2932 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 11, NO.

4, OCTOBER 2020

Virtual Inertia Emulator-Based Model Predictive

Control for Grid Frequency Regulation
Considering High Penetration of Inverter-Based
Energy Storage System
Nicolas Sockeel , James Gafford, Member, IEEE, Behnaz Papari , Member, IEEE,
and Michael Mazzola , Senior Member, IEEE

Abstract—As awareness of human footprint grows, solutions are TESS Energy storage system time constant [s]
investigated to reduce greenhouse gas emissions (GHG). To reduce GRC Generation rate constraint [p.u. MW/min]
the carbon intensity of electricity production, a massive deployment Vu Maximum valve gate limit [p.u.MW]
of renewable energy sources (RESs) has become a mandatory
goal. A significant challenge behind the deployment of RESs is the VL Minimum valve gate limit [p.u.MW]
frequency regulation of such systems due to the high penetration Ek Stored kinetic energy [J]
of inverters based on energy storage systems (ESS). To overcome J System inertia moment [kg.m2 ]
this issue, this manuscript proposes, as the main contribution, ω Angular frequency deviation [rad/s]
to use a virtual inertia emulator-based model predictive control. Tm Mechanical torque [N]
MPC is an optimization-based control strategy that aims at finding
the optimal control actions of a system by predicting its future Te Electrical torque [N]
behaviors. This control solution is compared to a proportional (P) Pm Mechanical power [W]
and a proportional integrator (PI) controller. The results of this Pe Electrical power [W]
study show that the MPC controller reduce the minimum ESS S Apparent power [V.A]
capacity by 55% and 2%, and the ESS energy throughput by 86% ΔPg Governor power variations [p.u. MW]
and 36% compared to a P and PI controller. Those results suggest
that the MPC controller can help to diminish the CAPEX and ESS ΔPm Thermal plant power variations [p.u. MW]
aging compared to a P or PI controller. ΔPACE Area control error system power variations [p.u.
MW]
Index Terms—Grid frequency control, model predictive control,
microgrid, renewable energy sources, proportional integrator
ΔPinertia Inertia power [p.u. MW]
control, virtual inertia emulator, energy storage system. ΔPess Energy storage system power [p.u. MW]
ΔPdist Disturbances power variations [p.u. MW]
Δf Frequency variations [Hz]
NOMENCLATURE p Soft constraint penalized weight [-]
Symbol Name Units q Weight on the output signals [-]
b Frequency bias factor [p.u MW/Hz] r Weight on the manipulated variable [-]
KI Integral control variable gain [-] Δfref Targeted frequency regulation [-]
Tg Governor time constant [s] ui Control output [-]
Tt Turbine time constant [s] ui_optimal Optimal control output [-]
R Droop constant [Hz/p.u MW] uref Targeted control output [-]
d Microgrid damping coefficient [p.u. MW/Hz] s Slack variable [-]
H Microgrid system inertia [p.u MW s] Nc Control step horizon [-]
KVI Virtual Inertia Emulator gain [-] Np Prediction step horizon [-]

Manuscript received September 20, 2019; revised December 20, 2019 and I. INTRODUCTION
January 31, 2020; accepted March 17, 2020. Date of publication March 20,
URRENTLY, respectively 35% and 28% of the US elec-
2020; date of current version September 18, 2020. This work was supported
by the Energy Production and Infrastructure Center at the University of North
Carolina at Charlotte. Paper no. TSTE-01028-2019. (Corresponding author:
C tricity is produced through natural gas and coal power
plant [1]. Those power plants always require the use of electric
Nicolas Sockeel.)
The authors are with the Energy Production and Infrastructure Center, Univer- generators that provide inertia to the grid through their rotating
sity of North Carolina at Charlotte, Charlotte, NC 28223 USA (e-mail: nsock- parts. This inertia help to manage the grid frequency by avoiding
eel@uncc.edu; jgafford@uncc.edu; bpapari@uncc.edu; mmazzola@uncc.edu). significant grid frequency deviation from the regular frequency
Color versions of one or more of the figures in this article are available online
at http://ieeexplore.ieee.org. operation (60 Hz in the US). However, those power plants are
Digital Object Identifier 10.1109/TSTE.2020.2982348 responsible for greenhouse gas emissions. In fact, according to
1949-3029 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.
SOCKEEL et al.: VIRTUAL INERTIA EMULATOR-BASED MODEL PREDICTIVE CONTROL 2933

