2022 4th International Conference on Smart Power & Internet Energy Systems

Adaptive Virtual Inertia Control Based on Nonlinear

Model Predictive Control for Frequency Regulation

Weimin Zheng Ruixu Liu Yangqing Dan

Electric Power Economic Research College of Electrical Engineering Electric Power Economic Research
2022 4th International Conference on Smart Power & Internet Energy Systems (SPIES) | 978-1-6654-8957-7/22/$31.00 ©2022 IEEE | DOI: 10.1109/SPIES55999.2022.10082140

Institute of State Grid Zhejiang University Institute of State Grid

Zhejiang Electric Power Co., Ltd. Hangzhou, China Zhejiang Electric Power Co., Ltd.
[email protected] [email protected] [email protected]

Zhen Wang
College of Electrical Engineering
Zhejiang University
Hangzhou, China
[email protected]

Abstract—The increasing penetration of low-inertia inertia, because it aims to emulate the outer characteristics of
renewable energy generation has exacerbated power system the conventional SG to compensate inertia support and
frequency. To address this issue, this paper proposes an improve system frequency performance [1][2]. As of VIC, the
adaptive virtual inertia control (VIC) for frequency regulation virtual inertia term and the damping term respectively
of power system with high renewable energy penetration. The correspond to the rate of change of frequency (ROCOF) and
energy storage system (ESS) is used to emulate the inertial and frequency excursions to exhibit inertia and damping
damping response of a conventional synchronous generator. properties similar to SGs. Thus, RESs with the control strategy
The inertia coefficient and damping coefficient of VIC are
of VIC can improve system stability. In [3], M. F. M. Arani et
adjusted dynamically by nonlinear model predictive control
al. proposed virtual inertia control based on the rotating mass
(NMPC) to optimize the frequency control effect. In addition, an
augmentation model with augmented state variables is
of doubly fed induction generator (DFIG) and supercapacitor
presented to enhance the robustness of adaptive VIC by NMPC. to reduce short-term oscillations. In [4], a virtual inertia
The performances of the proposed adaptive virtual synchronous controller of the optimized power point tracking method for
generator(VSG) control and augmentation model are compared WTGs was developed to provide fast-frequency response and
with no VSG and fixed parameters VSG on MATLAB/Simulink reduce frequency fluctuations. The above methods use the
platform. It is found that the proposed control strategy is more momentum of wind turbine rotors to participate in frequency
efficient in frequency regulation as well as enhancement of the regulation, but they are usually at the expense of wind power
system robustness. curtailment, and even worse, the rotor speed recovery process
may cause the secondary drop in frequency after the inertial
Keywords—adaptive control, frequency control, nonlinear response. Whereas auxiliary devices such as energy storage
model predictive control (NMPC), virtual inertia control (VIC) systems (ESSs) offer an alternative to implement VIC. In [5],
G. Delille et al. studied the short-term frequency support by
I. INTRODUCTION the fast-acting storage with a large number of RESs and the
Due to the energy crisis and environmental pollution, there results show that the ESS, by acting as synthetic inertia, can
is intense attention on integration of power generation from reduce the impact in the case of a major generation outage.
renewable energy sources (RESs) into power systems, such as Reference [6] proposed an ESS unit connected to a DC link to
wind turbine generators (WTGs) and photovoltaic (PV), etc. implement the strategy of VIC and improve the angular rotor
However, intermittency and volatility of RESs have speed stability.
exacerbated system instability. Meanwhile, most WTGs/PVs The VIC-based control schemes mentioned in the above
have converter interface with interconnected power grid and literature have fixed structures/parameters in the frequency
use the maximum power point tracking (MPPT) control response process. However, as discussed in [7] and [8], design
design, which exhibit characteristics of few rotating masses of VIC parameters has significant influence on system
and decouple from the grid, resulting in their weaker performance, so the control with adaptive VIC parameters
frequency support compared with conventional synchronous according to operating conditions of the system may achieve
generators (SGs). Hence, as penetration of WTGs/PVs better frequency improvement. One way of dynamically
increases, system inertia is reduced and frequency regulation adjusting the inertia and damping coefficient is based on
capability decreases, which have a great impact on system piecewise functions, such as a bang-bang control scheme with
reliability and security. In view of these, there is an urgent the alternating virtual inertia according to its virtual angular
need for converter-based sources to actively assist in velocity in each phase of oscillation in [9], a self-adaptive
maintaining the frequency stability of the power system. inertia and damping combination control based on the sign of
Virtual inertia control (VIC) is a solution towards the product of frequency deviation and ROCOF in [10], and
overcoming the frequency stability issues caused by lack of automatic adjustment of inertia constant by fuzzy logic
according to the penetration level of RESs and system
This work is supported by the Science and Technology Project of State frequency deviations in [11] etc. Another way of dynamically
Grid Zhejiang Electric Power Co., Ltd (No. 5211JY21N001). adjusting VIC parameters is by optimal algorithms, for
Corresponding author: Zhen Wang ([email protected]). example, reference [12] used the linear quadratic regulator to

