . . . . . . . . . . . . . . . 178
4.5 Dominant Modes of the Isolated MG During Inverters Droop
Constants Increment-Case 1 . . . . . . . . . . . . . . . . . . . . 184
4.6 Dominant Modes of the Isolated MG During Inverters Droop
Constants Increment-Case 2 . . . . . . . . . . . . . . . . . . . . 188

xii List of tables
4.7 Dominant Modes of the Isolated MG During Demand Increment
at Mean Wind Speed: 7 m/s . . . . . . . . . . . . . . . . . . . . 194
4.8 Dominant Modes of the Grid-Connected Mode During Demand
Increment at Mean Wind Speed: 7 m/s . . . . . . . . . . . . . . 198
4.9 Dominant Modes of the Isolated MG During Demand Increment
at Mean Wind Speed: 12 m/s . . . . . . . . . . . . . . . . . . . 202
4.10 Dominant Modes of the Grid-connected Mode During Demand
Increment at Mean Wind Speed: 12 m/s . . . . . . . . . . . . . 206

Nomenclature xiii
Nomenclature
Abbreviations
AGC Automatic generation control
AIC Akaike information criterion
ARMA autoregressive moving average time series model
AVR Automatic voltage regulator
BIC Bayesian information criterion
CCS Carbon capture and storage
CDF Cumulative distribution function
CF Capacity factor
CHP Combined heat and power
CIGRE International Council on Large Electric Systems
COE Cost of Energy
CPF Continuation power ow
DGs Distributed generators
DERs Distributed energy resources
DFIG doubly-fed induction generator

xiv Nomenclature
DNOs Distribution network operators
EMS Energy management system
ESS Energy storage systems
GSC grid side converter od doubly-fed induction generator
GRR Generator to rotor ratio
LHS Latin Hypercube Sampling
MCS Monte Carlo Sampling
MG Microgrid
MSUs Microgrid supporting units
NCP Nonlinear complementarity problem
NLP Nonlinear programming problem
NPV Net Present Value
NR Newton-Raphson
ODEs Ordinary dierential equations
PCC Point of common coupling
PDF Probability density function
P ll Phased locked loop
PI Proportional integral controller
p(w) Distribution probability of wind speed
RES Renewable energy sources
RSC rotor side converter od doubly-fed induction generator
SNB Saddle node bifurcation
SRU System reserve unit

Nomenclature xv
SPB Simple payback
SS Static switch
WECS Wind energy conversion systems
WTGs Wind turbine generators
Variables
AAR wind turbine average annual revenue
ACE area control error
AEPbat annual energy production of the battery
AEPwt annual energy production of the wind turbine
Cstbat energy stored in the battery bank
Cp wind turbine power coecient
Ebat Energy production of the battery
ei current error in current controller of droop inverter
ev voltage error in voltage controller of droop inverter
Ewt Energy production of the wind turbine
F (x) distribution function vertical axis
fCOE cost of energy function
FPg wind power cumulative distribution function
Fw wind speed cumulative distribution function
g algebraic variables vector
g(x) inequality constraints vector
h(x) equality constraints vector
ib input current to DC-link circuit

xvi Nomenclature
ic lter capacitor current
iCdc DC-link capacitor current
ig output current from doubly-fed induction generator to the
microgrid
ii inductance lter current
iline distribution line current
iload static load current
im doubly-fed induction generator magnetization current
io output lter current of droop inverter
ir doubly-fed induction generator rotor current
is doubly-fed induction generator stator current
K uncertain variables or parameters
n Sampling moment
N number of samples
P active operating inverter power
Pdemand demand power
Pg wind turbine power
Pg0 Nominal power of MSUs, where the frequency is nominal
Pgmax Nominal power of MSUs, where the frequency is nominal
PtEc battery power charge
PtEd battery power discharge
Ptie tie line power among dierent areas
Pwt instantaneous wind turbine power

Nomenclature xvii
p̃ instantaneous active power
Q reactive operating power
q̃ instantaneous reactive power
rr wind turbine rating ratio
rx,y cross-correlation between two variables
Te doubly-fed induction generator electromagnetic torque
Tm doubly-fed induction generator mechanical torque
u control variables vector
vc tension of capacitor lter
vdc , vb similar terms of DC-link tension
vg microgrid node tension
vi input lter tension of droop inverter
vL c coupling inductance voltage drop
vL f lter inductance voltage drop
vm doubly-fed induction generator magnetization circuit
tension
vo lter capacitor tension
vr doubly-fed induction generator rotor tension
vs doubly-fed induction generator stator tension
w instantaneous wind speed sample
wavg mean wind speed
x state variables vector
ximax maximum interval limit of distribution function

xviii Nomenclature
ximin minimum interval limit of distribution function
y output variables vector
δ angle between droop inverter synchronous frame and the
microgrid common synchronous frame
ζdc , ζib integral terms of dc-link circuit
ζi integral term of current controller state variable
ζib integral term of DC-link circuit
ζir integral term of doubly-fed induction generator rotor side
converter
ζpll integral term of phased-locked loop circuit
ζQ integral term of doubly-fed induction generator rotor side
converter state variable
θpll angle between doubly-fed induction generator synchronous
frame and the microgrid common synchronous frame
λi wind turbine blade tip speed ratio
µ average value
σ standard deviation
Φv integral term of voltage controller
Ψs doubly-fed induction generator stator ux linkage
ω operating frequency
ω0 Nominal value of frequency
ωpll doubly-fed induction generator pulsation frequency
ωr wind turbine mechanical speed

Nomenclature xix
Parameters
Cbat battery capital cost
Cbatstmax maximum energy stored
Cbatstmin minimum energy stored
Cf lter capacitor of droop inverter
Cwp initial capital cost of wind turbine
D wind turbine rotor diameter
F wind turbinefriction constant
F CR xed rate of charge
H wind turbine height
i discount rate
J wind turbine inertia constant
kii current controller integral gain
kiv voltage controller integral gain
kpi current controller proportional gain
kpv voltage controller proportional gain
L life time of wind turbines in years
LRC levelized replacement cost
Lc coupling inductance of droop inverter
Lf output lter inductance of droop inverter
Lline distribution line inductance
Lload static load inductance
Lm doubly-fed induction generator mutual inductance

xx Nomenclature
Lr doubly-fed induction generator rotor inductance
Ls doubly-fed induction generator stator inductance
mp active droop constant
nq reactive droop constant
O&M operation and maintenance cost
p doubly-fed induction generator poles number
PEcmax maximum charge rate
PEdmax maximum discharge rate
Pmax maximum real power of inverter droop
Prated rated power of wind turbine
Qmax maximum reactive power of inverter droop
Rc internal resistance of coupling inductance of droop inverter
Rf internal resistance of output lter inductance of droop
inverter
Rline resistance of distribution line
Rload static load resistance
Rr doubly-fed induction generator rotor resistance
Rs doubly-fed induction generator stator resistance
U Ce : Unit cost of input electricity for charging batteries
Vmax maximum voltage
wf wind turbine furling speed
wi wind turbine cut-in speed
wn wind turbine blending speed

Nomenclature xxi
wr wind turbine rated speed
Xm doubly-fed induction generator mutual reactance
ηc charge eciency
ηd discharge eciency
ϑ model parameter
ωmax maximum frequency
Matrices
A open loop state variables matrix
Ac closed loop state variables matrix
B control matrix
C output matrix
D feed-forward matrix
j1A state variables coecients matrix of state variables equations
j2A algebraic variables coecients matrix of state variables
equations
j1C state variables coecients matrix of output variables
equations
j2C algebraic variables coecients matrix of output variables
equations
j3 state variables coecients matrix of algebraic variables
equations
j4 algebraic variables coecients matrix of algebraic variables
equations
xxii Nomenclature
k system gains matrix
ϕ right eigenvector
ψ left eigenvector
Operators and reference systems

dq synchronous reference related to each droop inverter
DQ common synchronous reference of the microgrid
s laplace operator
∆ linearized variables operator
∆p synchronous reference dierential operator
∇ gradient vector
Superscripts and subscripts

∗ superscript indicates to reference variable
dq subscript means that the variable is referred to droop
inverter synchronous reference
DQ subscript means that the variable is referred to microgrid
common synchronous reference
i, j subscripts indicates to microgrid elements of the same type,
such as DFIGs but related to nodes i and j

Introduction 1
Introduction
As aordable and adequate energy is considered the main path towards eco-
nomic prosperity, by 21th century the EU countries are confronted with mov-
ing from centralized conventional generation methods, which depend mainly
on fossil fuels, to another trend of generation based on distributed generation.
Fossil fuels release damaging gases, the combustion of such sources creates
some acids as sulfuric and carbonic, which rise into the atmosphere and return
as acid rain, and subsequently damages soil and water. In addition, these re-
leased gases cause rapid heating and adverse eects on climate change.
According to the latest available information from EUROSTAT in 2005,
thermal power plants, fuelled mostly by fossil fuels, accounted for 54% of the
total installed capacity in EU, followed by hydro, nuclear generation and other
sustainable sources. As a result of the latest EUROSTAT statistics, some
important issues need to be addressed, such as global warming; average tem-
perature rose by 0.6C, because conventional generation fuelled by fossil fuels
are the main emitters of carbon dioxide (CO2 ) that is the basic contributor
to global warming. Another issue regarding security of energy supply, as EU
imports 50% of its energy requirements and the dependency on imports is to
be increased to 70% by 2030. More drawbacks of the centralized conventional
generation can be briefed in the following points: (1) The transmission and
distribution costs amount up to 30% of the delivered electricity cost, due to
high line transmission and conversion losses. (2) It is required to upgrade
the transmission and distribution networks by huge investment. (3) The cen-
tralized generation can oer adequacy mostly, but cannot ensure the required
power quality for certain network areas.
Several studies have been conducted to ensure the main shortfalls of the
centralized generation and to detail the motivation of the customers towards
a new architecture of generation which is distributed generation as a primary
source of electricity or as a backup generation source.
In order to overcome such drawbacks, the European Commission proposed
a new energy policy in January 2007 based on distributed generation, for re-
duction greenhouse gas emission, improving power quality and increase the
contribution of sustainable energy sources to 20%. Figure 1 shows the growing

2 Introduction
dependency of EU electricity generation on fossil fuel resources, followed by
nuclear energy, wind energy and other renewable energy sources (RES), ac-
cording to EUROSTAT.
In order to reduce the dependency on centralized fossil fuel generation
sources, improve power quality and increase the contribution of sustainable
energy sources, the study and analyses of the new developed generation trend
which depends on distributed energy sources are necessary.
This thesis concentrates on a new architecture of generation known as Mi-
crogrid (MG), which is supplied by distributed generators (DGs). In late 1990s,
the major issues related to the distributed generation have been extensively in-
vestigated by the working groups of the International Council on Large Electric
Systems (CIGRE). The new era of generation is under continuous investiga-
tion to overcome the negative sides of the centralized power utilities, in order
to enhance the electric and thermal energy, optimize power quality, minimize
pollutant deposition and contributes to ancillary services.
The main objective of the thesis is to conduct several technical and eco-
nomic analyses in such new architectures of generation, such as providing an
insight about site matching suitability for droop-regulated microgrids supplied
by wind generators, optimizing both energy production and economic perfor-
mance for isolated microgrids supplied by wind energy and dierent energy
storage systems and nally predicting some stability constraints for genera-
tion sources by small signal stability analysis.
This thesis is divided into ve chapters; Chapter 1 shows the state of the
art, to clear the main dierence of operation and control between the conven-
tional power utility grid and microgrids; moreover, the problems of supplying
microgrids by non-dispatchable generation as wind energy are highlighted, to
illustrate the main challenges of supplying microgrids by such intermittent
energy sources. Finally, thesis contributions are listed, and comparisons be-
tween thesis contributions and other recent researches regarding microgrids are
shown. In Chapter 2, capacity factor analyses of droop-regulated microgrids
are performed based on a reliable wind speed sampling strategy, by reduc-
ing the variance of wind speed samples, and also including the correlations
among wind speeds. In Chapter 3, the idea of site matching suitability of
droop-regulated microgrids supplied by wind generators is more detailed, by
considering both energy production and economic performance of isolated mi-

Introduction 3
crogrids supplied by wind energy and dierent energy storage systems. The
optimization function in this Chapter is conducted to minimize the cost of
energy, which is a function of both capital costs and energy production of
the generation sources. In Chapter 4, microgrid stability analyses are con-
ducted by the small signal stability criteria. The analyses are carried out
to illustrate the stability margin and to determine the critical eigenvalues of
isolated microgrids supplied by doubly-fed induction generators and inverter-
interfaced distributed generators, when applying common actions to share the
power among generation sources, and also through dierent wind speed levels.
Finally, the conclusions of the thesis are listed in Chapter 5.
4500
nuclear
4000 other RES
wind
3500 fossil
3000
electricity (TWh)
2500
2000
1500
1000
500
0
2005 2007 2010 2012 2015 2017
year
Figure 1: EU electricity generation by dierent energy

resources
4 Introduction
Chapter 1 5
Chapter 1
State of the Art
1.1 Introduction
The world's rst power station was built in the Bavarian town of Ettal and
went into operation in 1878. The rst public power station was the Edison
Electric Light Station, built in London, which started operation in January
1882. In September 1882, the Pearl Street Station was established by Edison
to provide electric lighting. Through the rst few decades of the 20th century,
central stations became larger, using higher steam pressures to provide higher
eciency. Interconnection by hydroelectric power was allowed to be in service
by 1906; however, a great generation capacity expansion has been achieved
through steam turbines in central stations, as they have played the primary
role in the generation level. Generators were no longer limited by slow speed
of reciprocating engines and could grow to greater sizes, represented mainly by
Synchronous Generators, which are considered the main sources of generation

level in centralized power utilities.
As the synchronous generators play the main role in the generation level of
the centralized power utility grids, they can be classied as either high-speed
generators, driven by steam or gas turbines called turbogenerators, or low-

speed generators driven by water turbines hydro generators. Both mentioned
generators have two main magnetic parts termed as the stator and rotor; both
of them are manufactured from magnetic steels. The armature winding car-
ries the load current and supplies power to the system. The rotor excitation
winding is supplied by a direct current to produce a rotating magnetic ux.
This rotating magnetic ux induces an electromotive force in each phase of the
6 Chapter 1
three-phase stator armature winding which forces alternating current to ow
out to the power system. Typically, the operation of synchronous generators
depends mainly on the electromechanical interaction between the stator and
rotor circuits. Thus, the rotor plays an important rule for power control and
stability. The details regarding control and stability of synchronous generators
are beyond the scope of this thesis. For more details, it is recommended to
refer to [2] and [3].
Through the last decades, the conventional power systems have faced some
problems related to gradual depletion of fossil fuel resources, poor energy e-
ciency, environmental pollution, low power quality and power market monopoly
policies. Firstly, the transmission and distribution costs amount up to 30% of
the delivered electricity cost on average. Secondly, it is required to upgrade the
transmission and distribution networks by huge investments [4]. Thirdly, the
centralized generation can supply the energy requirements of the consumers
at all times, but cannot ensure the required power quality for certain network
areas [4]. Fourthly, the catastrophic environmental impact of the centralized
generation is clear due to the dependence on fossil fuels [5].
These problems have led to a new trend of locally power generation at
distribution voltage level by using non-conventional/renewable energy sources
like natural gas, biogas, wind power, solar photovoltaic cells, fuel cells, com-
bined heat and power (CHP) systems, microturbines and stirling engines. This
type of power generation is termed as distributed generation (DG) and the en-
ergy sources are termed as distributed energy resources (DERs). The term
distributed generation has been devised to distinguish this concept of genera-
tion from centralized conventional generation. Distribution networks become
active with the integration of DGs, subsequently the are termed as active dis-
tribution networks.
In contrast to conventional power systems, the distributed generation sourc-
es have proved more positive impacts over the centralized generation. Firstly,
the distributed generation can alleviate the loading of the transmission and
distribution. Consequently, the network transmission and distribution losses
can be minimized, which leads to lower energy taris. Secondly, the distributed
generation can be employed to locally control the power quality according to
customer requirements even on an hourly basis. Thirdly, the distributed gener-
ation can reduce system expansion costs of the main power utility by limiting
Chapter 1 7
the utility transmission capacity expansion and oering the best utilization
of the current transmission capacity. Fourthly, the distributed generation can
eectively reduce the emissions and environmental warming, due to high de-
pendency on renewable energy sources. Fifthly, the distributed generation is
capable to restart its generation after a total system collapse without import-
ing any external power to restore, at least, a portion of the power system to
service after the collapse (black start capability).
The operational dierences between conventional generation and distribut-
ed generation can be claried by some main features, such as DGs have small
generation capacities, DGs are located near to customers and DGs are inter-
faced to networks via electronic inverters to synthesize their productions.
In order to reduce the dependency on centralized fossil fuel generation
sources, improve power quality and increase the contribution of sustainable
energy sources, this thesis concentrates on a new architecture for generation
known as Microgrid (MG), which is supplied by distributed generators. Mi-
crogrids are considered as clusters of distributed energy sources, loads and
controls organized to deliver the optimum energy service [6].
The proposed microgrid models in this thesis are supplied mainly by DGs
interfaced to the microgrids via electronic inverters. Moreover, intermitted
energy sources which are wind energy connected to the grids directly with-
out interfaces. Thus, the generation supply depends on inverter-interfaced
distributed generators and intermittent energy sources, unlike the centralized
generation. The droop control strategy is adopted in the thesis, to enhance
real-reactive power exchange among the DGs interfaced by inverters. The
droop control uses only local power to detect changes in the system and adjust
the operating points of the generators accordingly.
The main objective of this thesis is to conduct several technical and eco-
nomic analyses in microgrids; therefore, some pivotal aspects are illustrated,
such as providing an insight about site matching suitability of droop-regulated
microgrids supplied by wind generators, optimizing both the energy production
and economic performance of isolated microgrids supplied by wind energy and
dierent energy storage systems and nally predicting the stability constraints
of the generation sources by small signal stability analysis. The following text
outlines the main objectives of the thesis.
A conventional generation unit is capable of generating rated power during

8 Chapter 1
normal operation. Typically, the capacity factor of a conventional generator is
its actual average generation as a function of its rated power; the capacity fac-
tor can be dened as the ratio of the actual energy produced from a generation
unit, in a given period, to its potential output at full capacity. Thus, the ca-
pacity factor of a conventional generator is dependent only on power demand.
On the other hand, for intermittent wind energy sources, the capacity factor
depends on the wind speed probability distribution and on the characteristics
of wind turbines, and so the determination of the wind generation capacity,
which can share supply microgrids with other distributed generators, is rela-
tively complex. Moreover, the capacity factor for a droop-regulated microgrid
cannot be conducted by considering the microgrid as a large generator that
is built up from several minor wind generating units, because droop control
imposes restrictions on the maximum loadability. Thus, the thesis conducts a
capacity factor analysis of droop-regulated microgrids that serves to evaluate
the site-matching suitability of a combination of wind generators and wind
speeds, by an advanced technique, known as Latin Hypercube (LHS), based on
[7], for accurate wind speed sampling.
In conventional power grids, the uncertainties of system reserves are e-
ciently reduced and well-settled. Although, the conventional generation suers
from the volatility of conventional energy prices, their variability is still much
lower than uncertainty of the wind power. In contrast, in the case of microgrids
supplied by wind energy, some doubts emerge regarding the impact of wind
variability on the generation capacity and economic considerations. The char-
acteristics of such grids require scheduling more reserve for ensuring adequate
security and reliability levels, but the higher reserve requirements may sub-
stantially deteriorate the economy of these supply systems. Usually, electric
energy storage systems, including dierent types as batteries, capacitors and
ywheels are required to supplement the wind generators, during the moments
of demand shortfall. Hence, the energy storage systems as batteries can handle
such problems by managing the power control during the transient moments.
For this purpose, the thesis introduces a study based on optimizing both en-
ergy production and economic performance of isolated microgrids, supplied by
wind energy and batteries. Hence, the maximum possible energy production
at the lowest energy cost can be claried.
In conventional generation systems, the possibility of system restoration

Chapter 1 9
and reducing power swings among dierent generation units is high, in case of
disturbances, as an adequate reserve can be supplied from neighboring gener-
ation units to restore the operation. Moreover, in the conventional generation
systems, some standards and recovery strategies from power imbalances have
been set and handled from decades. In addition, the stability analysis of
conventional generation has been conducted in many studies, and the most
dominant variables which can violate the stability of networks are well-known.
In contrast, the situation is dierent in the case of isolated microgrids, as the
system reserve and capabilities of the whole system to recover from distur-
bances are limited, especially when the microgrid is supplied by intermittent
sources as wind energy. Moreover, microgrids suer from lack of standards to
overcome disturbance and power swing problems, as it is considered as a new
architecture of generation. Therefore, the thesis introduces the small signal
stability analysis of isolated microgrids, supplied by wind energy and also gen-
eration sources interfaced via inverters, under dierent scenarios. Thus, the
dominant variables which violate the microgrid stability can be recognized,
and so in future work, an appropriate control strategy can be developed to
keep the microgrid stable.
This chapter shows the state of the art to clear the main dierences of oper-
ation and control, between the conventional power utility grid and microgrids
supplied by distributed generators. In addition, the problems of supplying mi-
crogrids by non-dispatchable generators as wind energy are studied, to unfold
the challenges to microgrids development.
This Chapter is divided into nine sections. In Sections 1.2-1.5.2, the con-
cepts of microgrids operation and control are introduced, to reveal the opera-
tional dierences between microgrids and centralized power utilities. In Section
1.5.3, the integration of non-dispatchable generators to microgrids is explained,
to illustrate the challenges faced by microgrids, supplied by non-dispatchable
energy sources. In Section 1.6, the topic of capacity factor analysis of droop-
regulated microgrids is addressed; this topic is studied through Chapter 2.
Section 1.7 shows the necessity of optimizing both energy production and eco-
nomic performance in isolated microgrids; this topic is detailed in Chapter 3.
Section 1.8 illustrates the stability analysis of isolated microgrids, supplied by
non-dispatchable energy sources, as microgrid stability analysis is performed
in Chapter 4. Finally, Section 1.9 illustrates the thesis contributions.

10 Chapter 1
Through this thesis, the DGs are employed in microgrids either in the iso-
lated microgrid mode or grid-connected mode. Clearly, the modelling and
dynamics of DGs are far from synchronous generation units; hence, the men-
tality and concept of microgrids operation must be detailed to explain how
microgrids can benet from the diversity of generation sources, in order to
optimize their operations and overcome problems arisen by conventional gen-
eration units.
1.2 Operational Concepts of Microgrids

Electricity networks are in the era of major transition from stable passive
distribution networks with unidirectional electricity transportation, to active
distribution networks with bidirectional electricity transportation. Distribu-
tion networks without any DG units are passive since the electrical power is
supplied by the main utilities to the customers embedded in the distribution
networks. The grid becomes active when DG units are added to the distribu-
tion system leading to bidirectional power ow in the networks [8].
The present t-and-forget operational strategy of DG units must be changed

for management of active networks. Active networks incorporate DGs in distri-
bution networks and demand side management, by motivating the distribution
network operators (DNOs) towards better asset utilization and management
by deferral of replacement of age-old assets. The application of active network
management can greatly support more connections of DGs as compared to
passive networks.
A microgrid is essentially an active distribution network, because it is an
aggregation of DG systems and dierent loads at distribution voltage level.
The microgrid generators (microsources) are usually integrated renewable/non-
conventional DERs to generate power at distribution voltage [9]. Such genera-
tors must be equipped with power electronic interfaces to synchronize all DERs
together and provide the required power quality [10]. This control exibility
allows the microgrid to present itself to the main utility power system as a
single controlled unit that meets local energy needs for reliability and security.
Figure 1.1 shows the energy ow for both a centralized conventional grid
and a microgrid supplied by distributed resources, also the gure shows the sin-
gle point of connection to the utility called point of common coupling (PCC);
Chapter 1 11
moreover, the static switch (SS) is responsible for isolating or connecting the
microgrid from/to the utility power grid. Basically, the key dierences be-
tween the microgrid and the conventional power grid can be accounted in the
following points: (1) The small capacity of microsources in respect to power
generators in conventional networks. (2) Microsources can feed power directly
to distribution networks. (3) Microsources are installed close to customers,
and so the optimum local power quality can be ensured. (4) The operation
of generators employed in conventional power utilities depend mainly on elec-
tromechanical interaction, between the stator and rotor circuits of those gen-
erators, whereas most DERs depend mainly on power electronic interfaces, to
synthesize their productions and ensure stable level of power exchange.
The operation of a microgrid in dierent modes is controlled through a local
controller of each microsource and a central controller for the whole microgrid.
The main function of the local microsource controller is to independently con-
trol the power ow and load voltage prole of the microsource in response to
any disturbance and load change, without any communication with the central
controller, whereas the main function of the central controller is to perform
protection coordination and provides the power dispatch and voltage set points
for all the controllers of the microsources [11].
1.3 Impacts of Microgrids

Microgrids have remarkable impacts on existing electricity and gas markets.
In order to harness their impacts, their market participation must be encour-
aged. Suitable market reforms must be made to allow such participation and
nancial incentives must be provided for owners to invest in microgrids. Major
changes in conventional electricity markets have already been initiated in some
countries. Once market participation is assured, there is a vast opportunity for
microgrids to supply quality service to the main utility distribution system.
Microgrids can naturally provide signicant ancillary services to the utility.
The impacts of microgrid can be summarized in the following items:
• Heat Utilisation: By installing CHP microsources near to customer prem-
ises, then CHP can operate at very high energy eciencies and the gen-
erated heat can be eectively used to supply local heat loads [5];
12 Chapter 1
CENTRALIZED GENERATION
TRANSMISSION
ENERGY FLOW
DISTRIBUTION
SS PCC
CUSTOMERS
DISTRIBUTED GENERATION
DISTRIBUTION
ENERGY FLOW
Figure 1.1: Centralized vs distributed generation
• Power Quality: Microgrids can be employed to control the power quality
locally according to customers requirements even on an hourly basis.
It is quite dicult for a utility grid to be attentive to special power

Chapter 1 13
quality needs of a particular customer. The conventional utility grid
normally provides a uniform level of quality and service to all customers,
within a given area and does not easily dierentiate among their needs.
Microgrid power quality services can be accounted in some items, such
as reactive power and voltage control, supply of reserves, regulation and
load following [12];
• Optimizing Utility Expansion: Microgrids normally reduce system ex-
pansion costs of the main power utility, by limiting the utility transmis-
sion capacity expansion and oering the best utilization of current trans-
mission capacity, and consequently the customers benet well from this
impact, as this cost is realized from the tari imposed on them. Without
this benet, the customers have to pay heavily for power supply due to
interregional transmission congestion;
• Environmental Impact: Low-carbon microsources can eectively reduce
emissions and environmental warming. This is one of the major criteria
to support the microgrid operation, as the central controller of a micro-
grid is capable to be programmed for operational decisions based on the
lowest net emission production [4];
• Black Start Capability: The black start is dened as the capability of
a power system to restart its generation after a total system collapse,
without importing any external power to restore at least portion of the
power system to service again after the collapse. The microgrid black
start involves a sequence of control actions, dened through a set of rules
and conditions to be checked during the restoration stage. These rules
are identied in advance and embedded in the software of any microgrid
central controller.
For the best utilization of microgrid benets, a microgrid may be employed
by one of two modes. Firstly, the isolated mode, where the microgrid is fully
isolated from the power utility grid. Secondly, the grid-connected mode, where
the microgrid is connected to the power utility grid.
The operational modes of a microgrid are more detailed hereunder:
1. Stand-Alone: The microgrid may be isolated due to faults, maintenance
purposes or as predetermined control sequence. In the stand-alone mode,

14 Chapter 1
there is no connection to the host grid. Thus, the DERs control is respon-
sible for frequency regulation. In addition, a voltage regulation strategy
is required to maintain the network voltage. Moreover, the DERs control
is also responsible for managing the real and reactive power contribution
of DERs, such that load power is shared among dierent DERs in the
microgrid;
2. Grid-Connected: In the grid-connected mode, the magnitude and fre-
quency of the system terminal voltages are imposed by the utility grid.
Thus, the control task can be summarized as the regulation of the real
and reactive power is undertaken through the exchange among the DERs
and the host network.
From the previous mentioned microgrid modes, it is clear that the power
electronic interface of each microsource plays the main role to synchronize the
microgrid sources in the both operational modes. In addition, the presence
of DGs introduce an additional supplying nodes and change the power ow
in microgrids. The presence of DGs directly aect voltage proles along feed-
ers by changing the direction and the magnitude of real and reactive power
ows. The directional characteristics of voltage regulation circuitry must also
be considered. Up to now, it is evident that the main dierence between the
conventional centralized grid and the microgrid emerges from two points:
• Power ow analysis;
• Power exhange control concept, in order to exchange real-reactive power
and to synchronize dierent microsources.
Therefore, it is required to illustrate the classical distribution power ow
models; on the other hand, some solving techniques related to microgrid power
ow models. Furthermore, the control concepts of microsources which depend
mainly on power electronic interfaces.
1.4 Power Flow in Microgrids

The rst requirement for planning, operation, control, protection and man-
agement of microgrids is the availability of accurate power-ow analysis results.
The conventional power-ow analysis based on the system positive-sequence

Chapter 1 15
representation, which is widely used for large power transmission systems is
not directly applicable to microgrids, for some reasons can be termed in the
following points: (1) The presence of unbalanced three-phase loads, single and
two-phase loads. (2) The presence of microsources with electronically coupled
interfaces, which are managed by various control strategies. (3) The presence of
dierent types of three-phase rotating machines related to microsources, with
various control strategies. (4) The presence of non-dispatchable microsources,
as wind turbines and photovoltaic generators.
The power ow software tools, which have been developed for large-system
are not tailored for the microgrid analyses, because they lack the exibility to
accommodate DER models and operating characteristics, particularly those of
electronically-coupled DER units. In addition, another limitation is imposed
on the steady-state modelling of electronically-coupled microsources, as the
available models assume only positive-sequence microsources representation
for power-ow analysis.
Some authors have developed new updates to power ow theories, in order
to accommodate new features of the microgrid concept which can consider the
full sequence modelling of power electronic interfaces.
In [13] Arturo proposed an object-oriented power ow analysis, to include
dierent DGs models; numerical applications were presented to evidence the
features of the algorithm and the exibility of the software tool, to overcome
the drawbacks of the classical Newton-Raphson algorithm for power ow anal-
ysis.
In [14] Hassan proposed a two-step power ow analysis approach that can
represent electronically coupled DG units accurately, which is capable to solve
for the internal variables of each DG unit in a non-iterative fashion based on
the presented closed-form DER models.
In [15] Zakaria provided a three-phase power-ow algorithm in the sequence-
component frame for a microgrid, which can accommodate single-phase later-
als, unbalanced loads and three/four-wire distribution lines. The author con-
sidered the steady state sequence component frame models of DGs for the
developed power ow approach, under balanced/unbalanced conditions.
In [16] Zhang showed a capable algorithm, to cope with the disadvantages
of classical power ow models, which do not account for the unbalances. The
author introduced a decomposition model based on the sequence components

16 Chapter 1
frame. The three-phase unbalanced power ow was decomposed into three
separate subproblems with weak mutuality. The subproblem corresponding
to the positive sequence component was solved by using a Newton-Raphson
iterative scheme, while the other two subproblems were formulated into two
sets of linear simultaneous equations.
The proposed model of the microgrid power ow in this thesis accounts
for dissimilarities between single and three phase loads, also considers the
full dynamic modelling of loads, non-dispatchable generators. The proposed
power ow is based on decomposing the electrical phasor quantities into direct-
quadrature components. All loads and generators are referred to one master
source of generation. This master is the innite bus, in case of the grid con-
nected mode, while the master is considered to be a certain microsource, in
the microgrid isolated mode. The full details of the microgrid dynamic model
are shown in Chapter 4.
After setting the microgrid power ow model, which is proved as totally
dierent from the power ow models of centralized utility grids, it is necessary
to move to another important point that is the control concept of real-reactive
power exchange among microsources or between microsources and utility grids.
For this sake, a special attention must be paid to the microsource electronic
interface algorithm to handle the power exchange in the microgrid.
1.5 The Interface Control of Microgrids

When a microgrid is connected to a utility grid, then the utility grid is the
only responsible for determining the dynamics of the system through the direct
interface of synchronous generators. In contrast, during the isolated microgrid
mode, the power electronic interfaces and their sequences of control are the
only responsible for determining the dynamics of the microgrid, by controlling
the power production and synchronizing DGs with the microgrid.
1.5.1 Power Electronic Interface
The power electronic interface has some characteristics can be termed in the
following points: (1) The electronic interface contains the necessary circuitry
to convert power from one form to another one. This may include both a rec-
Chapter 1 17
tier and an inverter or just an inverter. (2) The power electronic interface is
compatible in voltage and frequency with the electric power system to which
it will be connected to and contain the necessary output lters.
Typically, the electronic interfaces are comprised of four basic components,
such as semiconductor switches, control system for switch gating, inductive
components and capacitive components. The inductive and capacitive compo-
nents are used to dynamically store energy for circuit power ow dampening,
ltering and transformation. By using electronic interfaces, the power quality
of DGs improves through the ability to control harmonic contents of the out-
put voltage and current; moreover, to provide fast operation to switch modes
between utility connected and stand-alone modes.
The electronic interfaces are self-commutated and can produce an AC-
voltage of arbitrary amplitude and phase; therefore, this allows the DGs to
produce any power amount at any power factor. The electronic interfaces
have a wider operating power factor range than synchronous generators. This
can be an extremely useful, if the DGs are allowed to regulate voltage and/or
supply reactive power to the system, whereas the voltage regulation in con-
ventional power systems is normally maintained using active-reactive power
capability of synchronous generators, load-tap-changing transformers at sub-
stations and line-voltage regulators.
Usually, microgrids combine capabilities of several DGs technologies and
can be extremely benecial when integrating dierent types of DGs together;
as an example, energy storage is often integrated with photovoltaic generators
or wind systems, to provide power when there is no solar or wind resource
available. Typically, there is a DC bus within the electronic interface that can
be used to integrate a variety of DGs technologies at a common point. The use
of a common DC bus can also allow the integration of several DGs technologies
inverters. The employed control strategies of microgrid power inverters can be
divided into two dierent strategies:
1. PQ-mode : The inverter only injects the power available at its input into
the grid. Usually, this mode is applicable to non-dispatchable generators
employed to track the maximum power point of sustainable sources.
2. Vf-mode : The inverter is controlled to feed the grid with reference values
of voltage and frequency, according to a specic control strategy that is

18 Chapter 1
similar to the behaviour of a synchronous machine.
Usually, distributed generators are interfaced to a microgrid by electronic
inverters, unlike synchronous generators, where their inertias contribute di-
rectly to conventional power utility, and so their responses contribute to the
system stability. In the microgrid case, the employed PQ-mode distributed
generators are usually interfaced to the microgrid by inverters and their in-
ertias are far from contributing to power system dynamics, in order to track
the maximum available power from the sustainable sources. However, there
is a trend to treat this problem of non-dispatchable generators, which operate
through the PQ-mode as wind turbines, by deloading their capacities. Thus,
some kinetic energy can be stored in the rotating parts, as the drive train part
of the wind turbine, in order to be used during voltage and frequency dips.
Consequently, this stored energy can contribute to improving the grid stability,
by restoring reference values of the voltage and frequency of the grid.
Some authors have discussed this important point. In [17] Morren proposed
a control strategy for wind turbines, to turn them more sensitive to power sys-
tem frequency dip, by means of an additional controller that adapts the torque
set point as a function of the grid frequency deviation. The emulated inertia is
proportional to the controller constant, and this loop is activated when the grid
frequency exceeds certain frequency deviation limits. In [18] Rogério adopted
another control strategy of a wind generator to supply the primary frequency
regulation by a mechanical strategy, which utilized the pitch control, by ad-
justing the rotor speed and active power according to the deloaded optimum
power extraction curve. In [19] Anaya showed a controller which can dynami-
cally manipulate the position of a wind turbine rotor, to slow down the wind
generator, allowing for a temporary power surge in case of a frequency dip.
However, The option of deloading the operating point of a wind turbine is
temporary option, as it can be employed for the moment of frequency dip only.
Otherwise, the wind turbine will suer from stalling as it loses all its kinetic
energy. The main strategy which must be adopted to control the dispatch-
able generators and even non-dispatchable generators interfaced by electronic
inverters is called droop control. The droop control strategy can enforce all
distributed generators to participate in a system restoration in case of distur-
bances, and also to enhance real-reactive power exchange in a microgrid.