the US Department of Energy (DOE), around 33% of the US an islanded microgrid by introducing a fractional-order integral
CO2 emissions are related to electricity production in 2018 [2]. cost function into the model predictive control (MPC) algorithm.
In 2017 in the USA, around 573g CO2 is produced for every In [24] and [25], using a VIE control based on the MPC method,
kWh of electric energy used [3]. To reduce the carbon intensity the stability and robustness performance of the microgrid is
of electricity production, a massive deployment of renewable investigated during the high penetration of RESs.
energy sources (RESs) has become a mandatory goal. As a According to the literature, authors mainly focused their
consequence, more and more microgrids appear [4]. A microgrid effort to develop a control method robustly and minimizing the
can be defined as a group of independent local grids composed frequency deviation. As far as our knowledge goes, no previous
of RESs, an energy storage system (ESS), and domestic loads study has analyzed and optimized the required size of the ESS.
having the capability to work in either isolated or grid-connected This is why the main contribution of this paper is to provide a
mode to reduce the utility grid burden [5], [6]. However, the comparison between VIE based MPC and a traditional P and
inverters used with RESs reduce dramatically the inertia of the PI controller in terms of frequency deviation, ESS size, power
grid leading to more frequency/ voltage instability compared to usage, and energy throughput.
the traditional grid using synchronous generators [7]–[10]. This The rest of this paper is organized as follows. In Section II, a
is why a significant challenge behind the deployment of RESs description of the microgrid system is given. In Section III, the
is the frequency regulation of such systems. model predictive control strategy is described. In Section IV,
To tackle this issue, virtual synchronous generators/machine the results and discussion of this study are provided. Finally, a
(VSG, VSM) is used to imitate the behavior of traditional virtual conclusion and future works are drawn in Section V.
synchronous generator. [11], [12]. Virtual inertia emulator (VIE)
is a particular case of VSG that offers better frequency regulation II. MICROGRID SYSTEM OVERVIEW
by emulating the rotating mass inertia of VSG. ESS is used
A. Frequency Control Based on Inertia Response
to emulate virtual inertia by providing or storing extra power
from the grid based on the rate of change of frequency (RoCoF) In this study, the frequency control is split into three different
strategy. VIE techniques offer a base to control the grid stability operations: the inertia, primary, and secondary response state.
and resiliency for high RESs penetration. When a frequency deviation occurs, the inertia response state
Numerous control techniques have been implemented us- acts first by balancing the power requirement by the kinetic
ing VIE to solve the microgrid frequency control problems, energy from a generating unit for bandwidth up to 5 s. Then,
improving frequency stability [13]–[18]. In [13] a traditional the primary response stabilizes frequency to a new steady-state
proportional-integral (PI) control technique is used. In [14], VIE condition for a time duration between 5 s to 30 s after the
based on the PI controller is applied for wind power support appearances of a disturbance. Finally, the secondary response,
microgrid frequency regulation. VIE based on fuzzy logic to also called load frequency control (LFC), recovers frequency
increase the frequency stability in an islanded microgrid, is to its nominal state of equilibrium for a bandwidth going from
developed in [15] and [16]. Paper [17] evaluates the frequency 30 s to 30 min after the appearances of a perturbation [26].
response estimation method based on VIE employed to improve The overall kinetic energy of the system rotational mass,
the stability of an extensive wind power system. An H con- including spinning loads, is computed as follow [11], [26], [29]:
troller, based on VIE, is presented in [18] and shows superior
Ek = 0.5 J ω 2 (1)
robustness compared to an optimal PI controller.
A more sophisticated control technique is of interest: model The rate of change of rotor speed depends on the torque
predictive control (MPC). MPC is an optimization-based control balance of spinning mass as follows:
strategy that aims at finding the optimal control actions of a Pm Pe dω
system by predicting its future behaviors. MPC has been recog- Tm −Te = − =J (2)
ω ω dt
nized as an efficient online general optimization-based control
strategy. In contrast to traditional feedback control strategies, The stored kinetic energy is commonly proportional to its
MPC can manage precise non-linear constraints and provide power rating and known as system inertia constant [25]:
adequate control solutions for event-driven and complex nonlin- Ek
ear dynamics systems. Several researchers have already studied H = (3)
S
VIE based MPC. In [19], authors have presented an adaptive The rate of change of frequency (RoCoF) is used to determine
model predictive load frequency control method for a multi-area the required power for controlling the system frequency against
interconnected power system with photovoltaic generation. In the variation in the power demand. The RoCoF can be computed
[20] is analyzed the impact of system parameters on the control as follows [26], [29]:
performance of MPC for frequency regulation in a stand-alone
MG. MPC is compared to the PI control strategy of the BESS dω ω (Pm −Pe )
= (4)
for microgrid frequency regulation in [21]. Considering electric dt 2HS
vehicles as energy storage, a decentralized MPC is used as a
grid frequency regulation method with RESs in [22]. In [23] B. Microgrid System
a novel fractional-order model predictive control (FOMPC) As shown in Fig 1, this paper deals with an islanded microgrid
method is proposed to achieve the optimal frequency control of composed of a 15 MW domestic load, 12 MW thermal power

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.
2934 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 11, NO. 4, OCTOBER 2020

component are detailed in the following sections. The prediction

horizon is limited to one time-step ahead due to the disturbances
predictor, as discussed in the further section, and the time-step
is 100 ms.

