2494 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 21, NO.

3, JULY 2024

A Multivariable, Adaptive, Robust, Primary Control

Enforcing Predetermined Dynamics of Interest in
Islanded Microgrids Based on Grid-Forming
Inverter-Based Resources
Amir Afshari , Masoud Davari , Senior Member, IEEE, Mehdi Karrari , Senior Member, IEEE,
Weinan Gao , Senior Member, IEEE, and Frede Blaabjerg , Fellow, IEEE

Abstract— This paper proposes a multivariable, adaptive, scheme introduced for the closed-loop stabilization and voltage
robust (MAR) control strategy for islanded inverter-based regulation in the presence of disturbances and a significant
resources (IBRs) operating as grid-forming inverters. The pro- amount of uncertainty under various case studies; moreover,
posed method is employed in the inner control loop of the comparative simulations by comparing the presented method
primary layer in the hierarchical or decentralized structures with other studies using sliding mode control are provided.
for the islanded operation of microgrids. The MAR control Finally, experiments verify the effectiveness and practicality of
scheme is responsible for stabilizing IBRs’ output voltage in the proposed MAR control scheme.
autonomous operations of microgrids, considering mismatched
input voltage disturbances from the grid side and a large amount Note to Practitioners—Inverter-based resources are integral
of system uncertainty. The control methodology introduced in this parts of current and especially future power and energy systems;
paper does not rely on the system’s physical parameters, such with increasing concerns about carbon footprints, the tendency to
as microgrid topology, load dynamics, LCL filters, and output substitute traditional synchronous generators with inverter-based
connectors. As a result, there is no need to know the nominal resources increases. This transition towards the widespread use
values or the bounds of uncertainties in system dynamics. of power electronics devices needs careful studies regarding
The MAR control method uses online adaptation rules first to the stability and control of power converters. Although existing
identify and then adjust the control parameters of the closed-loop studies are addressing potential control system challenges, they
system based on an arbitrary dynamic model. In other words, suffer from complex mathematical computations and the need
the MAR method replaces the actual dynamics of IBRs with for the system’s preliminary information. With this in mind, this
predetermined dynamics of interest. Simulation results in the study proposes a multivariable, adaptive, robust control strategy
MATLAB/Simulink environment confirm the capability of the for the inner voltage control loop of grid-forming inverters.
This method utilizes online estimation algorithms to identify
inverter-based resources’ parameters and tune control system
Manuscript received 11 August 2022; revised 20 January 2023; accepted parameters simultaneously, making it applicable even to cases
10 March 2023. Date of publication 4 May 2023; date of current version
with slow parameter variations caused by aging or environmental
8 August 2024. This article was recommended for publication by Associate
Editor M. Robba and Editor Q. Zhao upon evaluation of the reviewers’ changes. It can compensate for potential voltage disturbances
comments. The work of Masoud Davari was supported in part by the U.S.– from the grid side and enable the designer to replace undesirable
Denmark Program funded by the Office of International Science and Engi- dynamics of inverter-based resource units with arbitrary and
neering (OISE) through the U.S. National Science Foundation (NSF) through stable dynamics of interest. In fact, unlike traditional methods
the NSF-OISE-IRES (International Research Experiences for Students) under that need control parameters and gains to be tuned to achieve
Award 2152905, in part by the U.S. NSF Electrical, Communications and a proper dynamic response, the control system designer can
Cyber Systems-Energy, Power, Control, and Networks (ECCS-EPCN) under choose reference dynamics and enforce the closed-loop system
Award 1808279 and Award 1902787, in part by the dSPACE company, to imitate the dynamical model selected. This model is usually
in part by the Verivolt company, in part by the Discovery and Innovation
chosen based on established priorities, such as response time and
Award from the 2020–2021 University Awards of Excellence at Georgia
Southern University, and in part by the Impact Area Accelerator Grant funded other transient behaviors. Moreover, this method does not require
by Georgia Southern University—at which all experiments were conducted. complex mathematical and algebraic calculations to design and
(Corresponding authors: Masoud Davari; Weinan Gao.) implement. It can be easily applied to inverter-based resource
Amir Afshari is with the Department of Electrical and Computer Engineer- units after selecting the desired reference dynamics, as shown
ing, University of Wisconsin–Madison, Madison, WI 53706 USA (e-mail: through this study’s experimental result. The above points give
[email protected]). this method a competitive edge over the existing algorithms,
Masoud Davari is with the Department of Electrical and Computer Engineer- especially in practical applications.
ing, Georgia Southern University, Statesboro Campus, Statesboro, GA 30460
USA (e-mail: [email protected]; [email protected]). Index Terms— Grid-forming inverter-based resources (GFM
Mehdi Karrari is with the Department of Electrical Engineering, Amirkabir IBRs), multivariable, adaptive, robust (MAR) control, online
University of Technology (Tehran Polytechnic), Tehran 1591634311, Iran adaptation rule, online estimation, predetermined dynamics of
(e-mail: [email protected]). interest for islanded IBRs, primary control of islanded IBRs,
Weinan Gao is with the State Key Laboratory of Synthetical Automation voltage-sourced converters (VSCs).
for Process Industries, Northeastern University, Shenyang, Liaoning 110819,
China (e-mail: [email protected]).
Frede Blaabjerg is with the AAU Energy, Aalborg University, 9220 Aalborg, N OMENCLATURE
Denmark (e-mail: [email protected]). A. IBR’s Variables
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TASE.2023.3262852. Vodq , Iodq IBR’s output voltage and current in the dq
Digital Object Identifier 10.1109/TASE.2023.3262852 frame.
1545-5955 © 2023 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: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 AFSHARI et al.: MAR, PRIMARY CONTROL ENFORCING PREDETERMINED DYNAMICS OF INTEREST IN ISLANDED MICROGRIDS 2495