Chapter 1 19
1.5.2 Droop Control in Microgrids
Microgrid distributed generators can be assigned to regulate the voltage
and frequency at the point of common coupling. Usually, there are multiple
DGs operating in parallel and feeding the microgrid, and so DGs must partici-
pate in the grid stabilization and voltage regulation. The frequency-droop and
voltage-droop control strategies are used to manage the real-reactive power
share in a microgrid among various generation sources.
In case of a frequency or voltage deviation from nominal values, then frequency-
droop P −ω and voltage-droop v−Q must be activated. This is achieved by
dynamically changing the power sharing levels to set the frequency and voltage
at new values.
Some authors discussed this important point. In [20] Lopes showed the
feasibility of the droop control strategy for an isolated microgrid, by assigning
the microgrid operation to single/multi master operations.
In [21] Milan proposed wave form quality control, in terms of harmonics of
an isolated microgrid loaded by unbalanced and non-linear loads. This duty
is shared equally among dierent DGs. The proposed control separates the
tasks in the frequency domain, as power sharing and voltage regulation are
controlled by the droop strategy.
In [22] Moreira showed the capability of electronic interfaces to perform a
black start, by sequence of actions as the interfaces are controlled by the droop
strategy.
In [23] Barklund introduced an energy management strategy for a droop-
controlled isolated microgrid, which successfully implements optimal genera-
tion dispatch levels by selecting the droop constants from a region where a
stable operation of the microgrid can be guaranteed.
In [24] Seon proposed power exchange among dierent DGs by droop reg-
ulation based on feeder ow control.
The proposed control among DGs in this thesis is based on the droop
strategy without any need of communication among DGs. Therefore, connect-
ing/disconnecting any DG from/to the microgrid without changing the con-
trol algorithm is possible; moreover, the droop control is handled by Fischer-

Burmeister algorithm, which reduces the complexity of problem solving, by for-
mulating the problem as a complementarity problem, through substituting the
20 Chapter 1
piecewise-dened droop function by an only scalar function that makes com-
putations of operating points of microgrids to be solved by means of Newton-
Raphson-like methods. The algorithm is detailed in Chapter 3.
At this point, the main operational concepts of microgrids were illustrated
and the dierences between microgrids and conventional grids were detailed.
However, the full scope of integrating non-dispatchable generators as wind
turbines in microgrids is not clear so far. One of the benets in a microgrid
strategy is to fully benet from sustainable sources, as a trial to reduce carbon
and toxic emissions. Some vital terms as generation capacity, economic perfor-
mance and stability of such non-dispatachable generators must be taken into
consideration. Next Section will explain the challenges faced by microgrids,
when being supplied by non-dispatchable generators.
1.5.3 Microgrids Supplied by Non-Dispatchable Genera-
tors
Introduction of generation based on renewable sources becomes an impor-
tant topic at present; however, these renewable resources are usually given by
very uctuating and uncontrollable stochastic parameters. These uctuations
can lead to severe problems, such as system frequency oscillations and/or vio-
lations of power lines capability. The generation patterns resulting from these
renewable sources may have some similarities with the electricity demand pat-
terns, but they usually record deviations from the demand patterns. Conse-
quently, a supplemental production is required to keep the demand and supply
in balance.
In [22] Moreira showed that during the black start, the connection of un-
controllable generators, as photovoltaic and wind generators must be during
the last step of the black start in order to smooth voltage and frequency vari-
ations, due to power uctuations of the uncontrollable generators.
In [25] Emanuel showed a critical case may occur, as a load demand may
be lower than the available power from photovoltaic solar arrays, and sub-
sequently the battery bank can be overcharged with unrecoverable damage
consequences. Therefore, he proposed a control algorithm for ecient battery
charging based on the microgrid line frequency.
Albert in [26] proposed a methodology to improve the reliability of non-

Chapter 1 21
dispatchable generators based on predicting consumers demands for a day
ahead, then processing an objective function to enhance the power quality for
all consumers.
In [27] Marcelo highlighted the possibility of overcoming wind uctuation
by a superconducting magnetic energy storage system, which can share the
demand supply of a microgrid with wind generators; the results show a capa-
bility to meet the demand peaks without any shortfall.
In [28] Changhee proved that synchronization in an isolated microgrid is
disturbed by wind turbines uctuations.
Alba in [29] proposed a method for tracking the deviations between wind
uctuations and load demands. Thus, a set of batteries were installed to
backup such deviations of real-reactive power through electronic interfaces with
a microgrid.
Obviously, the major dierence of generation between the wind power and
conventional power is related to the intermittent nature of the wind. The ef-
fectiveness of the wind power control is highly dependent upon the available
generation margin, which is decided from the available wind speed. In order
to have an accurate margin assessment in hand, an accurate capacity factor
analysis of wind energy based on ecient wind speed sampling can play a key
role.
As mentioned from the last Section, the rst objective of this thesis is to
provide an insight about site matching suitability of droop-regulated micro-
grids supplied by wind generators. For this reason, capacity factor analyses of
droop-regulated microgrids are carried out in Chapter 2, based on a reliable
wind speed sampling strategy, by reducing the variance of wind speed sam-
ples, and also including the correlations among wind speeds. Thus, each wind
energy source status can be determined accurately.
Most researches related to wind power have used Monte Carlo for wind
speed sampling, taking into consideration that Monte Carlo technique accu-
mulates the samples in certain areas of the wind speed probability domain.
Moreover, this technique is time consuming; thus, a more advanced technique
is employed in Chapter 2, which is Latin Hypercube, for accurate wind speed
sampling. This technique is capable to spread wind speed samples equally
overall the wind speed probability domain. Consequently, a better determina-
tion of wind power can be ensured.

22 Chapter 1
Usually, the increase of wind energy penetration requires also attention to
reserve requirements. For isolated microgrids, sources of reserve are always
required, to share microgrids supply with intermittent energy sources, as gen-
eration must always be as close as possible to system loads. The problem of
keeping the power balance is still more dicult for stand-alone microgrids,
supplied by intermittent generators. The characteristics of such grids require
scheduling more reserve for ensuring adequate security and reliability levels,
but high reserve requirements may substantially deteriorate the economy of
these supply systems.
Figure 1.2 shows a sample of two interconnected areas. The tie line power
ow between both areas Ptie is managed by the automatic generation control
(AGC) of each area. In conventional power grids, the function of AGCs is to
control power ow among interconnected areas. In conventional power grids,
the control parameters of AGCs are updated according to annual load de-
mand changes. Thus, uncertainties of providing reserve are eciently reduced.
Although, conventional generation suers from the volatility of conventional
energy prices, their variability is still much lower than uncertainty of the wind
power.
In contrast, in the case of microgrids supplied by intermittent generators,
some urgent questions emerge, regarding the optimum economical sizing of
microgrids. Moreover, the debate about the capacity needed from intermit-
tent generators to contribute to supplying microgrids with constraints of power
quality is pivotal. For this sake, the purpose of Chapter 3 is to provide a study
based on optimizing both energy production and economic performance for
isolated microgrids, supplied by wind energy and batteries. Hence, the maxi-
mum possible energy production at the lowest energy cost can be claried.
However, the objective of optimizing the energy production and economic
feasibility for microgrids must be extended to include additional constraints
related to the operation of microgrids, as the objective is to enhance the min-
imum generation cost with stable power ow limits in the islolated mode or
interconnected mode of microgrids.
For the microgrid interconnected mode and during a stable operation in
each area, it is assumed that summation of the total power generation, load
demand and net tie-line interchange power is nearly zero. During an unstable
operation, the tie-line interchange power is assumed to ow from high capac-
Chapter 1 23
ACE
PI
AGC area 1
(tie line power)
AGC area 2
ACE
PI
Figure 1.2: Interconnected systems power exchange
ity areas to low capacity areas. When the tie-line capacity is small and power
imbalance among dierent areas is large, then large oscillations in the tie-line
power will occur and the possibility of resynchronization among dierent gen-
eration units will be small. Thus, some generation units will be disconnected
from the operation. On the other hand, in the isolated microgrid mode, the
large penetration of DG units in distribution networks may not only lead to
problems under faulty conditions, but also to instability under small distur-
bances, such as small changes in loads and power transfers that occur during
the normal power system operation.
Figure 1.3 shows the mechanical analogue of a disturbed multi-machine

24 Chapter 1
system, based on Elgerd (1982). The disturbance near to one of the power sta-
tions distorts the power balance at the neighbouring stations. The resulting
electromechanical swings in the power system can be compared with the way
that masses swing in a mechanical system.
For conventional generation systems, some standards and recovery meth-
ods from power imbalances have been set and handled from decades. These
standards can be briefed in the following terms: (1) Power system stabilizers
in rotor circuits. (2) Fast valving for steam turbines. (3) Braking resistors for
hydro turbines. (4) Defence plan strategy to handle the operation of turbines
and feeding pumps. (5) Recovery plan by series-shunt compensators.
stable generators
resynchronized generators
disturbed generator
out of service generator
Figure 1.3: Mechanical analogue of swings in a multi-

machine system
In order to clear the necessity of stability analysis of microgrids, Zhixin [30]
conducted a small signal stability analysis of an isolated microgrid, supplied
by a synchronous generator and distributed generators interfaced by electronic
inverters. The analysis proved that the dominant variables which could violate
Chapter 1 25
the system stability of the both synchronous generators and distributed gen-
erators interfaced by inverters are dierent. These result showed a necessity
for stability analysis of microgrids, because the generation sources in the case
of microgrids are non-conventional and their modellings are totally far from
synchronous generators. Stability analysis is required to identify the impacts
of microgrid controllers, system characteristics and the intermittent behaviour
of wind generation units on the stability of the whole network.
For this important matter, some recent articles have discussed the small
signal stability of power systems supplied by wind energy. In [31] Ostadi proved
that stability in a grid supplied by wind turbines and a grid utility varies under
dierent levels of wind penetration.
In [32] Yateendra showed that the stability of a network supplied by wind
turbines and a utility grid varies under dierent wind speeds, and proposed a
damping controller for the dominant variables.
In [33] and [34], both Lihui and Yang respectively proved that the stabil-
ity of a power system varies under dierent wind speeds, and determined the
dominant variables.
For the previous considerations, the small signal stability analysis are con-
ducted in Chapter 4, for isolated microgrids, supplied by doubly-fed induction
generators and also generation sources interfaced via inverters, under dierent
scenarios, such as dierent levels of wind penetration, power demand increment
and dierent levels of real-reactive power exchange by droop manipulation.
The three topics of this thesis, which were discussed briey in this Section
will be more detailed through the following Sections. The rst topic belongs to
capacity factor analysis of droop-regulated microgrids based on a reliable wind
speed sampling strategy will be introduced in Section 1.6. The second topic
belongs to optimizing both energy production and economic performance of
isolated microgrids supplied by wind energy and dierent energy storage sys-
tems will be shown in Section 1.7. The third topic belongs to small signal
stability analysis will be shown in Section 1.8. Finally, the thesis contributions
will be detailed in Sections 1.9.

26 Chapter 1
1.6 Capacity Factor Analysis of Droop-Regulat-

ed Microgrids
Conventional generation units are capable of generating rated power dur-
ing normal operation. Reliability evaluation techniques for these units are well
established and are routinely used in the capacity planning of electric power
utilities. Contradictory to Conventional generation units, the wind energy
sources are always dependent on climate stochastic parameters, and so wind
energy sources are hourly and seasonal uctuating. Thus, capacity determi-
nation of power systems including wind energy conversion systems (WECs),
which can share supply microgrids with other distributed generators is rela-
tively complex.
Through last decades, power networks in Europe and North America in-
volved in several blackouts; however, in each case the blackout was caused by
a specic technical problem. The unprecedented concentration of blackouts
has driven many observers to argue that there are underlying reasons for such
large number of disturbances. The increased liberalization of electricity supply
industry by the beginning of 1990 has resulted in a signicant increase of inter-
area problems. Those problems have been compounded by the penetration of
wind energy. Large changes of wind power due to changing weather patterns
mean that the actual network ow may be quite dierent from the predicted
one.
With the increase of wind energy penetration in power networks, it be-
comes important matter to conduct accurate capacity factor analysis of wind
energy. The reason for the interest in capacity factor analysis of a wind gen-
erator is that the analysis provides a smart, summarized indication about the
combined interaction between the generator and the site that is between wind
power curve and stochastic wind speed. Subsequently, the accurate wind en-
ergy capacity determination reduce the need for balancing energy and reserve
power, which are needed to integrate wind power into the balancing of sup-
ply and demand in electricity supply systems. Thus, power plant scheduling
optimization can be ensured. In addition, the grid operator needs to know
the current and future wind power feed-in at each grid connection point. This
leads to lower integration costs for wind power, lower emissions from conven-
tional power plants used for balancing, and subsequently to a higher value of
Chapter 1 27
wind power.
The capacity factor is widely studied in Chapter 2, for autonomous droop-
regulated microgrids, supplied by distributed wind generators. The main con-
cern about capacity factor analysis in such grids is that the proposal of consid-
ering the droop-regulated microgrid as a large generator that is built up from
several wind generation units, cannot be solved as a simple sum of individual
rated and average power of wind generation units. This fact can be explained
as the micogrid load is shared among the intervening generators by means of
the droop control, and so particular restrictions on the maximum loadability
of the microgrid may occur that make it possible for the maximum delivered
power to be less than the sum of power injections. However, before conducting
the capacity factor analysis of the proposed droop-regulated microgrids, an
accurate wind speed sampling strategy is needed to appropriately determine
the status of wind generators.
Figure 1.4a shows the wind time series during a day by dierent averaging
times in Germany, Ernst (ve minute and an hour) sampled wind time series.
It is observed that wind is more uctuating when the wind speed samples are
taken every ve minutes. The average value of the one hour sampled wind
series was 3.6 m/s, whereas the average value of the ve minutes sampled
wind series was 4.4 m/s. Thus, this gure explains well that the ve minutes
sampled wind series are spread over a wider area of wind speed values.
Figure 1.4b shows the relation of correlation coecients between dierent
wind farm turbines and the distance separating these turbines. Normally, the
wind is driven by weather fronts and a daily pattern caused by the sun; de-
pending on these dominants, there is a signicant diurnal pattern in the wind
production.
According to [35], the wind speed correlations among wind turbines based
on distances separating them can be modelled by an exponential tting with a
decay parameter ranges from 500 to 700. Regardless the issue of correlations,
it is clear that the variations smooth out faster when the time scale is small.
The conclusion of this gure coincides with the conclusion of Figure 1.4a that
the wind speed series of the same wind distribution parameters, but at dif-
ferent sampling periods are spread over dierent values. Both gures indicate
the impossibility that the wind speed sampling can concentrate on a specic
range of the wind speed probability domain.

28 Chapter 1
Most researchers have used the conventional sampling technique Monte

Carlo Sampling (MCS) to estimate wind speed samples. Typically, MCS has
some engrained weak points, such as disappearance of stratications through
the whole range of the event probability domain, and so the sampling process
can not cover all wind speed probabilities. Moreover, MCS is characterized by
an intensive computationally nature. These weak points of MCS applied for
wind power sampling typically lead to a degree of risk analysis and give the
superiority always to conventional generation units, and subsequently a less
dependency on wind generators.
For this reason, the wind speed sampling process based on LHS is adopted
in Chapter 2, in order to provide more accurate and reliable wind speed sam-
pling. Thereafter, the status of wind generators can be determined accurately.
The proposed LHS sampling technique can stratify equally all wind speed
samples overall the wind speed probability domain. In Chapter 2, some com-
paritive simulations are conducted to clear the superiority of the LHS over the
conventional MCS.
The reliability of wind generators does not depend only on ecient deter-
mination of wind speed samples, but also on a wind speed correlation which
must be considered among dierent wind farm turbines. Clearly, wind speed
does not maintain a specied stable level, and so multi-state model must be
used to estimate a reliable wind energy system.
In [36] Dimitrios used a neural network training to produce correlated wind
speed samples for a wind farm.
In [37] Wijarn employed an autoregressive moving average (ARMA) time
series model, which considers the time lagging eect among dierent wind tur-
bines.
In [38] Gao used a genetic algorithm method to adjust the ARMA models
and simulate hourly wind speed samples based on the degree of wind speed
correlations among wind sites.
A simple method is adopted in Chapter 2 to introduce correlations among
dierent wind turbines based on Cholesky decomposition technique. The
method has the following desired properties: (1) It can be applied to all types
of input distribution functions. (2) The simplicity, as there is no need for
complex mathematical computations. (3) The marginal distribution of wind
speed samples remain intact.

Chapter 1 29
10
5min−sample
wind speed (m/s)
8 1h−sample
6
0
0 5 10 15 20
time(hour)
(a)
1
correlation coefficient
12h−avg
0.8 4h−avg
0.6 1h−avg
30min−avg
0.4 5min−avg
0.2
0
0 100 200 300 400 500 600
distance(km)
(b)
Figure 1.4: An example of wind measurements at dier-

ent sampling rates in Germany, Ernst: (a) Wind speed. (b)
Correlation coecients vs distance.
After introducing a reliable wind speed sampling technique and introducing
a correlation among dierent wind turbines, several simulations are processed
in Chapter 2 by an optimized load ow, under dierent mean wind speeds
and dierent wind turbines, in order to determine the maximum loadability of
microgrids supplied by wind turbines. The maximum loadability provides an
insight about the capacity factor of droop-regulated microgrids. Consequently,
the capacity factor analysis can be used to evaluate the site-matching suitabil-
ity of a combination of wind generators and wind speeds.
Unlike conventional generation units, where the capacity and reserve of
each conventional generator is known in advance, the capacity and reserve of
a microgrid supplied by wind turbines are vital issues, as it aects both the
stability and economic performance of the whole microgird. In Chapter 3,

30 Chapter 1
the idea of site matching suitability for droop-regulated microgrids supplied
by wind generators and dierent energy storage systems is more detailed, by
considering analysis of energy production and economic performance.
1.7 Benet Evaluation of Microgrids Using Ca-

pacity Factor and Economic Analysis
Power system operators are always responsible for maintaining the system
reliability, as generation must always be as close as possible to the system load,
to ensure the system frequency is maintained at the nominal values. When
generation levels can be easily determined, then achieving this responsibility
is easy. However, in the case of intermittent generators integration as wind
energy, additional capacity is needed to meet actual load demands. The char-
acteristics of such grids require scheduling more reserve for ensuring adequate
security and reliability levels, but high reserve requirements may substantially
deteriorate the economy of these supply systems.
With the market liberalization and occurrence of competitive markets, vi-
tal questions are always conducted regarding the optimum economical sizing
of microgrids. Moreover, the debate about the capacity needed from inter-
mittent generators to contribute to supplying microgrids, considering power
quality constraints is pivotal. For this sake, technical and economic analyses
are necessary, in order to meet the requirements in competitive markets.
In general, the most ecient wind turbine might not be the most eective
for transforming wind power in electric power on a specic site, and so the
main goal is to maximize the wind annual energy production [39]. The energy
production from a wind turbine depends on several factors, such as wind speed
conditions from the area and the characteristics of the wind turbine itself, in-
cluding the cut-in, rated and cut-o wind speed parameters.
The selection of the optimal wind turbine has been discussed in some pa-
pers based on maximization of the capacity factor [40]. In addition, other
studies have been conducted regarding site matching of wind turbines. The
site matching is based on identifying optimum turbine parameters, to yield
higher energy production at higher capacity factor.
The matching of a wind turbine to a site using normalized power and

Chapter 1 31
capacity factor curves was discussed in [40], but in general the economic con-
siderations of the turbine are not provided through wind site matching studies.
Other studies have focused on the matching of wind turbines to wind farms
from the viewpoint of both performance and economic considerations, as in [41]
and [42]. In [43], [44] and [45] Sureshkumar, Razak and Srivastava respectively
presented economic analyses of microgrids by HOMER Energy software.
Usually, HOMER Energy software analyses the economic feasibility of hy-
brid energy systems, but it does not take into consideration the droop regu-
lation of microgrid interfaces, and also does not account for the reserve issue,
because it is based on historical data only. Therefore, the HOMER Energy
software cannot be reliable to determine the denite optimum model from the
viewpoint of both performance and economic considerations.
However, for isolated microgrids, storage systems are always required, to
share microgrids supply with the intermittent wind energy, as generation must
always be as close as possible to system loads. The problem of keeping the
power balance is still more dicult for stand-alone microgrids supplied by in-
termittent generators. The total requirements for these storage systems are
unclear at present, because most of the recent studies which have optimized
the capacity factor and economics of wind energy production, have not taken
into consideration the operation of storage systems, which share supply with
wind energy. Thus, the optimum point of operation is totally dierent, when
considering the dynamics of storage systems. The performance and economic
aspects of such storage systems are signicant and must be brought into the
economic evaluations.
Therefore, the proposed study in Chapter 3 is based on optimizing both
energy production and economic performance for isolated microgrids, supplied
by wind energy and various batteries via a multi-objective function. The op-
timization function is conducted to minimize the cost of energy, which is a
function of both capital costs and energy production. Hence, the maximum
possible energy production at the lowest cost of energy can be ensured, taking
into consideration the droop regulation of microgrid interfaces and dynamic
operation of batteries (charging-discharging). In addition, the economic anal-
yses are conducted also by simple payback and net present value.
However, the objective of optimizing the energy production and economic
feasibility for microgrids must be extended to satisfy the power balance and
32 Chapter 1
generation limit among dierent microgrid generators; a stable operation must
be ensured without power swings and resynchronization problems. Small sig-
nal stability analysis is conducted in Chapter 4, and the purpose of the analysis
will be more detailed in the following Section.
1.8 Microgrids Stability Analysis

As mentioned in the last Section, the objective of optimizing the energy
production and economic feasibility must be extended to include additional
constraints related to the operation of microgrids, such as ow limit between
two successive areas and stability constraints of each DG supplying microgrids,
because all these constraints may cause perturbations, and subsequently insta-
bility. Such perturbations in microgrids aect important operating variables,
such as frequency and voltage. Therefore, dierent microgrid perturbations
must be well studied, to avoid power network collapses.
The proposed microgrid in Chapter 4 is supplied by DGs interfaced through
electronic inverters and intermittent sources which are doubly-fed induction
generators (DFIGs). A large penetration of DG units interfaced by electronic
inverters in distribution networks may not only lead to problems under faulty
conditions, but also to instability under small disturbances, such as small
changes in loads and power transfers, which occur during normal power system
operation. Moreover, the modelling of DFIGs are totally far from synchronous
generators. Consequently, analyses and design frameworks of microgrid sup-
plied by intermittent sources are required to ensure stable performance and
robustness. Analyses are required to identify the impacts of microgrid con-
trollers, system characteristics and the behaviour of intermittent wind gener-
ation units on the stability of the whole system.
The most common perturbations in power networks can be analysed by
the small signal stability study. Small signal stability is the ability of a power
system to maintain stability, when it is subjected to small disturbances. A
disturbance is considered to be small, if the linearized system around the equi-
librium point still adequately represents dynamics of the original nonlinear
system under this disturbance [2].
The small signal stability analyses of conventional power systems have been
studied in many works. In [46] and [47], Wang and Evandro studied the small
Chapter 1 33
signal stability of conventional power systems, with respect to system param-
eters changes.
The production of electricity in a conventional utility is secured primarily
using synchronous generators, and for this reason, it is important to secure
their synchronism and parallel operation. Thus, the question of stability in
a conventional power system is mainly based on the stability of synchronous
machinery and on the relationship between the active power and rotor angle
of the generator. In contrast to conventional power systems, the proposed mi-
crogrids in this thesis are supplied by electronic interfaces operating by high
frequencies and intermittent energy sources which are DFIGs. Therefore, the
stability analysis in such microgrids must include all these elements.
Microgrid small signal stability analyses have been conducted in some re-
cent papers. In [33] and [34], both Lihui and Yang respectively presented
stability analyses of microgrids supplied by wind energy, in grid-connected
modes.
In [48] Pogaku conducted a stability analysis in an isolated microgrid sup-
plied by distributed generators, and interfaced to the microgrid through elec-
tronic inverters.
All the previous microgrid stability studies belong to either microgrid sup-
plied by wind generators and connected to utility grids, or to isolated micro-
grids supplied by distributed generators interfaced by electronic inverters and
without penetration of any intermittent energy sources.
When a microgrid is connected to the main utility, the possibility of system
restoration and reduction of power swings among dierent generation units is
high. In the case of grid-connected mode, the utility grid can provide an ade-
quate reserve, in the case of demand shortfall. Unfortunately, the situation is
dierent in the case of isolated microgrid, as the available reserve and capabil-
ities of the whole system to recover from disturbances are limited, especially
when the microgrid is supplied by intermittent sources. Hence, in Chapter 4,
the small signal stability analysis of isolated microgrids, supplied by DFIGs
and also DGs interfaced through inverters are conducted, under dierent sce-
narios, such as dierent levels of wind penetration, power demand increment
and dierent levels of real-reactive power exchange by droop manipulations.
Thus, in future work, the controllers of DFIGs and inverter-interfaced dis-
tributed generators can be set appropriately, and an adequate control strategy

34 Chapter 1
can be developed to keep the microgrids stable.
1.9 Thesis Contributions

The Chapter mentioned several doubts and weak points of isolated micro-
grids, supplied by intermittent energy sources. These weak points are ac-
counted by the following terms: (1) The uncertainties of the generation ca-
pacity for microgrids, supplied by intermittent energy sources. (2) The full
insight regarding the economic and technical feasibility evaluation for micro-
grids are unclear at present. (3) Usually, microgrids are supplied by DGs via
electronic interfaces, or through renewable energy sources; thus, the modelling
and the behaviour of such DGs are totally far from conventional generation
units. Therefore, stability analysis of microgrids supplied by such DGs is re-
quired to ensure stable performance and robustness.
For the all mentioned doubts, the thesis analyses and proposes solutions
for these problems regarding capacity factor analysis of droop-regulated mi-
crogrids, optimizing both energy production and economic performance for
isolated microgrids and nally small signal stability analysis.
Thesis contributions are summarized in the following points, also some com-
parisons between thesis contributions and other recent researches are shown
hereunder:
• Concerning a reliable wind speed sampling method to determine the
penetration and capacity of intermittent wind energy, in [49] Andréa
proposed a model for a probabilistic representation of a wind farm to
provide an annual estimation of energy production. This model included
all operational information of wind turbines, as failure and repair rates
regarding wind farm. The model was sampled by Monte Carlo Markov
chain.
In [50] Roy explained reliability eects of wind energy penetration on
a composite generation and a transmission system, using the state sam-
pling Monte Carlo simulation technique.
Vallée in [51] showed some clustering algorithms for wind turbines, in or-
der to group wind parks with close statistical behaviour, allowing highly
Chapter 1 35
correlated wind parks into the same cluster to be integrated in a realis-
tic way, and the evaluation was processed by the non sequential Monte
Carlo sampling.
In this thesis, a modied algorithm for wind speed sampling Latin Hy-
percube Sampling (LHS), based on [7] is adopted. The algorithm avoids
disappearance of stratications through the whole range of the event
probability resulting from Monte Carlo Sampling. Furthermore, the al-
gorithm avoids intensive computations, which result from Acceptance
Rejection and Markov chain Monte Carlo Sampling;
• The reliability of wind generators does not depend only on determination
of wind speed samples, but also on wind speed correlations among wind
turbines based on the distance separating them. The wind speed does
not maintain a specied stable level, and so a multi-state model must be
used to estimate a reliable wind energy system.
In [36] Dimitrios used a neural network training, to produce correlated
wind speed samples for a wind farm. In [37] Wijarn employed the ARMA
time series model, which considers the time lagging eect among dierent
wind turbines. In [38] Gao used a genetic algorithm method to adjust
ARMA models, and to simulate hourly wind speed samples based on the
degree of wind speed correlation among wind farm turbines.
A simple method is adopted in this thesis, to introduce correlations
among dierent wind turbines based on the Cholesky decomposition
technique. The proposed correlation method has some desired prop-
erties, as it can be applied on all types of input distribution functions,
simplicity as there is no need for complex mathematical computations
and the marginal distribution of wind speed remain intact;
• In order to determine the feasibility of wind farm installations from the
viewpoint of performance and economic considerations, some authors
have discussed this matter.
In [41] Tai determined the feasibility of a wind farm, via the capacity
factor of each wind turbine, under dierent values of tower height and
36 Chapter 1
rated wind speed, and determined the feasibility of the wind farm as a
function of the energy cost and mean wind speed.
In [42] Wang processed simulations for wind turbines under dierent hub
heights, diameters and mean wind speeds, to determine the most feasible
case. As wind turbine is considered an intermittent source of energy and
wind power always uctuates depending on stochastic parameters, it is
not logic to optimize the capacity factor and economic performance of
wind turbines without considering other storage systems or dispatchable
generators, which share the operation with wind farms.
In this thesis, the issue of optimizing both energy production and eco-
nomic performance is shown for microgrids supplied by dierent wind
turbines and dierent types of batteries. Several authors have optimized
the capacity factor and economics of wind turbines production, but have
not taken into consideration the dynamic operation of storage systems,
which share the supply with wind generators. Therefore, the optimum
point of operation is totally dierent when considering the dynamics of
storage systems, if both wind turbines and storage systems are employed
together in a multi-objective function.
This thesis provides benet evaluations for microgrids from the perspec-
tive of energy production and economic performance. Several analyses
are processed for microgrids supplied by composite generation of wind
energy units and batteries, all economic evaluations in the thesis are
conducted by several economic evaluation models, such as cost of en-
ergy, simple payback and net present value.
In addition, some comparative simulations among various batteries based
on their capital costs-eciencies are shown, in order to clear the possi-
bility of enhancing the microgrid economic performance and energy pro-
duction by high capital cost-high ecient batteries. The cost of a battery
is typically related to its performance, including the discharge rate, en-
ergy density, operational losses and life time. Through the thesis, some
simulations show the possibility of improving the economic performance
of high capital cost-high ecient batteries by employing them with wind

Chapter 1 37
turbines in a multi-objective function;
• Regarding an adequate real-reactive power exchange strategy in a mi-
crogrid, the droop regulation of each distributed generator supplying
a microgrid must be performed in a quick ecient manner, and also
the economical aspects must be taken into consideration. Some authors
proposed a centralized energy management in microgrids. Regardless
the high cost that entails communication establishment between all dis-
tributed generators and central controller of a microgrid, the concept of
plug & play cannot be conrmed. Through the thesis, the real-reactive
power exchange among dierent distributed generators in microgrids is
based on the droop control. Therefore, if power ow level changes in a
microgrid, the change can be met directly by droop regulation without
any need of central controller management;
• As the droop control strategy can enforce all distributed generators to
participate in a system restoration in case of disturbances, and also man-
age the real-reactive power exchange in the system, several authors have
discussed this strategy.
In [52] Katiraei addressed a real-reactive power management of elec-
tronically interfaced distributed generators, which is based on locally
measured signals and droop characteristic, in order to provide compli-
mentary frequency restoration.
Barklund in [23] proposed an energy management strategy for a stand-
alone droop-controlled microgrid based on frequency-droop gains of dif-
ferent distributed generators, which adjusts generators output power, to
minimize the fuel consumption and also ensures a stable operation.
In [53] Majumder explained a method for power ow control between
an utility and a microgrid through back-to-back converters, which facil-
itates the desired real and reactive power ow between the utility and
microgrid, based on droop characteristics of dierent generators.
All the previous authors proposed strategies to optimize the power ow
in microgrids through the droop control based on considering two opera-
tional constraints for each distributed generator. The rst constraint

38 Chapter 1
is for droop regulation, while the second constraint is for maximum
power limit of each distributed generator. Thus, realizing an equilib-
rium point requires employing heavy computations by the conventional
Newton-Raphson algorithm. In this thesis, another strategy based on
Fischer-Burmeister is proposed, to handle the droop control. The pro-
posed strategy is based on a nonlinear complementarity problem that
substitutes the piecewise droop function by an only scalar function, and
makes the problem to be solved easier by less iteration process;
• The objective of optimizing the energy production in microgrids must
be extended to satisfy the power balance and generation limit among
dierent microgrid generators, such a stable operation must be ensured
without power swings and resynchronization problems. Some authors
have discussed the stability analysis of microgrids.
In [31] Ostadi proved that the stability of a microgrid supplied by wind
turbines and a grid utility varies, under dierent levels of wind penetra-
tion.
In [32] Yateendra showed variations in the stability of a microgrid sup-
plied by wind turbines and a utility grid, under dierent wind speeds,
and proposed a damping controller for the dominant variables.
In [33] and [34], both Lihui and Yang respectively proved that the sta-
bility of a power system supplied partially by wind energy varies under
dierent wind speeds, and determined the dominant variables.
In [48] Pogaku conducted a stability analysis of an isolated microgrid sup-
plied by distributed generators, and interfaced to the microgrid through
electronic inverters, and without penetration of any intermittent energy
sources as wind turbines.
All the previous stability studies belong to either microgrids supplied
by wind energy and connected to utility grids, or to isolated microgrids
supplied by distributed generators and without penetration of any in-
termittent energy source. Through this thesis, the stability analyses are
performed for isolated microgrids supplied by intermittent energy sources
which are DFIGs and also other DGs interfaced by inverters. The reason
Chapter 1 39
of conducting such analyses is to determine the critical state variables
which can violate the stability of microgrids. The analyses are processed
under dierent scenarios, such as dierent levels of wind penetration,
power demand increment and dierent levels of real-reactive power ex-
change by droop manipulations.

40 Chapter 1
Chapter 2 41
Chapter 2
Capacity Factor Analysis of

Droop-Regulated Microgrids
2.1 Introduction
The main function of an electric power system is to meet its customers
demand with electrical energy by minimum possible cost; at the same time,
continuity and power quality must be ensured [54]. Advanced electric power
systems around the world have undergone de-regulation with major changes
in structure, operation and regulation.
Each part of the power chain, as power plant owners, transmission system
owners, operators, regulators and nally the customers are all involved in this
chain. The development process in any modern society is mainly dependent on
power availability and quality. Any modern society, can appreciate perfectly
how main necessities of daily life would be without energy. From statistics, it
is expected that necessity for available and high quality of power supply will
continue to increase.
The reliability associated with a power system is a measure of the overall
system ability to satisfy the customer demand for electrical energy. Power sys-
tem reliability can be further subdivided into two dierent categories, system
adequacy and system security [54], as shown in Figure 2.1.
The reliability of the energy infrastructure is indication of the whole in-
frastructures capacity to meet the customers demand of the electrical energy.
Power system reliability can be clearly categorized to dierent branches of sys-
tem adequacy and security [54] as shown in Figure 2.1.

42 Chapter 2
System Reliability
System Adequacy System Security
Figure 2.1: Categories of system reliability
System adequacy is considered as indication of sucient facilities within the
system to meet the customer demand. These facilities include those parties of
the power chain to generate energy starting from power plants to transmission
and nally distribution networks required to transport the energy to the actual
consumer load points. On the other hand, system security is related to the
ability of the system to respond to originating disturbances within that system
[54]. Therefore, to ensure a reliable power system, both reliability terms must
be well evaluated.
The proposed isolated microgrid in this Chapter is supplied by wind tur-
bines, interfaced to the microgrid by inverters, and so the main reliability
challenge results from intermittent nature of wind energy sources.
Wind energy is one of the fastest growing renewable energy sources; Fig-
ure 2.2 shows the total wind capacity installed around the world from 1980,
to 2008. Total of 120,791 MW of wind capacity was installed throughout the
world [55] by the year 2008. The cost of energy from wind has dropped to the
point that in some sites it is nearly competitive with conventional sources. The
current total installed wind capacity in Canada is 2,577 MW, which is about
1% of Canada's total electricity demand [56]. Saskatchewan city, currently
has 171.2 MW of installed wind capacity with the completion of the 150 MW
centennial wind project in 2006 [56].
The World Energy Council has estimated that the wind energy capacity
worldwide may reach as high as 474,000 MW by the year 2020 [55]. In Canada,
Chapter 2 43
Ontario, Nova Scotia and Prince Edward Island have committed to generate
10%, 5% and 15% respectively of the total electricity production from renew-
able energy sources by the year 2010 [56]. Many countries around the world
implement dierent policies to promote the growth of renewable energy.
Renewable energy policies, such as the Fixed Feed-in-Taris in Germany,
Denmark, Spain [57] and Renewable Obligation in the UK [58] have driven the
development of wind power in these countries.
Figure 2.2 explains the gradual increase in the wind energy production
around the world, as mentioned in [59].
4
x 10
14
12
10
wind production (MW)
0
1980 1985 1990 1995 2000 2005
year
Figure 2.2: World wind energy generation
As a result of continuing growth and penetration of the wind energy in the
traditional power systems, it has been urgent necessity to consider the relia-
bility evaluation for wind power production. Typically, the conventional gen-
erating units are capable of generating rated power during normal operation;
reliability evaluation for these units are well established and are routinely used
in the capacity planning of the electric power utilities. Wind energy sources
are always dependent on the climate stochastic parameters, and so it is usually
hourly and seasonal uctuating, unlike conventional generating units. Through
last decade, the power networks in Europe and North America involved in sev-
44 Chapter 2
eral blackouts, due to the increased penetration of wind energy.
The increased liberalization of electricity supply industry by the beginning
of 1990 has resulted in a signicant increase in the inter-area problems. Those
problems have been compounded by increased penetration of the wind gener-
ation.
Large changes of wind power due to changing weather patterns mean that
an actual network ow may be quite dierent from a predicted ow. Thus,
the reliability evaluation of power systems icluding the wind energy conversion
systems (WECS) is relatively complex.
Determination of non-dispatchable generator reliability, as in case of the
wind turbine generator (WTG) unit can be performed by dierent techniques.
Analytical methods evaluate the reliability indices by using numerical solu-
tions from mathematical models, which represent the power system.
A considerable research activity has been dedicated to the area of power
system reliability including wind energy in the past several years. The wind
generator unit has been modelled as a multi-state unit to evaluate the eect
of the wind power on the reliability indices of power systems by analytical
techniques, as shown in [60], [61] and [62].
In [61], a methodology was proposed to classify the impact of wind genera-
tion, which is dependent on time variation and climatic change on the reliability
of the power system.
In [63], the author proposed a method to clarify the multi-state of wind
turbines, due to dependency on wind speed variation.
The reliability study has been extended in [62] to investigate the eect of
the wind power generation by a probability theory on the power system.
On the other hand, the capacity factor is a measure of operating eciency,
which indicates the ability of a generating plant to deliver its full capacity;
thus, it is considered as an indicator of the reliability of supply.
The reason of the interest in capacity factor analysis of a wind generator is
that the analysis provides a smart, summarized indication about the combined
interaction between the generator and the site that is between the wind power
curve and stochastic wind speed. Subsequently, the accurate wind energy ca-
pacity determination reduce the need for balancing energy and reserve power,
which are needed to integrate wind power into the balancing of supply and
demand in electricity supply systems.