B. State Space Equations

The following state-space equations describe the linear time-
invariant (LTI) microgrid plant model included inside the MPC
controller. Those equations represents exactly the microgrid
plant depicted in Fig. 2.
ΔPACE − R1 Δf
Fig. 1. Schematic of the islanded microgrid. ΔPg = (5)
1 + sTg
−Ki
ΔPACE = (bΔf) (6)
plant, 16 MW RESs, and an undefined size of ESS. Concerning s
the ESS, no constraints have been considered for this manuscript. (ΔPt + ΔPess + ΔPdist )
Δf = (7)
Instead of designing a controller made for a specific type of ESS, 2Hs + d
the controller performances give the constraints that the ESS has ΔPg
to pursue. The system base is 12 MW. ΔPt = (8)
1 + sTt
The mathematical schematic of the islanded microgrid is pro- u
vided in Fig. 2. The microgrid parameters are given in Table I. To ΔPess = (9)
1 + sTess
simulate a scenario close to reality, this study considers realistic
constraints and conditions required by a physical system. For ΔPinertia = KVI s ΔPess (10)
instance, the thermal power plant is limited by the generation
From equation 5 to 10, the following state-space equation can
rate constraint (GRC) and the maximum/ minimum valve gate
be derived:
opening or closing for a turbine unit (Vu and Vl ). The GRC of 
the thermal generation is set as 12% p.u. MW/min. Vu and Vl Ẋ = AX + BU
are set to 0.5 and −0.5, respectively. (11)
Y = CX + DU

C. Energy Storage System Based Virtual Inertia Emulator With X = [ΔPg ΔPACE Δf ΔPt ΔPinertia ]T ,
Virtual inertia emulator (VIE) is a particular case of VSG that ⎡ ⎤
−1/Tg 1/Tg −1/(RTg ) 0 0
propounds better frequency regulation by emulating the rotating ⎢ ⎥
⎢ 0 0 −bKi 0 0 ⎥
mass inertia power of synchronous generator through an ESS ⎢ ⎥
[11]–[13]. The VIE power is computed based on the rate-of- A=⎢ ⎢ 0 0 −d/(2H) 1/(2H) −1/(2H Tess ) ⎥
⎥
⎢ ⎥
change of frequency (RoCoF) to add or store extra active power ⎣ 1/Tt 0 0 −1/Tt 0 ⎦
through the ESS to the microgrid while frequency disturbances 0 0 0 0 −1/Tess
appear. Such a system is highly sensitive to frequency noise ⎡ ⎤
measurement. This issue can be solved by emulating the ESS 0 0
dynamic behavior with a low pass filter [26]–[29]. ⎢ 0 ⎥
⎢ 0 ⎥
Consequently, the VIE system helps to improve the frequency ⎢ ⎥
B = ⎢ KVI /(2H Tess ) 1/(2H) ⎥ , C = [0 0 1 0 0], D = [0 0]
stability and resiliency as if RESs in the islanded microgrid ⎢ ⎥
⎣ 0 0 ⎦
would have inertia similar to the traditional generating units
(i.e., synchronous generators). In this study, it is assumed that KVI /Tess 0
the inertia power is emulated through the ESS. U = [u ΔPdist ]T , and Y = Δf

III. MODEL PREDICTIVE CONTROL For simplifying the mathematical notation, it can be noted
that U is composed of the control variable, u, and the power
A. Overview disturbance variations, ΔPdist .
In this section, the MPC-based control strategy is described. The following hard constraint can not be violated:
The plant and disturbances have been presented in the previous
−0.5 ≤ ΔPg ≤ 0.5 (12)
section. The primary module, presented in Fig. 3. includes a
linear time-invariant microgrid plant model an optimization
C. Objective Function
solver, an objective function, and a disturbances predictor. The
MPC controller aims at finding the optimal control output that For this research, the objective function of the single input
minimizes the objective function. Details concerning each MPC single output (SISO) optimization problem can be written as

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.
SOCKEEL et al.: VIRTUAL INERTIA EMULATOR-BASED MODEL PREDICTIVE CONTROL 2935

Fig. 2. Schematic of the microgrid plant model.

