Central Inverter Topology Issues in Large-Scale

Photovoltaic Power Plants: Shading and System Losses

Nicole Foureaux¹, Alysson Machado², Érico Silva¹, Igor Pires³, José Brito5 and Braz Cardoso F.4
¹*Graduate Program in Elect. Eng., ²Undergrad. Program in Control Eng., ³*Dept. of Electronic Eng.,
4
*Dept. of Elect. Eng. /UFMG, Belo Horizonte/MG, Brazil and 5COELBA S.A., Salvador/BA, Brazil

Abstract — Most of the large-scale PV power plants are based presented. The author’s goal is to exhibit that large panel
on central inverters, in which the PV panels are concentrated in associations diminish the injected power into the grid
single or few MPPTs and connected to the grid through three- compared to the total available power under irregular pattern
phase inverters. Although its high conversion efficiency, conditions of shading, reducing the overall PV array
extracting the total available power from an association of a large performance. In addition, it will be pointed out a more
number of PV panels represents an issue in these systems. The appropriate topology for large-scale photovoltaic inverters.
irregular shading effects added to the intrinsic limitations in
series/parallel panel associations may result in severe limits on II. TYPICAL LARGE SCALE PV POWER PLANT
the power injected into the grid. This paper is focused on
simulations and experimental results from a power plant The typical solar panel commercially available nowadays
subjected to controlled shading conditions. In addition of this has rated power around 300 W, whereas large-scale plants
work, an alternative multi-MPPT converter topology capable of have rated power from one until some hundreds of megawatts.
dealing with such conditions is enunciated and proper references Thereafter a huge number of PV panels are used in such
from it are also pointed out. installations and these are connected in some thousands of PV
Index Terms - Photovoltaic cells, Photovoltaic systems, Maximum modules groups of series/parallel associations, called arrays.
power point trackers, DC-AC power converters. The arrays are commonly used as the DC bus of a two-level
three-phase inverter, commercially available in blocks of
around 250kW rated power. The connection into the medium
I. INTRODUCTION
voltage grid is done by a 50/60Hz transformer. As an
Solar farms – large-scale photovoltaic (PV) power plants advantage, the semiconductor, the passive devices, and the
with installed capacity greater than 1MW – typically use a control system are conventional for inverter systems and also
consolidated topology: a central three-phase two-level inverter used for many applications, such as motor drives, hence
injecting power from a large amount of PV panels into the AC cheaper and well known topology.
medium voltage grid. PV panels in such installations are As an example, this paper has its tests made on a 1MW
arranged in series/parallel sets connected directly to the installation, fig. 1, located in São Lourenço da Mata,
inverter DC Bus. This is the simplest topology and also PE/Brazil near a soccer stadium, as a 2014 Soccer World Cup
commercially cheaper than any other converter, due to its investment, a typical investment in Brazil during last years.
usage in many other applications such as motor drives. The The plant is divided into an 840kWp conventional large scale
common used topology has good conversion efficiency power plant and a 50kWp experimental plant. Using a
(higher than 95%) and high reliability, since its technology is commercial inverter, the power plant characteristics are
already well established and has fewer components compared presented on the tab. 1.
to other options [1-2]. Despite those advantages, this topology
has poor maximum power point tracking (MPPT) capabilities
[3-4]. Since the PV-curve is highly affected by irradiance
levels and temperature, this paper shows that the maximum
available power from each panel may not be extracted by a
central inverter considering the technology aspects of the
series/parallel association and also the minimum DC bus
voltage limit despite the MPPT complexity. An array with
minor number of PV panels occupies a smaller area; therefore
it is minor affected under a same irregular pattern of shading.
The losses, even during the described situations, are smaller Source: Neoenergia Group Publicity
than in common large arrays.
A detailed study of panel arrays facing partial shading Figure 1 – Large-scale PV Power Plant (located in São
condition, using simulations and experimental test results are Lourenço da Mata,PE/Brazil).
*Graduate Program in Electrical Engineering - Federal University of Minas Gerais - Av. Antônio Carlos 6627, 31270-901, Belo
Horizonte, MG, Brazil

978-1-4799-7944-8/15/$31.00 ©2015 IEEE

 TABLE 1 – PV Plant Characteristics characterized irradiation shields. Next sections present those
840kWp conventional large scale power plant tests.