978-1-6654-8957-7/22/$31.00 ©2022 IEEE 1036

Authorized licensed use limited to: Universidad Nacional Autonoma De Mexico (UNAM). Downloaded on September 23,2024 at 18:54:46 UTC from IEEE Xplore. Restrictions apply.
 adaptively adjust the VIC parameters according to the constant. And it adopts VIC control strategy. Based on the
frequency disturbance. In recent years, to improve stability, measured frequency deviation Δf, the VIC output power ΔPVIC
numerous attempts have been made to apply MPC to the is calculated and used as ESS output power reference
operation and control of power system such as in [13] and [14]. (variation) ΔPESS, ref. The VIC output power is defined as
MPC can be considered as a control framework that uses the
system model to predict the future behavior of the system and d f
PVIC = J ESS + DESS f (2)
focuses on the receding-horizon optimal control problem [15]. dt
Based on the appropriate cost function and constraints, it has
where JESS is the virtual inertia coefficient of the ESS, DESS is
the ability to optimize the future performance of the system.
the virtual damping coefficient of the ESS. In this paper, these
However, there is no comprehensive study of the optimal
two system parameters are time-varying based on the
design of inertia and damping coefficients with MPC in the
proposed controller explained in Section III.
above literature and most of them do not strictly explain the
influence of adaptive VIC parameters on system stability. To In addition, the state of charge (SOC) which can imply the
deal with the issue of frequency stability, this paper proposes rest capacity of ESS is used to simulate the real-time response
the adaptive VIC based on nonlinear model predictive control of the ESS, shown in (3).
(NMPC) for frequency regulation of the microgrid with high
renewable energy sources penetration.  SOC (tk −1 ) + PESS (tk )  t c / CESS ,

The remainder of this paper is organized as follows.  ESS is charging
SOC (tk ) =  (3)
Section II describes the mathematical model of microgrids.  SOC (tk −1 ) + PESS (tk )  t / (d  CESS ),
Section III is devoted to adaptive VIC control design,  ESS is discharging
including the proposed control scheme and augmentation 
model. Section IV presents and discusses the simulation where CESS is the ESS nominal capacity, Δt is the sampling
results. Finally, the conclusion is given in Section V. interval, PESS (tk) is the ESS output power at time tk, ηc and ηd
are the charging and discharging efficiency respectively.
II. SYSTEM DESCRIPTION AND MODELING
A. System Structure
PG
Fig. 1 shows the structure of a small-scale microgrid. This Synchronous Generators
system is composed of SGs , an ESS, a RES station, and some
loads. The SGs and ESS are the major sources of primary PESS +
+ 1 f
frequency regulation. It is noted that the RES station does not Energy Storage System  Dsys + sJ sys
+
participate in system frequency regulation in this paper. PD
PW + - PL
Wind Power System  Loads
Disturbances

Fig. 2. System frequency dynamics model.

f + PVIC PESS.ref 1 PESS

DESS (t ) 
1 + sTESS
+ Limiter
Deadband
d
J ESS (t )
dt

Fig. 3. ESS model with VIC control.