I Ldq IBR’s generated current in the dq frame. In order to address these issues in the islanded mode or even
Vidq Terminal voltage (control input) in the dq grid-connected operations, various studies have been done
frame. in the literature [14]. For the case of grid-connected mode,
Vbdq Grid-side voltage (disturbance signal) in a robust control algorithm is developed for unbalanced-grid
the dq frame. conditions in [15]. A cascaded PI-based control strategy is
VDC DC-side voltage. employed to regulate the output voltage of voltage-sourced
IDC , IC Controllable dc-source current and dc cur- converters (VSCs) in [16]. A robust H∞ controller is proposed
rent delivered to IBRs. in [17] to reject load currents as disturbances—it deploys
the stationary reference frame to design the control algo-
rithm. In [18], a cascaded structure including sliding mode
B. IBR’s Parameters
and mixed H2 /H∞ controllers is designed for the primary
R f , RDC Output filter and dc-side resistances.
layer of ac microgrids, considering external load currents
Lf Output filter inductance.
as disturbances. The methods presented in [19] and [20]
C f , CDC Output filter and dc-side capacitances.
solve the voltage stability problem of islanded microgrids
ω Grid nominal frequency.
using decentralized robust state and output feedback control
Rc Output connector resistance.
algorithms, respectively. In [21], a decentralized sliding mode
Lc Output connector inductance.
controller is designed for both grid-connected and islanded
operation of microgrids considering nonlinear and unbalanced
C. Control System Parameters load conditions. The authors in [22] have proposed a robust
G(s), G m (s) System and reference dynamics. decoupling pre-compensator for the autonomous operation
χm (s) Left interactor matrix associated with sys- of microgrids. This control structure only deals with norm
tem dynamics. bounded uncertainties and depends on the system’s structure
K H , 0(t) High frequency gain matrix and its and topology. Voltage stabilization of islanded microgrids,
estimate. considering plug-and-play functionality and robustness against
r (t), ym (t) Reference input and output signals. microgrid topology changes, is addressed in [23]. A model
K 1∗ , K 2∗ , K 3∗ Nominal feedback gains. predictive control algorithm for the voltage regulation of
η System relative degree. islanded ac microgrids is introduced in [24]. Considering
∗
g(t), 9̃ , 9 ∗ Disturbance parametrization terms. uncertainties and load disturbances, a decentralized sliding
γ Estimate of nominal control parameters. mode controller for the islanded operation of microgrids is
61 , 62 Adaptation gains. developed in [25]—it uses a second-order sliding mode control
e(t), δ(t) Estimation errors. strategy based on a sub-optimal algorithm. It is noteworthy that
τDC DC-side time constant. the design of this control algorithm requires the structure and
KP DC-side proportional controller gain. physical topology of the microgrid to be “known.” The authors
in [26] investigated the robust performance of a microgrid
using a sliding mode control algorithm in both grid-connected
I. I NTRODUCTION and islanded operations.