Panels per Inverter

A. Modeling and Simulation
Series Connected 20 Inverter Rated Power 210kW
Parallel Strings 14 Power at MPP 222.6kW Based on a solar power plant in operation, a simulation
was performed on PSCAD software using a proper model of
Arrays 3 plant panels and the grid where it takes place. From the PV
Catalog Data of the AC Central Inverter Output panels simulation models, it was computed the sum of the
Number of Inverters 4 Total Rated Power 840kW individual maximum power from each panel (MPi) as well as
the theoretical power available for a single MPPT/inverter set
Total of Panels 3360 Total Power at MPP 889.8kW
(MParray) and the effective power extracted by the single
50kWp experimental plant MPPT/inverter (MPPT).
The test was performed using several controlled partial
Panels per Inverter
shading scenarios. One of those scenarios is presented in fig.
Series Connected 21 Inverter Rated Power 10kW 2, where all PV panels from one string are subjected to 1000
Parallel Strings 2 Power at MPP 11.13kW W/m² and the second string is subjected to partial shading,
ranging from 1000W/m² (no shading) down to two different
Arrays 1
shading levels, G=700W/m² and G=300 W/m2. The tests were
Catalog Data of the Experimental Plant based in two characteristics panel temperatures obtained from
Number of Inverters 5 Total Rated Power 50kW manufacturer catalog, T = 25ºC (STC – Standard Test
Condition) and T = 46ºC (NOCT – Nominal Operation Cell
As a characteristic of the PV system, the total PV panels’ Temperature). In this simulation it was used a moving shade,
power at maximum power point (Pmpp) is greater than the with the number of shaded panels starting from 0 until all 20
inverter rated power (Prated), tab 1. The catalog maximum panels in one of the two inverter panels’ strings.
power point is given at a standard thermal condition, STC, The results for the first scenario are shown in fig. 3 and
which means 1000W/m² and 25°C at panel’s surface. The fig. 4. It is observed that there is an important difference
STC catalog condition does not represent any real condition in between the total power available by adding MPi’s and the
the place where the plant is located, the modules temperature maximum power from each array (MParray). Such
is higher and the generated maximum power, in such discrepancy is not due to a problem in the MPP tracking
conditions, smaller. In order to reach the inverter rated power, algorithm used (MPPT), also presented on fig. 3. This is
a greater number of panels are used. caused by the fact that individual panels operate out of their
maximum power point, since the single MPPT/inverter sets
the operating point for one equivalent PV panel. The available
III. PV ARRAY OPERATING UNDER A SINGLE MPPT power not extracted due to the lumped effect can be visualized
This paper is focused on presenting the limitations in the by the area existing between the ‘Sum(MPi)’ curve and the
large array/single MPPT configuration commonly used in ‘MParray’ curve in fig. 3, and is computed by the difference
large-scale PV power plants, especially those located in places of these two curves for each shading step.
with low clearness index. The clearness index is defined by Condition
the quotient between the total irradiation outside of the planet 1 2 3 20
Not shaded
and the received amount of irradiation considering, for 1 2 3 20
example, cloud shading.
In order to avoid irregular irradiation during the One Panel
experiment, two smaller systems are chosen: only one three Shaded
phase two level inverter connected to a 20 panels x 2 strings
array. These studies aim to characterize the losses in one of
experimental plant inverters under controlled shading 2 Panels
conditions applied during a test routine. This will take place Shaded
using solar irradiation shields. The comparison will be stated
between this inverter under test and an inverter under natural 20 Panels
conditions. Shaded
The actual power plant was modeled and, assuming
controlled shading conditions, simulated. Experimental tests
Not shaded Panel Shaded Panel
were performed to validate the simulation ones; under similar
controlled shading conditions introduced using properly Figure 2 – Proposed shading scenario 1.

978-1-4799-7944-8/15/$31.00 ©2015 IEEE

In this case, one can see that shading affects the system at the maximum power point (Vmpp) goes down. When this
efficiency (1) quite significantly and the shading effects on the voltage is very close or evenly lower than the line peak
PV array may be characterized as part of the overall system voltage seen from two inverter terminals, the MPPT lacks in
losses. tracking the Vmpp. The operation point becomes a local
maximum, instead of a global maximum. Fig. 5 shows a plot
of the power versus voltage for each number of panels shaded
(1)
in scenario 1 with G = 300W/m2 and STC condition. The line
peak voltage is about 509V, and for five panels shaded, the
power curve is depicted in red. As can be seen, the global
Shading scenario 1. (G = 700W/m2, T = 25°C) maximum is located at a voltage lower than the line peak
10000 voltage, and when the fifth panel is shaded the MPPT hit this
MPPT
9500 MParray
down limit, oscillates and lost tracking. The oscillation cause
Power (W)