Fig. 1. A typical structure of microgrid
During the control process, there are several operation
B. System Frequency Dynamics Model rules, including the SOC limits of ESS as (4), the output
active power limits of ESS as (5). Moreover, the deadband
The system frequency dynamics model is shown in Fig. 2.
module with deadband width Dw is applied to avoid frequent
Thus the dynamic relationship between the changes of active
switching of charge/discharge states caused by small
power and frequency can be expressed by the system swing
frequency fluctuations.
equation as (1).
SOCmin  SOC(t )  SOCmax (4)
d f
PG + PESS +PD = J sys + Dsys f (1)
dt
where SOCmin and SOCmax are respectively the ESS lower and
where Δf=f−f0 is the frequency deviation of the system, ΔPD upper limits for SOC.
is system disturbances which may come from load PESS.min  PESS (t )  PESS.max (5)
disturbances ΔPL and wind power disturbances ΔPW, ΔPG is
the SG power change, ΔPESS is the power change of the ESS,
where PESS.min and PESS.max are respectively the ESS lower and
Jsys is the system inertia and Dsys is the load damping
upper limits for the output active power.
coefficient.
In addition, when the frequency deviation exceeds Δflimit,
C. ESS Model
in order to ensure the stability of the system, the ESS will
The model and control of ESS is shown in Fig. 3. To charge/discharge with the maximum rated power to provide
analyze frequency control, the ESS is simply represented by frequency support for the system.
the first-order transfer function, where TESS is the ESS time

1037
Authorized licensed use limited to: Universidad Nacional Autonoma De Mexico (UNAM). Downloaded on September 23,2024 at 18:54:46 UTC from IEEE Xplore. Restrictions apply.
 III. ADAPTIVE VIRTUAL INERTIA CONTROL the disturbances in lower frequency range are compensated
by the slow response of automatic generation control (AGC)
This paper proposes an adaptive VIC based on NMPC
and the disturbances in higher frequency range can be
framework to assist system frequency regulation. By solving
absorbed by the rotor inertia of the generator [19]. Thus,
the optimization problem with a moving time horizon
though the first-order low-pass or high-pass filters have the
window, it can automatically adjust the parameters of VIC
simpler structure, they are not applied in this paper for the
(inertia coefficient and damping coefficient) according to the
economical consideration to avoid unnecessary output of the
system frequency changes. Fig. 4 illustrates the microgrid and
ESS.
the proposed adaptive VIC controller. The proposed adaptive
VIC controller mainly consists of three parts: 1) the system Based on the above, an augmentation system model is
model; 2) the cost function and relevant constraints; 3) the constructed, shown in Fig. 5. Because of the second-order
optimization algorithm. function, the system state variables have increased by two. In
addition, the three items, frequency deviation, ROCOF, and
A. Nonlinear System Model
frequency deviation after filtering are used as the new outputs
1) Normal model of the system yA=[y1, y2, yF ]T. Combining (6) and (7), the
Combined with the content of the Section II, the system augmentation system can be expressed by (8).
can be expressed by the following state space function (6).
State variables are x=[Δf, ΔPESS, ΔPGV, ΔPCH, ΔPRH1, ΔPRH2, x A = f A ( x A , u p , w)
ΔPCO]T, output variables are y=[y1, y2]T=[Δf, dΔf/dt]T, and (8)
y A = g A ( x A , u p , w)
w=ΔPD is the uncontrollable external input disturbance. The
states about the SG such as ΔPGV, ΔPCH, ΔPRH1, ΔPRH2, and where subscript A represents the variables of augmentation
ΔPCO are gotten by the external state estimator which is model, fA and gA are respectively the augmentation model
beyond the scope of this paper. It is noted that the system is a functions.
nonlinear system rather than a linear time-varying system
because the adaptive parameters JESS (t) and DESS (t) are time- yF
varying depending on the system state x and input w
WF (s)
disturbance w and this artificially introduces nonlinear parts y1 yA
into the system. up=[JESS (t), DESS (t)]T is defined as the vector Normal Model
up
of controllable parameters (the inertia and damping y2
coefficient) and is transmitted to the system to change the
corresponding parameters after online calculation. For
Augmentation Model
simplicity, the nonlinear system is expressed as follows:
Fig. 5. Structure of augmentation system model.
x = f ( x, u p , w )
(6)
y = g ( x, u p , w ) B. Nonlinear Model Predictive Control Modelling
1) Nonlinear model predictive control
Past Future
Reference