R ECENT advances in distributed energy resources, renew-

ables, information theory, and communication tech-
nologies have led to significant utilization of microgrids
In this regard, some studies use adaptive algorithms for
controlling VSCs in microgrids. For example, an adaptive
algorithm for the current controller of grid-connected VSCs
[1], [2]—thereby integrating various inverter-based resources is developed in [27]—it utilizes indirect adaptive algorithms
(IBRs) [3]. Traditionally, the hierarchical control structure has based on identifying system parameters, which are needed
been applied to the control system design of microgrids, even to implement the proposed algorithm. Thus, the persistent
modern ones. This structure consists of the primary, secondary, excitation condition must be met, so convergence problems
and tertiary control layers; see [4], [5], [6], [7], and [8]. are inevitable in this method. The authors in [28] addressed
Using this structure, and due to the flexibility and scalability voltage stability in islanded microgrids using an adaptive
of microgrids, they can operate in both grid-connected and voltage controller. This control algorithm is solely responsible
islanded modes [9]. In the islanded operation, voltage and for a specific bound of systems uncertainty, so knowing the
frequency stability is the primary objective of the microgrid nominal values of system parameters is a must. Addition-
control systems, which are realizable by employing GFM ally, the existence of external disturbances is ignored in this
inverters and different types of GFM controllers operating in method.
the primary control layer [10], [11], [12]. The primary layer is Most studies mentioned above have investigated the robust
utilized to implement real-time voltage and frequency control stability and control of microgrids under different conditions
algorithms [13]. Because of the volatile and uncertain nature using robust control algorithms. But there are still some lim-
of the renewables, disturbances, and systems’ uncertainties, itations in these algorithms that should be addressed. Control
it is necessary to have a properly designed advanced control systems operate in various unknown operational environments
system for IBRs’ operating as GFM units to have a smooth with external disturbances and uncertainties. Additionally,
integration of IBRs into microgrids and a stable power supply component aging and even system failures can change system
for consumers. parameters. Thus, in practice, system parameters are expected

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 2496 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 21, NO. 3, JULY 2024

to change over time. In this regard, robust control algorithms 1) A multivariable, adaptive, robust control strategy is pro-
are designed offline and have fixed parameters. They are posed for the operation of VSC-based islanded micro-
usually based on prior knowledge of system parameters and grids. Unlike previous control strategies, the proposed
designed for the worst case of uncertainties. This issue makes approach is independent of system parameters. With
them often conservative; consequently, it can degrade the the slightest prior knowledge and the use of online
performance of the closed-loop system in instances—where learning mechanisms, it can estimate these parameters
uncertainties are not at their worst. and adjust itself to match the uncertain closed-loop
Additionally, it is always required to know the nominal system response to a selected reference dynamic model.
values of parameters and bounds of their uncertainties in the To put it another way, the proposed control strategy can
design procedure of these algorithms. In addition to the con- identify the unknown system parameters and simultane-
straints mentioned above, sliding mode controllers are usually ously adjust control parameters. As a result, unlike most
suitable for rejecting matched uncertainties—thereby suffering current studies—which are based on fixed controllers—
from problems related to the reaching phase [29]. Therefore, no previous calculations (such as solving LMIs and
according to the existing studies, the main drawbacks of the algebraic Riccati equations) are needed to obtain or tune
previously reported control algorithms are listed as: 1) they the control parameters. Typical robust control methods
are unable to compensate for a large amount of uncertainty are too conservative. Considering possible variations in
and mismatched disturbances; 2) system response may be too operating points, they may be subject to instability in
conservative for different operating conditions, especially in case of changes in system parameters. It is worthy of
robust H2 /H∞ algorithms; 3) due to the variation of system note that the proposed control algorithm is indepen-
parameters over time, it is needed to constantly “re-tune” dent of the system topology, line impedance, and load
the controller’s parameters to maintain the control system’s dynamics.
performance. 2) Besides the robustness against parameter variations and
In order to address the issues detailed above, this paper uncertainties, the proposed controller is robust against
presents a multivariable, adaptive, robust control strategy mismatched input disturbances. Through the complete
based on an “online” estimation methodology—in direct con- dynamic modeling of IBRs and their output filters,
trast to what scholars have already investigated thus far—to mismatched input disturbances will appear in the system
guarantee the stability and robust performance of islanded ac dynamics from the grid-side and affect system trajecto-
microgrids. With the least prior information about the system ries. The proposed strategy can estimate and reject these
and bounds of uncertainties, the proposed control algorithm external disturbances using online adaptation rules to
performs online estimations, thereby identifying the uncertain improve the power quality.
system’s parameters—which can change over time. It adjusts 3) The proposed MAR algorithm allows controller design-
the control parameters to achieve the desired performance. ers to choose the predetermined model according to
The method presented utilizes a dynamic reference model or their desired priorities. For instance, the system response
a predetermined dynamic model of interest (as an ideal and speed of the proposed method will be improved with-
arbitrary stable dynamic system) expressing the expected and out increasing the system sensitivity. Still, the pre-
desired performance of the closed-loop system. It changes sented MAR method can increase the response speed
the behavior of the actual system in a way that its dynamic via properly selecting the reference system and its
response matches the characteristics of the desired system properties—unlike many existing algorithms employing
selected using online adjustment rules; see Fig. 1 (left side) large feedback gains to increase it. Also, due to the
and [30], [31]. Therefore, there is a large degree of freedom to lack of knowledge of the system’s physical parameters,
design and select the reference model. It should be pointed out the proposed algorithm uses indirect identification to
that the control strategy introduced in this article is entirely adapt the controller parameters and match the system’s
different from the traditional adaptive droop studies. In exist- dynamic performance with the selected dynamic model’s
ing research, either droop gains are adaptively changed using performance. Specifically, the dynamics of IBR units
different measurements (e.g., see [32]) or the studies in which with unknown parameters will be replaced by the desired
various adaptive algorithms are employed in the secondary reference model. It is noteworthy that the persistent
layer of the hierarchical-based microgrids (e.g., see [33], [34], excitation condition to identify the parameters will be
and [35]). Indeed, it may be utilized as the inner control loop of relaxed via the algorithm introduced.
GFM inverters, whether in the hierarchical structure combined
with other control layers or even in a decentralized structure II. P RELIMINARIES AND P ROBLEM F ORMULATION
with a decentralized energy management system. In this section, some essential mathematical explanations
To the best of the authors’ knowledge, this paper is the first and notations are defined, which will be used throughout the
to present an adaptive robust algorithm based on a multivari- manuscript. Throughout this paper, χm (s) stands for the left
able control structure for the voltage control of islanded ac interactor matrix of a transfer function matrix like G(s), which
microgrids—without prior knowledge about nominal system specifies the properties of the infinite zeros of G(s). Readers
parameters considering the complete dynamics of IBRs and can refer to [36] for more details. Let L be the Laplace
unmatched input disturbances. The main contributions of this transform operator. Additionally, for ease of reference and
paper are as follows. simplicity, let’s define the inverse of the Laplace transform