Sum(MPi) it to drop and stabilize in the local maximum. This explains

9000 Power lost the behavior presented in fig. 3. For higher temperatures, the
8500 Vmpp is lower, and this process occurs with a few number of
panels shaded, like in scenario 1 in fig. 4, with G = 300W/m2
0 5 10 15 20 and NOCT temperature.
Number of shaded panels

10000 Shading scenario 1. (G = 300W/m2, T = 25°C)

9000
Power (W)

8000

7000

6000
0 5 10 15 20
Number of shaded panels

Figure 3 – Shading scenario 1 for two shading levels and panel

temperature in STC condition.

Shading scenario 1. (G = 700W/m2, T = 46°C)

MPPT
8000 MParray
Power (W)

Sum(MPi) Figure 5 – Shading scenario 1 for two shading levels and panel
7500 temperature in NOCT condition.

7000
Condition
0 5 10 15 20
Number of shaded panels Not shaded
2
Shading scenario 1. (G = 300W/m , T = 46°C)
8000
2 Panel
Power (W)

7000
Shaded

6000
4 Panels
5000
0 5 10 15 20
Shaded
Number of shaded panels

Figure 4 – Shading scenario 1 for two shading levels and panel 40 Panels
temperature in NOCT condition. Shaded

Another effect happens due the inexistence of a boost

converter between PV panels and DC bus of the converter. A
bigger difference of ‘MParray’ and ‘MPPT’ curve can exist,
because as the number of shading panels increases, the voltage Figure 6 – Proposed shading scenario 2.

978-1-4799-7944-8/15/$31.00 ©2015 IEEE

 B. Experimental Test Results
The other scenario is presented in fig. 6, where both PV
In order to validate the simulation models the presented
panels from the two strings are subjected to partial shading
simulation studies, measurements were performed on the
with the same ranges applied in the first scenario, increasing
described PV power plant, presented on fig.1. The array under
the number of panels shaded by one in each string. The results
test has 42 polycrystalline panels Yingli Solar YL265P-29b
for the second scenario are shown in fig. 7 and fig. 8. The
connected as an array with 21 series panels’ string connected
same behavior is seen in shading scenario 1. In fig.7 and in
in parallel with the other 21 series panels’ string. A similar
fig. 8 for a radiation of 700W/m2, the MPPT tracks a local
array using similar polycrystalline silicon panels, but from a
maximum, even when an achievable global maximum exists
different manufacturer, connected to other single
for the firsts eight – ten shaded panels. This problem in
MPPT/inverter was used as a power generated reference, as no
practice can be overcome doing a scan in the feasible voltage
shading was applied to this system.
range periodically and checking if there is a global maximum
[3].
Shading scenario 2. (G = 700W/m2, T = 25°C)
10000
MPPT
9000 MParray
Power (W)

Sum(MPi)
8000

7000

0 10 20 30 40 Source: SECOPA/PE
Number of shaded panels

Shading scenario 2. (G = 300W/m2, T = 25°C)

Figure 9 – Highlighted Arrays: used as a reference and under
10000
test
8000
Power (W)

These tests consisted of covering the PV panels using a

6000
previous characterized solar irradiation shield. Two types of
4000 shield are used: one is a white fabric (fig. 10), that provides
light shading on the panels, and second is a grey fabric, that
2000
0 10 20 30 40 provides heavy shading. We call the first as Shield 1 and the
Number of shaded panels
second as Shield Two in the results. So, as soon as the panels
were covered by a non-woven white fabric, the power injected
Figure 7 – Shading scenario 2 for two shading levels and panel
temperature in STC condition.
by the single MPPT/inverter connected the array under test
was measured. The generated power was also measured.
Shading scenario 2. (G = 700W/m2, T = 46°C) Using both measurements (shading and no shading system)
help to evaluate the decrease of power injected into the grid
8000 MPPT
due to partial shading.
MParray
Power (W)