Predicted Output y
Real Output y
Predicted Optimal Input u

Real Input u
Prediction Horizon

Control Horizon

t t + t t + Tc t +Tp

Fig. 6. Schematic diagram of NMPC.

Fig. 4. Overview of the microgrid and the proposed adaptive VIC controller.
TABLE I. PARAMETERS OF THE PROPOSED NMPC CONTROLLER
2) Augmentation model with weighting function
The input to the system may include various disturbances. Parameter Value
Therefore, as the outputs of the system, the system frequency
deviations will also have various frequency components. Sampling Interval Δt (s) 0.5
Though the grid disturbance signal is quite different in time Control Horizon NC (s) 10
domain, it is similar in the harmonic content [18]. Inspired by Prediction Horizon NP (s) 10
this, the weighting function WF (s) similar to the concept of
band-pass filter is considered being applied to decompose the A schematic diagram of NMPC is shown as Fig. 6. It’s
system outputs to reduce frequency disturbances and enhance well-known that NMPC takes account of the system future
the robustness of system, and it is shown as follows: output over a prediction horizon TP=NPΔt to determine the
optimal control for the system performance. In this paper, by
H s the current state of the system, an optimization problem is
WF ( s ) = (7)
s 2 + (L + H ) s + L H solved to investigate the optimal inertia coefficient and
damping coefficient of VIC and a sequence of optimal values
where ωL and ωH are respectively the lower cut-off frequency can be obtained at each time interval of a control horizon
and upper cut-off frequency of the band-pass filter. Generally, TC=NCΔt, but only the first value of the sequence is input to

1038
Authorized licensed use limited to: Universidad Nacional Autonoma De Mexico (UNAM). Downloaded on September 23,2024 at 18:54:46 UTC from IEEE Xplore. Restrictions apply.
 the system as control parameters. The parameters of the VN P ( f ( x (n), u p (n))) − VN P −1 ( f ( x (n), u p (n)))
proposed NMPC in this paper is shown in Table I.
 (1 −  )  l ( x (n), u p (n))
2) Optimization problem (15)
The intention of this paper is to achieve a dynamically
optimal control for frequency response of the microgrid. Thus, where   (0,1) is a constant.
a key issue is to reduce the frequency deviations during
frequency events. Besides, for the sake of economy and IV. SIMULATION STUDIES
stability, the ESS power flow should also be as small as In order to verify the effectiveness of the proposed
possible. In view of the above, the control objective is adaptive VIC control strategy and augmentation model,
twofold. On the one hand, for frequency response, the control several simulations are carried out on MATLAB/Simulink
errors of frequency deviation and ROCOF are essential terms platform. The control loop of the SG’s speed-governing
of the control objective. On the other hand, the magnitude of system is given in Fig. 7. The penetration rates of RESs and
inertia coefficient and damping coefficient value needs to be ESS are 20% and 20% respectively. The main parameters in
considered as well. Therefore, in light of these objectives, the this study are given in Table II.
optimal inertia and damping coefficient u p can be obtained
f 1 + sT2 + 1 1 PGV
through (9), RG  (1 + sT1 )

T3 s
-
Rate
uP = arg min J ( x, u p , w) (9) Limiter
Limiter

with (a) Speed-governing system model.