TABLE I TABLE II
PARAMETERS OF THE ISLANDED MICROGRID PARAMETERS OF THE OBJECTIVE FUNCTION

As q and r are positive weight values, it can be mathematically

proven that the problem is quadratic and convex [30].
The goal of the MPC is to minimize the ESS power and ESS
energy throughput while keeping the frequency deviation below
a certain trigger value. As the control variable is proportional to
the ESS power (cf eq. 9), both the targeted frequency deviation,
noted Δfref , and targeted control variable, noted uref , are set
to 0. Moreover, the following soft constraint has been specified
to avoid the frequency deviation get above 0.05 Hz.

−0.05 ≤ Δf ≤ 0.05 (14)

Such constraint can be violated under an added penalty to the

cost function. Finally, the objective function used in this study
can be simplified as follows:

k+Np −1 k+Nc −1
J = q (Δfi )2 + r (ui )2 + p s2 (15)
i=k i=k iZ

Fig. 3. Model Predictive Control based Power Management System.

where Z is the set of i such that |Δfi |> 0.05.
The parameter value of the cost function is given in Table II.
The coefficients q, r, and p have been selected to demonstrate
follow: that an MPC solution can be found that performs better than
k+Np −1 k+Nc −1 the optimal values for P and PI controllers (cf Section IV. B).
J = q (Δfi − Δfref )2 + r (ui − uref )2 (13) The methodology to find optimal q, r and p values should be
i=k i=k investigated in a future work.

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.
2936 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 11, NO. 4, OCTOBER 2020

Fig. 4. Disturbance signals for the simulation.

Fig. 5. Pareto-Front analysis of the ESS energy throughput as a function of

the P and I values.
D. Optimization Function
As the optimization problem is quadratic and convex, the
TABLE III
global minimum can be found quickly using a gradient-based SPECIFICATION FOR P AND PI CONTROLLER
algorithm. The one used is the default implemented by Matlab
for the MPC toolbox. It is an active set algorithm named KWIK
[31].

E. Disturbances Predictor
The disturbances predictor is an essential part of the MPC
controller. The ability to predict future disturbances precisely
has a significant impact on how far in the future the MPC
can forecast the system behavior and how efficiently the MPC
controller can select the control variable. For this study, a prac-
tical and straightforward approach has been considered. Over
the prediction step horizon, the disturbances are considered
constant and equal to the current one. Such consideration limits
the ability to predict far in the future, and explains why the
prediction (Np ) and receeding (Nc ) horizon have been set up
to 1 in Table II. However, interesting and feasible results are
still found, as discussed in the next section.
Fig. 6. Microgrid frequency deviation.

IV. RESULTS AND DISCUSSION

A. Scenario the P and I values such as the maximum frequency deviation is
MPC from Matlab/Simulink toolbox has been selected. The equal to 0.05 Hz.
different parameters have been described previously. The dis- This Pareto-Front curve goes from the case P = 0 (purely
turbances are composed of two parts: the power consumed by integrator controller) to the case I = 0 (purely proportional
the load (domestic and industrial), and the power production controller). Any P and I values that does not belong to this
by the RESs. Both have been created using a Gaussian noise curve leads to a constraint violation or a suboptimal solution
signal. To disturb the system highly, a square signal is added. The (higher ESS energy throughput). Consequently, the Pareto-front
different disturbances signals are presented in Fig. 4. The MPC analysis provides the unique set of optimal P and I values with
performance is compared to a proportional (P) and proportional the minimum energy throughput and following the maximum
integrator (PI) controller in the following section. The settings frequency deviation constraints of 0.05 Hz. It also gives the
for both controllers are presented in the next section. optimal value to set a P controller. The values of the P and PI
controller are provided in Table III.
B. Selection of P and PI
The P and PI controllers have been set up manually with the C. Results
aid of a trial error methodology to construct a Pareto-Front curve The simulation results are presented from Fig. 6 to 9 and
in order to estimate optimal parameters for the conventional Table IV. In Fig. 6, the frequency deviation of the microgrid for
controllers. Optimal conventional controllers’ performance is the three different controllers is shown. From this figure, it can
thus compared with the MPC alternative. Fig. 5 shows a Pareto- be observed that all the controllers have a maximum frequency
Front analysis of the ESS energy throughput as a function of deviation of around 0.05 Hz. Those frequency deviation peaks