Chapter 2 45
Moreover, the main concern about capacity factor analysis in droop-regulat-
ed microgrids that the proposal of considering the droop-regulated microgrid
as a large generator that is built up from several wind generation units cannot
be solved as a simple sum of individual rated and average power of wind gen-
eration units. This fact can be explained as the micogrid load is shared among
the intervening generators by means of the droop control, and so particular
restrictions on the maximum loadability of the microgrid may occur, which
make it possible for the maximum delivered power to be less than the sum of
power injections.
For this reason, the capacity factor analysis of droop-regulated microgrids
is performed in this Chapter based on a reliable wind speed sampling strat-
egy, by reducing the variance of wind speed samples, and also including the
correlations among wind speeds; thus, each wind energy source status can be
determined accurately. Finally, capacity factor analysis can provide an in-
sight about site matching suitability for droop-regulated microgrids supplied
by wind generators.
However, before conducting the capacity factor analysis of the proposed
droop-regulated microgrids, an accurate wind speed sampling strategy is needed,
to determine appropriately the status of wind generators.
Most wind energy reliability studies have been conducted by one of the
most well-known sampling methods which is Monte Carlo sampling (MCS).
Typically, MCS has some engrained weak points, such as disappearance of
stratications through the whole range of the event probability domain, and
so the sampling process can not cover all wind speed probabilities, as shown
in Figure 2.3 where the simulation is processed by ten samples, at wind scale
parameter equals to three. Wind speed samples by MCS are accumulated over
certain areas of the wind speed distribution probability. Moreover, MCS is
characterized by an intensive computationally nature.
The MCS weak points typically lead to a degree of risk analysis and give
the superiority always to conventional generator units, and a less dependency
on wind generators. For this reason, a modied strategy for wind speed sam-
pling is adopted in this Chapter; the proposed sampling technique is Latin

Hypercube (LHS), which can stratify equally all wind speed samples overall
the probability domain, by simple calculations and non-time consuming man-
ner.
46 Chapter 2
0.9
cumulative distribution function−Monte carlo 0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
2 3 4 5 6 7 8 9 10 11
wind speed sample (m/s)
Figure 2.3: Monte Carlo transformation method
Furthermore, the reliability of a wind generator does not depend only on an
ecient sampling technique for the wind speed, but also on a wind speed corre-
lation which must be considered. In this Chapter, a simple method is adopted
to introduce correlations among dierent wind turbines based on Cholesky de-
composition technique. The method has some desired properties, as it can be
applied on all types of input distribution functions, simplicity as there is no
need for complex mathematical computations and the marginal distribution of
wind speed samples remain intact.
The following Section, explains the organization of this Chapter. Starting
by sampling the wind speed by LHS strategy, then introducing the correla-
tion among wind turbines. Therefore, a microgird supplied by wind energy
can be processed by an optimized power ow, and consequently the maximum
loadability of the microgid can be accurately determined.
2.1.1 Chapter Outlines
As mentioned in Chapter 1, the rst concern of this thesis is to conduct
capacity factor analysis of droop-regulated microgrids based on a reliable wind
speed sampling strategy, by reducing the variance of wind speed samples,
and also including the correlations among wind speeds. Thus, the states of
Chapter 2 47
wind turbines can be obtained in an accurate manner and wind turbines ade-
quacy term can be determined later more precisely by capacity factor measures,
through an optimized power ow.
This Chapter concentrates mainly on the following items:
• Developing a reliable sampling strategy for wind speed samples. Through
this Chapter, a strategy based on Latin Hypercube Sampling is adopted;
• The reliability of wind generators does not depend only on a sampling
technique of the wind speed, but also on correlation among wind turbines,
which must be considered. Therefore, a simple method is adopted in this
Chapter, to introduce the correlation among dierent wind turbines;
• Furthermore, in this Chapter an ecient curve tting for a wind turbine
is proposed, taking into consideration the null wind speeds;
• A piecewise-denition of the wind power cumulative distribution function
(CDF) is proposed, and so wind power samples can be obtained directly
by the probability distribution of the wind speed;
• Another important issue is shown through this Chapter that is the maxi-
mum loadability of an isolated microgrid, as it provides an insight about
the capacity factor of microgrids. A global optimization technique Global

Search is used to calculate the maximum loadability of the microgrid;
• Measures of the relative goodness of t were employed by Akaike and
Bayesian to set the distribution tting of the injected power to the mi-
crogrid from the wind turbines, explaining the eect of seasonal or loca-
tional wind speed variations on the shape of the output power, and so
an accurate estimation of generators capacities can be claried;
• Capacity factor estimations for both wind turbines and microgrids are
conducted. Thus, the adequacy term can be evaluated according to
capacity factor measures;
• In addition, the bifurcation analysis is performed by the Continuation
Power Flow (CPF), to determine the feasibility of microgrids, under dif-
ferent scenarios;
48 Chapter 2
• Finally, several case studies are conducted to measure the capacity factor
of a microgrid, by suppling the microgrid by various wind turbines and
under dierent values of droop constants of the microgrid supporting
units (MSUs). Thereafter, some important conclusions can be obtained.
The main outlines of the Chapter are shown hereunder:
1. Providing LHS technique (Section 2.2);
2. Applying a correlation technique among wind turbines (Section2.3);
3. Conducting a curve tting for wind turbines (Section 2.4)
4. Performing a cumulative distribution function of wind power samples
(Section 2.5);
5. Microgrid layout (Section 2.6);
6. Maximize the microgrid loadability by the global search technique (Sec-
tion 2.7);
7. Processing simulations and conclusions (Section 2.8);
8. Applying Akaike and Bayesian techniques to the microgrid injected power
(Section 2.9);
9. Capacity factor estimation for both wind turbines and microgrid (Section
2.10);
10. Bifurcation analysis (Section 2.11);
11. Case studies (Section 2.12).

Chapter 2 49
2.2 Latin Hypercube Sampling

History and Denition
By 1975, W. J. Conover of Texas Tech University was asked to develop
a method for improving the eciency of simple Monte Carlo, to character-
ize the uncertainty of inputs to computer models. The research resulted in a
stratied model of Monte Carlo called Latin hypercube sampling (LHS), and
so LHS has been applied to some computer modelling applications at Sandia
National Laboratories [64].
The Latin Hypercube Sampling uses a stratied sampling to cover the whole
domain of a distribution function of a random variable Xj . The stratication
is processed by dividing the distribution function vertical axis F (x) into (N )

non-overlapping intervals of equal length, where N is the number of samples
to be processed. Therefore, by deriving F −1 (x), the distribution function hor-
izontal axis is stratied into N equiprobable and non-overlapping intervals.
The main four steps of LHS are listed as follows:
1. Denition of the probability density function (PDFs), or the Cumulative
Distribution Function (CDFs) of parameters and state variables, which
will be sampled;
2. Calculating the number of simulations (N );
3. Dividing each parameter into equiprobable intervals;
4. Creating the LHS table.
2.2.1 Distribution Functions of Parameters and State Va-
riables
Probability density functions (PDF) or cumulative distribution functions
(CDF) are assigned to uncertain parameters or state variables, depending on
the nature of the parameters or state variables.
As mentioned in [65], [66] and [67] regarding the wind speed distribution
probability, it has been assumed that wind speed distribution probability is
formed as weibull distribution function by two parameters, scale and shape
parameters.
50 Chapter 2
Other authors, as in [68] and [69] considered that the probability distribu-
tion of the wind speed of specic wind farms cant not be formed as weibull
distribution, due to geographical nature of such wind farms.
The only considered parameters in this Chapter is the wind speed distribu-
tion probabilities; also wind speed probability is assumed to be distributed as
weibull, by two parameters, scale and shape parameters, and sampled through
the CDF.
2.2.2 Calculating the number of samples
The LHS design involves sampling without replacement; therefore, if a
mathematical model of a problem contains uncertain variables (K), then the
minimum limit of samples must be K+1; In [70], the author showed a necessary
condition that (N > 4/3K). Moreover, the desired partial rank correlation co-
ecient is important factor to determine (N ). In this Chapter, N is assumed
to be 100.
2.2.3 Dividing the range of each of the parameters into
equi-probable intervals
By assuming the parameter being sampled is x and its CDF is f (x), and
must be divided into (N ) non-overlapping equiprobable intervals, as shown in
Figure 2.4, where the simulation is processed for ten samples.
Each interval is indexed by i, and so the limits of each interval (ximin and
ximax ).
The integral of f (x) is assumed to be F (x), and so the transformation of
the probability values into the sample x is conducted by using the inverse of
the cumulative distribution function F (x), as shown by equation (2.1).
xi = F (x)−1 (2.1)
The lower interval limit of the interval i is (ximin ) and the upper interval
i
limit (xmax ) is determined from equation (2.2).
ximax = F −1 [F (ximin ) + 1/N ] (2.2)
The minimum value for the next interval xi+1

min , is set to be equal to the
i
maximum value of the previous interval xmax , and this process is repeated for
Chapter 2 51
the remaining intervals.
Regarding the wind speed sampling, the CDF of the wind speed is divided
into N intervals, according to the proposed number of simulations, then above
equations are applied to generate the samples from CDF function as shown in
Figure 2.4, as the simulation is processed by ten samples, at wind speed scale
parameter equals to three.
0.9
cumulative distribution function−LHS
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
wind speed samples (m/s)
Figure 2.4: Stratied wind speed samples by LHS
2.2.4 Creating the LHS table
The LHS design involves random sampling without replacement; the equipr-
obable interval of each input variable is sampled once. LHS table is generated
as (N ×K) matrix, where N is the number of simulations and K is the number
of the sampled input variables.
The LHS design was rst proposed in [70]; a computer program based upon
this methodology is available for generating LHS tables [71].
The N sampling indices of the rst variable are combined randomly, with
the N sampling indices of the second variable. These N pairs are then paired
randomly with the N values of the third variable. The random pairing contin-
ues until all the K parameters are included and the N ×K matrix is generated
[71].
52 Chapter 2
Figures 2.5a and 2.6a represent the LHS sampling at wind speed scale pa-
rameters: 3 and 5; while, Figures 2.5b and 2.6b represent MCS sampling at
wind speed scale parameters: 3 and 5.
Clearly, Figures 2.5 and 2.6 show the superiority of the LHS over MCS at
wind speed scale parameters: 3 and 5.
The LHS ensures a better spread of the sample points over the sampling
space, while MCS lacks stratication and that is clear from MCS gures, where
some samples are accumulated over each other without equiprobable distribu-
tion.
After conducting such a reliable sampling technique, inducing correlations
among wind turbines is a must. Next Section will propose a correlation method
to be introduced among wind turbines.
1 1
0.9 0.9
cumulative distribution function−Monte carlo
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 2 4 6 0 2 4 6
wind speed sample (m/s) wind speed sample (m/s)
(a) (b)
Figure 2.5: Wind speed sampling at scale parameter (3)

processed by: (a) LHS. (b) MCS.
Chapter 2 53
1 1
0.9 0.9
cumulative distribution function−Monte carlo

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 5 10 0 5 10
wind speed sample (m/s) wind speed sample (m/s)
(a) (b)
Figure 2.6: Wind speed sampling at scale parameter (5)

processed by: (a) LHS. (b) MCS.
2.3 Correlation among Dierent Wind Turbines
The wind is driven by weather fronts and a daily pattern caused by the
sun; depending on these dominants, there is a signicant diurnal pattern in
the production. When combining the generation of variable sources, an av-
eraging eect occurs, if the time series of generators are uncorrelated or only
partly correlated.
For an analysis of this behaviour, a correlation coecient of (1) means that
time series are perfectly correlated, and hence go up and down in exactly the
same manner. A coecient of (0) means that time series are randomly dis-
tributed. On the other hand, when the correlation coecient is (-1) meaning,
if there is a tendency of decreasing a production at one site, a production in-
creases at the another site.
There is a signicant variation in the cross-correlation coecients for sim-
ilar distance, the correlation becomes weak when it reaches below 0.5, with
distance 200-500 km. Cross-correlation (rx,y ) is a measure of how two time

54 Chapter 2
series follow each other, as indicated by equation (2.3).
∑n
1
n i=1 (xi − µx )(yi − µy )
rx,y = (2.3)
σx σy
Where µ is related to the average value, σ is the standard deviation, n is
the number of points in the times series and rx,y is the coecient of correlation
between two variables.
When distributing the wind power production to larger area, the total pro-
duction will be smoother and less variable, if the correlations among the site
turbines are high.
There is no signicant change in correlation coecients calculated from dif-
ferent years. The cross correlation can be modelled by an exponential tting
as shown in Figure 2.7, with decay parameter ranges from 500 to 700, based
on [35].
0.8
corss correlation
0.6
0.4
0.2
0
0 500 1000 1500 2000 2500 3000
distance(km)
Figure 2.7: Cross-correlation vs distance.
As a result of the important eect of the wind speed correlation among
wind farms on power system reliability, this issue have been discussed through
many papers, as [72] and [73]. In [74], the correlation among wind farms based
on wind speed and wind directions was proposed.
Other papers have discussed the correlation issue for single time series, but
with time lags (auto-correlation). For the wind power, the auto-correlation
Chapter 2 55
decreases soon with increasing the time lag. Some authors discussed this im-
portant matter as in [37], by employing the ARMA time series model, which
considers the time lagging eect, noise and residual eects; thus, by choosing a
certain instant the correlation coecients among wind farms can be produced.
The sampling time was found to inuence signicantly the forecasting error
[75], as wind variation in one site aects the another site when a delay time
is considered. The Neural Network in [36] was used as a training technique to
produce the wind speed samples, after considering an auto-correlation.
All the previous correlation methods are based on intensive computational
techniques; a simple method is adopted in this Chapter to introduce correla-
tions among wind turbines based on Cholesky decomposition technique. The
method has some desired properties as it can be applied on all types of input
distribution functions, simplicity as there is no need for complex mathemati-
cal computations and the marginal distribution of wind speed samples remain
intact.
The LHS design is proposed for models, where all of input variables are
statistically independent. One approach to incorporate dependences is to con-
sider linear combinations of the independent random variables to achieve the
desired correlation structure. In case of normal random variables and random
sampling, this approach is well-known to produce multivariate normal input
vectors.
However, if the samples are stratied as in the LHS case, then this approach
destroys the integrity of the stratied samples, and so the values obtained from
linear combinations will no longer map back into each of the original structure.
Moreover, the linear combinations of non-normal variables will adversely
aect both the random samples and stratied samples, as the marginal distri-
bution may no longer resemble the original marginal distribution desired on
the input variables.
The adopted strategy in this Chapter for introducing the correlations among
wind turbines is based on [7], the methodology replaces the random pairing of
the (N ) values of each input parameter with a restricted pairing. This new
technique may be appropriate for both dependent and independent variables,
because the restricted pairing can reduce spurious correlations and induce the
desired rank correlation among the input variables.

56 Chapter 2
2.3.1 Correlation Steps
The method is based on the Cholesky decomposition of the correlation
matrix. Suppose matrix X is composed of independent random variables, with
correlation matrix I and C is the desired correlation matrix. The matrix C can
be written as C = P P ′ , where P is the lower triangular matrix. Similar to the
simple random sampling, multiplying vector into P′ yield random variables,
with correlation matrix C. Therefore, the objective is to rearrange the input
variables close to the target correlation matrix. The main steps of correlation:
1. Generating matrix R using Latin hypercube sampling of (K) variables
at sample size (N);
2. Calculating T, the correlation matrix of R;
3. Calculating P, the lower triangular matrix of the target correlation ma-
trix C using Cholesky factorization C = P P ′, and also Q, the lower
triangular matrix of T ; T = QQ′ ;
4. Solving to obtain matrix S such that ST S ′ = C , which is calculated from

S = P Q−1 ;
5. Calculating target correlation matrix R∗ = RS ′ , which has a correlation
matrix equal to C;
6. Rearranging the values of each variable in R, so they have the same rank
order as the target matrix R∗ .
The statistical dependencies can be performed by the rank sorting ap-
proaches as shown by Iman and Conover in [7] or by Stein in [76]. Stein's
method applies the rank sorting approach when the input variables are related
in a non-monotonic way, and requires producing samples with denite joint
distribution, by then it is not the purpose of the current study. Thus, Iman's
and Conover's method is applied in this Section to induce correlations among
wind speed samples. In order to clear the eectiveness of the proposed
method, the proposed correlation strategy is applied to correlate among wind
speed samples of separated wind turbines, as shown in Figure 2.8, according
to correlation values from Figure 2.7. By measuring the distance among the
dispersed wind turbines at nodes 8, 12 and 14 in Figure 2.7, the correlation

Chapter 2 57
Figure 2.8: Microgrid layout

wind speed (m/s)
10
0
0 20 40 60 80 100
samples
(a)
wind speed (m/s)
10
0
0 20 40 60 80 100
samples
(b)
Figure 2.9: Wind speed time series processed by scale

parameter (3): (a) Correlation included. (b) Correlation
excluded.
factor ranges from 0.75 to 0.9.
Figure 2.9 clears the deviation between each turbine sample and its corre-
spondent in the other turbines at the same time. Figure 2.9a represents the
deviations in time series among wind turbines samples, in case of applying
correlation. Figure 2.9b represents the deviations in time series among wind
58 Chapter 2
turbines samples without correlation. Uncorrelated samples show large devia-
tions.
In order to obtain accurate outputs from wind turbines, after conducting
ecient sampling and correlation techniques among separated wind turbines,
a curve tting for the wind turbines must be conducted, which can take into
consideration null wind speed samples. Therefore, steps for conducting curve
tting of the proposed wind turbine in this Chapter will be shown in the next
Section.
2.4 Curve Fitting of the Wind Turbine Enercon-

E40
A simple model of wind turbine curve tting derivation has been proposed
in [77], but this model did not account well for dierent regions of operation;
moreover, null wind speed region was not well considered.
In order to correctly t the power curve of a wind turbine, this Section
proposes a curve tting which can describe well the nonlinearity of the turbine
operation through all possible wind speed ranges, including the null wind speed
region.
The selected turbine for tting is Enercon-E40 600 kW rated power, cut-in
(wi ), cut-out (wf ) and rated speeds (wr ): 2.5, 28 and 12.5 m/sec respectively.
The collected wind turbine energy is null below wi ; also wind turbine energy
is null above wf , because of dramatic increase in wind energy, and so the
pitch control is activated to protect the mechanical parts of the turbine. The
blending speed (wm =10 m/s) is selected to separate two regions (II) and (III),
as shown in Figure 2.10, in order to reduce the tting errors which may occur
between the two regions.
The selection of the blending speed wm is based on nding a reduced least
square residuals of the turbine curve. Contrary to [78], where the author
applied two third polynomials to t the turbine curve and resulted in diculties
for curve inversion.
The dierent operating regions of the wind turbine are shown in Figure 2.10
and can be justied as follows:
• Regions I and V, where w < wi and w > wf ;

Chapter 2 59
• Regions II and III, where the wind power varies according to wind speed,
as wi ≤ w ≤ wr ;
• Region IV, where the turbine power is always constant and equals to the
rated power, as wr ≤ w ≤ wf .
The tting of the whole wind turbine is shown in Figure 2.10, manufacturer
data are shown by diamonds; also dierent operating ranges of the wind turbine
are cleared by equation (2.4).


 0 w < wi or w > wf ,


if

Pg = a3 w3 + a0 if wi ≤ w ≤ wr , (2.4)




 b1
if wm ≤ w ≤ wr .
1+e( −(w−b2 )b3 )
Considering, a3 = 6.847 × 10−4 , a0 = −9.332 × 10−3 , b1 = 1.001, b2 = 9.278

and b3 = 1.074.
After conducting the curve tting, a piecewise-denition of the wind power
cumulative distribution function (CDF) is presented in the next Section, in
order to obtain the wind power samples directly from wind speed samples by
an inversion process.
2.5 Cumulative Distribution of Wind Power

The piecewise-denition of the wind power cumulative distribution function
(CDF) is conducted in this Section, in order to inverse the operating regions of
the wind turbine. It can serve to determine directly the turbine power from the
probability function of the wind speed. Thereafter, the CDF can be employed
to calculate the capacity factor of each wind turbine that serves the microgrid.
In the previous Section, a direct relation to obtain the turbine power sam-
ples from wind speed samples was shown as Pg = g(w), such a relation makes
it dicult to infer a closed form of the (power CDF) FPg (y) from the (wind
−1
CDF) Fw (g (y)). As a result of the turbine power curve has separate and
dierent representations, the denition of the wind power cumulative distri-
bution function must be inferred by another way. As shown in Figure 2.10,
there are ve operating regions of the wind turbine, and so the power CDF
must be divided also into ve separate regions (FP I (y) , FP II (y) , FP III (y) , FP IV (y)
g g g g
60 Chapter 2
0.8
turbine power (p.u)
0.6
0.4
0.2
0
0 5 10 15 20 25 30
wind speed (m/s)
Figure 2.10: Curve tting of Enercon-E40 600 kW, man-

ufacturer data indicated by (asterisks).
and FPgV (y) ) as indicated by equation (2.5).


1 + Fw (wi ) − Fw (wf )

 if w < wi or w > wf ,




1 + Fw (g −1 (y) − Fw (wf )) if wi ≤ w ≤ wm ,
FPg (y) = (2.5)



1 + Fw (g −1 (y) − Fw (wf )) if wm ≤ w ≤ wr ,




1 if w ≥ wf
Figure 2.11a shows the cumulative distribution of the wind power at wind
speed scale parameter equals to three, while Figure 2.11b shows the cumulative
distribution of the wind power at wind speed scale parameter equals to eight.
Clearly, if the wind scale parameter is low, then a higher probability of calm
wind speed samples, and subsequently null wind power production results.
After conducting the formula of the probabilistic wind power CDF, now
the turn is for a wind power sample derivation (Pg ) from a given wind speed
Chapter 2 61
1
0.8
0.6
cdf
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
power samples (p.u)
(a)
1
0.8
0.6
cdf
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
power samples (p.u)
(b)
Figure 2.11: Cumulative distribution of Enercon-600kW:

(a) Processed at wind speed scale parameter (3). (b) Pro-
cessed at wind speed scale parameter (8).
probability sample ξ, as indicated by equation (2.6).


 0 ≤ ξ ≤ FPg (0),

 0 if




F −1II if FPg (0) ≤ ξ < FPg (g(wm )),
P (ξ)
g
FPg−1 (ξ) = (2.6)



 FP−1III (ξ) if FPg (g(wm )) ≤ ξ < FPg (g(wr )),

 g


1 if otherwise.
The proposed microgrid layout is detailed in the following Section, in order
to show the locations of the wind turbines where the wind speed samples are
applied, the static loads and the microgrid supporting units (MSUs).
2.6 Microgrid Layout

The proposed isolated microgrid is congured as shown in Figure 2.12, and
the main elements of this microgrid are briefed hereunder:

62 Chapter 2
• Fourteen nodes distribution system;
• Five microgrid supporting units (MSU), which are fed from Enercon-E40
wind turbines at nodes (1, 6, 8, 12 and 14);
• Distributed static loads at all microgrid nodes;
• Three dynamic loads, mainly induction motors at nodes (2, 4 and 9).
The power and frequency are regulated by the droop control constants
−6
(5 × 10 rad/W/s) of the MSUs, the nominal power of each MSU is 0.12 MW
and the base power of the microgrid is 1.27 MW. Thus, each wind turbine has
0.472 p.u, as a rated power. The balance between the demand and supply
is handled by the droop control strategy. Figures 2.13 and 2.14 show P −ω
and v − Q droop control by MSU interfaces. Equations (2.7) and (2.8) explain
the operational characteristic of the droop regulation by each MSU in the
microgrid, where ω is the operating frequency, ωmax is the maximum frequency,

P is the active operating MSU power, Q is the reactive operating MSU power,
mp is the active droop constant, nq is the reactive droop constant, ωmax is the
maximum frequency and vmax is the maximum voltage.
ω0 − ω1 = −mp (P1 − P0 ) (2.7)
v0 − v1 = −nq (Q1 − Q0 ) (2.8)

Chapter 2 63
The next step is to obtain the maximum loadability of the microgird, in
Figure 2.13: Active power vs frequency droop character-

istics
Figure 2.14: Reactive power vs voltage droop character-

istics
order to evaluate the capacity factor of the wind turbines. A global optimiza-
tion technique Global Search will be used in the following Section to obtain
64 Chapter 2
the maximum loadability of the microgrid, overcoming the shortcomings from
other optimization techniques, which usually attain only local maximum load-
abilties.
2.7 Nonlinear Optimization

The maximum loadabiltiy problem is aimed at determining the maximum
power demand that the microgird can meet. The maximum loadability can
provide an insight about the capacity factor of droop-regulated microgrids.
The maximum loadability can be conducted through non-linear optimization
problem.
The applied optimization function to attain the maximum loadability of
the microgrid is shown by equation (2.9). Where θ is the objective function
that represents load increment; the set x represents the microgrid variables; h
and g respectively represent the microgrid equality and inequality constraints.
maxθ θ
s.t h(x, θ) = 0 g(x, θ) ≤ 0 (2.9)
The mathematical optimization is usually considered as selection of a best
element from some set of available alternatives. This process must be employed
within nonlinear programming problem (NLP) [79].
The NLP is the process of solving a system of equalities and inequalities,
collectively termed constraints over a set of unknown real variables along with
an objective function, to be maximized or minimized, where some of the con-
straints or the objective function are nonlinear.
The general nonlinear programming problems are represented by the fol-
lowing form:
min f (x)
s.t h(x) = 0 g(x) ≤ 0 (2.10)
The constraints h(x) = 0, g(x) ≤ 0 are considered as the functional con-
straints. The inequality constraints g(x) ≤ 0 are said to be active at a feasible

point x, if g(x) = 0 and inactive at x, if g(x) < 0.
The necessary conditions to obtain the point of global optimization are
dened by two conditions. Therefore, the two main conditions must be met
Chapter 2 65
by an optimization solver.
The rst condition is dedicated to obtain the local point of optimization
and the second condition is to ensure the global point of optimization.
The First-Order Necessary Condition
∗
By assuming (x ) to be a point satisfying the constraints of equation 2.11,
and J is a set of indices j for which gj (x∗ ) = 0 , x∗ is regular point of the
constraints (2.11), if the gradient vectors ∇hi (x ), ∇gj (x∗ )

∗
are linearly inde-
pendent, then x∗ is considered as regular minimum point for the problem
(Karush-Kuhn-Tucker condition). The optimization function is considered by
the form shown in equation (2.12); also complementary slackness condition
must be met, as shown from equations (2.13) and (2.14). Stating gj (x∗ ) < 0
implies that µj = 0 and µj > 0 implies gj (x∗ ) = 0.
h(x∗ ) = 0, g(x∗ ) ≤ 0 (2.11)
min f (x)
s.t h(x) = 0 g(x) ≤ 0 (2.12)
∇f (x∗ ) + λT ∇h(x∗ ) + µT ∇g(x∗ ) = 0 (2.13)
µT g(x∗ ) = 0 (2.14)
The Second Order Condition
The sucient conditions for a point x∗ satisfying (2.11) to be a strict rel-
ative minimum is to realize equations (2.15)-(2.17). Moreover, the second
derivative of (2.17) must be positive denite.
µ≥0 (2.15)
µT g(x∗ ) = 0 (2.16)
∇f (x∗ ) + λT ∇h(x∗ ) + µT ∇g(x∗ ) = 0 (2.17)
2.7.1 Global Optimization Search
General nonlinear optimization problems are dicult to solve due to a large
number of local minima in the search space. A good local minima are dicult
66 Chapter 2
to be found by local search methods, because solver stops at each local min-
imum. Therefore, to obtain global optimal solutions, a global search method
is provided in this Section, to realize only global peaks.
2.7.2 Applied Global Search Optimization Methodology
The applied Global Search solver in this Section presents an attempt to
locate the optimum solution as shown in Figure 2.15. This methodology can
eciently realize the global solution; moreover, it converges with high dimen-
sional systems.
Figure 2.15 explains the steps of this technique. The main steps of Global
Search can be briefed as follows:
1. The global solver generates trial points, by employing a Scatter Search
method. These trial points are then ltered by checking an initial condi-
tion, as if the distance between the trial points and expected local solver
points are greater than the preset standard distance, then these trial
point are accepted and the optimization process starts from each of the
ltered points;
2. A continuous direction update from the trial points to local solvers is
processed, taking into consideration the following condition:
• If the penalty Factor, which is used to determine the update amount
of a set of rejected trial point threshold is greater than this threshold
value, then the threshold value must be updated.
3. Check if all trial points are rejected, if so the global solver generates
other starting trial points and repeats the optimization process again;
4. After analyzing a set of accepted trial points by the scatter search algo-
rithm, the best quality point is chosen.

Chapter 2 67
Generating and filtering trial

points by employing the
scatter search
Yes
The Maximum
Update the step size to No
number of trials
local solvers exists?
Penality factor > Yes

Update the threshold value
threshold?
No
Choosing the best trail point
End
Figure 2.15: Flow chart of the Global Search

68 Chapter 2
2.8 Processing Simualtions and Conclusions
Figure 2.16 shows histograms of 100 samples of injected power from each
wind turbine, interfaced by a droop-regulated MSU, at droop constants mp =

0.5 p.u. The wind distribution is assumed as Weibull of scale 8.0 and shape
1.4. As mentioned from Section 2.6, the base power of the whole microgrid
is 1.27 MW, while the rated power of each wind turbine is 600 kW, which
represents 0.47 p.u.
Figure 2.16 contains two arrows, arrow 1 indicates a bulk of ve histograms,
each histogram represents the frequency of occurrence of each wind turbine
power samples. Arrow 2 indicates the total injected power to the microgrid.
The group of power histograms represented by arrow 1 ranges from 0to
0.47p.u, while the total injected power represented by arrow 2 ranges from 0
to 2.74p.u.
It is observed that each histogram of the ve histograms related to arrow 1
in Figure 2.16 have two peaks at the beginning and the end of each histogram;
left peak represents the null power production, when the wind speed samples
are lower than the cut-in speed of the wind turbine, whereas the right peak
represents the rated power production, if the wind speed samples are higher
than the rated speed of the wind turbine.
Worthy to mention, the two peaks left and right ones of each histogram
depend on the stochastic parameters of the wind speed. If the wind scale pa-
rameter is high, then the histogram will shift to the right side much more and
the frequency of the injected power will be distributed all over the horizontal
axis of the injected power domain. On the other hand, if the wind scale pa-
rameter is low, then the left histogram peak will be more salient and the right
histogram peak will be nearly invisible.
An important conclusion must be mentioned regarding Figure 2.16 that
the distribution frequency of the total injected power to the grid diers from
the total injected power by all wind generators in microgrid. Therefore, the
injected power estimation into the microgrid cannot be measured by multi-
plying directly the injected power of one generator by number of generators
connected to the grid. Regardless the losses in each branch among the micro-
grid nodes, the correlations among generators plays an important role in such
discrepancies.
Chapter 2 69
6
1
frequency
2
2
1
0
0 0.5 1 1.5 2 2.5
real power (p.u)
Figure 2.16: Histograms of wind turbines and microgrid

power samples. Individual powers indicated by (thin lines)
and aggregated injected power indicated by (blue, thick
line).
Unlike many researchers who have concentrated on nding the distribu-
tion functions for wind speed samples only, the model selection criteria will
be applied in the next Section to obtain the distribution tting of the injected
power to the microgrid and to indicate that distribution ttings may change
depending on the mean wind speed and the wind turbine characteristic.
2.9 Model Selection

Usually, most papers have discussed the derivation of wind speed distribu-
tion functions by model selection criteria, in order to obtain the wind power
samples from wind turbines curve ttings.
In this Section, the model selection criteria are conducted to infer the dis-
tribution of the output power from the wind turbines and microgrid directly.
Therefore, the analysis can be used as an index for the total microgrid energy
production and capacity factor, according to mean wind speeds, wind turbine
70 Chapter 2
characteristics and droop constants of MSU interfaces. Consequently, an e-
cient model selection criterion must be adopted.
Many recent researches are concerned with the question of the model choice;
researchers usually collect data in form of measurements and observations; con-
sequently they are in need to study how such data aect the outcome. Both fre-
quentist and Bayesian schools have discussed this issue by dierent methodolo-
gies, such as Spawning methods, Akaike Information Criterion (AIC), Bayesian
Information Criterion (BIC) and others [80].
If there are K models, indexed by k, model k has parameters ϑk ∈ Ωk ,

then the whole parameter space is (M, ϑ1 , ϑ2 , . . . . . . ., ϑk ), where M denotes
the model. In each statistical model, the estimation may be thought as the
choice of a single value of the parameter chosen according to some criterion
to represent the distribution. There may be occasions just as with estimation
in general, when one model clearly dominates the others that the choice is
unobjectionable, and other occasions, when the choice is misleading.
Model selection is best seen as a way of approximating, rather than iden-
tifying full reality. The idea can be briefed by adding parameters to a model,
an improvement in a distribution tting is obtained to some degree, but at the
same time the parameter estimates are worse, because less data per parameter
exist.
Many techniques have been developed to cope with the diculty of dis-
tinguishing among true statistical dierences. Typically, any tted model is
never going to be a true model, and so the eciency of a statistical modelling is
based on its ability to compare large numbers of competing statistical models
simultaneously.
The Information criteria balance the model complexity with a goodness-of-
t, to select a simple model with the highest predictive power in a parsimonious
fashion [81]. Comparability and parsimony are basic traits in informational
modelling not found in classical statistical methods.
The model selection is a classical topic in statistics, the idea of selecting
a model via penalizing a log-likelihood type criterion goes back to the early
seventies with the pioneering works of Mallows and Akaike.
Akaike and Bayesian methods can compare multiple nested or non-nested
models, examine model uncertainty and estimate model parameters eciently.
Therefore, the two methods, Akaike Information Criterion [82] and the Bayesian
Chapter 2 71
Information Criterion [83] are applied through this Chapter; the aim of these
methods is to penalize the error of the model on the training data, to estimate
the error on unseen cases.
The objective is to estimate distribution ttings of the injected power to
the microgrid. Before proceeding in the both criteria, the error measure (like-
lihood) must be introduced, because both criteria AIC and BIC depend on the
likelihood measure.
2.9.1 The likelihood measure
The likelihood concept is used to measure the strength of statistical ev-
idence; it equates the amount of evidence against a null hypothesis. This
measure is a probability, and so it is bounded between [0, 1], one represents
the certainty that the model can generate the data set and zero represents the
impossibility that this model is able to generate the data.
The likelihood measure can be represented as shown in equation (2.18).
∏
N
l[P (X|Θ)] = P (xi |ϑ) (2.18)
i=1
Where l[•] is the likelihood measure, X is the data set, xi is a particular
datum, N is the number of observations and ϑ is the list of the model param-
eters. The likelihood is always a very small number which results in underow
in modern computers. A good solution to the underow problem is to use the
logarithmic transformation, as shown by equation (2.19).
∏
N
log(l[P (X|Θ)]) = log[ P (xi |ϑ)]
i=1
∑
N
= log([P (xi |ϑ)])
i=1
= ll[P (X|Θ)] (2.19)
Where ll[•] is the log-likelihood. This new function is bounded between
[-∞, 0], minus innity represents the incapacity of the model to generate the
data set and zero represents the certitude that this model can eciently gen-
erate the data. This criteria is to evaluate the error in measurements and is
called negative log-likelihood.