N P −1 PGV 1 PCH 1 PRH 1 1 PRH 2 1 PCO
J =  l ( x, u p ) (10) 1 + sTCH 1 + sTRH 1 1 + sTRH 2 1 + sTCO
n =0 FV HP FHP FIP FLP

l ( x, u p ) = eQe + u p Ru
T T
p (11) + + + P
G
  
+ + +
where J is the cost function, NP is the finite prediction horizon. (b) Steam system model
e is the tracking error of system output and all of the output Fig. 7. The control model of the SG’s speed-governing system
references are set to zero. Q and R are the positive definite
weighting factors matrices to determine the importance of TABLE II. SYSTEM PARAMETERS
each term and guarantee the balance between the frequency
regulation and the ESS power flow. Model Parameter Value Parameter Value

In addition, the solution process of (9) should also satisfy System Model Jsys 8 Dsys 2
some constraints including the box constraints (4) and (5) in T1 (s) 0 T2 (s) 0
Section II.C, numerical size constraints of up as (12), rate of T3 (s) 0.025 RG 0.05
Synchronous TCO (s) 0.3 TCH (s) 0.1
change constraints of up as (13). And there is also an
Generator TRH1 (s) 10 TRH2 (s) 10
inequality constraint as (15) about the system stability,
FVHP 0.22 FHP 0.22
illustrated in Section III.C. FIP 0.3 FLP 0.26
TESS (s) 0.5 SOC0 0.5
u p .min  u p  u p.max (12)
SOCmin 0.05 SOCmax 0.95
Pmin
ESS (p.u.) -1 Pmax
ESS (p.u.) 1
du p du p du p ESS
( )min  ( )max (13) ηc 0.95 ηd 0.95
dt dt dt Dw (Hz) 0.02 Δflimit (Hz) 0.2
CESS (MW•h) 1
where the subscripts min and max respectively represent the
minimum and maximum limits. Firstly, several indexes are proposed to comprehensively
The proposed optimization problem is solved with the quantify the effect of different control strategies from
sequential quadratic program (SQP) algorithm. different perspectives. Then the contrastive simulation of the
two scenarios (load step and wind power fluctuations) under
C. Stability Constraint four control strategies is conducted for testing the system
In this section, we focus on analyzing the system stability frequency response. The four control strategies are conducted
with the proposed adaptive VSG through Lyapunov stability for comparison: 1) no VIC, which means the conventional
theorem. control without VIC; 2) fixed parameters VIC (FP-VIC); 3)
adaptive VIC without augmentation model (A-VIC) and 4)
The cost function is used as the energy function VN (x), adaptive VIC with augmentation model (AA-VIC). For FP-
shown as (14). VIC, set the virtual inertia coefficient and damping
N P −1 coefficient to the average values of the A-VIC results.
VN P ( x ) =  l ( x, u
n=0
p ) (14) A. Indexes
In order to comprehensively quantify the effect of
If the control can meet the inequality constraint (15), the different control strategies, this paper uses the following four
system under the feedback law u p is Lyapunov evaluation indexes.
asymptotically stable, which is proved in [20] and [21].