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.
SOCKEEL et al.: VIRTUAL INERTIA EMULATOR-BASED MODEL PREDICTIVE CONTROL 2937

TABLE IV
COMPARISON AMONG THE DIFFERENT CONTROLLERS

Fig. 9. ESS energy throughput.

noted that the PI controlled ESS system does not necessarily

return back to its initial charge condition. When considering a P
controller in Fig. 2, it becomes clear that ΔPinertia is a function
of the derivative of Δf via the RoCoF equation. The integration
of ΔPinertia provides the ESS energy variation (cf Fig. 8).
Consequently, the ESS energy variation of the P controller is
proportional to Δf, as shown when comparing the P controller
results in Fig 6 and 8. Furthermore, the ESS energy variation
Fig. 7. ESS power provided to the microgrid. of the I controller is proportional to the integral of Δf. As a PI
controller is a sum of P and I controllers, it can be concluded that
the PI controller adds an open-loop integration of Δf, as shown
when comparing the PI controller results in Fig. 6 and Fig. 8. This
is why the PI controlled ESS system does not necessarily come
back to the initial charge condition and may stay almost fully
charged or discharged depending on the shape of disturbance
in a steady-state situation.. The red lines in Fig. 6 and 8 clearly
are proportional, while the green line in Fig. 8 has a temporal
shape similar to the integral of the green line in Fig. 6. Finally,
Fig. 9 shows the ESS energy throughput for the three different
controllers. Using Table IV, the energy going through the ESS
using the MPC controller is reduced by 86% and 36% compared
to a P or PI controller respectively.
Fig. 8. ESS energy variations.
D. Discussion
The overall goal of the MPC controller is to minimize the
appear when step changes happen. Also, it can be noted that the ESS power and energy throughput while keeping the frequency
frequency is oscillating significantly more for the P controller deviation below 0.05 Hz. In this regards, results show that the
than for the others. In Fig. 7, the power provided by the ESS for MPC has better performance than the optimal P and PI controller.
the different controllers is given. The same trend as Fig. 6 can Indeed, both are able to keep the frequency deviation below
be noticed: each controller has a peak power around 2.5 MW, 0.05 Hz, but the ESS energy throughput is lower when controlled
each peak power appears when the disturbance step happens, and by MPC than P and PI.
the ESS power using the P controller is oscillating significantly Many previous publications have correlated the increase of
more than the two others. As the ESS characteristics are kept battery [32]–[34] or supercapacitor [35], [36] ESS aging with
open, a State of Energy (SoE) or State of Charge (SoC) cannot an increase of the power, depth of discharge, and the number of
be computed. However, the ESS energy variations, provided in cycles, and consequently, the energy throughput [36]. This loss
Fig. 8, is the closest calculated value to an ESS SoE or SoC. of performance is called cycling aging. Results show that the
From this graph, the minimum required ESS capacity can be ESS managed by an MPC controller handles less power (Fig 7
found by inspecting the peak to peak value. Using Table IV, it and Table IV), a smaller depth of discharge (Fig 8 and Table IV),
can be calculated that the MPC controller can help to reduce and, most of all, less energy throughput (Fig 9 and Table IV)
by 55% and 2% the minimum ESS capacity compared to a than an ESS managed by a P or PI controller. If we consider
P or PI controller respectively. Also, from Fig. 8, it can be the same ESS design (capacity and technology) used under the

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.
2938 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 11, NO. 4, OCTOBER 2020