72 Chapter 2
2.9.2 Akaike Information Criterion
One of the most commonly used information criteria is AIC. The idea of
AIC [82] is to select the model that minimizes the negative likelihood penalized
by the number of parameters as shown from equation (2.20).
AIC = 2[−ll(data)] + 2d (2.20)
Considering N is the number of observations and d is the number of free
parameters. Only the error measure (−ll[•]) and the number of the free pa-
rameters in the model are needed. The model which has the minimum AIC
value will be chosen. AIC is aimed at nding the best approximating model
to the unknown true data, while penalizing the model for its complexity.
2.9.3 Bayesian Information Criterion
The BIC [83] is similar to the AIC [82], except that it is motivated by
the Bayesian model selection principle [84]. In addition to the information
required to compute the AIC, the number of observations must be provided.
The Bayesian has the following form:
BIC = 2[−ll(data)] + log[N ]d (2.21)
BIC diers from AIC only in the second term, which depends on the sample
size N. The model with the smallest BIC is equivalent to the model with the
highest probability. BIC can be used not only to choose the best model, but
also to assess the merit of each of the tested models.
In [85], the simulation results indicated the ability of AIC to select a true
model rapidly increases with the sample size, but at larger sample sizes it con-
tinued to exhibit a slight tendency to select complex models.
Contrary to AIC, BIC tends to outperform AIC in large sample size and
appears to perform relatively poorly in small samples size.
By employing both AIC and BIC for obtaining the best probabilistic rep-
resentation of the injected power to the microgrid, the models with the lowest
AIC and BIC are considered as the best distribution probability of the micorgid
power. By taking into consideration there is no unique distribution tting for
all scenarios, the distribution tting always depends on the mean wind speeds
and characteristics of wind turbines; thus, it may change from one scenario to
Chapter 2 73
another.
Figure 2.17 explains the goodness of t of the total injected power by 100
samples to the micorgrid, at scale and shape parameters of wind speed 8 and
1.4 respectively. The distribution tting is plotted by cumulative distribution
functions. Figure 2.18, explains the error between the empirical distribution
and each tted distribution.
The concept of the choice is based on choosing the minimum value of AIC
and BIC as shown from Table 2.1. The minimum values record always rst;
such tted distributions depend also on the stochastic parameters of the wind
speed.
However, the main purpose is beyond estimation of distribution ttings of
the injected power. Usually, the main reason of the wind energy analysis is
always to determine the capacity factor, which wind turbines can provide, in
order to evaluate their feasibilities for service. The capacity factor calcula-
tions for both wind turbines and the whole microgid will be shown in the next
Section.
0.9
0.8
0.7
0.6
cdf
0.5
0.4
0.3
empirical
0.2 generalized pareto
exponential
0.1 generalized extreme value
normal
0
0 0.5 1 1.5 2 2.5 3
active power (p.u)
Figure 2.17: Cumulative distribution functions of the mi-

crogrid injected power processed at wind speed scale and
shape parameters (8) and (1.4) respectively.
74 Chapter 2
Table 2.1: Model Selection
Input weibull
scale 8
shape 1.4
Generalized pareto (1)
BIC 171.66
AIC 163.84
shape -1.12
scale 2.77
threshold 0
Exponential (2)
BIC 212.70
AIC 210.09
scale 1.04
Generalized extreme value (3)
BIC 265.62
AIC 257.80
shape 1.89
scale 0.29
location 0.14
Normal (4)
BIC 276.40
AIC 271.19
location 1.04
scale 0.92
Chapter 2 75
generalized pareto
0.6 exponential
generalized extreme value
0.5 normal
0.4
0.3
error
0.2
0.1
−0.1
−0.2
−0.3
0 0.5 1 1.5 2 2.5
active power (p.u)
Figure 2.18: Errors between the empirical distribution

and a set of tted distirbution functions of the microgrid
injected power processed at wind speed scale and shape
parameters (8) and (1.4) respectively.
2.10 Microgrid Capacity Factor

As a result of continuous growth in the wind energy penetration in the
power systems around the world, it has been a must to consider the capacity
credits of the wind energy systems. In [86] and [87], the authors focused on
the capacity credit calculations of the wind energy system, which is associated
with the average wind power.
The increase in the penetration of the wind generation in the recent years
has led to a number of challenges for the planning and operation of the power
systems.
The power system reliability is divided into two basic aspects, system se-
curity and system adequacy as shown in Figure 2.1. The system is secure,
if it can be up to a loss of power supply as generators or transmission lines.
Adequacy can be attained by a set of generators probably have dierent char-
acteristics; the capacity of each individual generator is the contribution that

76 Chapter 2
this individual generator can ensure to overall system [88]-[89].
Adequacy can be dened as the amount of additional loads which can be
served due to the addition in generation capacity, and at the same time main-
taining the existing levels of reliability. The variability of the stochastic nature
of wind rises the challenges must be met by wind generators.
The capacity factor (CF) of a wind generator is the ratio of the actual out-
put turbine power over a period of time, to its full nameplate capacity output
for the same period of time. For wind generation, the capacity factor depends
on the probability distribution of the wind speed, also on the characteristics
of the wind turbine, such as cut-in, cut-out and rated wind speeds. It is a
percentage of the electrical power that a wind turbine is able to supply from
an available wind at a specic wind farm.
The CF provides a summarized indication about the combined interaction
between a wind generator and a site that is between the wind power curve and
the stochastic wind speed. CF was introduced in other researches as a simple
quotient between delivered and rated energy over a given period of time as
in [90], whereas in [40] and [91], it was conducted from the mean cubic wind
speed.
Through this Section, the capacity factor of the wind energy is studied to
serve an important issue which is the site-matching with the isolated micro-
grid. The comparison between the capacity factor of each wind turbine and
the capacity factor of the whole microgrid is a must, because the capacity fac-
tor of the droop-regulated microgrid cannot be considered as a large generator
that is built up from several minor wind generation units, because the droop
control imposes restrictions on the maximum loadability.
Table 2.2 shows a summary of the mean injected power to the microgrid,
the capacity factor of the individual wind turbines and the capacity factor of
the whole microgrid. The capacity factor of the wind turbines ranges from
42.3% to 45.6%, while the capacity factor of the whole microgrid records only
42.16%. Therefore, it can be concluded that the droop regulation of the MSU
interfaces causes a drop in the capacity factor of the microgrid. The droop con-
stants of MSUs play an important role in restricting the maximum power of the
micorgrid. This conclusion explains the behaviour of MSU when it reaches its
Pmax , which indicates the start of frequency degradation limit. Consequently,
whenever the bifurcation occurs, the microgrid can never support any more
Chapter 2 77
power.
Moreover, the capacity factor of each wind turbine is dierent from the
other turbines, and also dierent from the capacity factor of the whole micro-
grid, as shown in table 2.2. This conclusion indicates a possibility of dierent
bifurcation limits of the microgrid, under dierent mean wind speeds and MSUs
droop constants. The detailed bifurcation analysis will be conducted in the
following Section.
Table 2.2: Microgrid Capacity Factor
Input weibull
scale 8
shape 1.4
Enercon E40
Mean power (p.u) 0.19 ≤P ≤ 0.215
Capacity factor 42.3% ≤ CF ≤ 45.6%
Microgrid
Mean power (p.u) 1.04
Capacity factor 42.16%

78 Chapter 2
2.11 Microgrid Bifurcation Analysis
The load demand increase leads to negative consequences on voltage stabil-
ity; typically, a system enters a state of voltage instability when an increase in
load demand causes a progressive and an uncontrollable drop in voltage. Dur-
ing such instability moments, the reactive power demand may be higher than
the supply reactive power and the voltage starts to decrease. Consequently,
the dierence between the reactive power supply and reactive power demand
increases and the voltage falls to a very small value (Voltage Collapse).
Nowadays, power networks are characterised by heavy loading and opera-
tions at their maximum stable margins [92] and [93].
The continuation power ow (CPF) is a technique used to compute the
stability margin, as the distance to saddle node bifurcation (SNB) from the
current loading point, by increasing the loading level until a voltage, a current
or a voltage stability limit is detected in the power ow model. The CPF is
based on a predictor-corrector scheme to nd the complete equilibrium prole
or bifurcation (PV curve) of a set of power ow equations, with respect to a
given scalar variable. This scalar parameter is typically known as the bifurca-
tion parameter or loading factor.
The previous results of capacity factor in Section 2.10 can be interpreted by
another way via applying CPF to the maximum microgrid loadability samples
as shown in Figure 2.19. The gure explains the behaviour of MSU interfaces
when restricting their Pmax ; solid line curve shows MSUs without power limits;
dashed line curve shows one MSU with power limitation; and dotted line curve
shows two MSUs with power limitation.
The CPF explains well the behaviour of the microgrid when tightening the
power limitation of MSU interfaces, as each MSU can deliver power from zero
to Pmax , while when it reaches its Pmax , the voltage degradation limit starts
to occur. Consequently, the bifurcation occurs and the microgrid can never
support any more power.
It is clear that the more restriction on power limitation, the earlier bifurca-
tion will occur, because the voltage is continuously degraded down to a point
in which a saddle-node bifurcation occurs; the same results were shown in [94].
As bifurcation limit appears under dierent microgrid power limits, mainly
related to wind power availability, as shown in Figure 2.19. Thereby, several

Chapter 2 79
simulations under dierent scenarios, such as dierent wind turbines, dierent
wind speeds and dierent MSUs droop constants will be shown in the follow-
ing Section, in order to clear this phenomena and observe which is the most
reliable scenario, and so the microgrid can meet the increased demand.
0.95
0.82
voltage (p.u)
0.70
0.58
0.46
3.00 3.37 3.75 4.11 4.48
load (p.u)
Figure 2.19: Microgrid loadability by CPF. Solid line:

MSUs without power limits; dashed line: one MSU with
power limitation; dotted line: two MSUs with power limi-
tation. Source: [1]
2.12 Case Studies

Several cases are summarized in Figure 2.20, to highlight the eect of droop
regulation and wind speed stochastic parameters on the microgrid capacity
factor. All cases of simulations in Figure 2.20 are processed for 100 sample,
based on [1]; each box represents the lower and upper quartiles (Q1 and Q3);
the horizontal line within the box shows the median (Q1), while maximum and
minimum values are represented by whiskers; the average value of each case is
plotted by (⋄), while capacity factor of each case is plotted by (∗). The details
of each case are shown hereunder:
1. Case1−→ wind speed scale and shape parameters: 8, 1.4 respectively,
while the droop regulation constant is 0.5 of the base droop, and the
employed wind turbines are Enercon E40-600 kW;
80 Chapter 2
while the droop regulation constant is 0.5 of the base droop, Enercon
E40-600 kW turbines located at nodes 12 and 14 had been replaced by
Vestas V52-850 kW;
5. Case5−→ wind speedscale and shape parameters: 8, 1.4 respectively,
E40-600 kW turbines located at nodes 12 and 14 is replaced by Vestas
V52-850 kW;
V52-850 kW;
V52-850 kW;
E40-600 kW turbines located at nodes 12 and 14 is replaced by Vestas-
V52 850 kW.

Chapter 2 81
delivered power (p.u.)
case no.
Figure 2.20: Box plot of several scenarios.
The cases from one to three explain the gradual increase in the maximum
power limitation of each MSU; as shown, the highest limitation occurs when
the droop is high, as in the rst case. Thus, both the average power and ca-
pacity factor are lower than the third case, because the droop of MSUs is the
highest among the three cases.
The cases from four to six ensure again the same results obtained from
cases one to three, as some of wind turbines are replaced by higher capacity
wind turbines. Therefore, the results as shown from Figure 2.20 indicate the
mean power and the capacity factor of the cases four to six are slightly higher
than their correspondents of the cases from one to three. By decreasing the
MSUs droops, the capacity factor and the average power rise again.
One of the interesting results that the cases nine and ten, where the wind
scale parameter equals six have average power and capacity factor higher than
cases one and two, where wind scale parameter equals eight. The clue of this
conclusion can be veried, as both the rst and second cases show a higher
restriction on MSUs by high droops, and this conclusion clears the dramatic
eect of the high MSU droop constant.
The previous cases show the eect of changing the droops of MSU inter-
faces, as the highest droop leads to the highest restriction on the delivered
power. The conditions are dierent in cases seven to ten, as the parameters
82 Chapter 2
which are tested through these cases are the stochastic parameters of the wind
speed. The scale parameter is reduced to four in cases seven and eight, while
in cases nine and ten the scale parameter is six. Figure 2.20 shows the lowest
scale parameter causes a dramatic fall in the average power and the capacity
factor of the grid.
Regarding Figure 2.20, it is clear that the delivered power varies according
to the wind turbines characteristics, the mean wind speed and the MSUs droop
constants.
The results of this Chapter show that the capacity factor of droop-regulated
microgrids supplied by wind generators depends on some terms, such as wind
speed stochastic parameters, droop regulation values of MSUs and the char-
acteristic of wind turbines. Unlike conventional generation units, where the
capacity and reserve of each conventional generator is known in advance, the
capacity and reserve of a microgrid supplied by wind turbines are vital issues,
as it aects both the stability and economic performance of the whole micro-
gird.
In Chapter3, the idea of site matching suitability for droop-regulated mi-
crogrids supplied by wind generators and dierent energy storage systems is
more detailed, by considering analysis for both energy production and eco-
nomic evaluations.
Chapter 3 83
Chapter 3
Benet Evaluation of Microgrids

Using Capacity Factor and
Economic Analysis
3.1 Introduction
In rural areas, it is very dicult as well as uneconomical to transmit power
over long distances through transmission lines to supply such areas. The lack
of an electrical network to supply remote areas, high connection cost of grid
extension and rough topography often leads to other options to supply energy.
Stand-alone hybrid systems which are dependent on renewable sources are
found promising ways to satisfy the energy supply requirements for these areas
[95].
The need for ecient electric power sources in remote locations is a driving
force for research in hybrid energy systems [96].
In [97], the research focused on economical drawbacks of supplying energy
to rural areas, as in the case of Alaskan communities. Alaskan communities pay
economic penalties for electricity, because they import diesel as a primary fuel
for electric power production, paying heavy transportation costs. In addition,
the local negative environmental impact caused by excessive consumption of
fossil fuels. In these remote locations, renewable resources and advanced tech-
nologies, coupled with energy storage systems, can compete eciently with
conventional generation sources.
The wind power is considered as one of the main possible energy sources
84 Chapter 3
in operation of rural areas, similarly to microgrids. The huge growth in the
wind power utilization through last years, environmental concerns and the ris-
ing cost of fossil fuel generation in many developing parts of the world have
led the energy cost of the wind power to fall one seventh of the cost in the
early 1980s [98]. In the United States, the wind energy production has shown
average annual growth 24% during the past 5 years [99]. Also in Europe, the
growth of wind energy production has shown extraordinary records before 2010
[100].
Typically, the cost of conventional electricity production is determined by
three main components: (1) Fuel cost. (2) Operation and maintenance costs.
(3) Capital cost.
When conventional generation units are substituted by wind energy, the
avoided energy cost by wind power depends on the degree to which wind power
substitutes each of the three components.
In general, wind energy avoids the full fuel cost and a considerable portion
of Operation and maintenance costs of the displaced conventional units. How-
ever, the level of avoided capital costs depends on the extent to which wind
power capacity can displace the conventional units, and so this level is directly
related to the capacity factor of wind plants [55].
However, the drawbacks of the wind power production is based mainly on
the uncertainty nature that it is unpredicted source of energy, and so the wind
power production cannot be estimated accurately.
Wind speed is characterized by its high variability spatially and temporally.

Therefore, the planners and the operators of the wind plants are concerned
about the eect of the wind speed variation on the operating costs of the sys-
tems. The system must maintain the balance between the demand and total
generated power from the wind energy. The costs associated with maintaining
this balance are referred as to ancillary-services costs [101].
With the market liberalization and occurrence of competitive markets, vi-
tal questions are always conducted regarding the optimum economical sizing
of intermittent wind generators, to contribute to supplying microgrids with
constraints of power quality. For this sake, technical and economic analyses
of wind energy sources are necessary, in order to meet the requirements in
competitive markets.
In the new model of power markets, the plant dispatch is performed by

Chapter 3 85
independent system operators, and the contracts between generators and cus-
tomers are either negotiated bilaterally or accomplished through power pools,
this subsequently means a drastic increase in risk for generators [102]. Hence,
the investors tend to favour generation technologies characterised by low capi-
tal costs, such as gas turbines, despite their high operating costs. Consequently,
the wind energy, which is characterised by high capital cost, is hurt by the shift
to low capital cost technologies.
Moreover, in a competitive market, the generators must submit bids a week,
day or hour in advance, oering to provide a given quantity of generation at
a given price during a particular hour. The failure to meet the committed
generation capacity as declared on the previous day is penalised.
As the wind is a uctuating energy source, the viability of wind power in
day-ahead markets is highly dependent on several factors, such as accurate
wind speed forecasting, degree of penalties charged by the market operator to
generators which are unable to meet their commitments and nally the wind
generators eciencies to provide power reserves as much as possible [103].
As a result, there is a growing diculty for nancing wind power projects
in the absence of a guaranteed long-term revenue stream, leading to a serious
challenge for continued wind power development.
This chapter performs several economic evaluations and capacity factor
analysis for microgrids, supplied by wind energy and energy storage systems,
in order to show a clearer perspective of the most ecient generation proles
from both economic evaluation models and generation capacities views. These
proles can be adopted to enhance and promote the wind energy contributions
in competitive markets.
In this Chapter, the results of economic evaluations and energy productions
are illustrated; moreover, pie charts are shown, to explain the daily power gen-
eration percentage, distributed between the load demand and system reserve
unit, to clarify the ability of microgrids supplied these generation sources to
meet load demands and provide power reserves as well.
In general, the most ecient wind turbine might not be the most eective
for transforming wind power into electric power on a specic site, and so the
main goal is to maximize the wind annual energy production [39]. The energy
production of a wind turbine depends on several factors, such as wind speed
conditions from the area and the characteristics of the wind turbine itself, in-
86 Chapter 3
cluding the cut-in, rated and cut-o wind speed parameters.
The suitability of a wind turbine to a specic location is given by the ca-
pacity factor. The selection of optimal wind turbines has been discussed in
some papers based on maximization of the capacity factor [40]. The choice of
turbines involves choosing parameters that lead to maximizing this factor.
In conclusion, the optimization of the wind turbine components leads to
magnifying the wind turbine output and contributes to a reduction of the cost
of energy [104], which is the main parameter used to compare economic per-
formances of systems.
In addition, other studies have been conducted regarding site matching of
wind turbines. The site matching is based on identifying optimum turbine
parameters to yield higher energy production at higher capacity factor.
The matching of a wind turbine to a site using normalized power and
capacity factor curves were discussed in [40], but in general the economic con-
siderations of the turbine are not provided through wind site matching studies.
Other studies have focused on the matching of wind turbines to wind farms
from the viewpoint of both performance and economic considerations, as in [41]
and [42].
However, for isolated microgrids, storage systems are always required to
share microgrids supply with wind energy sources, as the generation must al-
ways be as close as possible to system loads. The problem of keeping the power
balance is still more dicult for stand-alone microgrids supplied by intermit-
tent generators. The characteristics of such grids require scheduling more
reserve for ensuring adequate security and reliability levels, but the higher
reserve requirements may substantially deteriorate the economy of these sup-
ply systems. The total requirements for these storage systems are unclear at
present, because most of the recent studies which have optimized the capacity
factor and economics of wind energy production, have not taken into consid-
eration the operation of storage systems that share supply with wind energy.
Thus, the optimum point of operation is totally dierent when considering the
dynamics of storage systems.
The performance and economic aspects of such storage systems are signif-
icant and must be brought into the economic evaluation, so the microgrid can
work at optimum conditions in terms of investment and power system reliabil-
ity requirement.
Chapter 3 87
Therefore, the proposed study of this Chapter is based on optimizing both
energy production and economic performance for isolated microgrids supplied
by wind energy and various batteries, by a multi-objective function. The opti-
mization function in this Chapter is conducted to minimize the Cost of Energy
(COE), which is a function of both capital costs and energy production of
the generation sources. Hence, the maximum possible energy production and
the lowest cost of energy can be ensured, taking into consideration the droop
regulation of microgrid interfaces and the dynamic operation of the batteries
(charging-discharging). Moreover, the economic analyses are performed also
by Simple Payback (SPB) and Net Present Value (NPV).
Before conducting the proposed simulations, the sources of data which are used
through this Chapter will be shown in the following Section.
3.2 Sources of Data

The sources of data include the layout of the proposed microgrid, the droop
regulation of MSUs by Fischer-Burmeister algorithm, the daily operation of
the batteries, the characteristics of several wind turbines and nally the em-
ployed economic evaluation models in the Chapter.
3.2.1 Microgrid Layout
The layout of the microgrid is shown in Figure 3.1. The power and frequency
of MSUs are regulated by droop constants: (9.4 × 10−6 rad/W/s); the base
power of the microgrid is 900 kW; the rated power of each MSU is 5 p.u. The
main conguration of the microgrid is briefed by the following points:
• Three wind turbine at nodes (6, 12 and 14), the rating of these turbines
will be set according to purpose of each simulation and will be specied
in the following Sections;
• Three microgrid supporting unints supplied by Lead Acid batteries at
nodes (1, 4 and 8);
• Distributed static loads at all microgrid nodes;

88 Chapter 3
• Group of aggregated motors at node (4);
• System Reserve Unit (SRU); the function of SRU is to store the excess
of generated power by wind turbines or MSUs, if the generated power is
higher than the load demand. The stored power in SRU can be used in
case of power shortfall or to charge the batteries.
MSU interface
MSU interface SRU
MSU interface
static load
static load motors static load
3.2.2 Droop Regulation by Fischer-Burmeister
Microgrids with a large number of MSUs may suer from reactive power
oscillations without proper voltage control. Voltage control function of MSUs
alleviates the large circulating reactive currents amongst MSUs. The circu-
lating currents can be controlled by using voltage-reactive power v−Q droop
controllers. In addition, the MSUs employ a local P −ω control to change the
operating point, in order to achieve a local power balance among MSUs.
Equations (3.1) and (3.2) explain the operational characteristic of the MSUs
droop regulation, where ω is the operating frequency, ωmax is the maximum
frequency, P is the active operating inverter power, Q is the reactive operating

inverter power, mp is the active droop constant and nq is the reactive droop
constant.
ω0 − ω1 = −mp (P1 − P0 ) (3.1)

Chapter 3 89
v0 − v1 = −nq (Q1 − Q0 ) (3.2)
The droop formulation depends on two possible states (with and without
power limit reached). In the rst state, the generated power Pg is less than the
max
maximum limit of power generation Pg , whereas the second state considers
the generated power as constant, because the power limit is reached. The
frequency can freely vary in both states, searching for an equilibrium in the
load share among all the generation units. The problem can be reformulated
by equation (3.3).


Pgmax − Pg if Pgmax ≥ Pg .
f= (3.3)

ω 0 − ω + mp (Pg0 − Pg ) if Pg < Pgmax .
By assuming the following variables:
f: Represents an equation of the microgrid model;
Pg0 : Nominal power of MSUs, where the frequency is nominal;
ω0: Nominal value of frequency;
ω: Equilibrium point of frequency.
The problem of droop formulation in a microgrid lies in that the equi-
librium power of a droop-regulated generation source that decides which is
the state that will be active upon the operating point computation is not
known in advance, because it is dependent on the equilibrium power of other
droop-regulated generation sources owning dierent settings and nominal val-
ues, which means that the solution by Newton-Raphson-like methods might
be hindered.
The applied methodology through this Chapter employs Fischer-Burmeis-

ter algorithm based on nonlinear complementarity problem (NCP), as in [105],
by substituting the piecewise droop function by only a scalar function. Thus,
the provided algorithm solves the problem easier by less iteration process to
nd the equilibrium point. The NCP reformed equation (3.3) to equation (3.4).
0 ≤ (Pgmax − Pg ) ⊥ [ω 0 − ω + mp (Pg0 − Pg ) ≥ 0] (3.4)
Equation 3.4 can be formulated by means of Fischer-Burmeister function,
as Ψ : R2 → R satisfying:
90 Chapter 3
Ψ(a, b) = 0 ⇔ a ≥ 0, b ≥ 0, ab = 0 (3.5)
Where b = Pgmax − Pg and a = ω 0 − ω + mp (Pg0 − Pg ). Equation (3.5)
which is general formula of NCP can be reformulated by Fischer-Burmeister
as shown in (3.6).
√
Ψ(a, b) = a2 + b2 − (a + b) (3.6)
3.2.3 Daily Operation of Batteries
The daily operation of the batteries must be veried to ensure that every
day the batteries start supplying loads with the same capacity as the previous
day.
Typically, each battery is rated to a certain capacity (kW) and feeds MSUs,
which are tuned by droop rules of (P − ω) and v − Q. Each MSU has a maxi-
mum power constraint Pmax , which the inverter can never surpass.
During the rst daily simulation sample, the maximum power constraint
Pmax must be equal to the rated capacity of each battery, but after each sam-
ple, if the battery discharges, the maximum power constraint of the inverter
Pmax must be subtracted from the consumed energy of the battery, till the bat-
tery depth of discharge is higher than 0.99 and the maximum power constraint
nearly equals zero. Thus, the battery must be out of service for the charging
process.
During the charging process of the battery, the wind turbines or other dis-
charging batteries can supply the charged battery till the maximum power
constraint equals to the rated capacity of the charged battery again.
Usually, the energy is stored in the form of electrochemical energy in the
batteries, in order to obtain the desired voltage and capacity. A Battery con-
sists of a set of cells connected together in series, parallel or both. Each cell
consists of two conductor electrodes and an electrolyte, placed together in a
special sealed container and externally connected to a source or load. The
electrolyte enables the exchange of ions between the two electrodes, while the
electrons ow through the external circuit.
The charging-discharging battery process is based on [106] and must com-

Chapter 3 91
ply with the following equations:
Ed /ηd
Discharging : Cstbat (t + 1) = Cstbat (t) − ∆tPt
Ec /ηc
Charging : Cstbat (t + 1) = Cstbat (t) + ∆tPt (3.7)
Subject to the following power Limits:
0 ≤ PtEd ≤ PEdmax
0 ≤ PtEc ≤ PEcmax (3.8)
Stored energy limits:
Cbatstmin ≤ Cbat (t) ≤ Cbatstmax (3.9)
PtEd : Power discharged by the battery bank during the time period (t);
PtEc : Power charged by the grid to the battery bank during the time
period (t);
Cstbat : Energy stored in the battery bank;
∆t: Duration time of each interval;
ηd : Discharge eciency;
ηc : Charge eciency;
PEdmax : Maximum discharge rate;
PEcmax : Maximum charge rate;
Cbatstmin : Minimum stored energy;
Cbatstmax : Maximum stored energy.
Worthy to mention that during the rst and last sample every day, the
consumed energy of each battery must be zero, and the remaining capacity
must equal to rated capacity of the corresponding battery.
The daily operation of the battery is shown by Figure 3.2, the gure ex-
plains the charging-discharging process of a simulated battery, taking into
consideration that the simulation is processed at base power equals to 150 kW.
92 Chapter 3
0.5
0
inverter power(p.u)
−0.5
−1
−1.5
−2
0 5 10 15 20 25
time(hours)
Figure 3.2: Daily battery charging-discharging process
3.2.4 Wind Turbines Characteristics
Typically, the wind turbine is classied by the cut-in speed, rated speed
and uring speed that complies with equation (3.10). Through this Chapter,
several turbines are simulated as shown in Table 3.1.
The daily operation of each turbine is shown in Figure 3.3. When the
wind turbine power is higher than demand load, the rest wind power will be
transferred to the SRU, to be used later in case of higher load demand, or to
charge batteries.



(w2 −w2 )
Prated w2 −wi2 if wi ≤ w ≤ wr .


 r i
Pwt = Prated if wr ≤ w ≤ wf . (3.10)





0 if otherwise.
Pwt : Instantaneous wind turbine power;
Prated : Rated power of wind turbine;
w: Instantaneous wind speed sample;
wi : Cut-in speed of the turbine;

Chapter 3 93
wr : Rated speed of the turbine;
wf : Fluring speed of the turbine.
Table 3.1: Wind Turbines Characteristics
Turbine Rated power kW Cut-in speed Rated speed uring speed
Enercon-E33 330 3 13 28
Enercon-E40 600 2.5 12.5 28
Enercon-E44 900 3 16 28
Enercon-E82 2000 2 13 28
Vestas-V80 1800 4 15 25
600 load power

wind power
stored power
500
400
power(kW)
300
200
100
0
0 5 10 15 20 25
time(hours)
Figure 3.3: Daily operation of wind turbines

94 Chapter 3
3.2.5 Economic Evaluation Models
Usually, the economic evaluation provides a framework for a specic de-
cision and considers both the eectiveness and cost data for two or more
comparative technologies being compared within the analysis. The choice to
evaluate an investment is determined by several factors, these factors include
the investor's perspective, regulation, risk, nancing, cash ow, comparison of
mutually exclusive alternatives, similarity of alternative benets and whether
the investment is fully dened in terms of its size and use.
Most economic measures are valid for most investments, and it is usually
a good idea to compute several of the measures to better evaluate the invest-
ment. Economic evaluation has usually been implemented by dierent models.
The selection criteria guide is based on suggesting which economic measures
are the most appropriate for a specic analysis. Dierent economic measures
apply to dierent situations, and it is generally a good idea to use several
measures when evaluating an investment.
In this Chapter, comprehensive economic evaluations for several case stud-
ies of microgrids supplied by wind turbines and batteries are explained, to
show the most feasible case for the microgrid. The main concern regarding
the economic evaluation belongs to cost of energy analysis (COE), but other
economic evaluation models, such as simple payback (SPB) and net present
value (NPV) are also conducted, in order to determine the protable gains of
each generating unit. Moreover, it is an asset, because most researches have
considered cost of energy issue only.
Levelized Cost of Energy
The levelized cost of energy is the cost of producing one kilowatt-hour
(kWh) of electricity including some items, such as the total cost of the gen-
erating plant, operating the plant over its life time, the nancing costs, the
return on equity and the depreciation factor. COE is a keen method to compare
technologies and designs of dierent generation infrastructures. The applied
COE model for wind turbines is based on [107] and is represented by equation
(3.11).
F CR × Cwp + LRC + CO&M

COE = (3.11)
AEPwt
Chapter 3 95
While, The applied COE model for batteries is represented by equation
(3.12).
F CR × Cbat + CO&M + U Ce
COE = (3.12)
AEPbat
F CR: Fixed rate of charge, assumed as 0.1;
Cwp : Capital cost of wind turbine;
AEPwt : Wind turbine annual energy production in kWh;
AEPbat : Battery annual energy production in kWh;
Cbat : battery capital cost;
U Ce : Unit cost of input electricity for charging batteries;
CO&M : Annual operation and maintenance cost;
LRC : Levelized replacement cost.
Through current study, LRC of wind turbines is assumed zero, as it is orig-
inally a little value and approximated to 0.7% of the Cwp [107]. Therefore, by
ignoring this term, the COE would result in the same design as minimizing
SPB. The xed charge rate F CR is the annual amount per dollar of the initial
capital cost needed to cover the capital cost items, such as debt payments and
return on equity. In this thesis, F CR is assumed as 0.1; moreover, the thesis
assumes a constant operation and maintenance cost of 0.007$/kWh for wind
turbines, based on [107]. In addition the annual operation and maintenance
cost for batteries ranges from 2$/kW-year to 15$/kW-year, according to the
battery type, based on [108] and [109], whereas the variable operation and
maintenance cost for batteries are neglected. The unit cost of input electricity
for charging batteries is assumed as 0.1$/kWh, according to [110].
For simplicity, the numerator terms of equations (3.11) and (3.12) will be
referred in the following Sections as Fwt (Ewt ) and Fbat (Ebat ) respectively.
The AEP corresponds to the integral of the power which is injected into
the grid over a year, taking into consideration that this thesis adopts an aver-
age availability of 0.9 for wind turbines. Thus, the annual energy production
of a wind turbine depends on the turbine power output and the probability of
96 Chapter 3
wind speed.
The levelized cost of energy mentioned in this Section has some limitations,
such as it does not allow for variable equity return and assumes that the debt-
term lifetime equals to the lifetime of equipments.
Simple Payback
Another economic measure of generation plants is to consider the time-
dependent valuation of the electricity in determining the optimal design; typ-
ically, COE is not indication to this issue. Thereby, another measure must
be considered. Simple payback (SPB) is the number of years which the plant
takes to recover the initial capital cost of an investment, without discounting
future prots [42]; SPB is calculated as shown in equation (3.13).
Cwp
SP B = (3.13)
AAR
Cwp : Initial capital cost;
AAR: Average annual revenue based on hourly production,
and calculated from equation (3.14).
AAR = CF × T × Prated × COE (3.14)
Considering the following variables:
CF : Capacity factor of wind turbine;
T: Time in hours;
Prated : Rated power of turbine;
COE : Levelized cost of energy.
The model assumes the wind turbine will produce the same amount of elec-
tricity each year. As the simple payback does not consider the discount rate or
life of the project, the optimal design obtained using simple payback analysis
will not be dependent on these values.

Chapter 3 97
Net Present Value
The Net present value (NPV) is another model for economic evaluation
which takes into consideration the time value of money, as money value of
today will worth more some years later. Any amount of money to be invested
backs a return higher than the rate of ination. Therefore, the future prots
must be discounted. NPV is considered as the present value of benets minus
the present value of costs [111].
The present value of the costs is the initial capital cost Cwp , assuming that
the wind power production is the same from year to year, considering the
produced electric energy and annual revenue are always constant from year
to year, and so the uniform cash ow must be discounted as it occurs in the
future. NPV is calculated as shown by equation (3.15).
( )
(1 + i)L − 1
N P V = AAR × − Cwp (3.15)
i(1 + i)L
By considering the following variables:
i: Discount rate;
L: Life time of wind turbine in years;
Cwp : Initial capital cost;
AAR: Average annual revenue.
Usually, the protable project is based on maximizing the value of NPV to
be greater than zero. If there is a comparison between two mutually exclusive
projects, the one with highest NPV will be chosen, taking into consideration
both discount rate and investment life of the project are eective items in any
protable project, as they directly aect the optimal design of a wind turbine.
All the mentioned economic evaluations models in this Section will be di-
rectly applied to the all simulated microgrids in this Chapter, for the purpose
of determining the economic performance of microgrids supplied by wind en-
ergy and energy storage systems.
The mentioned sources of data in this Section will be used later for the
economic evaluations and capacity factor analysis of several microgirds. As
mentioned before, the purpose of this Chapter is to optimize both the energy
98 Chapter 3
production and economic performance for isolated microgrids supplied by wind
turbines and batteries.
However, the optimization process must start by optimizing the character-
istics of wind turbines to be installed at microgrids rstly. Next Section will
show an interesting manner for optimizing the wind turbine specic rating.
3.3 Economic Evaluation for a Scaled Down Wi-

nd Turbine
The energy production from a wind turbine depends on some factors, such
as wind speed conditions from the area and the characteristics of the wind
turbine itself, including the cut-in, rated and cut-o wind speed parameters.
The turbine of high power rating might not be the most eective for trans-
forming wind power into electric power on a specic site, and so the main
goal becomes the maximization of the annual energy production [39], or the
minimization of the cost of energy [107].
In [112], the author proved that cost of energy optimization is possible
since an annual energy production can be obtained with dierent turbine de-
signs and the one should be chosen where the cost of energy is the minimum.
The turbine rotor diameter is the main element which determines the tur-
bine size, energy production, and subsequently its capital cost [107].
The cost model of the wind turbine is suitable for extrapolation in terms
of the rotor diameter (D ) and rated power, and is shown by Table 3.2, based
on [107]. The cost model is used to calculate the cost of turbines with rotor
diameters from D = 20 m to D = 200 m and rated power up to 10 MW [112].
Thus, if the components of the turbine can be scaled down, mainly the
rotor diameter, then minimizing the wind turbine capital cost will be possible.
The optimal wind turbine rating for a given application is determined by the
wind turbine rating ratio (rr ): the ratio of the turbine rated wind speed wr to
the mean speed of the wind regime wavg [113].
Figure 3.4 shows the power curve of turbines with reduced rating ratios
have more spacious area between the rated turbine speed and cut-out speed;
each zone shown in this gure explains a dierent value of the rating ratio
rr . Therefore, the scaled down turbine, which has lower rr spends more time
Chapter 3 99
to produce power at rated capacity and it is more capable to produce higher
capacity factor than other turbines of higher rr . Moreover, Figure 3.5 explains
that wind turbine capital cost decline with reduction of the rating ratio, based
on [113].
The objective of this Section is to scale down a wind turbine and to check
this eect on its cost of energy and annual energy production.
The simulation scenario consists of two steps:
1. Scaling down a wind turbine by minimizing its capital cost;
2. Minimization of the energy cost of both wind turbines and batteries,
which share supply the proposed microgrid, as shown Figure 3.1.

100 Chapter 3
1.2
1
wind output/rated output
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
wind speed/mean wind speed
Figure 3.4: Wind turbine output power according to sev-

eral rating ratio values
10
fixed block
9 torque block
thrust block
8 power block
7
per turbine cost (106$)
0
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
rating ratio r
r
Figure 3.5: Wind turbine capital cost vs rating ratio

Chapter 3 101
Table 3.2: Items of Wind Turbine Capital Cost
Item Equations
0.0703×D + 0.6781×D
3 2.5025
Blade - 29471
0.0817×D
2.53
Hub + 5680.3
0.4802×D
2.6578
Pitch system
−4 −4
1.2267×10 ×D - 3.0360×10 ×D
3.5 2.5
Bearing
Brake 1.9894×Prated - 0.1141
Generator 219.33×Prated
0.0678×D
2.964
Yaw system
627.28×D
0.85
Main frame
1.3354×D
1.953
Railing Platform
Electric connections 40×Prated
Cooling 12×Prated
Nacelle 11.537×Prated + 3849.7
0.2116×H×D
2
Tower + 2669
275.06×(H×D
2 0.4037
Foundation )
102 Chapter 3
Table 3.2: Continued-Items of Wind Turbine Capital Cost
Item Equations
−6
2.17×10 ×Prated - 0.0145×Prated + 69.54
2
Roads
1.965×(H×D)
1.1736
Assembly
Converter 79×Prated
Where:
D: Rotor diameter;
Prated : Rated power of the turbine;
H: Height of the turbine.
3.3.1 The First Step of Optimization
The rst optimization step in this Section is to scale down a wind turbine,
by minimizing its capital cost at a specic base load power. The objective
function for this purpose can be conducted as follows:
∑
min Cwp (D, Prated ) (3.16)
Cwp : Capital cost of wind turbine, as function of turbine diameter D and
the rated power;
Prated : Rated power of each wind turbine;
D: Diameter of each wind turbine.
The optimization solver is run at average wind speed 4 m/s, and reaches
to a turbine of 40 m diameter and rated power 600 kW, which coincides with
Enrecon-E40 600 kW.

Chapter 3 103
3.3.2 The Second Step of Optimization
It is important to evaluate the eect of scaling down a wind turbine on the
energy production and economic performance. Several elements must be taken
into consideration, as initial capital cost (ICC), operations and maintenance
cost CO&M and annual energy production (AEP ).

The impacts of scaling down a wind turbine are not clear; for example,
decreasing ICC of the wind turbine may reduce its AEP . If one step does not
balance out the another one, the proposed scaled down wind turbine may ac-
tually has a negative eect. It is necessary to study the impact of scaling down
a wind turbine on energy production and economic performance. Therefore, a
comparative simulation study between the scaled down turbine (Enercon-600
kW) and another turbine with higher rating which is chosen to be (Enercon-
900 kW) is conducted, to minimize the cost of energy.
A wind speed time series of average value 4 m/s and a duration curve of
load demand are applied for the proposed comparative simulation. Thus, the
cost of energy is minimized during every time series sample. The purpose of
this step is to force the wind turbine and batteries to produce the maximum
available power at the lowest cost of energy. The objective function of this
step is shown by equation (3.17).
Figure 3.6a shows the curve tting of Enercon-600 kW; Figure 3.6b shows
the curve tting of Enercon-900 kW. The diamonds in Figure 3.6 indicate the
manufacturer data.
( )
∑
n
Fwt (Ewt ) Fbat (Ebat )
min + (3.17)
i=1
Ewt Ebat
By assuming the following variables and parameters:
Ebat : Energy production of the battery;
Ewt : Energy production of the wind turbine;
Fwt (Ewt ): Wind turbine cost function;
Fbat (Ebat ): Battery cost function.