1039
Authorized licensed use limited to: Universidad Nacional Autonoma De Mexico (UNAM). Downloaded on September 23,2024 at 18:54:46 UTC from IEEE Xplore. Restrictions apply.
 1) Range of frequency deviation RΔf : This index as (16) TABLE III. PERFORMANCE COMPARISON UNDER SCENARIO I
can reflect the difference between the maximum and Type RΔf S Δf RSOC δSOC
minimum frequency deviation. No VIC 1 1 0 0
FP-VIC 0.7256 0.6613 0.5451 0.9768
Rf = f max − f min (16) A-VIC 06310 0.6057 0.6189 1
AA-VIC 0.5047 0.4141 1 0.9710
where Δfmax and Δfmin are respectively the maximum and
minimum frequency deviation. The results are given in Fig. 8 (a)-(c). Fig. 8 (a) depicts
the system frequency response among all four strategies. And
2) Standard deviation of frequency deviation SΔf : This variations of the virtual inertia and the damping coefficient
index as (17) is used to measure the degree of dispersion of are illustrated in Fig. 8 (b) and (c). It can be seen that the ESS
frequency deviation, indicating that the effect of smoothing adopting VIC control strategy can effectively respond to
frequency fluctuations. disturbances caused by load-step change. Moreover, facing
the same unforeseen load-step change, the frequency nadir of
1 N the system with AA-VIC is smaller, which means it has better
Sf =  (f (n) − f )2
N n =1
(17)
effect of frequency regulation. The inertia and damping
coefficients values of AA-VIC are larger, as indicated in Fig.
where  f is the average value of frequency deviation during 8 (b) and (c). This is understandable, because for the primary
frequency regulation. frequency regulation, control strategies such as droop control
and virtual inertia control essentially support the system
3) Range of SOC RSOC: This index can reflect the frequency by providing additional active power. Table III
difference between the maximum and minimum SOC. summarizes the four indexes of simulation results. Though
the range of SOC of AA-VIC is the largest as the ESS with
RSOC = SOCmax − SOCmin (18) AA-VIC provides more active power to support frequency,
the system with AA-VIC has the best performance in other
where SOCmax and SOCmin are respectively the maximum and
three indexes. It has smaller range of frequency deviation and
minimum SOC during the frequency regulation.
standard deviation of frequency deviation, which means that
4) Relative Variation of SOC δSOC: In order to evaluate it has better frequency regulation effect. Besides, the smaller
how much energy of the VIC is used to reduce the frequency relative variation of SOC indicates that the less energy is used
deviation, the index, relative variation of SOC, is defined as: to reduce the frequency deviation per herz, so it can be found
that AA-VIC is more efficient in frequency regulation. It
SOC − SOC0 should be noted that the system frequency cannot return to
 SOC = (19) the nominal value in a short time, because this will be
f − f 0
achieved through secondary frequency control, which is
beyond the scope of primary frequency control.
where SOC is the average value of SOC during frequency
regulation with VIC, SOC0 is the average value of SOC E. Scenario II: Wind Power Fluctuations
For this scenario, the microgrid operations are disturbed
without VIC, f 0 is the average value of frequency deviation
by the wind power fluctuations which are randomly generated
without VIC. The numerator and denominator respectively according to the general wind speed variation law. This
reflect the additional change of SOC and the reduction of scenario is set to verify the frequency regulation effect and
frequency deviation under different VIC control strategies robustness of the proposed control strategy under the
compared with no VIC control. The smaller the value of δSOC, disturbances.
the less additional output energy of the ESS required to
Frequency Deviation (Hz) Virtual Inertia Coefficient J
improve each hertz frequency deviation, which indicates the 0.10
No VIC
40 No VIC FP-VIC A-VIC AA-VIC

control strategy is more efficient and economical in FP-VIC

A-VIC
30
20
frequency regulation. 0.05 AA-VIC 10
0

D. Scenario I: Load Step 0.00

(b)
Time (s)
Damping Coefficient D
50
In this case, a load increase of 0.02 p.u. is happened at 40
t=5s on the microgrid. The same load change is imposed for -0.05 30
20
the above four strategies. -0.10 Time (s)
10
0
0 10 20 30 40 0 10 20 30 40
Frequency Deviation (Hz) Virtual Inertia Coefficient J
0.05 30 (a) (c)
No VIC 25 No VIC FP-VIC A-VIC AA-VIC
FP-VIC 20
0.00 A-VIC 15 Fig. 9. The system responses under scenario II: (a) frequency response; (b)
AA-VIC 10
5 Time (s) changes of the VIC coefficients J; (c) changes of the VIC coefficients D.
0
-0.05 -5
(b)
TABLE IV. PERFORMANCE COMPARISON UNDER SCENARIO II
Damping Coefficient D
-0.10 50
40 Type RΔf S Δf RSOC δSOC
30
-0.15 20 No VIC 1 1 0 0
10 FP-VIC 0.7939 0.7662 0.4362 1
Time (s) Time (s)
-0.20 0
0 10 20 30 40 0 10 20 30 40 A-VIC 0.7691 0.7354 0.7777 0.8170
(a) (c) AA-VIC 0.7481 0.7043 1 0.6295