7000 Sum(MPi) One of the major challenges of this experiment arises from
the fact that this test is done in an uncontrolled environment,
6000 subjected to climatic variations, especially fast variations of
irradiance due to the passage of clouds. In order to mitigate
0 10 20 30 40 these effects, some actions should be taken to improve test
Number of shaded panels quality. As the panels covering procedure is very time
Shading scenario 2. (G = 300W/m2, T = 46°C) consuming, it is not advisable gradually covering the panels
8000
and measuring power because solar radiation naturally
changes from time to time.
Power (W)

6000
Aiming to reduce the test time, and consequently
4000 decreasing the variation of irradiation, all panels were initially
covered using clips, guaranteeing the correct positioning of
2000
the irradiation shields without the presence of folds or covers
0 10 20 30 40
Number of shaded panels of unwanted positions. The use of clips allows irradiation
shields to be removed quickly. Thus, the testing procedure
Figure 8 – Shading scenario 2 for two shading levels and panel consisted of covering all panels, using clips to fix the
temperature in NOCT condition. irradiation shields, waiting the conditions of no clouds in the
sky, and gradually uncovering the panels and measuring the

978-1-4799-7944-8/15/$31.00 ©2015 IEEE

power. By doing so, the variation of irradiation during the test test array.
period was mitigated. An example of the procedure is shown As presented in fig. 3 and fig. 4, a small number of shaded
in fig. 10 and fig. 11. panels cause a greater difference in the power injected into the
grid. That can be tesfitied more experessively in the Shield 1
(fig. 12) results. When the number of shaded panels increases,
the power loss has a lower slope, also testified by the both
shield types used in the test. Since these are experimental test
results, there is an uncontrolled factor that has to be
recognized: the continuously varying solar irradiation. And
that is why just relative and normalized results is showed.

0.9 Shield 1

Normalized Power
Shield 2
0.8

0.7

Figure 10 – All panels of one string covered by the shield 1 – 0.6

White fabric (Array under test – fig.9).
0.5

0.4
0 5 10 15 20
Number of Shaded Panels

Figure 12 – Experimental test results: PV array under test power

(under controlled shading) normalized to PV array reference power.
Shield 1 – White fabric and Shield 2 – Grey Fabric.

III. CENTRAL SOLAR INVERTERS LIMITATIONS AND

ALTERNATIVES MULTI-MPPT TOPOLOGY
The common used topology in large-scale photovoltaic
power operates as the array was one equivalent panel,
regardless of the MPPT algorithm employed. The PV panels
association, in the presented case, is the origin of the reduced
power generation capacity (or increased system losses). The
PV panels’ connection directly into the DC bus incurs in a
Figure 11 – Partial string covered by the shield 1 - White fabric higher voltage minimum limit, restricting the MPPT operation
(Array under test – fig.9). as presented. Also, the central inverter uses a high-power low-
frequency transformer (50/60Hz) which is considered heavy,
The PV array connected to one inverter was subject to a bulky and difficult to replace, but necessary due to galvanic
controlled irradiation shielding, according to the procedure isolation issues, step up the voltage to the power grid levels,
used in the simulation model. The experimental results, and to avoid leakage currents [8-9].
present on fig. 12, show the generated power of the PV array Suitable alternatives for large-scale PV systems connected
under test (under controlled shading) normalizes to the to the medium voltage grid already exists [4-7], as presents fig
generated power from PV array reference. In this figure, the 13 and 14. The cascaded-type topology accommodates a
reference array always generates 1p.u. of power in all of reduction in the number of PV panels connected to each
conditions. Two shields are used and presented in this result, MPPT/inverter cell. These topologies have several
in which Shield 1represents the white fabric and Shield 2, the independent high frequency dc-dc links, which allows
grey fabric. As expected, as the number of shaded panels independent voltage control, and therefore a greater number of
increases, also the relative power generated decreases. Also, MPP trackers. Since the panels are distributed among a larger
as the grey fabric represents a heavy shading pattern, the number of tracking algorithms, a higher performance ratio is
deeper is the difference between the reference array and under obtained, thanks to the minimization of the effects caused by

978-1-4799-7944-8/15/$31.00 ©2015 IEEE

shaded strings. shading situations.
It was demonstrated that the MP from an array is lower
than the sum of the MP from each panel under the same
irradiation/shading conditions, due to association effects
alone. Experimental tests and simulations using well
established computer models were used to this end. More
details of the PV panel, converter and grid models, description
of the test procedures and further simulation and experimental
tests will be presented in final paper. A suitable alternative for
large-scale photovoltaic power plants is discussed, addressing
the multiple MPPT requirements for maximizing the PV array
utilization.