104 Chapter 3
turbine power (p.u) 1

0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30
wind speed (m/s)
(a)
turbine power (p.u)
1
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30
wind speed (m/s)
(b)
Figure 3.6: Wind turbines curve tting: (a) Enercon-600

kW. (b) Enercon-900 kW
3.3.3 Optimization Results
From Table 3.3, it can be concluded that the decision of choosing (Enrecon-
E40 600 kW) is more optimal than choosing (Enrecon-E44 900 kW), from the
energy production and economic evaluation view.
The capital cost, cost of energy and simple pay back of the 600 kW turbine
are lower than the correspondent results of the 900 kW turbine. Moreover, the
net present value, capacity factor and annual energy production of the 600 kW
turbine are higher than the correspondent results of the 900 kW turbine.
These results coincide with the results of [113] and [40], as the authors
proved if a turbine has higher (wi /wr ) and (wf /wr ) ratios taken together than
other turbines, then it proves higher capacity factor, because the turbine works
closer to its rated speed. From Table 3.1, the (Enrecon-E40 600 kW) has higher
(wi /wr ) and (wf /wr ) than (Enrecon-E44 900 kW).
The same comparative simulation procedure is repeated at wind speed time
series of average value 5 m/s. In the case of average wind speed 5 m/s, the
behaviour of (Enrecon-E44 900 kW) starts to improve and its annual energy
Chapter 3 105
production is higher than (Enrecon-E40 600 kW), although its cost of energy is
still higher than (Enrecon-E40 600 kW). This result indicates that (Enrecon-
E44 900 kW) can provide higher energy production than (Enrecon-E44 600
kW), in case of installation at sites of high mean wind speeds.
Table 3.3: Enercon-600 kW Versus Enercon-900 kW at

Mean Wind Speed: 4 m/s
Item 900 kW turbine 600 kW turbine
4.5710×10 3.2381×10
5 5
Capital Cost
Capacity factor 0.0698 0.1168
Diameter 44 40
Cost of energy 0.1010 0.0642
Simple pay back 8.4202 8.3283
-3.9891×10 -2.8259×10
4 4
Net present value (15 years)
1.6286×10 1.8169×10
6 6
Annual wind turbine production
106 Chapter 3
Table 3.4: Enercon-600 kW Versus Enercon-900 kW at

Mean Wind Speed: 5 m/s
4.5710×10 3.2381×10
5 5
Capital Cost
Capacity factor 0.0952 0.1247
Diameter 44 40
Cost of energy 0.0741 0.0601
Simple pay back 8.3308 8.3371
-3.9891×10 -2.8259×10
4 4
Net present value (15 years)
2.2214×10 1.9388×10
6 6
Annual wind turbine production
Chapter 3 107
Figure 3.7a shows the power generation percentage of both wind turbines
and batteries, distributed between the loads and the SRU in the case of
Enercon-600 kW turbine, at mean wind speed 4 m/s. Figure 3.7b shows the
power generation percentage of both wind turbines and batteries, distributed
between the loads and the SRU in the case of Enercon-900 kW, at mean wind
speed 4 m/s. It is clear that the energy production in the case of the 600 kW
turbine is higher than the case of 900 kW. Therefore, this power shortage in
the case of 900 kW turbine is compensated by the batteries, as the batteries
power production is higher in the case of 900 kW than 600 kW scenario.
The economic evaluation was conducted in this Section, assuming that the
microgrid is supplied by wind turbines of the same power rating. However,
in microgrids it is common to supply power by dierent ratings of wind tur-
bines, and so the economic evaluations must be performed in such cases. Next
Section will show the economic evaluation of a microgrid supplied by wind
turbines of dierent power ratings.

108 Chapter 3
supplied wind power to SRU 35.25%

supplied wind power to load 35.25%
supplied battery power to load 35.25%
supplied battery power to SRU 35.25%
supplied battery power to SRU
supplied battery
supplied wind power power to load
to SRU
supplied wind power to load

(a)
supplied wind power

to SRU supplied battery
power to load

(b)

Figure 3.7: Daily power generation percentage: (a)

Enercon-600 kW. (b) Enercon-900 kW.
Chapter 3 109
3.4 Economic Evaluation for a Microgrid Sup-

plied by Wind Turbines of Dierent Power
Ratings
Through this Section, the microgrid is supplied by wind turbines of dierent
power ratings and also by Vanadium Redox batteries interfaced through MSUs,
as shown in Figure 3.8. The microgrid is simulated to minimize the cost of
energy and maximize the energy production, according to equation (3.17).
Simulations are processed at wind speed time series of average values 2, 4 and
5 m/s. The analyses of each turbine are performed, according to the cost of
energy, the simple pay back and the net present value.
the curve tting of Enercon-900 kW; Figure 3.9c shows the curve tting of
Enercon-2000 kW; diamonds indicate the manufacturer data. Power ratings of
the contributing wind turbines are shown hereunder:
1. Enercon-E33-330 kW
MSU interface
SRU
MSU interface
MSU interface
Enercon 330 kW
Enercon 2000 kW static load
static load motors Enercon 900 kW static load
Figure 3.8: Layout of a microgrid supplied by wind tur-

bines of dierent power ratings
110 Chapter 3
turbine power (p.u) 0.4
0.2
0
0 5 10 15 20 25 30
wind speed (m/s)
turbine power (p.u) turbine power (p.u)
(a)
1
0.5
0
0 5 10 15 20 25 30
wind speed (m/s)
(b)
2
0
0 5 10 15 20 25 30
wind speed (m/s)
(c)
Figure 3.9: Curve tting of contributing wind turbines:

(a) Enercon-330 kW. (b) Enercon-900 kW. (c) Enercon-
2000 kW.
The economic analysis is carried out to evaluate the economic performance
of each turbine. The results are shown in Figures 3.10, 3.11 and 3.12. The
main conclusions of these simulations are shown as follows:
1. The capacity factor of both 330 kW and 2000 kW turbines are higher
than the capacity factor of the 900 kW turbine. As concluded previously
the turbines which have higher (wi /wr ) ratios, have higher generation
capacities. From Table 3.1, the 900 kW turbine has the lowest (wi /wr )
ratio;
2. The cost of energy decreases with increasing the wind speed and still
the 900 kW shows the highest cost of energy, even though at high wind
speeds;
3. The simple payback improves slightly at higher wind speeds;

Chapter 3 111
4. From Figures 3.13, 3.14 and 3.15, the net present value of the 330 kW
turbine is the lowest after 10 years, but the 2000 kW turbine has the
best record from economic view after 20 years, indicating more protable
gains;
5. The optimum cases from the economic evaluation view are related to the
330 kW and 2000 kW turbines, as they have the highest capacity factor
and the lowest cost of energy. Moreover, if the protable gains are taken
into consideration, the 2000 kW turbine is the best choice, because it has
the highest NPV after 20 years.
0.15
330 kW turbine
900 kW turbine
2000 kW turbine
0.1
capacity factor
0.05
0
0 1 2 3 4 5 6
wind speed (m/s)
Figure 3.10: Capacity factor vs wind speed
The previous results through this Section were aimed at analysing the eco-
nomic evaluation of the wind turbines only. However, an interesting result of
the batteries daily production which entails the simulations at wind speed time
series of average values 2 and 5 m/s respectively, must be considered, to show
the role of the multi-objective optimization function, which is used to minimize
the cost of energy for both wind turbines and batteries. The optimization func-
tion was shown by equation equation (3.17) and will be mentioned hereunder
again for convenience.

( )
∑
n
min + (3.18)
i=1
Ewt Ebat
112 Chapter 3
900 kW turbine
2000 kW turbine
330 kW turbine
levelized cost of energy($/kWh) 0.2
0.15
0.1
0.05
0
0 1 2 3 4 5 6
wind speed (m/s)
Figure 3.11: Wind turbines levelized cost of energy vs

wind speed
8.337
900 kW turbine
2000 kW turbine
8.336 330 kW turbine
8.335
simple pay back (years)
8.334
8.333
8.332
8.331
8.33
0 1 2 3 4 5 6
wind speed (m/s)
Figure 3.12: Simple payback vs wind speed
The rst half of the equation (3.17) is dedicated to the wind turbine term,
while the second half is dedicated to the batteries. In the case of wind speed
time series of high average value as 5 m/s, the energy production of the wind
( )
Fwt (Ewt )
turbines increase, and so the term of the wind turbines is much
Ewt
Chapter 3 113
4
x 10
−2
−4
NPV ($)
−6
−8
−10
after 20 years
after 15 years
−12 after 10 years
1 2 3
cases
Figure 3.13: Net present value of Enercon-900 kW tur-

bine measured at dierent periods of time
5
x 10
−0.5
−1
NPV ($)
−1.5
−2
−2.5
after 20 years
−3
after 15 years
after 10 years
1 2 3
cases

( )
Fbat (Ebat )
lower than the term of the batteries, but both terms of the wind
Ebat
turbines and batteries must have the same weight to get minimized, because
the both terms are employed in the multi-objective function. Subsequently,

114 Chapter 3
4
x 10
−1
NPV ($)
−2
−3
−4
after 20 years
−5 after 15 years
after 10 years
1 2 3
cases

the optimization solver search for more optimal equilibrium points, in order to
balance the both terms.
Figures 3.16 and 3.17 explain the daily production of the wind turbines
and batteries in the case of wind speed time series of average values 2 m/s and
5 m/s respectively. In Figure 3.17, the slope of the wind production part is
higher than the slope of wind production in Figure 3.16, due to wind speed
increase. Moreover, the axis of the battery production in Figure 3.17 reaches
higher values than the axis of the battery production in Figure 3.16.
Chapter 3 115
2000
total production(kW)
1500
1000
500
0
600
1500
400
1000
200
500
battery production(kW) 0 0
wind production(kW)
Figure 3.16: Daily power production at average wind

speed 2 m/sec
5000
4000
total production(kW)
3000
2000
1000
0
3000
1500
2000
1000
1000
500
battery production(kW) 0 0
wind production(kW)
Figure 3.17: Daily power production at average wind

speed 5 m/sec
116 Chapter 3
These results are translated to levelized cost of energy as shown in Fig-
ure 3.18; it is clear that the total cost of energy for both the wind turbines
and batteries in the case of wind speed time series of average value 5 m/s is
much lower than the total cost of energy in the case of wind speed time series
of average value 2 m/s. The levelized costs of energy for both wind turbines
and batteries are summed, taking into consideration that the cycle life of the
Vanadium Redox are high and equal to 12,000 and subsequently their life time
are appreciated as 30 years, according to [114].
To this end, the higher mean wind speed does not provide a higher energy
production and economic performance for the wind turbines only, but for the
storage systems as well, if the both generation units are employed in a multi-
objective function.
The objective of the Chapter is the economic evaluation for both the wind
turbines and energy storage systems; the previous interesting results of the
daily battery production, which show the possibility of reducing its cost of
energy through the performance of wind turbines. Thus, the next Section
will conduct the economic evaluation for microgrids supplied by various wind
turbines and batteries. The batteries will be analysed under dierent capital
costs, according to their eciencies.
3.5 Economic Evaluation for Microgrids Supplie-

d by Various Wind Turbines and Batteries
The microgrid energy management systems (EMS) are responsible for dis-
patching the energy among dierent generators in microgrids. EMS makes
decision to which energy storage devices to dispatch for various types of con-
tingencies.
Electric energy storage systems including dierent types as batteries, ca-
pacitors and ywheels are required to supplement the DER generators during
low voltage transients on the distribution system, or when operating in iso-
lated mode, or actual electric outages as a result of motor starts or other short
term overloads. Some types of DERs as microturbines have little inertia for
successful starting motors, unless some sort of storage is supplied. Hence, the
energy storage systems as batteries can handle such problems by managing the
Chapter 3 117
1.5
COE at 5m/s
COE at 2m/s
total levelized cost of energy ($/kWh)
0.5
0
1 2
cases
Figure 3.18: Total levelized cost of energy at dierent

wind speed levels
power control during the transient moments. If a network is supplied by syn-
chronous machines, which are characterized by high inertia, they cannot ride
through the high power demand in a little time, leading to some power quality
problems, such as frequency dips and dierent ranks of harmonics included
in the voltage waveforms. In some cases, the harmonics are extreme that an
equipment being powered from these generators may fail. Therefore, the en-
ergy storage devices can be employed to overcome such urgent problems, by
providing power during sag moments, and to keep the voltage at buses always
constant.
The continuous risk results from high penetration of intermittent wind en-
ergy sources in power systems, as shown in [115], has been a crucial principle of
the energy storage systems supplying power grids during moments of low wind
speed. Most analyses regarding wind integration in the power system have
focused on more engineering aspects, such as grid stability, load-balance and
system security. Some of these analyses have shown benets from using energy
storage as an alternative to other dispatchable generators, in order to manage
wind uncertainty, but the economic evaluations for power systems supplied by
wind energy in parallel with energy storage systems have not been cleared.
118 Chapter 3
Through this Section, the energy storage devices are represented by various
types of batteries. The economic evaluations of the proposed microgird sup-
plied by wind turbines and various batteries are performed. The simulations
considers batteries of the most widely used for small-medium scale storage ap-
plications, such as Lead-Acid, Nickel-Cadmium, Lithium-Ion and Vanadium
Redox batteries.
The Lead-Acid battery is one of the most mature types of batteries and
they are widely used for small-medium scale storage applications. However,
one of the main disadvantages of the Lead-Acid battery is the poor perfor-
mance at low and high ambient temperatures. In addition, the necessity for
periodic water maintenance for ooded batteries and low specic energy and
power. Furthermore, Lead-Acid batteries show diculties in providing fre-
quent power cycling [116].
The Nickel-Cadmium batteries compete with Lead-Acid batteries because
they have a higher energy density; the technology of the Nickel-Cadmium bat-
tery has been under development since 1950. On the other hand, the main
disadvantage of Nickel-Cadmium batteries is the memory eect. If Nickel-
Cadmium batteries are not fully discharged before being recharged, the battery
will start losing its capacity. Consequently, the Nickel-Cadmium batteries can-
not operate economically, in parallel with non-dispatchable generators [117].
The Vanadium Redox batteries are suitable for small and medium scale
applications. There are currently over 20 MWh of installed Vanadium Re-
dox batteries in the world. The installed Vanadium Redox batteries are used
for remote area power systems and renewable energy stabilization. Vanadium
Redox plants can be upgraded at a low incremental cost, by increasing the
volume of electrolytes for more stored energy or by adding new cell stacks for
additional power. The drawback of Vanadium Redox battery is referred to its
low specic energy and energy density [118].
The Lithium-Ion batteries are under growing demand, due to their high
eciencies of over 95%, long life cycle of and fast discharge capabilities. The
main problem of Lithium-Ion batteries are their high costs [119].
This Section tries to answer a question regarding the possibility to reduce
energy cost of a high capital cost-high ecient battery supplying a grid, in par-
allel with wind turbines. For this reason, the microgrid shown in Figure 3.8 is
four times simulated at wind speed time series of average value 4 m/s; through
Chapter 3 119
each simulation the microgrid is supplied by a dierent type of battery, such
as Nickel Cadmium, Lithium, Vanadium-Redox or Lead-Acid.
Figure 3.19 shows the capital costs of the proposed batteries, as mentioned
in [120]. The battery price diers from an article to another, depending on the
manufacturer, the cycle eciency, the life time and the discharge rate. Each
battery is processed in separate simulation, in order to minimize the energy
cost of the whole microgrid through the multi-objective function as shown by
equation (3.19). The charge-discharge equations for all the simulated batter-
ies, according to [106], are shown by equations (3.7)-(3.9).
The characteristics of the proposed batteries are detailed in Table 3.5, ac-
cording to [121] and [122].

( )
∑
n
min + (3.19)
i=1
Ewt Ebat
700
Vanadium Redox
Lithium
600 Lead Acid
Nickel Cadmium
capital cost of batteries ($/kW)
500
400
300
200
100
0
1 2 3 4
Figure 3.19: Capital cost of batteries
Figure 3.20 express a comparison of generation percentage between Lithium
and Vanadium batteries respectively, as the both gures explain the contri-
120 Chapter 3
Table 3.5: Characteristics of the Simulated Batteries
Energy storage Cycle eciency Cycle life Energy density
technology % (Wh/kg)
Lead-Acid battery 70-90 500-1000 30-50
Nickel-Cadmium battery 72 2000-2500 50-75
Lithium-Ion battery almost 100 1000-10,000 75-200
Vanadium-Redox battery 85 12,000 10-30
bution between the wind turbines and batteries at supplying the microgrid,
whereas the annual energy production of the two batteries is shown in Fig-
ure 3.22.
Both pie charts in Figure 3.20 brief the conclusion of the current simula-
tion, as it is a comparison between the lowest capital cost battery (Vanadium)
and the highest capital cost one (Lithium). The wind power production is the
same, because the ratings of the wind turbines and the mean wind speed are
the same in the both cases, but the battery power ow in the both cases are
dierent. The optimization solver which is used in the simulations must bal-
( ) ( )
ance between the wind turbines term and battery term .
Ewt Ebat
Clearly, the capital cost in the case of Lithium batteries is higher than
Vanadium case, and so the optimization solver searches for another equilib-
rium point to reach the minimum cost of energy.
In the case of Lithium batteries, more power ows to the SRU. By end of
the day, the SRU backs this energy again to the Lithium batteries to achieve
the battery daily operation as shown in Figure 3.2. Therefore, the total power
which ows to the load is the same in case of Lithium and Vanadium batteries
to achieve Pgen =Pdemand . The total power ow from Lithium batteries to the
SRU is much higher than the case of Vanadium batteries.
This result indicates the ability of the high capital cost-high ecient bat-
tery to provide optimum economic performance and high energy production,

Chapter 3 121
if the problem is performed by a multi-objective function.
The previous conclusion could be translated to cost of energy and annual
energy production, as shown in Figures 3.21 and 3.22. The energy cost of the
Lithium battery which has capital cost (600 $/kw) equals to the cost of energy
of the Nickel-Cadmium battery which has capital cost (400 $/kw). This can
be explained by Figure 3.22, as the annual energy production in the case of
Lithium battery is higher than the Nickel-Cadmium case, due to higher power
ow to the SRU in the case of Lithium battery.
Until now, some important results were concluded regarding supplying iso-
lated microgrids by wind turbines and batteries. Next Section will show the
aggregation eect of wind turbines capacities on the economic performance
and energy production for microgrids.

122 Chapter 3

supplied battery
supplied wind power power to SRU
to SRU
supplied wind power

to load supplied battery power to load
(a)
(b)
supplied battery
power to SRU
supplied wind power supplied battery

to SRU power to load

Figure 3.20: Daily power generation percentage: (a)

Lithium battery case. (b) Vanadium battery case.
Chapter 3 123
0.8
Vanadium Redox
Lithium
0.7 Lead Acid
Nickel Cadmium
levelized cost of energy($/kWh)
0.6
0.5
0.4
0.3
0.2
0.1
0
1 2 3 4
cases
Figure 3.21: Batteries levelized cost of energy
6
x 10
7
Vanadium Redox
Lithium
6 Lead Acid
Nickel Cadmium
5
annual production (kW)
0
1 2 3 4
cases
Figure 3.22: Batteries annual energy production

124 Chapter 3
3.6 The Aggregation Eect of Wind Turbines on

the Economic Performance
Most conducted analyses regarding the wind turbine aggregation have con-
centrated mainly on technical issues, as dierent simulation algorithms have
been proposed to represent aggregated models of wind turbines without unnec-
essarily increasing the wind farm model. After occurrence of several aggregated
wind farm models, the eect of such models on steady state and dynamic anal-
yses have been shown in many papers. However, the economic evaluation of
aggregated wind farm models on power systems have not been considered.
Through this Section, the economic evaluations of aggregated wind farms
are performed. The economic evaluations of the aggregated wind farms are
provided by the cost of energy, simple pay back and net present value.
The comparison is conducted among three cases:
• The rst case employs three dispersed Enercon-600 kW wind turbines,
to supply the microgrid with a set of batteries;
• The second case employs one Enercon-2000 kW wind turbine only to
supply the microgrid with a set of batteries (Aggregation case);
• The third case employs one Vestas-1800 kW wind turbine only to supply
the microgrid with a set of batteries (Aggregation case).
The microgrid layout for the rst case is shown in Figure 3.1, while the
layout of the second and third cases is shown by Figure 3.24. Moreover, the
micorgrid is supplied in the rst, second and third cases by Vanadium Redox
batteries interfaced through MSUs, in order to share power supply with the
proposed wind turbines of each case. The simulations are performed for each
case at wind speed time series of average values 4 and 5 m/s respectively, by
the same function shown in equation (3.19).
the curve tting of Vestas-1800 kW; Figure 3.23c shows the curve tting of
Enercon-2000 kW; diamonds indicate the manufacturer data.

Chapter 3 125
turbine power (p.u) turbine power (p.u)

1
0.5
0
0 5 10 15 20 25 30
wind speed (m/s)
(a)
2
0
0 5 10 15 20 25 30
wind speed (m/s)
turbine power (p.u)
(b)
2
0
0 5 10 15 20 25 30
wind speed (m/s)
(c)
Figure 3.23: Curve tting of contributing wind turbines:

(a) Enercon-600 kW. (b) Vestas-1800 kW. (c) Enercon-
2000 kW.
MSU interface
MSU interface SRU
MSU interface
static load
motors
static load static load
aggregated wind turbine

126 Chapter 3
Figures 3.25, 3.26 and 3.27 show that the turbines with higher wi /wr ratio
still have the highest capacity factor, the highest annual energy production
and the lowest cost of energy. Enercon-E40-600 kW turbines have the lowest
wi /wr ratio, and so the case of supplying the microgrid by Enercon-E40-600
kW turbines shows the lowest capacity factor and the highest energy cost.
The annual wind turbine energy production is higher in both cases of
Vestas-1800 kW and Enercon-E82-2000 kW than the Enercon-E40 600 kW
turbines case, as shown from Figure 3.26.
The benets of aggregation are not for wind turbines only, as the annual en-
ergy production for batteries in the both cases of Vestas-1800 kW and Enercon-
E82-2000 kW, are higher than the Enercon-E40 600 kW case, as shown from
Figure 3.28. Therefore, the batteries energy cost in both cases of Vestas-1800
kW and Enercon-E82-2000 kW are less than the Enercon-E40 600 kW case, as
shown from Figure 3.29. Consequently, the total energy cost in both cases of
Vestas-1800 kW and Enercon-E82-2000 kW are less than the Enercon-E40 600
kW case, as shown in Figure 3.30.
It could be concluded from the previous simulations, in order to provide a
better economic performance and energy production, the aggregation of wind
turbines capacities can be a better choice than dispersing the same capacities
of wind turbines at dierent microgrid nodes. Moreover, the cases of wind tur-
bines aggregation show more protable gains than smaller capacity dispersed
turbines (600 kW), as shown Figure 3.31, 3.32 and 3.33.
Figure 3.34 shows the percentage of power generation for Enercon-600 kW
at mean wind speeds 4 and 5 m/s respectively. Figure 3.35 shows the per-
centage of power generation for Vestas-1800 kW at mean wind speeds 4 and
5 m/s respectively. Figure 3.36 shows the percentage of power generation for
Enercon-2000 kW at mean wind speeds 4 and 5 m/s respectively. Finally, Ta-
bles 3.6 and 3.7 show the results of the economic evaluation models and energy
production.
Up to this point, all the possible scenarios for economic evaluation of mi-
crogrids, supplied by wind energy and batteries were conducted; however, the
objective of optimizing the energy production and economic feasibility for mi-
crogrids must be extended to include additional constraints related to the

Chapter 3 127
operation of microgrids, as the objective is to enhance the minimum genera-
tion cost with stable power ow limits. For this sake, the small signal stability
analyses will be conducted in Chapter 4.
E−40 at 5m/s
0.18
E−40 at 4m/s
0.16 E−82 at 5m/s
E−82 at 4m/s
cpacity factors of wind turbines
0.14 V−80 at 5m/s

V−80 at 4m/s
0.12
0.1
0.08
0.06
0.04
0.02
0
1 2 3 4 5 6
cases
Figure 3.25: Wind turbines capacity factor

128 Chapter 3
6
x 10
3
E−40 at 5m/s
E−40 at 4m/s
E−82 at 5m/s
2.5
E−82 at 4m/s
V−80 at 5m/s
V−80 at 4m/s
2
AEP (kW/year)
1.5
0.5
0
1 2 3 4 5 6
cases
Figure 3.26: Wind turbines annual energy production
0.3
E−40 at 5m/s
E−40 at 4m/s
E−82 at 5m/s
0.25
E−82 at 4m/s
V−80 at 5m/s
V−80 at 4m/s
0.2
LCOE($/kWh)
0.15
0.1
0.05
0
1 2 3 4 5 6
cases
Figure 3.27: Wind turbines levelized cost of energy

Chapter 3 129
6
x 10
E−40 at 5m/s
4
E−40 at 4m/s
E−82 at 5m/s
3.5 E−82 at 4m/s
V−80 at 5m/s
3 V−80 at 4m/s
AEP(kW/year)
2.5
1.5
0.5
0
1 2 3 4 5 6
cases
Figure 3.28: Batteries annual energy production
1.4 E−40 at 5m/s

E−40 at 4m/s
E−82 at 5m/s
1.2
E−82 at 4m/s
V−80 at 5m/s
1 V−80 at 4m/s
LCOE ($/kWh)
0.8
0.6
0.4
0.2
0
1 2 3 4 5 6
cases
Figure 3.29: Batteries levelized cost of energy

130 Chapter 3
1.6
E−40 at 5m/s
E−40 at 4m/s
1.4 E−82 at 5m/s
E−82 at 4m/s
1.2 V−80 at 5m/s
V−80 at 4m/s
1
LCOE ($/kWh)
0.8
0.6
0.4
0.2
0
1 2 3 4 5 6
cases
Figure 3.30: Total levelized cost of energy
4
x 10
−1
−2
−3
NPV($)
−4
−5
−6
−7
after 20 years
−8 after 15 years
after 10 years
1 2 3
cases
Figure 3.31: Net present value, Enercon-600 kW

Chapter 3 131
5
x 10
0.5
−0.5
−1
NPV($)
−1.5
−2
−2.5
after 20 years
−3
after 15 years
after 10 years
−3.5
1 2 3
cases
Figure 3.32: Net present value, Enercon-2000 kW
5
x 10
0.5
−0.5
−1
NPV($)
−1.5
−2
−2.5
after 20 years
after 15 years
−3 after 10 years
1 2 3
cases
Figure 3.33: Net present value, Vestas-1800 kW

132 Chapter 3

supplied wind supplied battery power

power to SRU to load

(a)
(b)
supplied wind
power to SRU supplied battery power
to load

Figure 3.34: Daily power percentage, Enercon-600 kW:

(a) mean wind speed (4 m/s). (b) mean wind speed (5
m/s).
Chapter 3 133

supplied battery power

to load
stored wind power in SRU

(a)
(b)

stored wind power in SRU to load

Figure 3.35: Daily power generation percentage, Vestas-

1800 kW: (a) mean wind speed (4 m/s). (b) mean wind
speed (5 m/s).
134 Chapter 3

supplied wind power to SRU supplied battery power

to load

(a)
(b)

supplied wind to load
power to SRU

Figure 3.36: Daily power percentage, Enercon-2000 kW:

(a) mean wind speed (4 m/s). (b) mean wind speed (5
m/s).
Table 3.6: Economic Evaluation at Mean Wind Speed: 4m/s-Aggregation Scenario
Chapter 3
Item Vestas-1800 kW Three Enercon-600 kW Enercon-2000 kW
6 5 6
Capital Cost 1.1694×10 9.7143×10 1.2759×10
Capacity factor 0.1173 0.1168 (for each turbine) 0.1086
Diameter 80 40 (for each turbine)
Cost of energy 0.0769 0.1926 0.0816
Simple pay back 8.3333 8.3283(for each turbine) 8.3333
5 4 5
NPV(15 years) -1.0205×10 -2.8259×10 (for each turbine) -1.1135×10
6 6 6
AEP (wind turbines) 1.8239×10 5.4507×10 1.8764×10
135
136
Table 3.7: Economic Evaluation at Mean Wind Speed: 5m/s-Aggregation Scenario
Item Vestas-1800 kW Three Enercon-600 kW Enercon-2000 kW
6 5 6
Capital Cost 1.1694×10 9.7143×10 1.2759×10
Capacity factor 0.1398 0.3741 0.1333
Diameter 80 40 (for each turbine) 82
Cost of energy 0.0646 0.1803 (for each turbine) 0.0665
Simple pay back 8.3333 8.3371 (for each turbine) 8.3333
5 4 5
NPV(15 years) -1.0205×10 -2.8259×10 (for each turbine) -1.1135×10
6 6 6
AEP (wind turbines) 2.1735×10 5.8164×10 2.3033×10
Chapter 3
Chapter 4 137
Chapter 4
Microgrid Stability Analysis
4.1 Introduction
The previous Chapter discussed the economic evaluation and capacity fac-
tor analysis of microgrids based on maximizing the annual energy production.
However, the objective of optimizing the energy production and economic fea-
sibility for microgrids must be extended to satisfy the power balance and gen-
eration limit among dierent microgrid generators. The electric power system
instability can be interpreted using various methods depending on the system
conguration and operational status.
Traditionally, the question of stability has been connected to maintaining
synchronous operations. The production of electricity in the conventional util-
ity is secured primarily using synchronous generators, and for this reason it
is important to secure their synchronism and parallel operation. Therefore,
the question of stability in conventional power systems is mainly based on the
stability of synchronous machinery and on the relationship between the active
power and rotor angle of the generator.
Usually, the operation of conventional generation systems is based on cen-
tralized control of utility generators, delivering power through an extensive
transmission and distribution system, to meet the given demands of widely
dispersed users. Nowadays, the justication for the large centralized station
is weakening due to depleting conventional resources, increased transmission
and distribution costs, deregulation trends and environmental concerns. Dis-
tributed generation is commonly used for small scale generations and can oer
solution to many of centralized generation challenges.

138 Chapter 4
The proposed microgird in this Chapter is supplied by inverter-interfaced
generators and doubly-fed induction generators (DFIGs). Typically, the in-
crease of inverter-interfaced distributed generators penetration might not only
lead to problems under faulty conditions, but also to instability under small
disturbances, such as small changes in loads and power transfers that occur
during normal power system operation. This type of stability is known as the
small-signal stability. Moreover, the modelling and operation of the DFIGs
are totally dierent from the synchronous generators; as they are intermittent
sources and depend on stochastic wind parameters.
Many recent microgrid small signal stability studies belong to either iso-
lated microgrids supplied by wind energy and connected to utility grids [33]
and [34], or to isolated microgrids supplied by inverter-interfaced distributed
generators without penetration of any intermittent energy sources have been
conducted, but the stability analysis of isolated microgrids supplied by the
both sources of generation have not been studied before. Thus, in this Chap-
ter, the small signal stability analysis of isolated microgrids supplied by DFIGs
and inverter-interfaced distributed generators are conducted.
The small signal stability is the ability of a power system to maintain
stability when it is subjected to small disturbances. A disturbance is consid-
ered to be small, if the linearized system around the equilibrium point, still
adequately represents dynamics of the original nonlinear system under this
disturbance [2]. Small-signal stability problems are normally divided into two
types, one type of problems is associated with the steady-state stability limits
of the power system elements (voltage-frequency stability). The power system
is usually controlled in such a way that an increase of the voltage angle leads
to active power transfer increase and an increase of reactive power injection
by a generator leads to load voltage magnitude increase. The other type of
small-signal stability problems is related to the oscillatory instability, due to
inecient generators control strategies. The other possible reason for oscilla-
tory problems is related to large power transfers through weak links, and so
inter-area oscillations may appear. The dierent types of small signal insta-
bility are directly related to the system eigenvalues.
In this Chapter, eigenvalue analyses are used to reveal the information of
dierent stability modes of microgrids small signal stability problems, such as
wind speed variations, power demand increment and dierent power exchange
Chapter 4 139
levels among DGs interfaced by electronic inverters.
In this Chapter, the small signal dynamic model of a microgrid is con-
structed, based on [123], for some important purposes, such as realizing the
reliable operating points of DGs, identifying the most dominant DGs control
parameters which can aect the microgrid stability. Therefore, in future work,
the most dominant DGs control parameters can be set appropriately and an
adequate control strategy can be developed to keep the microgrid stable. The
following Section shows the main outlines of this Chapter.
4.2 Chapter Outlines

In order to conduct the small signal stability analysis of a microgrid, the
microgrid state space modelling is conducted, where eigenvalue analyses are
used to reveal the data of the microgrid stability modes. Thereafter, the sta-
bility analysis is conducted under dierent wind speed levels, droop constants
increment of inverter-interfaced generators, power demand increment.
This Chapter is divided into the following Sections:
• Microgrid layout, (Section 4.3);
• Microgrid state space modelling, (Section 4.4);
• Conducting participation factor and time response (Section 4.5);
• Wind variation impact on the microgrid stability, (Section 4.6);
• Droop constants variation impact on the microgrid stability, (Section
4.7);
• Power demand increment impact on the microgrid stability, (Section 4.8).
4.3 Microgrid Layout

The base power of the microgrid is 2.97 MW, and the proposed microgrid
for the isolated mode and grid-connected mode is congured and briefed as
follows:

140 Chapter 4
• Static loads at nodes (2, 4, 6, 8, 9, 11, 12 and 14);
• Four doubly-fed induction generators rated at 1.27 MW at nodes (6, 9,
12 and 14).
In the case of isolated mode, the microgrid is supplied by additional four
voltage-frequency inverters at nodes (8, 10, 11 and 13), as shown in Figure 4.1.
The power and frequency are regulated from voltage-frequency inverters by
droop control constants, which are valued to be 9.4 × 10−7 rad/W/s, whereas
in the case of grid-connected mode, the microgrid is connected and supplied
by the innite bus at node 1, as shown in Figure 4.2.
Figure 4.1: Isolated micorgrid layout
4.4 Microgrid State Space Modelling

4.4.1 Introduction
The small signal stability is dened as the capability of system to return to
a stable operating point after the occurrence of a disturbance that leads to an
incremental change in one or more of the state variables of the power system.
The power system state variables are usually dierent from one state space
model to another one. Some examples of power system state variables can be
considered, such as generator load angles, magnetic ux linkages, generator

Chapter 4 141
Figure 4.2: Layout of the microgrid connected to the

innite bus
currents and voltages.
The microgrid state variables in this Chapter are chosen based on micro-
grid elements which are capable of storing energy. These elements can be
accounted, such as inductive, capacitive and integrator elements [124], which
are included in the proposed microgrid plant model to be controlled.
If a disturbance causes a change in the value of one or more of these state
variables, the system is driven from the equilibrium. If thereafter the system
returns to its steady state, then it is stable; whereas, if the initial deviation
from the steady state becomes larger, then it is unstable.
The main dierence between transient stability and small signal stability
is that, if a steady state is reached after a disturbance leading to a transient
phenomenon, producing a change in the system topology, such as tripping of
a generator or a line, the new steady state might be dierent from the initial
state; this case is known as transient stability. In contrast, if a system returns
to a steady state after an incremental change in a state variable, this steady
state is identical to the initial state, because no change in the network topology
has occurred; this case is known as small signal stability.
In this Chapter, eigenvalue analyses are used to reveal the informations
of dierent stability modes. The aim is to illustrate the correspondence be-
tween the eigenvalues of the microgrid and their dynamic behaviour. The
142 Chapter 4
linearization of the state equations of the microgrid is applied. Thereafter, the
microgrid dominant eigenvalues can be detected, and also the relation between
the behaviour of such dominant eigenvalues and the time domain can be ob-
tained.
The behaviour of an electrical power system dynamics can be described
based on [124] by equations (4.1) and (4.2).
ẋ(t) = f (x(t), u(t)) (4.1)
y(t) = g(x(t), u(t)) (4.2)
By assuming that:
f: Vector contains the microgrid rst-order non-linear
dierential equations;
x: Vector contains the microgrid state variables;
u: Vector contains the microgrid control variables;
g: Vector contains the microgrid non-linear algebraic equations;
y: Vector contains the microgrid algebraic output variables;
t: The proposed time of a state variable change.
By assuming that the system indicated by equations (4.1) and (4.2) is time
invariant, then the time part can be excluded from these equations.
Equations (4.1) and (4.2) can be linearized based on [2], and the resulting
linearized description of the system can be used to investigate its response to
small variations in the input or state variables, starting from an equilibrium
point. To this end, equations (4.1) and (4.2) are expressed in terms of their
Taylor's series expansion. The second and higher orders of the partial deriva-
tives of f and g are excluded and only taking into account the rst-order terms.
Consequently, the nal form can be described by equations (4.3) and (4.4).
∆ẋ(t) = A∆x(t) + B∆u(t) (4.3)
∆y(t) = C∆x(t) + D∆u(t) (4.4)
Where A, B, C and D are the matrices containing the coecients of the
state variables and system output equations. The proposed microgrid dynamic
Chapter 4 143
model in this Chapter dierentiates between the open-loop and closed-loop
system. The open loop includes microgrid plant and controller models, whereas
the closed loop system applies feed-back concept. As mentioned before that
the microgrid state variables in this Chapter are chosen based on microgrid
elements which are capable of storing energy. These elements can be accounted,
such as inductive, capacitive and integrator elements, which are included in
the proposed microgrid plant model to be controlled.
A model of RLC lter will be highlighted in the following part, to clear how
the state variables are obtained, and how the linearization model is conducted.
Thus, by the same manner, the state space model can be conducted for all the
microgrid elements, such as voltage-frequency inverters, doubly-fed induction
generators, distribution lines and static loads.
By considering the RLC lter as shown in Figure 4.3, the single phase
current and voltage of the inductive part can be regarded as ii (t) and vL (t),
while the same correspondent values of capacitive part can be regarded as as
ic (t) and vc (t). Applying the voltage drop equation for the inductive part of
the lter as function of time as shown in equation (4.5).
d ii (t)
vL (t) = Rf ii (t) + Lf (4.5)
dt
The proposed state space modelling in this Chapter considers all the mi-
Figure 4.3: RLC lter
crogird state variables are constructed on a synchronous reference frame whose
rotation frequency is ω. Therefore, the dierentiation considers both the time

d
and rotation frequency, as follows: (
dt
+ jω ). Consequently, equation (4.5) can
144 Chapter 4
be rewritten as follows:
d
vL (t) = Rf ii (t) + ( + jω)Lf ii (t) (4.6)
dt
The dynamic equations of the lter inductive part can be obtained as shown
in equation (4.7).


i̇id = −Rf
i + ωiiq + 1
v
Lf i d Lf Ld
(4.7)

i̇iq = −ωiid + −Rf
i + 1
v
Lf iq Lf Lq
Where:
ii : Inductance lter current;
ω: Operating frequency;
vL : Filter inductance voltage drop.
Typically, the microgrid frequency is variable with respect to time, during
normal cases of power exchange among the microgrid generators and during
transient cases. Thus, equation (4.7) is considered non-linear equation. The
considered non-linear terms in equation (4.7) are (ωiiq ) and (−ωiid ). The
linearized equation for the lter inductive part is shown by equation (4.8). The
operator ∆ indicates the linearized variables, while the subscript 0 indicates
the linearization around the considered system operating point.
         