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 AFSHARI et al.: MAR, PRIMARY CONTROL ENFORCING PREDETERMINED DYNAMICS OF INTEREST IN ISLANDED MICROGRIDS 2497

Fig. 1. Structure of the proposed MAR control (left) and the islanded IBR as an uncertain dynamic system with the proposed controller (right).

1
operator with L−1 such that L−1 {P(s)b(s)} = P(s)⟨b⟩(t), stand for the state, input control signal, output, and mis-
where the outcome is a function of time, “t.” In the defined matched input disturbance vectors, respectively. It is note-
operator, P(s) and b(t) stand for a transfer function matrix worthy that the control matrix B and the input disturbance
(or a filter in the frequency domain) and a continuous signal, matrix Bw are linearly independent (B ̸= α Bw ), so the input
respectively. disturbances affect the system states through a different input
channel.
A. Mathematical Modeling The disturbance vector w(t) stands for the grid-side voltage
Consider an islanded microgrid with an arbitrary topology. harmonics—which should be considered in the controller
A typical IBR in that is comprised of a source of electrical design. It is assumed that the disturbance signal w(t) has an
energy on the dc side. It is interfaced with the microgrid upper bound which is not needed to be known. The disturbance
through a VSC, which is generally connected to the point is actually a voltage signal; therefore, it is reasonable to con-
of common coupling through an LCL filter [22]. Let ω = sider it bounded. The output LCL filter can reduce higher-order
2 × π × 60 rad/s be the nominal frequency of the system— harmonics, but a proper control strategy needs to eliminate
which can be obtained via either a GFM control method, lower-order harmonics. In this regard, applying the dq-frame,
such as the droop technique, or an internal oscillator. Thus, the lower-order grid-side voltage harmonics (3rd , 5th , and 7th )
according to the single-line diagram shown in Fig. 1 (right are expressed as
w1 (t)
 
side), each IBR’s mathematical model in the dq-frame is
w(t) = (3)
expressed as w2 (t)
Rf 1 1

where

 I˙ Ld (t) = − I Ld (t) + ωI Lq (t) + Vid (t) − Vod (t)
L L L

f f f w1 (t) = Vds2 sin(2ωt) + Vds4 sin(4ωt) + Vds6 sin(6ωt)



Rf 1 1


I˙ Lq (t) = − I Lq (t) − ωI Ld (t) + Viq (t) − Voq (t) w2 (t) = Vq0 + Vqc2 cos(2ωt) + Vqc4 cos(4ωt)





 Lf Lf Lf

 1 1 + Vqc6 cos(6ωt). (4)
 V̇ od (t) = ωVoq (t) + I Ld (t) − Iod (t)


Cf Cf Consequently, the grid-side voltage harmonics can be pre-
1 1
V̇ oq (t) = −ωVod (t) + I Lq (t) − Ioq (t) sented as a summation of sinusoidal signals in the dq-frame.