ACKNOWLEDGEMENT
This work was supported by Companhia de Eletricidade
do Estado da Bahia (COELBA). Project: “Transformadores
Eletrônicos”, Research Grant: PD-0047-0060/2011. This work
Figure 13 – Cascaded Multilevel Topology with LF isolation has been also supported by the Brazilian agency CAPES.
[6].
MV Grid V. REFERENCES
Ls Ls Ls [1] A. Elasser, M. Agamy, J. Sabate, R. Steigerwald, R. Fisher,
Rs Rs Rs M. Harfman-Todorovic, “A Comparative Study of Central and
Distributed MPPT Architectures for Megawatt Utility and Large
Scale Commercial Photovoltaic Plants,” in 36th Annual Conference
PV Vc1 of IEEE Industrial Electronics, pp. 2753 – 2758, 2010.
PV Vb1 [2] J.A. Villarejo, A. Molina-Garcia, E. De Jodar, “Comparison
PV String String Va1
String
of Central vs Distributed Inverters: Application to Photovoltaic
Systems,” IEEE International Symposium on Industrial Electronics,
PV Vc2 pp 1741 – 1746, 2011.
PV String Vb2
PV String Va2 [3] E. Koutroulis and F. Blaabjerg, “A new technique for tracking
String the global maximum power point of PV Arrays operation under
partial-shading conditions,” IEEE Journal of Photovoltaics, vol. 2,
PV Vcn no. 2, pp. 184–190, Apr. 2012.
PV String Vbn [4] N. C. Foureaux, B. J. C. Filho, and J. A. S. Brito, “Cascaded
PV String Van Multilevel SST Medium Voltage Converter for Solar Applications,”
String
9th Int. Conf. Power Electron. - ECCE Asia (to be Publ. ), 2015.
[5] E. Villanueva, P. Correa, J. Rodriguez, and M. Pacas, “Control
Figure 14 – Cascaded Multilevel Topology using Solid State of a Single-Phase Cascaded H-Bridge Multilevel Inverter for Grid-
Transformer Isolation - Proposed topology [4]. Connected Photovoltaic Systems,” Ind. Electron. IEEE Trans., vol.
56, no. 11, pp. 4399–4406, 2009.
[6] H.S. Krishnamoorthy, S.Essakiappan, P.N. Enjeti, R.S. Balog,
IV. CONCLUSIONS S. Ahmed, “A New Multilevel Converter for Megawatt Scale Solar
Photovoltaic Utility Integration,” in IEEE Applied Power Electronics
As shown in the previous sections, hundreds or even Conference and Exposition (APEC), Pg. 1431 – 1438, 2012.
thousands of PV panels in common large-scale installations [7] R. Chattopadhyay , S. Bhattacharya , N. Foureaux ,S. Silva , B.
are arranged in arrays and connected to a single inverter. Cardoso, H. Paula, I. Pires, P. Cortizio , L. Moraes , J. Brito “Low
Voltage PV Power Integration into Medium Voltage Grid Using
Despite high reliability, simple topology and good conversion
High Voltage SiC Devices,” in Proceedings of the IEEE
efficiency, the central converter topology has poor maximum International Power Electronics Conference, 2014
power point tracking (MPPT) capabilities, which impacts the [8]H. L. Jou, W.J. chiang, J.C. Wu “Voltage-Mode Grid-Connected
overall system performance. Since the PV-curve is highly Solar Inverter with High Frequency Isolated Transformer,” in IEEE
affected by irradiance levels and temperature, a central International Symposium on Industrial Electronics, Pg 1087 - 1092,
2009.
inverter cannot extract the maximum power available from
[9] T. Tran-Quoc, C. LE Thi Minh, H. Colin, C. Duvauchelle, S.
each individual panel. PV panel behavior, the effect of the Bacha, S. Aissanou, B. Gaiddon, C. Kieny, G. Moine, Y. Tanguy
bypass and series diodes and the minimum dc voltage “Transformerless Inverters and rcd: What’s the problem?,” in 25th
requirement for proper inverter operation prevent that. There European Photovoltaic Solar Energy Conference and Exhibition/5th
will be some power loss from the array connections in partial World Conference on Photovoltaic Energy Conversion, pg 4554 –
4559, 2010.

978-1-4799-7944-8/15/$31.00 ©2015 IEEE