−Rf
∆i̇id ω 0 ∆i i
1
0 ∆v Ld I i [ ]
  =  Lf   +  Lf  +  ∆ω
d q
−R (4.8)
∆i̇iq −ω0 Lff ∆iiq 0 L1f ∆vLq −Iid
The linearized equation of the lter capacitive part can be derived by the
same manner. The current in the lter capacitive part can be indicated by
equation (4.9). By assuming that internal resistance Rc = 0, the nal lin-
earized equation of the capacitive part is shown by equation (4.10)
d
ic (t) = Rc vc (t) + ( + jω) Cf vc (t) (4.9)
dt
         
∆v̇cd 0 ω0 ∆vcd 1
0 ∆i cd Vc [ ]
 =   +  Cf  +  ∆ω
q
(4.10)
∆v̇cq −ω0 0 ∆vcq 0 C1f ∆icq −Vcd
Chapter 4 145
Where:
vc : Tension of capacitor lter;
ic : Filter capacitor current.
The state space modelling of the rest microgrid elements can be obtained
by the same manner of the RLC lter. The following Section shows the state
space modelling of the voltage-frequency inverter based on the droop control
strategy as shown in equations (2.7) and (2.8).
4.4.2 Voltage-Frequency Inverter State Space Modelling
The proposed model of the voltage-frequency inverter is shown in Figure
4.4, as it consists of LCL output lter and controllers of power, voltage and
current. The inverter dc bus dynamics are neglected with respect to high
switching frequencies (4-10 kHz).
The reason for LCL lter choice is that the droop control is mainly based
on the assumption that the output impedance of the generator is pure induc-
tive, due to the large impedance of the lter and power lines. However, the
impedance of power lines in low voltage application, as the case of microgrids is
mainly resistive. This problem can be overcome by adding an output inductor,
leading to LCL output lter.
inverter
voltage current
controller controller
power
controller
Figure 4.4: Voltage-frequency inverter
As mentioned in the previous Section, in order to derive the state space
modelling, the separation between open-loop and closed-loop circuits must be
cleared. Thus, in case of voltage-frequency inverter, the open-loop is consid-
ered as the LCL lter and the controllers of the inverter. The feed-back part
146 Chapter 4
considers the introduced control variables to the inverter controllers, as shown
in Figure 4.4.
LCL Filter
The detailed LCL lter is shown in Figure 4.5. By considering the same
manner for modelling the RLC lter, the state variables of LCL lter are cho-
sen as the inductance currents ii , io and the voltage across the capacitor vo .
The linearized model of the LCL lter is shown by the following equation.
Figure 4.5: LCL lter
   −R    
f −1 1
∆i̇id ω0 0 0 0 ∆iid 0 0 0
   Lf Lf
   Lf
 
 ∆i̇  −ω −Rf 0 −1 0 0  ∆i   1
0
 iq   0 Lf   iq   0 0  ∆v id
   Lf
   Lf
 
∆v̇   1 0 ω0 −1  
0  ∆vod    
0   ∆viq 
 od   Cf 0   0 0 0 
   Cf
 +  
∆v̇  =  0 1
−ω0 0 0 −1     0  
 oq     ∆v oq  0 0 0  ∆vgd 
   Cf Cf
    
      −1 
 ∆i̇od   0 0 L1c 0 0 ω0   ∆iod   0 0 0  ∆vgq
      Lc

−1
∆i̇oq 0 0 0 L1c −ω0 0 ∆ioq 0 0 0 Lc
 
Iiq
 
 −I 
 id 
 
 Vo [ ]
 q 
+

 ∆ω

−V od 
 
 
 Ioq 
 
−Iod
(4.11)
Chapter 4 147
Where:
ii : Inductance lter current;
io : Output lter current of the inverter;
vo : Filter capacitor tension;
vg : Microgrid node tension.
Power Controller
The external power control loop sets the magnitude of the frequency and
inverter output voltage, according to the droop characteristics for the real and
reactive power. The instantaneous active and reactive power components p̃ and
q̃ respectively are calculated from the measured output voltage and current as
shown by equation (4.12).


p̃ = 1.5(vod iod + voq ioq )
(4.12)

q̃ = 1.5(voq iod − vod ioq )
Where:
q̃ : Instantaneous reactive power;
p̃: Instantaneous active power.
The instantaneous power components are passed through low-pass lters of
cut-o frequency ωc , to limit the high transient variation, as shown in equation

(4.13) and Figure 4.6.


P = ωc
1.5(vod iod + voq ioq )
s+ωc
(4.13)

Q = ωc
1.5(voq iod − vod ioq )
s+ωc
The linearized equation of the power controller is indicated by equation(4.14),
where the variables P and Q are considered as state variables. The operator
148 Chapter 4
Figure 4.6: Inverter power controller
∆ indicates to the linearized variables, while variables Iod , Ioq , Vod , Voq are the
variables of the considered operating point.


∆Ṗ = −ωc ∆P + 1.5ωc (∆vod Iod + ∆voq Ioq + Vod ∆iod + Voq ∆ioq )
(4.14)

∆Q̇ = −ωc ∆Q + 1.5ωc (∆voq Iod − ∆vod Ioq + Voq ∆iod − Vod ∆ioq )
Voltage and Current Controller
The output voltage and current are processed by the voltage-current con-
troller. The output voltage control is achieved through PI controllers, taking
into consideration that the inverter voltage is decomposed into dq coordinates,
and the voltage across the q coordinate is assumed to be zero. The proposed
current controller is employed to limit the high transient current, and improves
the control dynamics of the system; therefore, the semiconductor protection
can be ensured.
Figure 4.7 shows the inverter voltage-current controller; the dynamic equa-
tions of the voltage-current controller are shown by equation (4.15), where

d
s= dt
. The state variables in this case output from the controller integrators.
Thereby, the terms Φvd , Φvq , ζid, ζiq are considered as the state variables.


 i∗id − iid = eid = sζid =⇒ ζ̇id = i∗id − vid






i∗ − i = e = sζiq =⇒ ζ̇iq = i∗ − v
iq iq iq iq iq
(4.15)

 ∗
− vod = evd = sΦvd =⇒ Φ̇vd

 vod




voq
∗
− voq = evq = sΦvq =⇒ Φ̇vq
Chapter 4 149
voltage controller current controller
Figure 4.7: Inverter voltage-current controller
Microgrid Reference Frame
In the case of isolated microgrid, the network state space modelling is rep-
resented on the reference frame of one of the individual inverters (the master
inverter), whereas in the case of grid-connected mode, the reference frame is
considered as the innite bus. This reference frame is regarded as the common
reference frame, and all the other inverters are referred to it. The common
reference frame must be chosen to rotate at the same frequency of the micro-
grid. The output voltage of each inverter is related directly to the microgrid.
Therefore, it is a good choice to consider the output voltage of the master
inverter as the common reference for the whole microgrid state variables.
Figure 4.8 shows the common reference frame of the master voltage-frequen-
cy inverter. This frame is denoted by DQ coordinates, and pulsating by the
same frequency of the master inverter in dq1 coordinates by frequency ω = ω1 .

If another inverter is decomposed into dq2 coordinates, and to be connected
150 Chapter 4
to the microgrid, this inverter will be aligned to the master inverter frame by
an angle δ2 , as shown in equation(4.16).
By considering the droop relation between the frequency and active power,
which is indicated by ∆ω = −mp ∆P , and in order to relate each inverter ro-
tating frame to the common reference frame, each inverter reference frame can
be referred to the master one by equation (4.17). Consequently, each inverter
output voltage can be referred to the common reference frame by transforma-
tion as shown in equation(4.18). Finally, the linearized equations of the new
derived inverter voltage can be rewritten as indicated by equation (4.19).
δ̇2 = ω2 − ω1 (4.16)
∆δ̇j = ωj − ω1 = −mpj ∆Pj + mp1 ∆P1 (4.17)



voDj = vodj cos(δj ) − voqj sin(δj )
(4.18)

voQj = vodj sin(δj ) + voqj cos(δj )


∆voDj = cos(δjo ) ∆vodj − sin(δjo ) ∆voqj − Vodj sin(δjo )∆δj − Voqj cos(δjo )∆δj

∆voQj = sin(δjo ) ∆vodj + cos(δjo ) ∆voqj + Vodj cos(δjo )∆δj − Voqj sin(δjo )∆δj
(4.19)
Figure 4.8: Reference frame transformation
As mentioned in the previous Section, the proposed model in this Chapter

Chapter 4 151
dierentiates between the open-loop and closed-loop system. The controller
gains are dened by a separate matrix included by the feedback system. There-
fore the eect of the controller gains change can be observed. The state space
modelling can be dened by the following form:

      
ẋ J1A J2Ax J5A [ ]
 =   +   u
0 J3 J4 z J6
      
y J1C J2C x J5C [ ]
 =   +   u
0 J3 J4 z J6
Where x is the state variables matrix, z is the algebraic variables matrix,
u is the control variables matrix and y is the output variables matrix. The
proposed model of the inverter in this Chapter considers the following matrices:
∆x = [∆iid ∆iiq ∆vod ∆voq ∆iod ∆ioq ∆ζid ∆ζiq ∆Φvd ∆Φvq . . .
∆P ∆Q ∆δ]T
∆u = [∆vid ∆viq ∆iid ∆iiq ]T
∆z = [∆VgD ∆VgQ ]T
4.4.3 Doubly-Fed Induction Generator State Space Mod-
elling
The doubly-fed induction machine is basically a wound rotor induction ma-
chine with its stator windings directly connected to the grid and its rotor
windings connected to the grid through a converter. The AC/DC/AC Con-
verter is divided into the rotor side converter and grid side converter. These
converters use force commutated power electronic devices to synthesize an AC
Voltage from a DC source. A capacitor connected on the DC side acts as the
DC voltage source.
Figure 4.9 shows the Power ow in the DFIG. The electro-magnetic torque
Te is positive for power generation and the electrical frequency ωs is positive
and constant for a constant grid frequency. Pr is positive, if the DFIG me-
chanical speed is greater than the synchronous speed, and it is negative, if the
152 Chapter 4
mechanical speed is lower than the synchronous speed. The rotor side and
the grid side converters have the capability of generating or absorbing reactive
power and can be used to control the reactive power or the voltage at the
grid terminals. The rotor side converter is used to control the wind turbine
output real power, whereas the reactive power control is assumed deactivated
in the proposed DFIG model in this Chapter, as the reactive power control is
managed only by the voltage-frequency inverters. The grid side converter is
used to regulate the voltage of the DC bus capacitor.
By the same manner of the inverter state space modelling, the DFIG model
can be derived. The model separates between open-loop and closed-loop sys-
tems. The open-loop includes the induction generator part and the regulators
regarding grid side converter (GSC) and the rotor side converter (RSC), while
the feedback system includes the control variables which are introduced to
GSC and RSC controllers.
RSC GSC
DC
AC
AC
Rotor
Figure 4.9: DFIG power ow
Induction Generator
The induction machine can be represented by the transformer per phase
equivalent circuit model, where Rr and Xr are the rotor resistance and reac-
tance before being referred to the stator side, as shown in Figure 4.10. The
DFIG induction machine steady state model is decomposed into the dq coor-
dinates, and can be briefed by equations (4.4.3)-(4.4.3).

Chapter 4 153
' '
'
Figure 4.10: Induction machine equivalent circuit





 Vmd = Rs isd + ωpll isq Lsd + Vsd


Vmq = Vsq − Rs isq − ωpll isd Lsd (4.19)

 √



Ψ s =
(Vsd −Rs isd )2 +(Vsq −Rs isq )2
ωpll


Vmd = Vrd + ωpll Lrd irq − Rr ird − ωr Lm imq
(4.19)

Vmq = Vrq − ωpll Lrd ird − Rr irq + ωr Lm imd


Vmd = −ωpll Lm imq
(4.19)

Vmq = ωpll Lm imd
ωr
Tm = Te + F . (4.19)
p
Where:
im : DFIG magnetization current;
ir : DFIG rotor current;
is : DFIG stator current;
Te : DFIG electromagnetic torque;
Tm : DFIG mechanical torque;
vm : DFIG magnetization circuit tension;

154 Chapter 4
vr : DFIG rotor tension;
vs : DFIG stator tension;
ωr : Wind turbine mechanical speed;
ωpll : DFIG pulsation frequency;
F: Wind turbine friction constant;
p: DFIG poles number.
The linearized model of the induction generator can be derived by the same
manner of the RLC lter. The induction generator linearized equations are
briefed by equations (4.4.3)-(4.4.3).


∆isd
˙ = − Rs ∆isd + ωpll ∆isq + 1
∆Vsd − 1
∆Vmd + Isq ∆ωpll
Lsd Lsd Lsd
(4.19)

∆isq
˙ = −ωpll ∆isd − Rs
∆isq + 1
∆Vsq − 1
∆Vmq − Isd ∆ωpll
Lsd Lsd Lsd


 ˙ = − Rr ∆ird + (ωpll − ωr )∆irq + 1
∆Vrd∗ − 1
− Lm
∆ird
 Lrd Lrd Lrd
∆Vmd ω ∆imq
Lrd r




+(−Irq − LLm Imq )∆ωr + Irq ∆ωpll
rd

 ˙ = − Rr ∆irq − (ωpll − ωr )∆ird + 1
∆Vrq∗ − 1 Lm

 ∆irq ∆Vmq + ω ∆imd
Lrd r


Lrd Lrd Lrd


+(Ird + LLm Imd )∆ωr − Ird ∆ωpll
rd


∆imd
˙ = ωpll ∆imq + 1
∆Vmd + Imq ∆ωpll
Lm
(4.19)

∆imq
˙ = −ωpll ∆imd + 1
∆Vmq − Imd ∆ωpll
Lm
p . Tm F p . Te
∆ωr˙elec = − . ωr − (4.19)
J J J
Where:
J: Wind turbine inertia constant.
Rotor-Side Converter
The proposed control scheme of the rotor-side converter is illustrated in
Figure 4.11. In order to decouple the electromagnetic torque and rotor excita-
tion current, the induction generator is controlled in a synchronously rotating

Chapter 4 155
stator-ux-oriented reference frame, with its d-axis oriented along the stator-
ux vector position. The PI controllers are used for rotor current regulation.
The rst part of the RSC controller aims at controlling the active power,
so as to track the Pref , which is shown by Figure 4.11b. The reference active
power is determined by the wind turbine power speed characteristic (Cp − λ)

curve for maximum power extraction. The second part of the RSC controller
is shown by Figure 4.11a, and aims at controlling the reactive power, but it is
deactivated in the proposed model.
(a)
(b)
Figure 4.11: Control scheme of the rotor-side converter:

(a) Reactive power controller. (b) Active power controller.
156 Chapter 4
The linearized model of the RSC is shown by equation (4.20).


 ∆ζ˙Q = −∆Qg = 1.5Igq ∆Vsd − 1.5Igd ∆Vsq + 1.5Vsd ∆iigq − 1.5Vsq ∆igd



˙ = ∆i∗ − ∆ird = − Lm +Lsd 2 T ∗ . ∆Ψs + Lm +Lsd ∆T ∗ − ∆ird
∆ζird


rd 1.5pLm . Ψs e 1.5pLm . Ψs e


∆ζirq
˙ = ∆i∗ − ∆irq
rq
(4.20)
Where:
Ψs : DFIG stator ux linkage;
ζQ : Integral term of DFIG rotor side converter state variable;
ig : Output current from DFIG to the microgrid;
ζir : Integral term of DFIG rotor side converter;
Q: Reactive power.
DC-link capacitor
The equation which describes the energy balance of the DC-link capac-
itor is indicated in equation (4.21), while the DC-link circuit is shown by
Figure 4.12.


 iCdc Vdc = 1.5(ird Vrd + irq Vrq )



iCdc Vdc = 1.5(ibd Vbd + ibq Vbq ) (4.21)





The linearized model of the DC-link circuit is shown by equation (4.22)
∆ib ∆ir
∆V˙dc = − (4.22)
Cdc Cdc
Chapter 4 157
Where:
ib : Input current to DC-link circuit;
iCdc : DC-link capacitor current;
vb : DC-link tension.
Figure 4.12: DC-link circuit
Grid-Side Converter
Figure 4.13a shows the DFIG reference frame transformation, as the DFIG
stator voltage must be synthesized with the microgrid frequency. Thus, the
stator voltage reference frame is transformed to the common reference frame,
by the same manner as mentioned in the voltage-frequency inverter modelling.
Figure 4.13b shows the DFIG phase-locked loop circuit, as the pulsation of the
DFIG reference frame is obtained from the DFIG stator voltage.
Figures 4.14a and 4.14b show the control scheme of the grid-side converter.
The converter control operates in the synchronously rotating grid voltage-
oriented reference frame with its d-axis oriented along the grid voltage vector
position. The PI controllers are used for DC-link voltage regulation. The ref-
erence signal for the DC voltage Vdc∗ is set to a constant value independent of
the wind speed.

158 Chapter 4
(a)
(b)
Figure 4.13: DFIG reference frame: (a) DFIG reference

frame transformation. (b) DFIG phase-locked loop circuit.
Chapter 4 159
(a)
(b)
Figure 4.14: Control scheme of the grid-side converter:

(a) Control scheme in the coordinate-d. (b) Control
scheme in the coordinate-q
160 Chapter 4
The linearized model of the GSC is shown by equation (4.23)


 ˙ bd ) = − Rb ∆ibd + ωpll ∆ibq + 1
− 1

∆(i ∆Vsd ∆Vbd + Ibq ∆ωpll


Lb Lb Lb



 ˙ bq ) = − Rb ∆ibd − ωpll ∆ibd +
∆(i 1
∆Vsq − 1
∆Vbq − Ibd ∆ωpll

 Lb Lb Lb



 ˙ = −∆Vdc

∆ζdc


˙ = ∆ibd∗ − ∆ibd
∆ζibd (4.23)





 ˙ = −∆ibq
∆ζibq








˙ = ∆Vsq
∆ζpll


pll



∆θ̇ = ∆ωpll − ∆ω1
The proposed model of the DFIG in this Chapter considers the following
matrices:
∆x = [∆isd ∆isq ∆ird ∆irq ∆imd ∆imq ∆ωr ∆ζQ ∆ζird . . .
∆ζirq ∆ibd ∆ibq ∆vdc ∆ζdc ∆ζibd ∆ζpll ∆θpll ]T
∆u = [∆vrd ∆vrq ∆ibd ∆vbd ∆vbq ∆irq ∆ωpll ]T
∆z = [∆vmd ∆vmq ∆vsd ∆vsq ∆igd ∆igq ∆iCdc ∆Te . . .
∆Tm ∆Te∗ ∆ψs ∆λi ∆Cp ]T
4.4.4 Distribution Line State Space Modelling
In conventional power systems, the network dynamics are neglected, due to
the high time constant of rotating machines and their controls. In contrast,
DGs in microgrids are interfaced by inverters and their response times are
small. Therefore network dynamics inuence the system stability.
Figure 4.15 shows the modelling of the proposed short distribution line in this
Chapter, based on [125]. The model considers both the inductive and resistive
terms of the distribution line.
By the same manner, the linearized form of the microgrid distribution lines
Chapter 4 161
can be derived as shown in equation(4.4.4). Where ∆ operator corresponds
to the linearized variables, whereas Ilined and Ilineq are the distribution line
variables corresponding to the considered operating point.
       
˙ d −Rline 1
∆iline ω0 ∆ilined 0 ∆v gi
  =  Lline   +  Lline  
−Rline
˙
∆ilineq −ω0 Lline
∆i lineq 0 1
Lline
∆v gi
   
−1
0 ∆vgj Ilineq [ ]
+    ∆ω
Lline
−1
0 Lline
∆v gj −I line d
Where:
iline : Distribution line current.
Figure 4.15: Microgrid distribution line
The proposed model of the distribution line in this Chapter considers the
following matrices:
∆x = [∆ilineD ∆ilineQ ]T
4.4.5 Static Load State Space Modelling
The proposed static load model is shown in Figure 4.16; each static load is
connected between its correspondent microgrid node and a neutral point. The
162 Chapter 4
state space modelling is shown by equation (4.4.5).
       
˙ d −Rload 1
∆iload ω 0 ∆i loadd 0 ∆vgj
 = Lload
 + Lload
 
˙ q
∆iload −ω0 −R load
Lload
∆i load q 0 1
Lload
∆v gj
 
Iloadq [ ]
+  ∆ω
−Iloadd
Where:
iload : Static load current.
Figure 4.16: Microgrid static Load
The proposed model of the static load in this Chapter considers the follow-
ing matrices:
∆x = [∆iloadD ∆iloadQ ]T
The state space modelling of all proposed microgrid elements were detailed
until now. The aim of the next Section is to conduct the eigenvalues of the
state matrix. Thereafter, the inuence of each dynamic state in the microgrid
can be measured through the participation factor index. Also, the correspon-
dence between the eigenvalues of the state matrix and the time domain will
be dened.
Chapter 4 163
4.5 Conducting Participation Factor and Time

Response
Equations (4.3) and (4.4) can be Laplace transformed; thus, the state equa-
tions in the frequency domains are obtained as follows:



s∆x(s) − ∆x(0) = A∆x(s) + B∆u(s)
(4.23)

∆y(s) = C∆x(s) + D∆u(s)
Rearranging equation (4.5), a solution to the state equations can be ob-
tained as:
(sI − A)∆x(s) = ∆x(0) + B∆u(s) (4.23)
Where I is an identity matrix.
Eigenvalues
The eigenvalues of the matrix A, which was mentioned by equation (4.3) are
indicated by λ. The eigenvalues are obtained from the characteristic equation
of the matrix A as follows:
det(A − λI) = 0 (4.23)
The eigenvalues of the matrix A provide information about the system
response to a small disturbance, such as any oscillatory mode that exists in
the system. The eigenvalues consist of real and imaginary components, and
can be indicated as shown by equation (4.5).
λ = σ ± jω (4.23)
Typically, the real eigenvalues correspond to a non-oscillatory mode. A
negative real eigenvalue represents a decaying mode, while a positive eigenvalue
represents aperiodic instability. Complex conjugate eigenvalues correspond to
an oscillatory mode.
164 Chapter 4
Participation Factor
The participation matrix provides a measure of the inuence of each dy-
namic state on a given mode. The normalized participation factor Pni of the
nth state variable into mode i is shown as follows:
|ϕni ||ψin |
Pni = ∑n (4.23)
i=1 |ϕni ||ψin |
Where n is the number of state variables, ϕin is the ith element of the nth
right-eigenvector and ψin is the nth element of the ith left-eigenvector.
Time Response
According to [2], the time response of ith state variable can be shown
as follows:
∆xi (t) = ϕi1 c1 eλ1 t + ϕi2 c2 eλ2 t + · · · + ϕin cn eλn t (4.23)
Where cn = ψn ∆x(0). The derived form of this time response complies
with the open loop systems only, but in current state space model, the closed
loop system is applied. Therefore, another form of time response based on
[123] is adopted to dene the time response in this Chapter, and can be shown
by equation (4.5).
∆ẋ(t) = (A − BKC)∆x(t) (4.23)
Where K is the system gains matrix, and the term (A−BKC) is the closed
loop state variables matrix and can be indicated by Ac .
Thus far, the state space modelling of all proposed microgrid elements
were detailed and the eigenvalues, participation factor and time response were
dened. The next Sections will show the small signal stability analyses of
microgrid under dierent scenarios.

Chapter 4 165
4.6 Wind Variation Impact on the Microgrid Sta-

bility
4.6.1 Introduction
The wind energy is characterized by its uctuating nature, and so the gen-
erated power of wind turbines is completely determined by the wind speed
and not controlled any further. Typically, there are two impacts of connecting
wind energy to the grid, local and wide impacts of wind power. The local
impacts of wind power are the impacts that occur in the electrical vicinity of a
wind turbine or wind park that can be attributed to a specic turbine or park.
These eects occur at each turbine or park, independently of the overall
wind power penetration level in the system as a whole. Local impacts can
be realized in some main points, such as power ow, voltage of power system
nodes, harmonics and ickers. The system wide impacts, on the other hand are
the impacts which cannot be localized, they cannot be attributed to individual
turbines or parks. Nevertheless, they are strongly related to the penetration
level in the system as a whole. The wide impacts can be realized in some other
points, such as dynamic stability and voltage-frequency control.
Although wind power exploitation appear particularly attractive under on-
going energy policies, the integration of a substantial amount of wind power
in isolated electrical systems needs careful consideration, in order to maintain
a high degree of reliability and security of the system operation. The main
problems concern operational scheduling, due to high production forecasting
uncertainties, as well as steady state and dynamic operation disorder [126].
These problems may considerably limit the amount of wind generation that
can be connected to the power systems, increasing the complexity of their op-
eration. Thus, beyond voltage stability concerns, frequency stability must be
ensured [127]. This depends on the ability of the system to restore balance
between generation and load.
The doubly-fed induction generator (DFIG) is one of the most popular wind
conversion systems due to its high energy eciency, reduced mechanical stress
on the wind turbine and low power rating of the connected power electronics
converter [128]. As the penetration level of DFIG wind power generation into
the grid increases, the stability issue of the DFIG wind turbine system is of
166 Chapter 4
particular increasing concern.
The dynamic behaviour of DFIGs based on time-domain simulations have
been studied to investigate the performance of the DFIG system and its im-
pact on power system dynamics in [129] and [130]. In [131] and [132], detailed
dynamic models of DFIGs have been presented and small signal stability anal-
yses have been conducted. The small signal stability analyses can identify the
eects of the system parameters and operating points on the system modes.
The purposes of the behaviour and stability analyses of the DFIG systems
can oer better understanding of the system intrinsic dynamics, which can be
useful for robust control design of such systems.
The main purpose of this Section, is to study the stability of microgrids
supplied by DFIGs and other DGs interfaced through inverters, with respect
to wind speed increment. The simulations are processed for the isolated mi-
crogrid, and also for grid-connected mode.
4.6.2 Processing Simulations
Through this Section, the wind speed increment is simulated, in order to
determine the critical eigenvalues. The proposed microgrid is shown in Fig-
ure 4.1, where the microgrid is isolated. The wind speed is sampled from 1.5
m/s to 8 m/s for the rst DFIG only at the 6th node of the microgrid, while
the wind speeds at other DFIGs are assumed constant at 7 m/s; two hundred
samples are taken, also the proportional and integral dc-link gains of all DFIGs
supplying the microgrid kpdc , kidc are set as 1.1.
Figure 4.17 shows the loci of dominant eigenvalues of the isolated micro-
grid during wind speed increment. Figure 4.17a is related to the rst DFIG
magnetization current imdq1 ; Figure 4.17b is related to the mode dominated by
the dc-link state variables of the rst DFIG vdc1 , ζdc1 ; Figure 4.17c is related to
the mode of the voltage-frequency inverters aected by angle of rotation state
variables δ2,3 ; Figure 4.17d belongs to the dc-link mode of the second, third
and fourth DFIGs vdc2,3 , ζdc2,3,4 .

Table 4.1 shows the dominant modes of current simulation, participation
factors and their damping ratios. In Table 4.1, the rst three columns belong
to the status of the dominant state variables at the rst simulation sample,
while the last column of remarks belong to the status of the dominant state
Chapter 4 167
variables at the last simulation sample.
Table 4.1 and Figure 4.17 show that the system is stable through all the
range of wind speed increment, because there are not any mode cross the imag-
inary axis creating a saddle node or Hopf bifurcations, but at the same time,
its observed that oscillatory modes aected mainly by magnetization currents
either of the rst DFIG where the wind speed variations occur, or other DFIGs
of constant wind speed overall simulation, move towards the imaginary axis.
The modes dominated by magnetization currents are divided into coupled
modes, which are dominated by all DFIGs magnetization currents and other
decoupled modes aected by magnetization current of rst DFIG only, and
the latter has lower damping ratio.
The modes dominated by dc-link state variables are decoupled, as ζdc1 of
the rst DFIG is a dominant mode, but its response clearly improves through
moving to the left part at the real axis as shown in Figure 4.17b, while other
DFIGs dc-link modes are out of interest, because of being far from the imagi-
nary axis; thus, their analyses are ignored.
The mechanical speed variables of the DFIGs aect some modes, all of
them lay at the real axis without any oscillation, and these modes are decou-
pled, as the modes dominated by the rst DFIG of the changeable speed is
separated from the rest DFIGs, and all the mechanical modes of the DFIGs
are nearly xed, they do not show better or worse behaviour during the wind
speed increase.
The modes aected by inverters angle of rotation are oscillatory at low
frequency (11.1 rad/sec) and far from the imaginary axis, and so they are not
included in the analysis. These modes as shown in Figure 4.17c, and move by
little values to higher frequencies far from the imaginary axis, such a move-
ment is logic, due to dierent levels of power exchange during the wind speed
variation, and so the inverters try to resynchronize and recover from power
swings.
Clearly, the dominant modes through current simulation are dedicated to the
mechanical speed, dc-link and magnetization currents of the DFIGs.

168
Table 4.1: Dominant Modes of the Isolated MG During Wind Speed Increment-Case 1
λ = σ ± jω f, HZ Damping (ξ ) Dominant states Participation factor Remarks
-0.28+j311.05 49.49 0.0009 electrical, imdq1 0.37 stable
-1.16 0 1 electrical, vdc1 , ζdc1 1 stable
-2.24 0 1 electrical, ζdc2 0.96 stable
-0.96+j311.1 49.5 0.0031 electrical, imdq1,2,3,4 0.13 stable
-0.28+j311 49.4 0.00099 electrical, imdq2,3 0.29 stable
-0.0019 0 1 mechanical, wr1 1 stable
-0.009 0 1 mechanical, wr2,3,4 0.47 stable

Chapter 4
Chapter 4 169
imag
(a) (b)
imag
(c) (d)
Figure 4.17: Loci of eigenvalues of the isolated microgrid

during wind variation-case 1: (a) Magnetization current
mode of the 1st DFIG imdq1 . (b) DC-link mode of the 1st
DFIG vdc1 andζdc1 . (c) Inverters angle of rotation mode
θpll . (d) DC-link mode of the 2nd, 3rd and 4th DFIGs
vdc2,3 andζdc2,3,4 .
170 Chapter 4
The Eect of Reducing the DC-Link Control Gains on Isolated Microgrids
The previous simulation is processed for the isolated microgrid through
dierent dc-link proportional and integral gains of DFIGs kpdc , kidc , as both of
them are set as 0.1, in order to clear the eect of the dominant dc-link modes
on the microgrid stability. The wind speed of the rst DFIG at the 6th node
is sampled from 1.5 m/s to 8 m/s by 200 samples, while the other DFIGs have
constant wind speed 7 m/s.
Table 4.2 shows the dominant modes, as they are dedicated to DFIGs state
variables, such as magnetization currents, dc-link and mechanical speed. In
Table 4.2, the rst three columns belong to the status of the dominant state
variables at the rst simulation sample, while the last column of remarks be-
long to the status of the dominant state variables at the last simulation sample.
Figure 4.18a shows the magnetization current mode of the rst DFIG imdq1 ;
Figure 4.18b shows the mechanical modes aected by the contributing DFIGs
wr2,3,4 ; Figure 4.18c shows the dc-link mode vdc2,4 of the second and fourth
DFIGs; Figure 4.18d shows the dc-link mode vdc1 , ζdc1 dominated by the rst
DFIG.
The obvious dierence that occurs with decreasing the proportional and
integral control gains of dc-link control circuit that the system turned to insta-
bility. The modes dictated by magnetization currents, are totally decoupled,
as the magnetization current of the rst DFIG aects an oscillatory mode,
while the magnetization currents of the rest DFIGs inuence other oscillatory
modes, and the frequencies of all magnetization current modes increase during
wind speed increment.
The trajectories of the mechanical modes involved by the DFIGs are nearly
constant and do not move either to the right or to left side.
An interesting conclusion occurs, as the modes dominated by dc-link vari-
ables are the only responsible for instability in the microgrid, as shown in
Figure 4.18c and 4.18d. The dc-link state variables of the second, third and
fourth DFIGs dominate the unstable mode, shifting it to the right part, while
the rst DFIG dc-link state variables motivate a Hopf bifurcation by crossing
the imaginary axis to the right side. Finally, the modes related to the invert-
ers angle of rotation δ, are still out of interest, because they lay far from the
imaginary axis.
Chapter 4 171
Figure 4.19 shows the deteriorated behaviour at low dc-link control gain
values versus the wind speed, as the dc-link tension reaches terrible values,
which could damage the dc-link circuit.

172
Table 4.2: Dominant Modes of the Isolated MG During Wind Speed Increment-Case 2
-0.28+j311.04 49.5 0.0009 electrical, imdq1 0.39 stable
-0.28+j311.04 49.5 0.0009 electrical, imdq2,3,4 0.36 stable
3.3 0 1 mechanical, wr2,3,4 0.56 stable
-0.3+j1.05 0.1671 0.27 electrical, ζdc1 , vdc1 0.52 unstable (
Hopf )
3.38 0 1 electrical, vdc2,4 0.56 unstable

Chapter 4
Chapter 4 173
−4
x 10
5
311.14
311.12
311.1
imag
imag
0
311.08
311.06
311.04
−5
−0.285 −0.28 −0.275 −0.27 −9.124−9.122 −9.12 −9.118
real real x 10
−3
−7 (a) (b)
x 10
1
1
0.5 0.5
imag
imag
0 0
−0.5
−0.5
−1
−1.5 −1
3.39 3.395 3.4 0 1 2 3
real real
(c) (d)

during wind variation-case 2: (a) Magnetization current
mode of the 1st DFIG imdq1 . (b) Mechanical speed mode
of the 2nd, 3rd and 4th DFIGs wr2,3,4 . (c) DC-link mode
of the 2nd and 4th DFIGs vdc2,4 . (d) DC-link mode of the
1st DFIG ζdc1 , vdc1 andζdc1 .
Wind Speed Increment, Grid-Connected Mode
Another simulation is conducted by connecting the microgrid to the in-
nite bus; the proposed microgrid is shown in Figure 4.2. The wind speed is
sampled from 3 m/s to 8 m/s by 150 samples, and these samples are applied to
the rst DFIG only at the 6th node, while the wind speed of the other DFIGs
are constant at 7 m/s, also the dc-link proportional and integral gains kpdc , kidc
are set as 1.1.
Table 4.3 shows the dominant modes of the current simulation, participa-
tion factors and their damping ratios. In Table 4.3, the rst three columns
belong to the status of the dominant state variables at the rst simulation
sample, while the last column of remarks belong to the status of the dominant
state variables at the last simulation sample.
Figure 4.20a shows the dc-link mode of the rst DFIG vdc1 , ζdc1 ; Fig-
174 Chapter 4
6
dc−link tension (p.u)
5
0
1.5 2.5 3.5 4.5 5.5 6.5 7.5
wind speed (m/s)
Figure 4.19: DC-link tension at low dc-link control gains

versus the wind speed
ure 4.20b shows the dc-link mode related to the rest DFIGs vdc2,3,4 ; Figure 4.20c
shows the mechanical speed mode of the rst DFIG wr1 ; Figure 4.20d shows
the mechanical speed mode related to the rest DFIGs ωr2,3,4 .

Unlike the isolated microgrid case, the modes aected by DFIGs magne-
tization currents disappear, and lay far from the imaginary axis, while the
oscillatory modes dictated by dc-link state variables are more damped with
wind speed increase in the current case than the isolated microgrid case, and
they step to the left, acquiring more stability.
In addition, the stability of the mechanical speed mode related to the rst
DFIG improves during wind speed increase, while in the isolated microgrid
simulations, this mode lays at the real axis near to the saddle node bifurcation
limit and without any improvement.

Table 4.3: Dominant Modes of the Grid-Connected Mode During Wind Speed Increment-Case 1

Chapter 4
-1.25+j0.89 0.1416 0.8146 electrical, vdc1 , ζdc1 0.8 stable
-1.68+j0.76 0.1210 0.9111 electrical, vdc2,3,4 0.9 stable

175
176 Chapter 4
0.75
0.95
imag
imag
0.7
0.9
0.65
0.85
-1.6 -1.5 -1.4 -1.3 -1.8 -1.75 -1.7
real real
-4 (a) -4 (b)
x 10 x 10
5 5
imag
imag
0 0
-5 -5
-6 -4 -2 0 -5.4 -5.2 -5 -4.8 -4.6
real real -3
x 10
(c) (d)
Figure 4.20: Loci of eigenvalues of the grid-connected

mode during wind variation-case 1: (a) DC-link mode of
1st DFIG vdc1 andζdc1 . (b) DC-link mode of 2nd, 3rd and
4th DFIGs vdc2,3,4 . (c) Mechanical speed mode of 1st DFIG
wr1 . (d) Mechanical speed mode of 2nd, 3rd and 4th DFIGs
ωr2,3,4 .
Chapter 4 177
The Eect of Reducing the DC-Link Control Gains, Grid-Connected Mode
The previous simulation is conducted regarding the grid-connected mode,
by setting the DFIGs proportional and integral gains of the dc-link circuits
as 0.1, in order to clear the eect of these gains on the stability of the whole
system.
Table 4.4 explains the dominant modes, their participation factors and
damping ratios. In Table 4.4, the rst three columns belong to the status
of the dominant state variables at the rst simulation sample, while the last
column of remarks belong to the status of the dominant state variables at the
last simulation sample.
Figure 4.21a shows the rst DFIG dc-link mode vdc1 ,ζdc1 ; Figure 4.21b shows
the magnetization current dominant mode of the second and third DFIGs
imdq2,3 ; Figure 4.21c shows the dc-link modes of the second and the fourth
DFIGs vdc2,4 .
The grid turns totally unstable; also the magnetization current modes ac-
quire their dominant oscillatory nature again in the current case, and step
towards the imaginary axis with increasing wind speed, raising a Hopf bifur-
cation possibility as shown in Figure 4.21b. Moreover, the modes inuenced
by dc-link state variables cross the imaginary axis with wind speed increase,
creating a Hopf bifurcation, and lay at the unstable part.
The dominant modes shown in the simulations of this Section coincide with
other studies as in [132] and [133], where the dominant modes are related to
the dc-link and mechanical speed of DFIGs.