 Cf Cf It is worthy of mention that the amplitudes of the disturbance
 Rc 1 1
I˙od (t) = − Iod (t) + ωIoq (t) + Vod (t) − Vbd (t)



 signals detailed above are “unknown” and “not” required to


 Lc Lc Lc be measured.
R 1 1


 I˙oq (t) = − c Ioq (t) − ωIod (t) + Voq (t) − Vbq (t) Although the main control objective of the studies in the


Lc Lc Lc literature is the same, unlike the modeling method used in [19],
(1) [22], and [25], the modeling procedure utilized in (1) does
not need knowledge regarding the physical topology, loads’
where I Ldq , Vodq , and Iodq are the dq-frame signals of the
and/or IBRs’ locations, and line impedance. Additionally, its
current injected to the ac-side filter, output voltage, and output
application is not limited to a specific topology of microgrids
current, respectively. Therefore, the state space equations of
in which loads are connected to IBRs outputs and there are
each IBR is written as
( only simple transmission lines between every two neighboring
ẋ(t) = Ax(t) + Bu(t) + Bw w(t) IBRs. The considered model is appropriate for the inner loop
(2)
y(t) = C x(t) of GFM inverters, whether in a decentralized or hierarchical
T structure.
I Ld , I Lq , Vod , Voq , Iod , Ioq , u(t) =

where x(t) = The dc side of generation units can be considered similar
T T T
Vid , Viq , y(t) = Vod , Voq , and w(t) = Vbd , Vbq
  
to the topology illustrated in Fig. 2. Accordingly, one can write

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 2498 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 21, NO. 3, JULY 2024

voltage stability with a specified frequency considering param-

eters uncertainty and external disturbances. The stable perfor-
mance of a GFM IBR on the ac side relies on the stability of
its dc link. In other words, a GFM IBR stabilizes the ac-side
voltage if the stability of the dc side is ensured—as the stability
of the dc side is an initial assumption for the operation of GFM
inverters. Still, grid-following (GFL) inverters can contribute
to the stability of the dc-link voltage if they are connected to a
stable ac grid with properly operating phase-locked units [39].
With this in mind, the detailed analysis of dc-link stability is
out of the scope of this study; it is the primary objective of
Fig. 2. DC side of the IBRs studied in this paper. other studies where the dynamics and operation of GFL IBRs
have been investigated in detail; see [40], [41], and [42].
the IBRs’ dc-link dynamics as follows.
1 RDC III. M ULTIVARIABLE , A DAPTIVE , ROBUST C ONTROL
V̇ DC (t) = IDC (t) − VDC (t) − IC (t) (5)
CDC CDC This section designs the MAR control strategy for the
where VDC , IDC , IC , RDC , and CDC are the voltage across primary layer of islanded microgrids; the proposed MAR
the dc-link capacitance, the dc-source current, the current control is robust against parametric uncertainties and input
delivered to the converter, dc-side resistance, and dc-side disturbances. Before going deep into the controller design
capacitance, respectively. The structure of the dc side and the procedure, some definitions are required.
variables are shown in Fig. 2. IDC acts as the control input As stated earlier, the main objective is to make the output
of the system—which needs to be appropriately designed. of the uncertain system to track the output of the selected
The following equation is employed to emulate the dynamic reference system. In other words, an arbitrary predetermined
response (or behavior) of the primary energy source (e.g., model to achieve our desired dynamic performance replaces
an energy storage system), where τDC is the time constant the actual dynamics (stable or unstable ones) of IBR units with
associated with the dynamics of the energy resource (see [37] unknown parameters. To this end, the output voltage of each
and [38]). IBR should track ym (t), i.e.,

ym (t) = G m (s)⟨r ⟩(t), (8)

Z
1
IDC (t) = ref
(t) − IDC (t) dt.


IDC (6)
τDC
where r (t) ∈ ℜ2 and G m (s) ∈ ℜ2×2 are the reference signal
The primary objective of the dc-side controller is to keep and the reference dynamics, respectively; for the meaning
the dc voltage stable according to its reference value. In this of the G m (s)⟨r ⟩(t) expression—or generally speaking, the
ref
regard, IDC is designed by a proportional controller detailed mathematical operator of P(s)⟨b⟩(t)—readers are referred to
in (7). the preliminaries described at the beginning of Section II.
In (8), the reference system is chosen to be G m (s) =
ref
(t) = K DC VDC
ref
− VDC (t)