same operating temperature for the three control algorithms, [6] M. Marzband, R. Ardeshiri, M. Moafi, and H. Uppal, “Distributed gen-
the results of this study suggest that the ESS cycling aging is eration for economic benefit maximization through coalition formation
based game theory concept,” Int. Trans. Elect. Energy Syst., vol. 27, no. 6,
decreased when using an MPC compared to a PI or P controller. pp. 1–16, 2017.
Literature often expresses the cost of an ESS in dollars per [7] H. Bevrani, M. Watanabe, and Y. Mitani, Power System Monitoring and
unit of installed capacity ($/kWh) [38]–[39]. As Table IV shows, Control. Hoboken, NJ, USA: Wiley, Jun. 2014.
[8] G. Yao, Z. Zhang, T. Tang, and M. Benbouzid, “Small signal models
the minimum ESS capacity is smaller for an MPC compared to based stability and controller parameters sensitivity analysis of microgrid
a P or PI controller. Therefore, it can be deduced that the ESS in islanded mode,” in Proc. 40th Annu. Conf. IEEE Ind. Electron. Soc.,
CAPEX is reduced when using an MPC compared to a P or PI Dallas, TX, USA, Oct. 29.–Nov. 1, 2014, pp. 4995–5001.
[9] E. Rakhshani, D. Remon, A. Cantarellas, and P. Rodriguez, “Analysis of
controller. derivative control based virtual inertia in multi-area high-voltage direct
Fig. 8 also suggests that the ESS required the need to have the current interconnected power systems,” IET Gener. Transmiss. Distrib.,
capability to deplete itself fully in a couple of minutes (around vol. 10, no. 6, pp. 1458–1469, May 2016.
[10] H. Bevrani, T. Ise, and Y. Miura, “Virtual synchronous generators: A survey
3–4 min). Moreover, the maximum Prate , meaning the ratio and new perspectives,” Int. J. Elect. Power Energy Syst., vol. 54, pp. 244–
between the absolute maximum peak power and the minimum 254, Jan. 2014.
size of the ESS, is largely above 10 for any controller and [11] D. Li, Q. Zhu, S. Lin, and X. Y. Bian, “A self-adaptive inertia and damping
combination control of VSG to support frequency stability,” IEEE Trans.
reaches the highest value for the MPC. With that respect, a Energy Convers., vol. 32, no. 1, pp. 387–398, Mar. 2016.
supercapacitor is much more adapted than a battery for such [12] H. Zhao, Q. Yang, and H. Zeng, “ Multi-loop virtual synchronous generator
an application. In terms of computation time, as the controller control of inverter-based DGs under microgrid dynamics,” IET Gener.
Transmiss. Distrib., vol. 11, no. 3, pp. 795–803, Feb. 2017.
is looking only one step ahead and the controller step time is [13] P. F. Frack, P. E. Mercado, and M. G. Molina, “Extending the VISMA
100 ms, the controller is likely to work fine on a real-time. concept to improve the frequency stability in microgrids,” in Proc. IEEE
Intell. Syst. Appl. Power Syst. Conf., Sep. 2015, pp. 1–6.
[14] E. Chikuni, F. F. Gonzalez-Longatt, and E. Rashayi, “Effects of the syn-
V. CONCLUSION thesis inertia from wind power on the total system inertia after a frequency
disturbance,” in Proc. IEEE Ind. Technol. Conf., Feb. 2013, pp. 826–832.
As the main contribution, this paper proposes to use a vir- [15] K. Mentesidi, R. Garde, M. Aguado, and E. Rikos, “Implementation of a
tual inertia emulator based model predictive control for grid fuzzy logic controller for virtual inertia emulation,” in Proc. IEEE Smart
Elect. Distrib. Syst. Technol. Conf., Sep. 2015, pp. 606–611.
frequency regulation considering high penetration of inverters [16] Y. Hu, W. Wei, Y. Peng, and J. Lei, “Fuzzy virtual inertia control for
based on ESS for managing RESs. MPC is an optimization- virtual synchronous generator,” in Proc. Chin. Control Conf., Jul. 2016,
based control strategy that aims at finding the optimal control pp. 8523–8527.
[17] R. Yan and T. K. Saha, “Frequency response estimation method for high
actions of a system by predicting its future behaviors. This wind penetration considering wind turbine frequency support functions,”
control solution is compared to a P and PI controller. The results IET Renew. Power Gener., vol. 9, no. 7, pp. 775–782, Aug. 2015.