Table 4.4:
178
Dominant Modes of the Grid-Connected Mode During Wind Speed Increment-Case 2
-0.2815+j314.12 49.5 0.0008 electrical, imdq2,3 0.43 stable
-0.004 0 1 mechanical, wr1 0.99 stable
3.58 0 1 electrical, vdc2,4 0.68 unstable
0.4+j1.1 0.17 -0.34 electrical, ζdc1 , vdc1 0.52 unstable

Chapter 4
Chapter 4 179
0.5
imag
-0.5
-1
0 0.5 1 1.5 2 2.5 3 3.5 4
real
(a) x 10
-4
5
314.12
314.115
imag
imag
0
314.11
314.105 -5
-0.28 -0.278 -0.276 3.58 3.5805 3.581
real real
(b) (c)

mode during wind variation-case 2: (a) DC-link mode of
1st DFIG ζdc1 andvdc1 . (b) Magnetization current mode of
2nd and 3rd DFIGs imdq2,3 . (c) DC-link mode of 2nd, 3rd
and 4th DFIGs vdc2,4 .
180 Chapter 4
The conclusions of this Section are briefed hereunder:
• The stability of the microgrid vanishes with decreasing the DFIGs dc-
link proportional and integral gains, either in the isolated mode or grid-
connected mode;
• By increasing the wind speed, the oscillatory dominant modes move to-
wards the imaginary axis, raising the possibility of a Hopf bifurcation,
this result coincide with other papers as in [132] and [33];
• For both the stable isolated microgrid and grid-connected mode, the
stability of the mechanical speed modes of the DFIGs improve with in-
creasing the wind speed;
• The oscillatory modes in the case of isolated microgrid are related to the
magnetization currents of the DFIGs, and move towards the imaginary
axis with wind speed increase, whereas in the case of grid-connected
mode, the oscillatory modes are related to the dc-link state variables
and move far from the imaginary axis with higher damping ratio with
increasing the wind speed;
• In the case of stable isolated microgrid, the dc-link mode of the DFIG,
where wind speed increments, is decoupled from dc-link modes of the
other DFIGs, where their wind speeds are constant. Moreover, the me-
chanical mode of the DFIG, where wind speed increments, is decoupled
from the other mechanical modes of DFIGs, where their wind speeds are
constant.
The stability analysis of microgrids either in the isolated mode or connected
to the innite bus were conducted in this Section during wind speed increment.
The simulated microgrid through this Chapter is supplied by both DGs inter-
faced by inverters and DFIGs. Therefore, the stability study during inverters
droop regulation is necessary, because they play the main role for stable power
exchange and safe resynchronization. Simulations regarding microgrid stability
during inverters droop regulations will be conducted in the following Section.

Chapter 4 181
4.7 Droop Constants variation Impact on the

Microgrid Stability
4.7.1 Introduction
The microgrids are usually built using low-inertia generators that can easily
adapt to a rapidly changing environment. Although the benets of low-inertia
generation, the drawback is that large demand steps can cause a system to
become unstable and losing synchronism. A control system is necessary to
bring stability while providing ecient electricity to the microgrid.
The droop control uses only local power to detect changes in the system and
adjust the operating points of the generators accordingly.
The droop control uses the real power out of a generator to calculate the
ideal operating frequency.
In [134] it was concluded that DGs may have signicant impact on the
power system stability, if the power system is not properly compensated by
reactive power. In [135] it was shown that DGs could have signicant impacts
on transmission system stability at heavy penetration levels. Typically, a DG
unit aects the system stability by generating or consuming active and reac-
tive power. Thus, the power control performance of a DG unit determines
its impact on the microgrid, whether in the isolated or grid-connected mode.
When the power control strategy can handle successfully the load demand, the
DG unit can be used as a mean to enhance the system stability and improve
the power quality; otherwise it could undermine the system stability.
Due to the large demand uncertainties, the DGs interfaced by inverters
suer more challenges which call for the needs of stability analyses. First of
all, the DGs interfaced by inverters have much less over current capability
compared to rotation machine units [136]. Secondly, the intermittent nature
of non-dispatchable generators supplying power systems, as in the case of the
propsoed microgrid in this Chapter. Thirdly, the microgrid normally has a
wide-band dynamics due to the presence of fast response of DGs interfaced by
inverters and multiple small DGs with dierent power capability [137].
Due to all previous challenges, there is a strong need for stability analyses
to guarantee proper and stable system operations in the microgrid. Through
the rst trials to study the stability of DGs interfaced by inverters, the stability
182 Chapter 4
analyses have been conducted by Bode plots using the phase margin and gain
margin. In single LCL grid-tie inverter systems, bode plots have been used
to study the eects of changing some important parameters, such as feedback,
plant parameters and controller gains as shown in [138] and [139].
The stability of large scale DG systems with droop control is usually anal-
ysed by the state-space models. The purpose of this Section is small signal
stability analysis of the isolated microgrid, supplied by wind energy and DGs
interfaced through inverters operating by the droop control, under droop con-
stants increment. Therefore, the stability of the isolated microgrid can be
evaluated.
Through this Section, the droop constants of the inverter-interfaced gener-
ators supplying the isolated microgrid are increased gradually through eighty
samples, in order to determine the critical eigenvalues.
Figure 4.22 explains the dominant modes trajectories of the isolated mi-
crogrid during inverters droop constants increment, when the simulation is
processed at high wind speed equals to 10 m/s. Figure 4.22a shows the trajec-
tory of DFIGs magnetization current mode imdq1,2,3,4 ; Figures 4.22b show the
trajectories of the modes related to the inverters angle of rotation δ2,3 ; Fig-
ure 4.22c shows the trajectories of the modes dominated by the integral terms
of inverters voltage controllers Φvd1,2,3,4 ; Figure 4.22d shows the dc-link mode
related to the DFIGs ζdc1,2 .

Table 4.5 shows the dominant modes of the current simulation, participa-
tion factors and their damping ratios. In Table 4.5, the rst three columns
belong to the status of the dominant state variables at the rst simulation
sample, while the last column of remarks belong to the status of the dominant
state variables at the last simulation sample.
It is clear that the system is stable, as no critical eigenvalues cross the imag-
inary axis. The modes aected by DFIGs magnetization current shift towards
the imaginary axis. In Figure 4.22b, the angle of rotation of the inverters in-
uence modes near to the imaginary axis and turn them to oscillatory modes,
moving clock wise towards the imaginary axis again to cross it creating a Hopf
bifurcation with incrementing the droops. In Figure 4.22c, other modes related
Chapter 4 183
to voltage controller variables of the inverters move at the real axis towards
the imaginary axis, by incrementing the droops, these modes turn unstable.
The Modes dominated by DFIGs dc-link variables step towards the imagi-
nary axis as in Figure 4.22d; other modes dominated by dc-link variables move
far from the imaginary axis with incrementing the droops. Mechanical speed
modes of the DFIGs shift far from the imaginary axis by little values.
Table 4.5:
184
Dominant Modes of the Isolated MG During Inverters Droop Constants Increment-Case 1
-0.58+j313 49.8 0.0019 electrical, imq1,2,3,4 0.127 stable
-3+j11.5 1.83 0.2524 electrical, δ2,3 0.414 unstable (Hopf-
bifurcation
-0.26 0 1 electrical, ζdc1,2 0.35 stable

Chapter 4
Chapter 4 185
imag 200 100
imag
0 0
−200 −100
−0.5 −0.4 −0.3 −15 −10 −5 0

real real
(a) −8
x 10 (b)
0.01
0.005 5
imag
imag
0 0
−0.005 −5
−0.01
−5.8−5.6−5.4−5.2 −5 −4.8−4.6 −0.265−0.26−0.255−0.25−0.245
real real
(c) (d)

during inverters droop increment-case 1: (a) Magnetization
current mode of DFIGs imdq1,2,3,4 . (b) Inverters angle of
rotation modes δ2,3 . (c) Integral term of inverters voltage
controller modes Φvd1,2,3,4 . (d) DC-link mode of DFIGs
ζdc1,2 .
186 Chapter 4
The Eect of the Droop constants Increment at Low Wind Speeds
The previous simulation is conducted at mean wind speed equals to 3 m/s,
by incrementing the droops of the inverter-interfaced generators. The system
turn to be unstable as shown in Figure 4.23. Figure 4.23a shows the trajec-
tory of the DFIGs magnetization current mode imdq1,2,3,4 ; Figures 4.23b and
Figure 4.23c show the trajectories of the inverters angle of rotation modes δ2,3 ;
Figure 4.23d shows the dc-link mode related to the third DFIG vdc3 , ζdc3 .
Table 4.6 shows the dominant modes of the current simulation, participation
factors and their damping ratios. In Table 4.6, the rst three columns belong
to the status of the dominant state variables at the rst simulation sample,
while the last column of remarks belong to the status of the dominant state
variables at the last simulation sample.
Clearly, the modes dominated by the inverters angle of rotation turn un-
stable and cross the imaginary axis by incrementing the droops, producing a
Hopf bifurcation; other modes aected by angles of rotation as shown in Fig-
ure 4.23c step towards the imaginary axis through dierent trajectories, but
they are still stable. The modes dominated by dc-link variables of the third
DFIG move towards the imaginary axis by incrementing the droops as shown
in Figure 4.23d. DFIG magnetization current modes shift towards the imagi-
nary axis as in Figure 4.23a.
It can be observed that DFIG magnetization current modes have the same
response, by incrementing the droops of inverters, during the low and high wind
speed cases (3 and 10 m/s) respectively, as they move towards the imaginary
axis. Moroever, The inverters angle of rotation modes have the same trajectory
for all wind speed cases, as they inuence oscillatory modes and move clock
wise towards the imaginary axis again to cross it causing a Hopf bifurcation,
with incrementing the droops. In addition, during the low wind speed case (3
m/s), the inverters angle of rotation modes turn unstable more rapidly. Fur-
thermore, for the dc-link and mechanical speed modes of the DFIGs, it is clear
that during the high wind speed case (10 m/s), the stability of such modes
improve by incrementing the droops, and during the low wind speed case (3
m/s), these modes shift towards the imaginary axis and their stability worsen.
The dominant modes of the current simulations coincide with the dominant
modes shown in [140], [23] and [48].

Chapter 4 187
Figure 4.24 shows the behaviour of the microgrid during the inverters
droops increment through the low wind speed case (3 m/s). Figure 4.24a ex-
plains that the inverters keep stable, till certain droop limit, and the increment
over this droop limit motivates synchronization problems and the inverter-
interfaced generators start to experience power swing moments as shown in
Figure 4.24b. Figure 4.24c clears the frequency collapse in the microgrid,
which is created from inverters droops increment. Figure 4.24 shows clearly
that the instability case experienced from the droop increment is (oscillatory)
instability.
Table 4.6:
188
Dominant Modes of the Isolated MG During Inverters Droop Constants Increment-Case 2
-0.99+j311 49.49 0.0032 electrical, imdq1,2,3,4 0.127 stable
-3+j11.6 1.83 0.2524 electrical, δ2,3 0.39 unstable (Hopf-
bifurcation
-1.3 0 1 electrical, vdc3 , ζdc3 1 stable

Chapter 4
Chapter 4 189
200 100
imag
imag
0 0
−200 −100
−1 −0.5 0 −20 −10 0

real real
−3
(a) x 10 (b)
4 2
2 1
imag
imag
0
0
−1
−2
−2
−4 −3
−8 −6 −4 −1.35 −1.3 −1.25 −1.2 −1.15
real real
(c) (d)

during inverters droop increment-case 2: (a) Magnetization
current mode of DFIGs imdq1,2,3,4 . (b) Inverters angle of
rotation modes δ2,3 . (c) Inverters angle of rotation modes
δ2 . (d) DC-link mode of the 3rd DFIG vdc3 , ζdc3 .
The conclusions of this Section are briefed hereunder:
• The magnetization current modes of the DFIGs have the same response
during the low and high wind speed cases, as by incrementing the invert-
ers droops, these modes have oscillatory motion towards the imaginary
axis;
• The inverters angle of rotation modes have the same trajectory during
the low and high wind speed cases, as they inuence oscillatory modes
near to the imaginary axis and move clock wise, to cross it, causing a
Hopf bifurcation by incrementing the droops;
• The stability of dc-link and mechanical speed modes related to the DFIGs
worsens by incrementing the droops during the low wind speed case;
• The stability of the microgrid through inverters droop increment worsens
earlier during the low wind speed case than the high wind speed case.
190 Chapter 4
inverter power (p.u) of rotation(degree)

3
inverter angle 2
1
0
9.4e−7 9.9e−7 2e−6 5e−6 2e−5 5e−5
droop constant
5
0
9.4e−7 9.9e−7 2e−6 5e−6 7e−6 9.4e−6 2e−5 4e−5 5e−5
droop constant
frequency (p.u)
0.8
0.6
9.4e−7 9.9e−7 2e−6 5e−6 7e−6 9.4e−6 2e−5 4e−5 5e−5
droop constant
Figure 4.24: Eect of inverters droop increment: (a) In-

verter angle of rotation. (b) Inverter power. (c) Frequency.
Till now, the stability analyses of microgrids were conducted, as a result
of inverters droop increment and wind energy increase, but the analysis of
the power demand increment is necessary, in order to clear the behaviour of
microgrid generators in response to such variations. In the following Section,
the demand increment impact on the microgrid stability will be conducted.

Chapter 4 191
4.8 Power Demand Increment Impact on the Mi-

crogrid Stability
4.8.1 Introduction
As the loading of a power system increases, the system reaches a critical
point, where it can no longer deliver the expected demand power. This point is
known by the Saddle-Node Bifurcation limit (SNB), where the system is char-
acterized by a singular Jacobian, and so the system loses stability in terms of
voltage. The bifurcation usually occurs in single parameter families of ordinary
dierential equations (ODEs), which are mainly related to Saddle Node and
Hopf Bifurcations.
The bifurcation eect becomes clearer by the limitation of the reactive
power that can be generated by the system due to voltage collapse, as it is a
well-established fact that voltage collapse in power systems is associated with
system demand increase beyond certain limits. Typically, if the reactive power
demand is greater than the supply reactive power, the voltage will decrease.
As the voltage decreases, the dierence between reactive power supply and
demand increases and the voltage falls even more until it eventually falls to a
very small value. The bifurcation limit is critical for the operation and plan-
ning of a power system, to be able to prevent instability.
Typically, Voltage stability is dened as the ability of the system to main-
tain steady voltages after being subjected to a disturbance. Therefore, possible
outcome of voltage instability is loss of load in an area, or tripping of transmis-
sion lines and other elements by their protective systems leading to cascading
outages. Consequently, generators loss of synchronism may result from these
outages.
Power system maximum loadability has been discussed in many papers as
in [141], where maximum loading point was estimated in normal steady state
condition and also in dierent line outage conditions. In [142], the sensitivi-
ties of the load parameter with respect to the load bus power injections are
evaluated for each optimal power ow solution, also the participation of the
load buses in both active power generation and critical loadability has been
conducted.
The proposed small signal stability analyses in this Section are conducted
192 Chapter 4
for microgrids supplied by DGs interfaced through inverters and doubly-fed
induction generators, under continuous power demand increment for the mi-
crogird, whether in the isolated mode or grid-connected mode. Doubly-fed
induction generators are typically nonlinear systems; bifurcation phenomena
in such nonlinear systems may occur under certain conditions, leading to os-
cillatory instability. Therefore, a special attention must be paid to bifurcation
phenomena through DFIG stability analysis, because the stability analysis of
DFIG from a bifurcation perspective has not been cleared well in most papers,
which have focussed on DFIGs stability.
Based on the proposed DFIG model in this Chapter, the dominant eigen-
values are analyzed, revealing that a Hopf bifurcation is likely to happen in
the system under load demand increase.
Through this Section, the load demand is increased gradually through four
hundred samples, from 7.5 p.u to 12 p.u, and simulations are processed for the
isolated microgrid mode and grid-connected mode, in order to determine the
critical eigenvalues.
Table 4.7 shows the dominant modes, participation factors and their damp-
ing ratios in the isolated microgrid. The simulation is processed at mean wind
speed equals to 7 m/s. In Table 4.7, the rst three columns belong to the
status of the dominant state variables at the rst simulation sample, while the
last column of remarks belong to the status of the dominant state variables at
the last simulation sample.
Figures 4.25a and 4.25b show the dierent DFIGs magnetization current
mode trajectories; Figures 4.25c and 4.25d show dierent trajectories of the in-
verters angle of rotation modes; Figure 4.26a shows the DFIGs dc-link modes;
Figure 4.26b shows the trajectory of DFIGs mechanical speed mode.
From the previous gures, it is clear that the instability originates from
a mode inuenced by DFIGs dc-link state variables, as it moves to the right
at the real axis and turn oscillatory to cross the imaginary axis and causing
a Hopf bifurcation, as shown in Figure 4.26a. The rest modes have dierent
response to the load demand increase, but all of them are stable.
Some modes related to DFIGs magnetization currents move counter clock wise
Chapter 4 193
towards the imaginary axis, while other modes inuenced by DFIGs magne-
tization currents step far from the imaginary axis. Inverters angle of rotation
modes have two dierent trajectories, but both of them step towards the imag-
inary axis. Finally, DFIGs mechanical speed mode shows a slight response to
the load demand increment and moves far from the imaginary axis by little
values.
Table 4.7:
194
Dominant Modes of the Isolated MG During Demand Increment at Mean Wind Speed: 7 m/s
-0.31+j311 49.49 0.0009 electrical, imdq1,2,4 0.16 stable
-3+j11.5 1.83 0.0096 electrical, δ2,3 0.39 stable
-5.7 0 1 electrical, vdc2,4 0.73 unstable (Hopf-
bifurcation)

Chapter 4
Chapter 4 195
200 200
imag
imag
0 0
−200 −200
−1 −0.8 −0.6 −0.4 −0.32 −0.31 −0.3 −0.29 −0.28

real real
(a) x 10
−3 (b)
10
10
5 5
imag
imag
0 0
−5
−5
−10
−4 −3 −2 −1 0 −4 −3 −2
real real
(c) (d)
Figure 4.25: Loci of eigenvalues of the isolated micro-

grid during demand increment at mean wind speed 7 m/s:
(a) Magnetization current mode of DFIGs imdq1,2,3,4 . (b)
Magnetization current mode of DFIGs imdq1,2,4 . (c) Invert-
ers angle of rotation modes δ2,3 . (d) Inverters angle of
rotation modes δ2 .
196 Chapter 4
2
imag
−2
−6 −5 −4 −3 −2 −1 0 1 2
real
(a)
0.4
0.2
imag
−0.2
−0.4
−9.6 −9.5 −9.4 −9.3 −9.2 −9.1
real x 10
−3
(b)
Figure 4.26: The rest of critical eigenvalues trajectories

of the isolated microgrid during demand increment at mean
wind speed 7 m/s: (a) DC-link mode vdc2,4 andζdc2,4 . (b)
DFIG mechanical speed mode ωr2,3,4 .
Chapter 4 197
Load Demand Increment at Mean Wind Speed 7 m/s, Grid-Connected Mode
ing ratios in the case of load demand increment from 7.5 p.u to 12 p.u for the
grid-connected mode. The simulation is processed at mean wind speed equals
to 7 m/s. In Table 4.8, the rst three columns belong to the status of the
dominant state variables at the rst simulation sample, while the last column
of remarks belong to the status of the dominant state variables at the last
simulation sample.
Figures 4.27a and 4.27b show dierent DFIGs magnetization current modes
trajectories; Figure 4.27c shows the trajectory of modes dictated by DFIGs in-
tegral terms of phased-locked loop circuits; Figures 4.27d and 4.28a show the
DFIGs dc-link modes; Figure 4.27b shows the trajectory of DFIGs mechanical
speed mode.
The modes dictated by DFIGs dc-link variables are still the source of in-
stability in the system, by crossing the imaginary axis to the unstable part.
Some modes related to DIFGs magnetization currents move towards the imag-
inary axis and their frequencies increase with incrementing the load demand as
in Figure 4.27a, while other modes aected by DIFGs magnetization currents
step far from the imaginary axis as in Figure 4.27b. In Figure 4.28b, DFIGs
mechanical speed mode shows a slight response to the load demand increment
and moves far from the imaginary axis by little values. DFIGs integral terms
of phased-locked loop circuit modes are far from the imaginary axis and out
of the analysis scope, as shown in Figure 4.27c.

Table 4.8:
198
Dominant Modes of the Grid-Connected Mode During Demand Increment at Mean Wind Speed: 7
m/s
-26.2 0 1 electrical, ζpll1,2,3,4 0.3 stable
-3.45+0.3j 0.047 0.99 electrical, vdc1,2,3,4 1 unstable
-2.48 0 1 electrical, ζdc1,2 1 unstable
-0.009 0 1 electrical, wr1,4 0.54 stable

Chapter 4
Chapter 4 199
200 200
imag
imag
0 0
−200 −200
−4 −3 −2 −1 0 −0.29 −0.285 −0.28

real real
x 10
−3 (a) (b)
2
2
1
imag
imag
0
0
−1 −2
−28 −27 −26 −3 −2 −1 0

real real
(c) (d)

mode during demand increment at mean wind speed 7 m/s:
(a)Magnetization current mode of DFIGs imdq1,2,3,4 . (b)
Magnetization current mode of DFIGs imdq1,4 . (c) DFIGs
integral terms of phased-locked loop ζpll1,2,3,4 . (d) DC-link
mode vdc1,2,3,4 .
200 Chapter 4
2
imag
−2
−3.5 −3 −2.5 −2 −1.5 −1 −0.5 0 0.5

real
(a)
0.4
0.2
imag
−0.2
−9.26 −9.24 −9.22 −9.2 −9.18 −9.16 −9.14 −9.12

real x 10
−3
(b)

of the grid-connected mode during demand increment at
mean wind speed 7 m/s: (a) DC-link mode vdc1,2 . (b)
DFIG mechanical speed mode ωr1,4 .
Chapter 4 201
Load Demand Increment at Mean Wind Speed 12 m/s, Isolated Microgrid
ing ratios, in the case of load demand increment from 7.5 p.u to 12 p.u in the
isolated microgrid, the simulation is processed at mean wind speed equals to
12 m/s. In Table 4.9, the rst three columns belong to the status of the dom-
inant state variables at the rst simulation sample, while the last column of
remarks belong to the status of the dominant state variables at the last simu-
lation sample.
Figure 4.29a shows the DFIGs magnetization current mode trajectories;
Figures 4.29b and 4.29c show the modes related to the inverters angle of rota-
tion modes; Figures 4.29d and 4.30a show the trajectories of the DFIGs integral
terms of dc-link modes; Figure 4.30b shows the trajectory of the DFIGs me-
chanical speed.
Through the simulation of low wind speed case (7 m/s), where the source of
instability in this case was related to the modes dominated by dc-link variables
only, but in the current case of wind speed equals to 12 m/s, the instability
source of the system originates from two dominant modes, those inuenced by
the inverters angle of rotation and the DFIGs dc-link state variables. DFIGs
magnetization currents mode steps towards the imaginary axis, while DFIGs
mechanical speed mode shifts slightly far from the imaginary axis through load
demand increment.
Table 4.9:
202
Dominant Modes of the Isolated MG During Demand Increment at Mean Wind Speed: 12 m/s
-3+j11.6 1.84 0.2504 electrical, δ2,3 0.39 stable
-0.68 0 1 electrical, ζdc2 0.68 unstable (Hopf-
bifurcation)
-0.68 0 1 electrical, ζdc1,4 0.63 unstable (Hopf-
bifurcation)
-0.0157 0 1 electrical, wr1,4 0.5 stable

Chapter 4
Chapter 4 203
10
200
5
imag
imag
0 0
−5
−200
−10
−0.8 −0.6 −0.4 −3 −2 −1

real real
(a) (b)
2
2
1 1
imag
imag
0 0
−1 −1
−2
−2
−5 0 5 −2 −1 0 1 2
real real
(c) (d)

during demand increment at mean wind speed 12 m/s:
(a) Magnetization current mode of DFIGs imdq1,2,3,4 . (b)
Inverters angle of rotation modes δ2,3 . (c) Inverters angle
of rotation mode δ2 . (d) DC-link mode ζdc3,4 .
204 Chapter 4
1
imag
−1
−2
−4 −3 −2 −1 0 1 2 3 4
real
(a)
1
0.5
imag
−0.5
−1
−0.0175 −0.017 −0.0165 −0.016
real
(b)
Figure 4.30: The rest of critical eigenvalues trajectories of

the isolated microgrid during demand increment at mean
wind speed 12 m/s: (a) DC-link mode ζdc1,4 . (b) DFIG
mechanical speed mode ωr1,4 .
Chapter 4 205
Load Demand Increment at Mean Wind Speed 12 m/s, Grid-Connected Mode
Table 4.10 shows the dominant modes, participation factors and their
damping ratios, in the case of load demand increment from 7.5 p.u to 12
p.u for the grid-connected mode. The simulation is processed at mean wind
speed equals to 12 m/s. In Table 4.10, the rst three columns belong to the
status of the dominant state variables at the rst simulation sample, while the
last column of remarks belong to the status of the dominant state variables at
the last simulation sample.
Figures 4.31a and 4.31b show the trajectory of the DFIGs magnetization
current modes; Figure 4.31c shows the modes related to the DFIGs integral
terms of phased-locked loop circuit; Figures 4.31d and 4.32a show the trajec-
tory of the DFIGs integral terms of dc-link circuit; Figure 4.32b shows the
mode inuenced by the DFIGs mechanical speed.
The current case is totally stable and no mode is observed crossing the
imaginary axis to the unstable part creating Hopf or saddle node bifurcations.
The modes dominated by all DFIGs magnetization currents still move towards
the imaginary axis with lower frequencies. The DFIGs mechanical speed mode
still show the same response as previous cases, and it moves slightly far from
the imaginary axis. The modes dictated by dc-link variables vdc step towards
the imaginary axis as shown in Figure 4.31d, while the modes dominated by
ζdc move far from the imaginary axis as shown in Figure 4.32a, but still the
both modes are stable.
Figure 4.33 shows a comparison between the isolated microgrid mode and
grid-connected mode. It is observed that in the isolated microgrid case, the
load demand increment leads to a frequency collapse, while in the grid-connect-
ed mode, the frequency is stable because a sucient reserve coming from the
innite bus.
206
Table 4.10: Dominant Modes of the Grid-connected Mode During Demand Increment at Mean Wind Speed:
12 m/s
-26.2 0 1 electrical, ζpll1,2,3,4 0.28 stable
-14.7 0 1 electrical, vdc1,2,3,4 0.256 stable
-0.87 0 1 electrical, ζdc1,2,3,4 0.26 stable
-0.015 0 1 electrical, wr1,2,3,4 0.25 stable

Chapter 4
Chapter 4 207
200 200
imag
imag
0 0
−200 −200
−2 −1.5 −1 −0.5 0 −0.32 −0.3 −0.28

real real
(a) −5
x 10 (b)
5
0.1
0
imag
imag
0
−5
−0.1
−10
−28.5 −28 −27.5 −27 −26.5 −15 −10 −5
real real
(c) (d)

mode during demand increment at mean wind speed 12
m/s: (a) Magnetization current mode of DFIGs imdq1,2,3,4 .
(b) DFIGs magnetization current mode of imdq1,4 . (c)
DFIGs integral terms of phased-locked loop mode ζpll1,2,3,4 .
(d) DC-link mode vdc1,2,3,4 .
208 Chapter 4
−3
x 10
1
0.5
0
imag
−0.5
−1
−1.5
−1.7 −1.6 −1.5 −1.4 −1.3 −1.2 −1.1 −1 −0.9

real
(a)
0.3
0.2
0.1
imag
0
−0.1
−0.2
−0.0163−0.0162−0.0161−0.016−0.0159−0.0158−0.0157−0.0156−0.0155−0.0154
real
(b)

of the grid-connected mode during demand increment at
mean wind speed 12 m/s: (a) DC-link mode ζdc1,2,3,4 . (b)
DFIG mechanical speed mode ωr1,2,3,4 .
1.05
0.95
0.9
grid frequency (p.u)
0.85
0.8
0.75
0.7
0.65
isolated microgrid
microgrid connected to Inf bus
0.6
7.5 8.4 9.2 9.8 10.5 11 11.5 12
demand power (p.u)
Figure 4.33: Frequency vs demand increment

Chapter 4 209
The conclusions of all the load demand increment scenarios are briefed
hereunder:
• For the both cases of isolated microgrid and grid-connected mode, and
during the low wind speed case (7 m/s), the source of instability is the
mode inuenced by the DFIGs dc-link state variables, which strikes the
imaginary axis and produces a Hopf bifurcation;
• For the isolated microgrid and during the high wind speed case (12 m/s),
two sources of instability can be considered which are the DFIGs dc-link
and the inverters angle of rotation modes, as both of them cross the
imaginary axis motivating a Hopf bifurcations;
• The best response is shown in the grid-connected mode, and during the
high wind speed case (12 m/s), as the system is totally stable, and the
frequency is sti and not aected by load demand increment.

210 Chapter 4
Chapter 5 211
Chapter 5
Conclusions, Published Articles

and Future Work
5.1 Conclusions
In this thesis, several technical and economic analyses were conducted, re-
garding microgrids, such as providing an insight about site matching suitability
for droop-regulated microgrids supplied by wind generators, optimizing both
energy production and economic performance for isolated microgrids supplied
by wind energy and batteries and nally predicting some stability constraints
for generation sources by small signal stability analysis. The conclusions of
this thesis are briefed hereunder:
• The applied LHS sampling technique serves to sample the wind speed,
by ensuring a better spread of sample points through the overall wind
speed probability domain in comparison with conventional MCS tech-
niques. Therefore, more precise estimation of wind turbines states and
capacity factor can be ensured. Moreover, the conducted correlation
strategy based on Cholesky decomposition technique ensures an ecient
correlation among dierent wind turbines, without intensive computa-
tions;
• Thesis provided that the wind turbines aggregation modelling in micro-
grids is not accurate; the estimation of injected power into a microgrid
cannot be measured by multiplying directly the injected power of one

212 Chapter 5
generator by number of generators of the grid. Regardless losses in each
branch among the microgrid nodes, the correlations among wind turbine
generators play an important role in such discrepancies. Moreover, if the
dynamic response is to be considered, the small signal stability analysis
of this thesis show that varying the wind speed input for only one wind
generator has a stability eect on other microgrid wind generators;
• Thesis provided that the maximum loadability of a droop-regulated mi-
crogrid is restricted, and so the total maximum power of the microgrid
can be less than the sum of power injections by each generation unit,
due to existence of DGs interfaced by electronic inverters, which oper-
ate by the droop control. The more restriction on the maximum power
constraint of each inverter, the earlier bifurcation will be, because the
voltage is continuously degraded down to a point in which a saddle-node
bifurcation occurs;
• The wind turbine cut-in to rated speed (wi /wr ) ratio is a vital element to
enhance the annual energy production of the wind turbine. The higher
(wi /wr ) ratio, the higher annual energy production and minimum cost of
energy can be ensured. Moreover, high wind turbines (wi /wr ) ratios do
not provide a better economic performance and higher energy production
for wind turbines only, but for storage systems which operate in parallel
with wind turbines as well, if both generation units are employed in a
multi-objective function, in order to minimize the cost of energy;
• For a microgrid supplied by wind energy and batteries, there is a pos-
sibility of employing high capital cost-high ecient batteries to provide
optimum economic performance, if the problem is employed in a multi-
objective function, including both wind generation and battery units;
• It could be concluded that in order to provide optimum economic per-
formance and high energy production, the aggregation of wind turbines
capacities can be a better choice than dispersing the same turbines ca-
pacities at dierent microgrid nodes;
• Several small signal stability analyses were conducted in this thesis, in
order to study the dominant variables which can violate the stability of
microgrids. These dominant modes are related mainly to DFIGs state

Chapter 5 213
variables, such as mechanical speed, dc-link and magnetization currents
state variables. Moreover, other modes related to DGs interfaced by
inverters, which are angle of rotation state variables;
• The experienced bifurcations through the dierent scenarios of the small
signal stability analysis are related to Hopf bifurcations. Thus, to elim-
inate or damp those variables, which violate the stability of microgrids,
an appropriate control strategy must be developed to keep microgrids
stable.
214 Chapter 5
5.2 Published Articles

Author: Guzmán Díaz; Ahmed M. Abd-el-Motaleb; Verónica Mier.
Title: On the Capacity Factor of Distributed Wind Generation in Droop-
Regulated Microgrids
Journal: IEEE Transactions on Power Systems, vol.PP, pp.1 September 2012

DOI: 10.1109/TPWRS.2012.2222941
Abstract: This paper presents the methodological basis for the computa-
tion of the capacity factor (CF) of the power delivered by an o-grid droop-
regulated microgrid when the power is injected by distributed stochastic gen-
eration. More particularly, the paper is focused on wind generation and its
modeling particularities. It shows how the availability of power can be e-
ciently sampled by a low variation Latin Hypercube Sampling (LHS) method,
supplemented with a restricted pairing technique to account for correlation
among generating units. The paper also presents a minimization problem that,
combined with a Fischer-Burmeister-based formulation, permits obtaining the
maximum deliverable power at each sample. This paper additionally shows a
rst numerical analysis of a 14-node microgrid supplied by ve Enercon E40.
The analysis includes an investigation into the best model that represents the
CF of the microgrid, and highlights the eects of correlation and maximum
loadability.
Chapter 5 215
5.3 Future Work

In Chapter 4, several small signal stability analyses were conducted, in order
to determine the dominant variables which can violate the stability of micro-
grids. The critical state variables were determined, and the proposed future
work will be dedicated to develop power ow models with stable constraints
for isolated microgrids. The proposed future work will extend to the following
items:
• Possible solutions may include reconstruction of demand and generation
models, to ensure the dispatching and loading within the small signal
stability constraints. Probabilistic stability analyses must be taken into
consideration, which can ensure the robustness of the solution for various
system operation modes. In order to cover this point, more uncertainties
and rene of probabilistic models of the demand and generation with
electricity market considerations must be conducted;
• Participation factor analyses of the dominant modes, which were per-
formed in Chapter 4 can be used to select the most eective input signal
to enhance the damping of relevant oscillation modes. Thereafter, adding
damping controllers for the dominant state variables is necessary. The
dominant state variables of Chapter 4 are related to the DFIGs dc-link
and inverters angle of rotation state variables.
Regarding the DFIGs, the impact of damping controllers on such domi-
nant modes must be studied. Thus, an ecient tuning method must be
adopted to enhance the damping of the DFIGs dominant state variables.
In [32] Mishra showed a coordinated tuning of a damping controller for a
DFIG, using bacterial foraging technique. In [34] Yang proposed a multi-
objective controller design for a DFIG, including the damping ratios and
real parts of the dominant eigenvalues. Thus, restricting the damping
ratios of the dominant modes within pre-dened values and sucient
stability margins could be ensured.
Regarding the generators interfaced by inverters, an ecient method

216 Chapter 5
must be applied to tune the inverter droops according to stability con-
straints. In [23] Barklund proposed a tuning algorithm for droop-regulat-
ed generators by inverters, which takes into consideration the power de-
mand level and economic dispatch of the interfaced generators. In [143]
Guerrero introduced a method to locally tune droop-regulated batteries
by inverters, taking into consideration the batteries charge levels and an
equal harmonic sharing among the inverters;
• It is also possible to add additional controllers into the microgrid, such
as FACTS devices for security enhancement. Chapter 4 proved that
the disturbances caused mainly by a power demand increment provide
catastrophic results on DFIGs especially, as the dc-link tension could
reach terrible values that could damage the dc-link circuit. In addition,
the power quality could be aected due to voltage-frequency sags, and so
grid code requirements cannot be ensured. Therefore, in order to protect
DFIGs from isolation, the implementation of FACTS devices as grid code
requirements restorer can be adopted.
In [144] and [145] Kenneth and Qiao showed that proper coordinated
reactive power control strategies between the wind farm and the FACTS
can be adopted to reduce the level of sags and comply with grid code
requirements.
Chapter 5 217
Chapter 5
Conclusiones, Artículos Publicados

y Trabajos Futuros
5.1 Conclusiones
En esta tesis se han realizado diversos estudios técnicos y económicos sobre
microredes y sobre la elección del emplazamiento de las mismas, alimentadas
por generadores eólicos y reguladas mediante droop. Buscando la optimización
tanto de la producción de energía como del rendimiento económico y por úl-
timo, la predicción de algunas restricciones de estabilidad para las fuentes de
generación, mediante el análisis de estabilidad de pequeña señal. Las conclu-
siones de la tesis se resumen a continuación:
• La técnica de muestreo aplicada, LHS, sirve para muestrear la velocidad
del viento, al garantizar una mejor difusión de los puntos de la muestra a
través del dominio global de probabilidad de la velocidad del viento, en
comparación con las técnicas convencionales MCS. Asimismo, la estrate-
gia de correlación llevaba a cabo, basada en la técnica de descomposición
de Cholesky, garantiza una correlación ecaz entre las diversas turbinas
eólicas, evitando cálculos laboriosos.
• La tesis determinó que los modelos de agregación de turbinas eólicas en
microredes no son correctos. La estimación de la potencia inyectada en
una microred no puede calcularse multiplicando directamente la potencia
inyectada de un generador por el número de generadores de la red.
• La tesis concluyó que la capacidad de carga máxima de una microred

218 Chapter 5
regulada mediante droop es limitada y por lo tanto, la potencia máx-
ima total de la microred puede ser inferior a la suma de inyecciones de
potencia de cada unidad de generación, debido a la existencia de DG
conectadas mediante convertidores electrónicos que operan a través del
control droop.
• El índice de conexión-desconexión (wi /wr ) de la turbina eólica es un ele-
mento de suma importancia para mejorar la producción anual de energía
de la turbina eólica. Cuanto mayor sea el índice (wi /wr ), mayor será
la producción anual de energía y menor el coste de energía. Además,
los altos índices (wi /wr ) de turbinas eólicas no sólo proporcionan un
mejor rendimiento económico y una mayor producción energética para
las turbinas eólicas, sino también para los sistemas de almacenamiento
que operan en paralelo con turbinas eólicas, si ambas unidades de gen-
eración se emplean en una función multiobjetivo para minimizar el coste
de energía.
• Podría concluirse que para proporcionar un rendimiento económico óp-
timo y una alta producción energética, la agregación de capacidades de
turbinas eólicas puede resultar mejor que la dispersión de las mismas
capacidades de las turbinas en diferentes nodos de la microred.
• En la tesis se han llevado a cabo diversos análisis de estabilidad de pe-
queña señal para estudiar las variables dominantes que pueden amenazar
la estabilidad de las microredes. Estos modos dominantes se relacionan
fundamentalmente con variables de estado de los DFIG, como las vari-
ables de estado de velocidad mecánica, de enlace de continua y de cor-
rientes de magnetización. Asimismo, otros modos relacionados con las
DG se conectaron mediante convertidores, que constituyen ángulos de
rotación de las variables de estado.
• Las bifurcaciones experimentadas a través de diferentes escenarios de
análisis de estabilidad de pequeña señal están relacionadas con las bi-
furcaciones de Hopf. Por consiguiente, para eliminar o amortiguar esas
variables que amenazan la estabilidad de las microredes, debe desarrol-
larse una estrategia adecuada de control para mantener la estabilidad de
las microredes.
Chapter 5 219
5.2 Artículos Publicados

Author: Guzmán Díaz; Ahmed M. Abd-el-Motaleb; Verónica Mier.
Title: On the Capacity Factor of Distributed Wind Generation in Droop-
Regulated Microgrids
Journal: IEEE Transactions on Power Systems, vol.PP, pp.1 September 2012

DOI: 10.1109/TPWRS.2012.2222941
Abstract: This paper presents the methodological basis for the computa-
tion of the capacity factor (CF) of the power delivered by an o-grid droop-
regulated microgrid when the power is injected by distributed stochastic gen-
eration. More particularly, the paper is focused on wind generation and its
modeling particularities. It shows how the availability of power can be e-
ciently sampled by a low variation Latin Hypercube Sampling (LHS) method,
supplemented with a restricted pairing technique to account for correlation
among generating units. The paper also presents a minimization problem that,
combined with a Fischer-Burmeister-based formulation, permits obtaining the
maximum deliverable power at each sample. This paper additionally shows a
rst numerical analysis of a 14-node microgrid supplied by ve Enercon E40.
The analysis includes an investigation into the best model that represents the
CF of the microgrid, and highlights the eects of correlation and maximum
loadability.
220 Chapter 5
5.3 Trabajos Futuros

En el Capítulo 4, se han llevado a cabo varios análisis de estabilidad de señal
pequeña para determinar las variables principales que pueden amenazar la
estabilidad de las microredes. Se determinaron las variables de estado críticas
y el trabajo futuro propuesto se centrará en desarrollar modelos de ujo de
potencia con restricciones estables para las microredes aisladas. Los trabajos
futuros propuestos abarcarán los siguientes aspectos:
• Soluciones posibles, entre las que se pueden incluir la reconstrucción de
modelos de demanda y generación para garantizar los límites estables
del ujo de potencia. Los análisis de estabilidad probabilísticos deben
tenerse en cuenta ya que pueden garantizar la robustez de la solución
para varios modos de operación de los sistemas.
• Los análisis sobre el factor de participación de los modos dominantes que
se presentaron en el Capítulo 4, pueden utilizarse para seleccionar la señal
de entrada más ecaz con el n de mejorar el amortiguamiento de los
modos de oscilación pertinentes. Por lo tanto, es necesario añadir contro-
ladores de amortiguamiento para las variables de estado dominantes. Las
variables de estado dominantes del Capítulo 4 se relacionan con el enlace
de corriente continua de los DFIG. Además, los ángulos de rotación de
los generadores operan a través de convertidores.
• Por último, también es posible añadir controladores adicionales en la
microred, como los dispositivos FACTS, para la mejora de la seguridad.
El Capítulo 4 demostró que las perturbaciones causadas principalmente
por un incremento en la demanda de potencia, ocasionaron resultados
devastadores especialmente en los DFIG, ya que la tensión de enlace de
corriente continua podría alcanzar valores nefastos que podrían dañar el
circuito de enlace de continua. Asimismo, la calidad de la potencia podría
verse afectada por los huecos de tensión-frecuencia y por lo tanto, no
pueden garantizarse los requisitos de código de la red. Para proteger a los
DFIG del aislamiento, puede adoptarse la implementación de dispositivos
FACTS, como el restaurador de requisitos de código de red.