Fig. 8. The system responses under scenario I: (a) frequency response; (b) Fig. 9 (a) shows the system frequency response results
changes of the VIC coefficients J; (c) changes of the VIC coefficients D.
under different control strategies. It can be seen that,

1040
Authorized licensed use limited to: Universidad Nacional Autonoma De Mexico (UNAM). Downloaded on September 23,2024 at 18:54:46 UTC from IEEE Xplore. Restrictions apply.
 compared with no VIC, the ESS adopting VIC control generation on isolated power system's inertia," IEEE Transactions on
strategy can effectively respond to disturbances caused by Sustainable Energy, vol. 3, no. 4, pp. 931-939, 2012.
wind power fluctuations. Throughout the frequency [6] M. P. N. v. Wesenbeeck, S. W. H. d. Haan, P. Varela, and K. Visscher,
"Grid tied converter with virtual kinetic storage," in 2009 IEEE
regulation process, in most cases, the VIC with adaptive Bucharest PowerTech, 2009, pp. 1-7.
parameter performs better, and the frequency regulation [7] S. D’Arco, J. A. Suul, and O. B. Fosso, "A virtual synchronous machine
effect of AA-VIC is better than A-VIC. Thus the system with implementation for distributed control of power converters in
AA-VIC shows stronger anti-disturbances ability and smartgrids," Electric Power Systems Research, vol. 122, pp. 180-197,
robustness. Fig. 9 (b) and (c) show the simulation results for 2015.
the variations of the virtual inertia and damping coefficient. [8] F. Wang, L. Zhang, X. Feng, and H. Guo, "An adaptive control strategy
Certainly, it has larger inertia and coefficients to provide for virtual synchronous generator," IEEE Transactions on Industry
Applications, vol. 54, no. 5, pp. 5124-5133, 2018.
more active power. Table IV summarizes the four indexes of
simulation results. It can be found that AA-VIC has the best [9] J. Alipoor, Y. Miura, and T. Ise, "Power system stabilization using
virtual synchronous generator with alternating moment of inertia,"
frequency response and it is more efficient in frequency IEEE Journal of Emerging and Selected Topics in Power Electronics,
regulation as well as enhancement of the system robustness. vol. 3, no. 2, pp. 451-458, 2015.
[10] D. Li, Q. Zhu, S. Lin, and X. Y. Bian, "A self-adaptive inertia and
V. CONCLUSION damping combination control of VIC to support frequency stability,"
IEEE Transactions on Energy Conversion, vol. 32, no. 1, pp. 397-398,
With the increasing penetration of RESs, the inertia level 2017.
of system decreases and the frequency stability problem
[11] T. Kerdphol, M. Watanabe, K. Hongesombut, and Y. Mitani, "Self-
becomes more and more serious. This paper introduces a adaptive virtual inertia control-based fuzzy logic to improve frequency
concept of adaptive VIC for the microgrid with high RES stability of microgrid with high renewable penetration," IEEE Access,
penetration. Based on NMPC, a novel method for vol. 7, pp. 76071-76083, 2019.
dynamically adjusting the inertia and damping coefficient of [12] U. Markovic, Z. Chu, P. Aristidou, and G. Hug, "LQR-based adaptive
VIC is proposed to improve the frequency support capability. virtual synchronous machine for power systems with high inverter
The control objectives involve the frequency deviation, penetration," IEEE Transactions on Sustainable Energy, vol. 10, no. 3,
pp. 1501-1512, 2019.
ROCOF and the magnitude of inertia and damping
[13] P. Kou, D. Liang, L. Yu, and L. Gao, "Nonlinear model predictive