of this study show that the MPC controller can help to reduce by [18] T. Kerdphol, F. S. Rahman, and M. Watanabe, “Robust virtual inertia
control of an islanded microgrid considering high penetration of renewable
55% and 2% the minimum ESS capacity and by 86% and 36% energy,” IEEE Access, vol. 6, pp. 625–636, 2018.
the energy throughput of the ESS compared to a P or PI controller [19] G. Q. Zeng, X. Q. Xie, and M. R. Chen, “An adaptive model predictive load
respectively. The first point implies that the MPC can help to frequency control method for multi-area interconnected power systems
with photovoltaic generations,” Energies, vol. 10, 2017, Art. no. 1840.
diminish the ESS CAPEX required for frequency regulation. [20] T. T Nguyen, H. J. Yoo, and H. M. Kim, “Analyzing the impacts of system
The second point suggests that the MPC can decrease the ESS parameters on MPC-based frequency control for a stand-alone microgrid,”
aging and so increase the lifetime of the system. Energies, vol. 10, no. 4, pp. 417–433, Feb. 2017.
[21] T. T Nguyen, H. J. Yoo, and H. M. Kim, “Application of model predictive
As a future work, optimality of the MPC parameters values control to BESS for microgrid control,” Energies, vol. 8, pp. 8798–8813,
should be investigated. Indeed, the purpose of this work is to 2015.
find a MPC solution that provide better results than the optimal [22] A. Ali, B. Khan, C. A. Mehmood, Z. Ullah, S. M. Ali, and R. Ullah,
“Decentralized MPC based frequency control for smart grid,” in Proc.
P and PI controllers. Moreover, the MPC predicting capability IEEE Int. Conf. Energy Conservation Efficiency, Lahore, Pakistan, 2017,
could be further developed to assess the trade-off between the pp. 1–6.
MPC computational complexity and performance. [23] M. R. Chen, G. Q. Zeng, Y. X. Dai, K. D. Lu, and D. Q. Bi, “Fractional-
order model predictive frequency control of an islanded microgrid,” En-
ergies, vol. 12, pp. 84–104, 2019.
REFERENCES [24] T. Kerdphol, F. S. Rahman, Y. Mitani, K. Hongesombut, and K. Kufeoglu,
“Virtual inertia control-based model predictive control for microgrid
[1] EIA, “What is U.S. electricity generation by energy source?” frequency stabilization considering high renewable energy integration,”
[Online]. Available: https://www.eia.gov/tools/faqs/faq.php?id=427& Sustainability, vol. 9, no. 5, pp. 773–793, May 2017.
t=3. Accessed: Nov. 22, 2019. [25] T. Kerdphol, F. S. Rahman, and Y. Mitani, “Virtual inertia control applica-
[2] EIA, “How much of US carbon dioxide emissions are associated with tion to enhance frequency stability of interconnected power systems with
electricity generation?” [Online]. Available: https://www.eia.gov/tools/ high renewable energy penetration,” Energies, vol. 11, no. 4, pp. 981–996,
faqs/faq.php?id=77&t=11. Accessed: Aug. 2, 2019. Apr. 2018.
[3] EPA, “Greenhouse gases equivalencies calculator – calculations and [26] H. Bevrani, Robust Power System Frequency Control, 2nd ed. Gewerbe-
reference,” [Online]. Available: https://www.epa.gov/energy/greenhouse- strasse, Switzerland: Springer, 2014.
gases-equivalencies-calculator-calculations-and-references. Accessed: [27] D. J. Lee and L. Wang, “Small-signal stability analysis of an autonomous
Aug. 2, 2019. hybrid renewable energy power generation/ energy storage system Part I:
[4] T. Kerdphol, Y. Qudaih, and Y. Mitani, “Optimum battery energy storage Time-domain simulations,” IEEE Trans. Energy Convers., vol. 23, no. 3,
system using PSO considering dynamic demand response for microgrids,” pp. 1339–1347, Jul. 2013.
Int. J. Elect. Power Energy Syst., vol. 83, pp. 58–66, 2016. [28] H. Bevrani, F. Habibi, P. Babahajyani, M. Watanabe, and Y. Mitani, “In-
[5] M. Marzband, S. S. Ghazimirsaeid, H. Uppal, and T. Fernando, “A real telligent frequency control in an AC microgrid: Online PSO-based fuzzy
time evaluation of energy management systems for smart hybrid home tuning approach,” IEEE Trans. Smart. Grid, vol. 3, no. 4, pp. 1935–1944,
microgrids,” Elect. Power Syst. Res., vol. 143, pp. 624–633, 2017. Dec. 2012.