IDC (7) χm−1 (s), where χm (s) is the left interactor matrix of G(s) =
ref C(s I − A)−1 B. The high-frequency gain matrix associated
where K DC is the controller gain, and VDC shows the reference
with χm (s) is defined as K H = lim χm (s)G(s). It should
value of dc voltage. s→∞
Based on the model derived earlier, the designed control be noted that if the left interactor matrix χm (s) is diagonal,
system should provide three objectives, as described next. the high-frequency gain matrix will also be diagonal.
1) The output voltage of each unit should track the reference Assumption 1: G(s) is full rank such that K H is
signal (the output of the predetermined model) asymptotically, non-singular.
2) the proposed controller should be able to compensate for Lemma 1: if the matrix of
the mismatched input disturbances and also meet the IEEE C1 Aη1 −1 B
 
recommendations regarding the output voltage harmonic lim- KG =  .. M×M
(9)
. ∈ℜ
 
its, 3) and the stability of the closed-loop system (considering
internal signals) should be guaranteed. In order to meet these C1 Aη M −1 B
objectives, the proposed MAR control technique is designed exists for a specific relative degree like ηi > 0 (for
in the following section. i = 1, 2, . . . , M) and is non-singular, it is concluded that
Remark 1: As stated previously, the MAR control algo- the left interactor matrix is diagonal such that χm (s) =
rithm applies to the inner ac voltage control loop of IBRs diag(h 1 (s), . . . , h M (s)) and K H = K G [36]. h i (s) = s η1 +
operating as GFM inverters in grids. The dynamic response (or biη1 −1 s η1 −1 + . . . + bi1 s + bi0 are arbitrary stable polynomials.
behavior) of renewable energy resources influences the dc-link Assumption 2: The zeros of G(s) are stable, and there exists
dynamics—but renewables have their own control systems a matrix like VH ∈ ℜ2×2 such that
functioning separately from the ac-side control system. The
designed control system is solely responsible for the ac-side K HT VH−1 = 6 P = 6 PT > 0. (10)

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 AFSHARI et al.: MAR, PRIMARY CONTROL ENFORCING PREDETERMINED DYNAMICS OF INTEREST IN ISLANDED MICROGRIDS 2499

TABLE I the second derivative of the output is written as

T EST S YSTEM ’ S PARAMETERS
ÿ i (t) = Ci A2 x(t) + Ci ABu(t) + Ci ABw w(t) + Ci ABw ẇ(t).
(13)

Substituting the control signal u(t) into (13) results in

ÿ i (t) = Ci A2 x(t) + ϑi (t) + Ci ABw w(t) + Ci ABw ẇ(t).

(14)

Consider the auxiliary control signal ϑi (t) defined as

ϑi (t) = −Ci A2 x(t) − bi1 ẏ i (t) − bi0 y(t)

− Ci ABw w(t) − Ci Bw ẇ(t) + ri (t).
(15)

Therefore, (14) is written as

By using Lemma 1, the high-frequency gain matrix is ÿ i (t) + bi1 ẏ i (t) + bi0 y(t) = ri (t). (16)
calculated with ηi = 2 as
  As a result,
1
0 1
 Cf Lf yi (s) = ri (s), y(s) = G m (s)r (s) (17)
1 .

C AB = 
 (11) h i (s)
0
Cf Lf where the reference system G m (s) is the inverse of left
interactor matrix. The auxiliary signal ϑi (t) is considered as
It is observed that K H is always diagonal and full rank
ϑ(t) = K̄ x(t) + K̄ 1 ω(t) + K̄ 2 ω̇(t) + r (t) with
independent of system parameters. According to the dynamic
( (
models in [17], and considering the parameters in Table I, K̄ = k̄ 1T k̄ 2T
 T  T T T
K̄ 1 = k̄ 11 k̄ 12
it is concluded that the zeros of G(s) are stable. Finally, T 2
,
k̄ i = −Ci A − bi1 Ci A − bi0 Ci k̄ 1i = −Ci ABw
considering the properties of the high-frequency gain matrix, (  T T T
VH can be chosen as VH = 61 = 61T to satisfy (10). K̄ 2 = k̄ 21 k̄ 22
(18)
k̄ 2i = −Ci Bw .
A. Nominal Control Algorithm
Thus, the nominal control signal is expressed as u(t) =
K H−1 K̄ x(t) + K̄ 1 w(t) + K̄ 2 ẇ(t) + r (t) with the nominal

This part presents the nominal structure of the proposed
controller in order to determine the algebraic feasibility of gains of K 1∗ = K H−1 K̄ , K 2∗ = K H−1 , and K 3∗ (t) =
the selected algorithm. Note that in this part, it is assumed
K H−1 K̄ 1 w(t) + K̄ 2 ẇ(t) . It is observed that the nominal
 