Bibliography 221
Bibliography
[1] G. Diaz, A. M. Abd-el Motaleb, and V. Mier. On the capacity factor
of distributed wind generation in droop-regulated microgrids. Power

Systems, IEEE Transactions on, PP(99):1, 2012.
[2] P. Kundur. Power system stability and control. Epri Power System
Engineering Series. McGraw-Hill, 1994.
[3] Paul M. Anderson and A.A. Fouad. Power system control and stability.
IEEE Press power engineering series. IEEE Press, 2003.
[4] C. Marnay, H. Asano, S. Papathanassiou, and G. Strbac. Policy making
for microgrids. Power and Energy Magazine, IEEE, 6(3):66 77, may-
june 2008.
[5] R.H Lasseter. Smart distribution: coupled microgrids. Proceedings of

the IEEE, 99(6):10741082, june 2011.
[6] B. Kroposki, R. Lasseter, T. Ise, S. Morozumi, S. Papatlianassiou, and
N. Hatziargyriou. Making microgrids work. Power and Energy Magazine,

IEEE, 6(3):40 53, may-june 2008.
[7] Ronald L. Iman and W. J. Conover. A distribution-free approach to
inducing rank correlation among input variables. Communications in

Statistics - Simulation and Computation, 11(3):311334, 1982.
[8] A. Chuang and M. McGranaghan. Master controller requirements speci-
cation for perfect power systems. Technical report, EPRI for The Galvin
Electricity Initiative, 2007.
[9] P. Piagi and R.H. Lasseter. Autonomous control of microgrids. In Power

Engineering Society General Meeting, 2006. IEEE, page 8 pp., 0-0 2006.
222 Bibliography
[10] H. Nikkhajoei and R.H. Lasseter. Distributed generation interface to
the certs microgrid. Power Delivery, IEEE Transactions on, 24(3):1598
1608, july 2009.
[11] R.H. Lasseter. Extended certs microgrid. In Power and Energy Society
General Meeting - Conversion and Delivery of Electrical Energy in the
21st Century, 2008 IEEE, pages 1 5, july 2008.
[12] R.H. Lasseter, J.H. Eto, B. Schenkman, J. Stevens, H. Vollkommer,
D. Klapp, E. Linton, H. Hurtado, and J. Roy. Certs microgrid labo-
ratory test bed. Power Delivery, IEEE Transactions on, 26(1):325 332,
jan. 2011.
[13] A. Losi and M. Russo. Dispersed generation modeling for object-
oriented distribution load ow. Power Delivery, IEEE Transactions on,

20(2):1532 1540, april 2005.
[14] Hassan Nikkhajoei and Reza Iravani. Steady-state model and power
ow analysis of electronically-coupled distributed resource units. Power

Delivery, IEEE Transactions on, 22(1):721 728, jan. 2007.
[15] M.Z. Kamh and R. Iravani. Unbalanced model and power-ow analysis
of microgrids and active distribution systems. Power Delivery, IEEE

Transactions on, 25(4):2851 2858, oct. 2010.
[16] K.L. Lo and C. Zhang. Decomposed three-phase power ow solution
using the sequence component frame. Generation, Transmission and

Distribution, IEE Proceedings C, 140(3):181 188, may 1993.
[17] J. Morren, S.W.H. de Haan, W.L. Kling, and J.A. Ferreira. Wind tur-
bines emulating inertia and supporting primary frequency control. Power

Systems, IEEE Transactions on, 21(1):433 434, feb. 2006.
[18] R.G. de Almeida and J.A.P. Lopes. Participation of doubly fed induction
wind generators in system frequency regulation. Power Systems, IEEE

Transactions on, 22(3):944 950, aug. 2007.
[19] O. Anaya-Lara, F.M. Hughes, N. Jenkins, and G. Strbac. Contribution
of dg-based wind farms to power system short-term frequency regu-

Bibliography 223
lation. Generation, Transmission and Distribution, IEE Proceedings-,

153(2):164 170, march 2006.
[20] J.A.P. Lopes, C.L. Moreira, and A.G. Madureira. Dening control strate-
gies for microgrids islanded operation. Power Systems, IEEE Transac-

tions on, 21(2):916 924, may 2006.
[21] M. Prodanovic and T.C. Green. High-quality power generation through
distributed control of a power park microgrid. Industrial Electronics,

IEEE Transactions on, 53(5):1471 1482, oct. 2006.
[22] C. L. Moreira, F. O. Resende, and J. A. P. Lopes. Using low voltage
microgrids for service restoration. Power Systems, IEEE Transactions

on, 22(1):395 403, feb. 2007.
[23] E. Barklund, N. Pogaku, M. Prodanovic, C. Hernandez-Aramburo,
and T.C. Green. Energy management in autonomous microgrid using
stability-constrained droop control of inverters. Power Electronics, IEEE

Transactions on, 23(5):2346 2352, sept. 2008.
[24] Seon-Ju Ahn, Jin-Woo Park, Il-Yop Chung, Seung-Il Moon, Sang-Hee
Kang, and Soon-Ryul Nam. Power-sharing method of multiple dis-
tributed generators considering control modes and congurations of a
microgrid. Power Delivery, IEEE Transactions on, 25(3):2007 2016,
july 2010.
[25] E. Serban and H. Serban. A control strategy for a distributed power gen-
eration microgrid application with voltage- and current-controlled source
converter. Power Electronics, IEEE Transactions on, 25(12):2981 2992,

dec. 2010.
[26] A. Molderink, V. Bakker, M.G.C. Bosman, J.L. Hurink, and G.J.M.
Smit. Management and control of domestic smart grid technology. Smart

Grid, IEEE Transactions on, 1(2):109 119, sept. 2010.
[27] M.G. Molina and P.E. Mercado. Power ow stabilization and control
of microgrid with wind generation by superconducting magnetic en-
ergy storage. Power Electronics, IEEE Transactions on, 26(3):910 922,

march 2011.
224 Bibliography
[28] Changhee Cho, Jin-Hong Jeon, Jong-Yul Kim, Soonman Kwon, Ky-
ongyop Park, and Sungshin Kim. Active synchronizing control of a mi-
crogrid. Power Electronics, IEEE Transactions on, 26(12):3707 3719,
dec. 2011.
[29] A. Colet-Subirachs, A. Ruiz-Alvarez, O. Gomis-Bellmunt, F. Alvarez-
Cuevas-Figuerola, and A. Sudria-Andreu. Centralized and distributed
active and reactive power control of a utility connected microgrid using
iec61850. Systems Journal, IEEE, 6(1):58 67, march 2012.
[30] Zhixin Miao, A. Domijan, and Lingling Fan. Investigation of microgrids
with both inverter interfaced and direct ac-connected distributed energy
resources. Power Delivery, IEEE Transactions on, 26(3):1634 1642,
july 2011.
[31] A. Ostadi, A. Yazdani, and R.K. Varma. Modeling and stability
analysis of a dg-based wind-power generator interfaced with a series-
compensated line. Power Delivery, IEEE Transactions on, 24(3):1504
1514, july 2009.
[32] Y. Mishra, S. Mishra, Fangxing Li, Zhao Yang Dong, and R.C. Bansal.
Small-signal stability analysis of a dg-based wind power system under
dierent modes of operation. Energy Conversion, IEEE Transactions

on, 24(4):972 982, dec. 2009.
[33] Lihui Yang, Zhao Xu, J. Ø andstergaard, Zhao Yang Dong, Kit Po Wong,
and Xikui Ma. Oscillatory stability and eigenvalue sensitivity analysis
of a dg wind turbine system. Energy Conversion, IEEE Transactions

on, 26(1):328 339, march 2011.
[34] Yang G.Y. Yang, L and Z. Xu. Optimal controller design of a doubly-
fed induction generator wind turbine system for small signal stability
enhancement. Generation, Transmission Distribution, IET, 4(5):579
597, may 2010.
[35] H. Holttinen. Hourly wind power variations in the nordic countries. Wind
Energy, 8(2):173195, 2004.
Bibliography 225
[36] D.A. Bechrakis and P.D. Sparis. Correlation of wind speed between
neighboring measuring stations. Energy Conversion, IEEE Transactions

on, 19(2):400 406, june 2004.
[37] Wijarn Wangdee. Considering load-carrying capability and wind speed
correlation of wecs in generation adequacy assessment. IEEE Transac-

tions on Energy Conversion, 21(3):734 741, 2006.
[38] Y. Gao. Adequacy assessment of generating systems containing wind
power considering wind speed correlation. IET Renewable Power Gen-

eration, 3(2):217226, 2009.
[39] Hui Li and Zhe Chen. Design optimization and evaluation of dier-
Electrical Machines and Systems, 2008.

ent wind generator systems. In
ICEMS 2008. International Conference on, pages 2396 2401, oct. 2008.
[40] S.H. Jangamshetti and V.G. Ran. Optimum siting of wind turbine gen-
erators. Energy Conversion, IEEE Transactions on, 16(1):813, 2001.
[41] Tai-Her Yeh and Li Wang. A study on generator capacity for wind
turbines under various tower heights and rated wind speeds using weibull
distribution. Energy Conversion, IEEE Transactions on, 23(2):592 602,

june 2008.
[42] We-Jen Lee Zhe Chen Li Wang, Tai-Her Yeh. Benet evaluation of
wind turbine generators in wind farms using capacity-factor analysis and
economic-cost methods. IEEE Transactions on Power Systems, 24(2),
may 2009.
[43] U. Sureshkumar, P.S. Manoharan, and A.P.S. Ramalakshmi. Economic
cost analysis of hybrid renewable energy system using homer.Ad- In
vances in Engineering, Science and Management (ICAESM), 2012 In-

ternational Conference on, pages 94 99, march 2012.
[44] N.A.b.A. Razak, M.M. bin Othman, and I. Musirin. Optimal sizing and
operational strategy of hybrid renewable energy system using homer. In
Power Engineering and Optimization Conference (PEOCO), 2010 4th

International, pages 495 501, june 2010.
226 Bibliography
[45] A.K. Srivastava, A.A. Kumar, and N.N. Schulz. Impact of distributed
generations with energy storage devices on the electric grid. Systems

Journal, IEEE, 6(1):110 117, march 2012.
[46] K.W. Wang, C.Y. Chung, C.T. Tse, and K.M. Tsang. Multimachine
eigenvalue sensitivities of power system parameters. In Power Engineer-

ing Society Winter Meeting, 2000. IEEE, volume 2, page 1434 vol.2,
2000.
[47] E.E. Souza Lima and L.F. De Jesus Fernandes. Assessing eigenvalue
sensitivities [power system control simulation]. Power Systems, IEEE

Transactions on, 15(1):299 306, feb 2000.
[48] N. Pogaku, M. Prodanovic, and T.C. Green. Modeling, analysis and
testing of autonomous operation of an inverter-based microgrid. Power

Electronics, IEEE Transactions on, 22(2):613 625, march 2007.
[49] A.P. Leite, C.L.T. Borges, and D.M. Falcao. Probabilistic wind farms
generation model for reliability studies applied to brazilian sites. Power

Systems, IEEE Transactions on, 21(4):1493 1501, nov. 2006.
[50] R. Billinton, Yi Gao, and R. Karki. Composite system adequacy assess-
ment incorporating large-scale wind energy conversion systems consid-
ering wind speed correlation. Power Systems, IEEE Transactions on,

24(3):1375 1382, aug. 2009.
[51] F. Vallee, G. Brunieau, M. Pirlot, O. Deblecker, and J. Lobry. Optimal
wind clustering methodology for adequacy evaluation in system genera-
tion studies using nonsequential monte carlo simulation. Power Systems,

IEEE Transactions on, 26(4):2173 2184, nov. 2011.
[52] F. Katiraei and M.R. Iravani. Power management strategies for a micro-
grid with multiple distributed generation units. Power Systems, IEEE

Transactions on, 21(4):1821 1831, nov. 2006.
[53] R. Majumder, A. Ghosh, G. Ledwich, and F. Zare. Power manage-
ment and power ow control with back-to-back converters in a utility
connected microgrid. Power Systems, IEEE Transactions on, 25(2):821
834, may 2010.

Bibliography 227
[54] R. Billinton. Power System Reliability Evaluation. Gordon and Breach,
Science Publ., 1970.
[55] R.Y. Redlinger and P.D Andersen. Wind energy in the 21st century: eco-
nomics, policy, technology and the changing electricity industry. Palgrave
Macmillan, 2002.
[56] C. Abbey, F. Katiraei, C. Brothers, L. Dignard-Bailey, and G. Joos.
Integration of distributed generation and wind energy in canada. In
Power Engineering Society General Meeting, 2006. IEEE, page 7 pp.,
0-0 2006.
[57] B. Fox. Wind power integration: connection and system operational

aspects. IET power and energy series. The Institution of Engineering
and Technology, 2007.
[58] M. P. Connor. The UK Renewable Obligation, chapter 7 of book: Switch-
ing of Renewable Power. 2005.
[59] J. Moccia and A. Arapogianni. Pure power. Technical report, European
Wind Energy Association, 2011.
[60] I Abouzahr and R Ramakumar. An approach to assess the performance
of utility-interactive wind electric conversion systems. Energy Conver-

sion, IEEE Transactions on, 6(4):627 638, Dec 1991.
[61] P. Giorsetto and K.F. Utsurogi. Development of a new procedure for
reliability modeling of wind turbine generators. Power Apparatus and

Systems, IEEE Transactions on, PAS-102(1):134143, 1983.
[62] P. Hu R. Karki and R. Billinton. Reliability evaluation of a wind power
delivery system using an approximate wind model. In41st International

Universities Power Engineering Conference, Newcastle, UK, 2006.
[63] R. Karki and Po Hu. Wind power simulation model for reliability eval-
uation. In Electrical and Computer Engineering, 2005. Canadian Con-

ference on, pages 541544, 2005.
[64] Iman R. L. Steck, G. P. and Dahlgren. Probablistic analysis of lOCA , an-

nual report. SAND76-0535, Sandia National Laboratories, Albuquerque,
NM, 1976.
228 Bibliography
[65] Ali Naci Celik. A statistical analysis of wind power density based on the
weibull and rayleigh models at the southern region of turkey. Renewable

Energy, 29(4):593604, July 2003.
[66] P. Ramírezb J.A. Cartaa and S. Velázquezc. A review of wind speed prob-
ability distributions used in wind energy analysis: Case studies in the ca-
nary islands. Renewable and Sustainable Energy Reviews, 13(5):933955,

June 2009.
[67] B. Stephen S. Gill and S. Galloway. Wind turbine condition assessment
through power curve copula modeling. Sustainable Energy, IEEE Trans-

actions on, 3(1):94 101, jan. 2012.
[68] O.A. Jaramillo and M.A. Borja. Wind speed analysis in la ventosa,
mexico: a bimodal probability distribution case. Renewable Energy,

29(10):16131630, August 2004.
[69] M. Cabello andJ.A.G. Orza. Wind speed analysis in the province of ali-
cante, spain. potential for small-scale wind turbines. Renew Sust Energ,
14(9):31853191, 2010.
[70] R. J. Beckman M. D. Mckay and W. J. Conover. A comparison of three
methods for selecting values of input variables in the analysis of output
from a computer code. Technometrics, 42(1):5561, 2000.
[71] R.L. Iman and M.J. Shortencarier. A FORTRAN 77 program and user's
guide for the generation of latin hypercube and random samples for use
with computer models. NUREG/CR. The Division, 1984.
[72] A. E. Feijo. Wind speed simulation in wind farms for steady-state se-
curity assessment of electrical power systems. IEEE Transactions on

Energy Conversion, 14(4):1582 1588, 1999.
[73] Roy Billinton. Composite system adequacy assessment incorporating
large-scale wind energy conversion systems considering wind speed cor-
relation. IEEE Transactions on Power Systems, 24(3):1375 1382, 2009.
[74] James R. Salmon and John L. Walmsley. A two-site correlation model for
wind speed, direction and energy estimates. Journal of Wind Engineering

and Industrial Aerodynamics, 79(3):233268, 1999.
Bibliography 229
[75] M.C. Alexiadis, P.S. Dokopoulos, and H.S. Sahsamanoglou. Wind speed
and power forecasting based on spatial correlation models. Energy Con-

version, IEEE Transactions on, 14(3):836 842, sep 1999.
[76] Michael Stein. Large sample properties of simulations using latin hyper-
cube sampling. Technometrics, 29(2):143151, 1987.
[77] M.H Albadi and E. F. El-Saadany. Wind turbines capacity factor model-
inga novel approach. Power Systems, IEEE Transactions, 24(3):1637

1638, Aug 2009.
[78] Cristina L. Archer and Mark Z. Jacobson. Supplying base load power
and reducing transmission requirements by interconnecting wind farms.
Applied Meteorology and Climatology, 46, november 2007.
[79] S. Bazaraa, H.D. Sherali, and C.M. Shetty. Nonlinear programming:

theory and algorithms. Wiley-Interscience, 2006.
[80] G. Claeskens and N.L. Hjort. Model selection and model averaging. Cam-
bridge Series on Statistical and Probabilistic Mathematics. Cambridge
University Press, 2008.
[81] Hamparsum Bozdogan. Akaike's information criterion and recent devel-
opments in information complexity. Mathematical Psychology, 44:6291,

2000.
[82] H Akaike. A new look at the statistical model identication. Automatic

Control, IEEE Transactions, 19(6)(6):716723, Dec 1974.
[83] Gideon E Schwarz. Estimating the dimension of a model. Annals of

Statistics, 6(2):461464, 1978.
[84] R. Tibsharani Hastie, T and J. Friedman. The elements of statistical

learning. Springer, New York, 2001.
[85] Markon KE and Krueger RF. An empirical comparison of information-
theoretic selection criteria for multivariate behavior genetic models. Be-

haviour Genetics, 34(6):593610, 2004.
230 Bibliography
[86] R. Billinton and A.A. Chowdhury. Incorporation of wind energy con-
version systems in conventional generating capacity adequacy assess-
ment. Generation, Transmission and Distribution, IEE Proceedings C,

139(1):4756, 1992.
[87] A.J.M. van Wijk and N. Halberg. Capacity credit of wind power in the
netherlands. Electric Power Systems Research, 23(3):189200, May 1992.
[88] R.M.G. Castro and L.A.F.M. Ferreira. A comparison between chronolog-
ical and probabilistic methods to estimate wind power capacity credit.
Power Systems, IEEE Transactions on, 16(4):904909, 2001.
[89] M Milligan and K. Porter. Determining the capacity value of wind:
an updated survey of methods and implementation. Technical report,
National Renewable Energy Laboratory, 2008.
[90] Aynur Ucar and Figen Balo. Evaluation of wind energy potential
and electricity generation at six locations in turkey. Applied energy,

86(10):18641872, October 2009.
[91] Z.M. Salameh and I. Safari. The eect of the windmill's parameters
on the capacity factor. Energy Conversion, IEEE Transactions on,

10(4):747751, 1995.
[92] K.R. Padiyar. Power System Dynamics: Stability and Control. Anshan
Pub, 2004.
[93] Jan Machowski, J.W. Bialek, and J.R. Bumby. Power system dynamics
and stability. John Wiley, 1997.
[94] G. Diaz, C. Gonzalez-Moran, J. Gomez-Aleixandre, and A. Diez. Com-
posite loads in stand-alone inverter based microgrids-modeling proce-
dure and eects on load margin. Power Systems, IEEE Transactions on,
25(2):894905, 2010.
[95] A.P. Agalgaonkar, C.V. Dobariya, M.G. Kanabar, S.A. Khaparde, and
S.V. Kulkarni. Optimal sizing of distributed generators in microgrid. In
Power India Conference, 2006 IEEE, page 8 pp., 0-0 2006.

Bibliography 231
[96] R.W. Wies, R.A. Johnson, A.N. Agrawal, and T.J. Chubb. Simulink
model for economic analysis and environmental impacts of a pv with
diesel-battery system for remote villages. Power Systems, IEEE Trans-

actions on, 20(2):692 700, may 2005.
[97] William Isherwooda and J.Ray Smith. Remote power systems
with advanced storage technologies for alaskan villages. Energy,

25(10):10051020, 2000.
[98] DF Ancona and P.R. Goldman. Wind program technological develop-
ments in the united states. Renew Energy, 10:2538, 1997.
[99] Mark Bolinger Lori Bird and Troy Gagliano. Policies and market fac-
tors driving wind power development in the united states. Energy Pol,
33(1):397407, 2005.
[100] Jager-Waldau and Heinz Ossenbrink. Progress of electricity from
biomass, wind and photovoltaics in the european union. Renew Sustain

Energy, 8(1):5782, 2004.
[101] Smith JC and DeMeo EA. Wind power impacts on electric power system
operating costs: summary and perspective on work to date. Presented

at the American wind energy association global wind power conference,
2004.
Power system restructuring and deregulation: trading, perfor-

[102] L.L. Lai.
mance and information technology. Wiley, 2001.
[103] Sally Hunt and Graham Shuttleworth. Competition and choice in elec-
tricity. John Wiley, may 1996.
[104] A. Kusiak and Z. Song. Design of wind farm layout for maximum wind
energy capture. Renewable Energy, 35:658694, 2010.
[105] G. Diaz and C. Gonzalez-Moran. Fischer-burmeister-based method for
calculating equilibrium points of droop-regulated microgrids. Power Sys-

tems, IEEE Transactions on, 27(2):959967, 2012.
[106] S.X. Chen, H.B. Gooi, and M.Q. Wang. Sizing of energy storage for
microgrids. Smart Grid, IEEE Transactions on, 3(1):142151, 2012.

232 Bibliography
[107] L. M. Hand Fingersh and A. Laxson. Wind turbine design cost and
scaling model. Technical report, National Renewable Energy Laboratory,
june 2006.
[108] F.A. Chacra, P. Bastard, G. Fleury, and R. Clavreul. Impact of energy
storage costs on economical performance in a distribution substation.
Power Systems, IEEE Transactions on, 20(2):684 691, may 2005.
[109] M. Kintner-Meyer and Chunlian Jin. Energy storage for variable re-
newable energy resource integration-a regional assessment for the north-
west power pool. In Power Systems Conference and Exposition (PSCE),

IEEE/PES, pages 1 7, march 2011.
[110] Susan Schoeenung and Jim Eyer. Benet and cost comparisons of energy
storage technologies for three engineering value propositions. In presented

at EESAT, San Francisco, California, 2005.
[111] D.G. Newnan, J.P. Lavelle, and T.G. Eschenbach. Essentials of engi-
neering economic analysis. Oxford University Press, 2002.
[112] M. Preindl and S. Bolognani. Optimization of the generator to rotor
ratio of mw wind turbines based on the cost of energy with focus on
low wind speeds. In IECON 2011 - 37th Annual Conference on IEEE

Industrial Electronics Society, pages 906 911, nov. 2011.
[113] Samir Succar and David C. Denkenbergerc. Optimization of specic
rating for wind turbine arrays coupled to compressed air energy storage.
Applied energy, 96:222234, December 2011.
[114] Patrick Mazza Christopher Schaber and Roel Hammerschlag. Utility-
scale storage of renewable energy. Electricity Journal, 17(6):2129, 2004.
[115] G. Desrochers, M. Blanchard, and S. Sud. A monte-carlo simulation
method for the economic assessment of the contribution of wind energy to
power systems. Energy Conversion, IEEE Transactions on, EC-1(4):50

56, 1986.
[116] Parker CD. Lead-acid battery energy-storage systems for electricity sup-
ply networks. Power Sources, 100:1828, 2001.

Bibliography 233
[117] Y. Moriokaa Y and S. Narukawab. State-of-the-art of alkaline recharge-
able batteries. Power Sources, 100(1):1071161, 2001.
[118] C. Ponce de Leon and A. Frias-Ferrer. Redox ow cells for energy con-
version. Power Sources, 160:716732, 2006.
[119] M. Wakihara. Recent developments in lithium ion batteries. Materials

Science and Engineering, 33:109134, 2001.
[120] Anthony Schellenberglabe Marc Beaudin, Hamidreza Zareipour and
William Rosehart. Energy storage for mitigating the variability of re-
newable electricity sources: an updated review. Energy for sustainable

development, 14(4):302314, September 2010.
[121] Haisheng Chen and Thang Ngoc Cong. Progress in electrical en-
ergy storage system: A critical review. Progress in Natural Science,

19(3):291312, March 2009.
[122] SusanmM Schoenung. Characteristics and technologies for long- vs.
short-term energy storage: A study by the doe energy storage systems
program. Technical report, Sandia National Labs, 2001.
[123] Cristina González Morán.Optimizacion en el espacio de estaods del

funcionamiento en modo autonomo de una microred. PhD thesis, Uni-
versidad de oviedo, 2010.
[124] E. S. Kuh and R. A. Rohrer. The state-variable approach to network
analysis. Proceedings of the IEEE, 53(7):672686, 1965.
[125] IEEE Std 399-1997. IEEE recommended practice for industrial and com-
mercial power systems analysis. 1998.
[126] E.N. Dialynas and N.D. Hatziargyriou. Eect of high wind power pen-
etration on the reliability and security of isolated power systems. In
presented at 37th session, CIGRE, Paris, august 1998.
[127] N.D. Hatziargyrious and E.S. karapidakis. Frequency stability of power
systems in large islands with high wind power penetration. In Bulk Power
Systems dynamics and Control Symposium, august 1998.
234 Bibliography
[128] A.D. Hansen. Generators and power electronics for wind turbines, chap-
ter 5 of book: Wind power in power systems, pages 5378. Wiley, 2004.
[129] Slootweg J.G. Wind power modelling and impact on power systems dy-
namics. PhD thesis, Delft University of Technology, Delft, The Nether-
lands, 2003.
[130] M.V. Nunes and A.J Lopesa. Inuence of the variable-speed wind gen-
erators in transient stability margin of the conventional generators inte-
grated in electrical grids. IEEE Trans. Energy Convers, 19(4):692701,
2004.
[131] F. Mei and B.C. Pal. Modelling and small-signal analysis of a grid con-
nected doubly-fed induction generator. In Power Engineering Society

General Meeting, 2005. IEEE, pages 2101 2108 Vol. 3, june 2005.
[132] F. Mei and B Pal. Modal analysis of grid-connected doubly fed induction
generators. Energy Conversion, IEEE Transactions on, 22(3):728 736,
sept. 2007.
[133] B.C. Pal and F. Mei. Modelling adequacy of the doubly fed induction
generator for small-signal stability studies in power systems. Renewable

Power Generation, IET, 2(3):181 190, september 2008.
[134] R. J. Thomas and A. G. Phadke. Operational characteristics of a large
wind-farm utility system with a controllable acdcac interface. IEEE

Trans. Power Syst, PS-3(1):220225, Feb 1988.
[135] M. K. Donnelly and J.E. Dagle. Impacts of the distributed utility on
transmission system stability. IEEE Trans. Power Syst, 11(2):741746,
May 1996.
[136] R C. Dugan and T. S. Key. Distributed resources standards. IEEE

Industry Applications Magazine, 12:2734, Feb 2006.
[137] F. Katiraei and M. R. Iravani. Power management strategies for a mi-
crogrid with multiple distributed generation units. IEEE Trans. Power

Syst, 21:18211831, Nov 2006.
Bibliography 235
[138] Guoqiao Shen and Xuancai Zhu. A new feedback method for pr cur-
rent control of lcl-lter-based grid-connected inverter. IEEE Trans. Ind.

Electron, 57(6):20332041, June 2010.
[139] Fei Liu and Yan Zhou. Parameter design of a two current-loop controller
used in a grid-connected inverter system with lcl lter. IEEE Trans. Ind.
Electron, 56(11):44834491, Nov 2009.
[140] Y. Mohamed and E.F. El-Saadany. Adaptive decentralized droop con-
troller to preserve power sharing stability of paralleled inverters in dis-
tributed generation microgrids. Power Electronics, IEEE Transactions

on, 23(6):2806 2816, nov. 2008.
[141] A. Anbarasan. Identication of maximum loading point in power sys-
tem under critical line outage condition. European Journal of Scientic

Research, 68(2):274282, 2012.
[142] C.F. Moyano, R. Salgado, and L.V. Barboza. Calculating participation
factors in the maximum loadability. In Power Tech Conference Proceed-

ings, 2003 IEEE Bologna, volume 2, page 6 pp. Vol.2, june 2003.
[143] J.M. Guerrero, J.C. Vasquez, J. Matas, M. Castilla, and L.G. de Vicuna.
Control strategy for exible microgrid based on parallel line-interactive
ups systems. Industrial Electronics, IEEE Transactions on, 56(3):726
736, march 2009.
[144] Wei Qiao, R.G. Harley, and G.K. Venayagamoorthy. Coordinated re-
active power control of a large wind farm and a statcom using heuris-
tic dynamic programming. Energy Conversion, IEEE Transactions on,

24(2):493 503, june 2009.
[145] Kenneth E. Okedu. Stability enhancement of dg-based variable speed
wind turbine with a crowbar by facts device as per grid requirement.
International Journal of Renewable Energy Ressearch, 2(3), 2012.
88.0pt

2013abd Eestab

Uploaded by

Copyright:

Available Formats

2013abd Eestab

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2013abd Eestab

Uploaded by

Copyright:

Available Formats

Departamento de Ingeniería Eléctrica, Electrónica, de

Technical and Economical Analysis of Isolated

Ahmed Mohamed Abd el Motaleb

guidance, patience, encouragement and support throughout the course of this

research work and in the preparation of this thesis.

the research work.

The author would like to thank all colleagues at Escuela Politécnica de

Ingeniería de Gijón for their support and help.

nancial support by the scholarship FPI-ene2010-14899.

Makram. The author presents this thesis as a gift to them.

1 State of the Art 5

1.2 Operational Concepts of Microgrids . . . . . . . . . . . . . . . . 10

1.3 Impacts of Microgrids . . . . . . . . . . . . . . . . . . . . . . . . 11

1.4 Power Flow in Microgrids . . . . . . . . . . . . . . . . . . . . . 14

1.5 The Interface Control of Microgrids . . . . . . . . . . . . . . . . 16

1.5.1 Power Electronic Interface . . . . . . . . . . . . . . . . . 16

1.5.2 Droop Control in Microgrids . . . . . . . . . . . . . . . . 19

1.5.3 Microgrids Supplied by Non-Dispatchable Generators . . 20

1.6 Capacity Factor Analysis of Droop-Regulat-ed Microgrids . . . . 26

1.7 Benet Evaluation of Microgrids Using Capacity Factor and

1.8 Microgrids Stability Analysis . . . . . . . . . . . . . . . . . . . . 32

1.9 Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . . 34

2 Capacity Factor Analysis of Droop-Regulated Microgrids 41

2.1.1 Chapter Outlines . . . . . . . . . . . . . . . . . . . . . . 46

2.2 Latin Hypercube Sampling . . . . . . . . . . . . . . . . . . . . . 49

2.2.1 Distribution Functions of Parameters and State Va-riables 49

2.2.2 Calculating the number of samples . . . . . . . . . . . . 50

2.2.3 Dividing the range of each of the parameters into equi-

2.2.4 Creating the LHS table . . . . . . . . . . . . . . . . . . . 51

2.3 Correlation among Dierent Wind Turbines . . . . . . . . . . . 53

2.3.1 Correlation Steps . . . . . . . . . . . . . . . . . . . . . . 56

2.4 Curve Fitting of the Wind Turbine Enercon-E40 . . . . . . . . . 58

2.5 Cumulative Distribution of Wind Power . . . . . . . . . . . . . 59

2.6 Microgrid Layout . . . . . . . . . . . . . . . . . . . . . . . . . . 61

2.7 Nonlinear Optimization . . . . . . . . . . . . . . . . . . . . . . . 64

2.7.1 Global Optimization Search . . . . . . . . . . . . . . . . 65

2.7.2 Applied Global Search Optimization Methodology . . . . 66

2.8 Processing Simualtions and Conclusions . . . . . . . . . . . . . . 68

2.9 Model Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

2.9.1 The likelihood measure . . . . . . . . . . . . . . . . . . . 71

2.9.2 Akaike Information Criterion . . . . . . . . . . . . . . . 72

2.9.3 Bayesian Information Criterion . . . . . . . . . . . . . . 72

2.10 Microgrid Capacity Factor . . . . . . . . . . . . . . . . . . . . . 75

2.11 Microgrid Bifurcation Analysis . . . . . . . . . . . . . . . . . . . 78

2.12 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

3 Benet Evaluation of Microgrids Using Capacity Factor and

3.2 Sources of Data . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

3.2.1 Microgrid Layout . . . . . . . . . . . . . . . . . . . . . . 87

3.2.2 Droop Regulation by Fischer-Burmeister . . . . . . . . . 88

3.2.3 Daily Operation of Batteries . . . . . . . . . . . . . . . . 90

3.2.4 Wind Turbines Characteristics . . . . . . . . . . . . . . . 92

3.2.5 Economic Evaluation Models . . . . . . . . . . . . . . . 94

3.3 Economic Evaluation for a Scaled Down Wi-nd Turbine . . . . . 98

3.3.1 The First Step of Optimization . . . . . . . . . . . . . . 102

3.3.2 The Second Step of Optimization . . . . . . . . . . . . . 103

3.3.3 Optimization Results . . . . . . . . . . . . . . . . . . . . 104

3.4 Economic Evaluation for a Microgrid Supplied by Wind Tur-

bines of Dierent Power Ratings . . . . . . . . . . . . . . . . . . 109

nancial support by the scholarship FPI-ene2010-14899.

1.7 Benet Evaluation of Microgrids Using Capacity Factor and

2.3 Correlation among Dierent Wind Turbines . . . . . . . . . . . 53

3 Benet Evaluation of Microgrids Using Capacity Factor and

bines of Dierent Power Ratings . . . . . . . . . . . . . . . . . . 109

3.6 The Aggregation Eect of Wind Turbines on the Economic Per-

1 EU electricity generation by dierent energy resources . . . . . . 3

1.4 An example of wind measurements at dierent sampling rates

in Germany, Ernst: (a) Wind speed. (b) Correlation coecients

2.4 Stratied wind speed samples by LHS . . . . . . . . . . . . . . . 51

2.10 Curve tting of Enercon-E40 600 kW, manufacturer data indi-