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.
SOCKEEL et al.: VIRTUAL INERTIA EMULATOR-BASED MODEL PREDICTIVE CONTROL 2939

[29] P. Kunder, Power System Stability and Control. New York, NY, USA: Behnaz Papari (Member, IEEE) received the Ph.D.
McGraw-Hill, 1994. degree in electrical engineering from Florida State
[30] O. Mikulas, “Quadratic programming algorithms for fast model-based University, Tallahassee, FL, USA. In December 2018,
predictive control, Chapter 2,” M.S. thesis, Czech Tech. Univ. Prague, she joined EPIC as an Assistant Professor of Practice
Prague, Czechia, May 2013. in Power Electronics and Engineering, and the Man-
[31] C. Schmid and L. T. Biegler, “Quadratic programming methods for reduced ager of the Duke Energy Smart Grid Lab. She has
Hessian SQP,” Comput. Chem. Eng., vol. 18, no. 9, pp. 817–832, 1994. expertise in power systems with an emphasis on mod-
[32] G. Suri and S. Onori, “A control-oriented cycle life model for hybrid eling, analysis, control, planning, and optimization.
electric vehicle lithium-ion batteries,” Energy, vol. 96, pp. 644–653, 2016. Her specialties are distributed controls and decision
[33] L. Serrao, S. Onori, G. Rizzoni, and Y. Guezennec, “A novel model-based analysis under uncertainty, applications on energy
algorithm for battery prognosis,” in Proc. 7th IFAC Symp. Fault Detection, system modeling, and Stochastic optimization. Her
Supervision, Saf. Tech. Process., 2009, vol. 42, no. 8, pp. 923–928. research interests include renewable energy sources, power and energy manage-
[34] M. Jafari, K. Khan, and L. Gauchia, “Deterministic models of li-ion battery ment, control of stand-alone and utility-interactive energy systems, distributed
aging: It is a matter of scale,” J. Energy Storage, vol. 20, pp. 67–77, 2018. control of smart grids, stochastic analysis, and real-time power distribution
[35] L. Zhang, X. Hu, Z. Wang, F. Sun, and D. G. Dorrell, “A review of super- system simulation and hardware-in-the-loop instantiation.
capacitor modeling, estimation, and applications: A control/ management
perspective,” Renew. Sustain. Energy Reviews, vol. 81, pp. 1868–1878,
2018.
[36] P. Kreczanik, P. Venet, A. Hijazi, and G. Clerc, “Study of supercapacitor
aging and lifetime estimation according to voltage, temperature, and RMS
current,” IEEE Trans. Ind. Electron., vol. 61, no. 9, pp. 4895–4902,
Sep. 2014.
[37] P. Haessig and H. B. Ahmed, “Energy storage control with aging limita- Michael Mazzola (Senior Member, IEEE) received
tion,” in Proc. IEEE Eindhoven Power. Tech., Jun. 29.–Jul. 2, 2015. the Ph.D. degree in electrical engineering from Old
[38] M. Faisal, M. A. Hannan, P. J. Ker, A. Hussain, M. B. Mansor, and Dominion University, Norfolk, VA, USA. He is the
F. Blaabjerg, “Review of energy storage system technologies in microgrid Director of the Energy Production and Infrastructure
applications: issues and challenges,” IEEE Access, May 28, 2018. Center (EPIC) and the Duke Energy Distinguished
[39] F. Ran, T. Remo, and R. Margolis, 2018 U.S. Utility-Scale Photovoltaics- Chair in Power Engineering Systems with UNC Char-
Plus-Energy Storage System Costs Benchmark. Golden, CO, U.S.: Na- lotte.
tional Renewable Energy Laboratory, NREL/TP-6A20-71714, 2018. After three years in government service with the
Naval Surface Warfare Center in Dahlgren, Virginia,
in 1993, he joined the faculty with Mississippi State
University, where he became known for his research
Nicolas Sockeel received the M.E. degree from the in the areas of silicon carbide power semiconductor device prototyping and
EPF Graduate School of Engineering, Montpellier, semiconductor materials growth and characterization. For ten years, he was
France, in 2015, and the Ph.D. degree in electrical and with the Mississippi State University Center for Advanced Vehicular Systems
computer engineering from Mississippi State Univer- as the Associate Director for advanced vehicle systems, where he lead research
sity, Starkville, MS, USA, in 2018. He is a French in high-voltage engineering, power systems modeling and simulation, the ap-
Postdoctoral Fellow with UNC Charlotte in EPIC. plication of silicon carbide semiconductor devices in power electronics, and
His main areas of focus are energy storage system the control of hybrid electric vehicle power trains. In addition, he served two
testing and modeling, model predictive control algo- years as the Chief Technology Officer of SemiSouth Laboratories, a company
rithm applied to hybrid electric vehicle powertrain, he cofounded.
and electric grid distribution systems.

James Gafford (Member, IEEE) received the mas-

ter’s degree in electrical engineering from Mississippi
State University, Starkville, MS, USA, and the bach-
elor’s degree in electrical engineering from Christian
Brothers University, Memphis, TN, USA. In 2018,
he joined EPIC as the Assistant Director of Research
and Special Projects. Prior to joining EPIC, he was a
Senior Research Engineer with the Advanced Vehicle
System Group, Mississippi State University’s Center
for Advanced Vehicular Systems.
He has more than 15 years of experience in uni-
versity research and development laboratories. His previous work includes the
development of advanced power electronics and vehicle autonomy. His current
research interests include wide bandgap semiconductor device applications in
power electronics, predictive control systems, as well as vehicle electrification
and the resultant impact on grid utility service.

Authorized licensed use limited to: Sultan Qaboos University. Downloaded on October 06,2024 at 12:35:29 UTC from IEEE Xplore. Restrictions apply.