coefficients. In addition, we built an augmentation model control of wind farm for system frequency support," IEEE Transactions
with the weighting function similar to a band-pass filter to on Power Systems, vol. 34, no. 5, pp. 3547-3561, 2019.
enhance system robustness. [14] A. M. Dizqah, A. Maheri, K. Busawon, and A. Kamjoo, "A
multivariable optimal energy management strategy for standalone DC
The performances of four control strategies, no VIC, FP- microgrids," IEEE Transactions on Power Systems, vol. 30, no. 5, pp.
VIC, A-VIC, and AA-VIC, were evaluated for scenarios of 2278-2287, 2015.
load step and wind power fluctuations. It was found that the [15] L. Wang, Model predictive control system design and implementation
system with adaptive VIC has better performance than using MATLAB. Berlin, Germany: Springer, 2009.
conventional control strategies. Moreover, the AA-VIC is [16] I. C. Report, Dynamic Models for Steam and Hydro Turbines in Power
more efficient in frequency regulation as well as System Studies, Nov. 1973. [Online] Available:
enhancement of system robustness as it used less energy to https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4075292,
Accessed on: Jan. 12, 2020.
reduce the frequency deviation per herz.
[17] P. Pourbeik, Dynamic models for turbine-governors in power system
REFERENCES studies, IEEE Task Force on Turbine-Governor Modeling, Jan. 2013,
Tech. Rep. PES-TR1. [Online] Available: https://site.ieee.org/fw-
[1] R. Liu, Z. Wang, and H. Xing, "Virtual inertia control strategy for pes/files/2013/01/PES_TR1.pdf, Accessed on: Jan. 12, 2022.
battery energy storage system in wind farm," in 2019 IEEE PES Asia- [18] F. Arrigo, E. Bompard, M. Merlo, and F. Milano, "Assessment of
Pacific Power and Energy Engineering Conference (APPEEC), 2019, primary frequency control through battery energy storage systems,"
pp. 1-5. International Journal of Electrical Power & Energy Systems, vol. 115,
[2] Q.-C. Zhong and G. Weiss, "Synchronverters: Inverters that mimic 2020.
synchronous generators," IEEE Transactions on Industrial Electronics, [19] W. Li, G. Joos, and C. Abbey, "Attenuation of wind power fluctuations
vol. 58, no. 4, pp. 1259-1267, 2011. in wind turbine generators using a DC bus capacitor based filtering
[3] M. F. M. Arani and E. F. El-Saadany, "Implementing virtual inertia in control scheme," in Conference Record of the 2006 IEEE Industry
DFIG-based wind power generation," IEEE Transactions on Power Applications Conference Forty-First IAS Annual Meeting, 2006, vol.
Systems, vol. 28, no. 2, pp. 1373-1384, 2013. 1, pp. 216-221.
[4] D. Ochoa and S. Martinez, "Fast-frequency response provided by [20] L. Grune and A. Rantzer, "On the Infinite Horizon Performance of
DFIG-wind turbines and its impact on the grid," IEEE Transactions on Receding Horizon Controllers," IEEE Transactions on Automatic
Power Systems, vol. 32, no. 5, pp. 4002-4011, 2017. Control, vol. 53, no. 9, pp. 2100-2111, 2008.
[5] G. Delille, B. Francois, and G. Malarange, "Dynamic frequency control [21] P. Giselsson and A. Rantzer, "Distributed model predictive control with
support by energy storage to reduce the impact of wind and solar suboptimality and stability guarantees," in 49th IEEE Conference on
Decision and Control (CDC), 2010, pp. 7272-7277.

1041
Authorized licensed use limited to: Universidad Nacional Autonoma De Mexico (UNAM). Downloaded on September 23,2024 at 18:54:46 UTC from IEEE Xplore. Restrictions apply.