that the system’s parameters are entirely known. As a result,
just the algebraic feasibility of the nominal controller will be control signal (12) causes the output to track the reference
investigated. After investigating the feasibility of the nominal signal asymptotically in the presence of unmatched input
structure, in the next section assuming that the nominal param- disturbances. This observation concludes the proof.
eters are not known, the online adaptation rules will be used to
construct the actual controller. The nominal control algorithm B. Matching Condition
is presented as
From the previous subsection, it is concluded that for the
u(t) = K 1∗ x(t) + K 2∗r (t) + K 3∗ (t) (12) condition of
 −1
where K 1∗ ∈ ℜ2×6 and K 2∗ ∈ ℜ2×2 are nominal feedback G m (s)K ∗2 −1 K 3∗ (s) + C s I −A − B K ∗1 T Bw w(s) = 0
gains (not functions of time) to track the reference signal,  −1
and K 3∗ (t) ∈ ℜ2 is the disturbance rejection term; K 3∗ (t) is a C s I − A− B K ∗1 T B K ∗2 T =G m (s),
function of time, “t,” as indicated in parentheses.
(19)
Theorem 1: The output of the system in (2) tracks the
reference signal ym (t) = G m (s)⟨r ⟩(t) asymptotically in the there exist non-singular matrices K 1∗ ∈ ℜ2×6 , K 2∗ ∈ ℜ2×2 ,
presence of the disturbance vector w(t), using the nominal and disturbance rejection term K 3∗ (t) ∈ ℜ2 to satisfy (19).
control algorithm (12) such that lim (y(t) − ym (t)) = 0. The input disturbances are summation of some sinusoidal
t→∞
 Proof: Consider
T the output of the system (2) as y(t) = signals, so their derivatives are well defined, and one can
y1 (t) y2 (t) . Let’s define the control signal as u(t) = parameterize them [31]. Consider w(t) = 9 ∗ T g(t) and ẇ(t) =
∗
K H−1 ϑ(t) with ϑ(t) = [ϑ1 (t) ϑ2 (t)]T and K hi = Ci Aηi −1 B 9̃ ∗ T g(t) where 9 ∗ ∈ ℜ8×2 , 9̃ ∈ ℜ8×2 , and g(t) ∈ ℜ8
for i = 1, 2. According to the relative degree of the system, are parameterization terms associated with w(t) and ẇ(t).

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 2500 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 21, NO. 3, JULY 2024

Let’s assume that g(t) is a continuous function. Therefore, 0̃(t) = 0(t) − 0 ∗ . In the following, let’s define λ(t) and σ (t)
the disturbance rejection term K 3∗ (t) is parameterized as as
λ(t) = γ T (t)σ (t) − p(s)⟨u⟩(t)
h ∗
i
K 3∗ (t) = K H−1 K̄ 1 9 ∗ g(t) + K̄ 2 9̃ g(t) = ϕ3∗ g(t)
h ∗
i σ (t) = p(s)⟨κ⟩(t). (26)
ϕ3∗ = K H−1 K̄ 1 9 ∗ + K̄ 2 9̃ ∈ ℜ2×8 . (20)
Using (24), it is concluded that

Remark 2: It should be emphasized that we just know the δ(t) = 0 ∗ γ̃ (t)σ (t) + 0̃(t)λ(t). (27)
structure of disturbance signal w(t) and its derivative ẇ(t), not It is observed that the estimation error δ(t), λ(t), and σ (t)
their actual value. According to the sinusoidal structure of the are all based on measurable signals.
disturbance signal in (3), the structure of the first derivative of Remark 3: Although the structure of the controller pro-
the signal can be obtained from w(t). Therefore, the adaptation posed in (21) is linear, it is classified as a nonlinear controller.
mechanism discussed in the next subsection can estimate ϕ3∗ The existence of time varying adaptive parameters K 1 (t),
in order to compensate for the disturbance term. K 2 (t), and K 3 (t) that are functions of system states and
measured errors, such as κ(t) and δ(t), makes the control
C. Proposed MAR Algorithm algorithm as well as the closed-loop system nonlinear.
Similar to the nominal design in (12), the proposed MAR
algorithm is expressed as E. Online Adaptation Rules
This section constructs the online adaptation rules required
u(t) = K 1 (t)x(t) + K 2 (t)r (t) + K 3 (t) (21) to estimate the unknown parameters using measurable signals.
where K 1 (t), K 2 (t), and K 3 (t) are the estimates of the nominal In this regard, it is assumed that K 1 (t), K 2 (t), and K 3 (t) are
parameters. In the following, online adaptation rules will be unknown, so it is needed to identify them using measurable
obtained in order to estimate the nominal parameters of the signals. Let’s define 61 = 61T > 0 and 62 = 62T > 0 as
closed-loop system. some arbitrary design parameters which are used to tune the
adaptation speed. Thus, the adaptive laws are written as
T 61 δ(t)σ T (t)
D. Tracking and Estimation Errors γ̇ (t) = −
n 2 (t)
The construction of error signals is one of the critical stages
62 δ(t)λT (t)
in adaptive control systems. The first step is to obtain the 0̇(t) = − (28)
output tracking error. According to the conditions in (19), the n 2 (t)
output tracking error is written as where n 2 (t) = 1 + σ T (t)σ (t) + λT (t)λ(t).
Lemma 2: Using the online adaptation rules obtained
e(t) = y(t) − ym (t)
in (28), one can guarantee that γ (t), 0(t) ∈ L ∞ and
= G m (s)K ∗2 −1 γ̃ κ (t) + Ce( A+B K 1 )t x(t)
δ(t)
D T E ∗T
(22) n(t)
, γ̇ (t), 0̇(t) ∈ L 2 ∩ L ∞ .
Proof: Consider the energy function of
where γ̃ (t) = γ (t) − γ ∗ is the parameter estimation error, and 1h    T i
T
for κ(t) ∈ ℜ16 , γ ∗ T , and γ T (t) ∈ ℜ2×16 , we have V (t) = tr K H γ̃ 61−1 γ̃ + tr 0̃ 62−1 0̃ . (29)
2
T
κ(t) = x T (t) r T (t) g T (t)
K H = kh I for kh > 0. Considering the fact that

Let’s define
γ ∗ T = K 1∗ K 2∗ ϕ3∗
  trace x y T = x T y for two arbitrary vectors x, y ∈ ℜn , the
time derivative of the energy function is written as
γ T (t) = K 1 (t) K 2 (t) ϕ3 (t) .
 
(23)
δ T (t)δ(t)
As the exponential term in (22) decays to zero, the output V̇ (t) = − ≤ 0. (30)
n 2 (t)
estimation error is rewritten as δ(t)
D T E As a result, one deduces that γ (t) ∈ L ∞ , 0(t) ∈ L ∞ , n(t) ∈
e(t) = G m (s)K 2 ∗ −1 γ̃ κ (t), (24) L 2 ∩ L ∞ , γ̇ (t) ∈ L 2 ∩ L ∞ , and 0̇(t) ∈ L 2 ∩ L ∞ —thereby
concluding the proof.
which is completely measurable.
Note: Given the length of complementary details regarding
In the next step, it is required to construct the estimation
1 the stability analysis and page limitations, more details can be
error. In this regard, consider p(s) = q(s) with q(s) = s 2 +
found in the 9th chapter of [43].
q1 s + q0 as an arbitrary stable polynomial. As a result, the Remark 4: In summary, the main objective of the presented
estimation error δ(t) is defined as MAR control algorithm is to conform the dynamical behavior
δ(t) = χm (s) p(s)⟨e⟩(t) + 0(t)λ(t) (25) of the system (with unknown parameters) to an arbitrary
reference system. Considering designers’ priorities, it is first
where 0(t) is the estimation of the high-frequency gain matrix required to select a dynamical model (usually a decoupled
K H . Let 0 ∗ be the nominal value of K H . As a result, the system with fast dynamics). This model demonstrates the
estimation error of high-frequency gain matrix is defined as desired dynamics that replace the original dynamics of IBR

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 AFSHARI et al.: MAR, PRIMARY CONTROL ENFORCING PREDETERMINED DYNAMICS OF INTEREST IN ISLANDED MICROGRIDS 2501

Fig. 4. Scenarios 1, 2, and 3 (Step Load Change, Short-Circuit Fault, and

Nonlinear Load Connection): IBRs #1’s and #2’s output (a) active power and
(b) reactive power.
Fig. 3. Single-line diagram of the test microgrid.

units. Then, it is needed to form the online adaptation rules in Besides, it is worth noting that the proposed control scheme
(28) using available measurements without tuning or obtaining is completely independent of the system’s size and topology
any control parameters through solving LMIs or any other and the number of IBRs. As a result, it can be applied to
algebraic equation. All that is needed is to choose adaptation a microgrid with arbitrary size and physical topology. The
gains that only affect the adaptation speed of the control nominal frequency of the system is 60√Hz, and the voltage
ref ref
parameters [K 1 (t), K 2 (t), and K 3 (t)] to their nominal values. reference is chosen to be Vod = 208 2 and Voq = 0 for
The designed control system regulates the amplitude of IBRs all IBRs. GFM control methods or decentralized management
terminal voltages with a specific frequency; it also guarantees systems provide these reference values for the inner control
ref
the stability of all other state variables regardless of parameter loop. The nominal dc voltage is set to VDC = 900, and the
uncertainty and grid-side disturbances with prescribed perfor- proportional gain is chosen to be K DC = 150. The reference
mance. As a result, the voltage and frequency stability of the system is selected as any arbitrary stable 2×2 transfer function
system is ensured as long as the designed control system has matrix with η = 2. Therefore, the predetermined dynamics of
a solid performance. interest are chosen to be
The proposed structure uses an online adaptation mecha-  9900 
0
G m (s) =  (s + 90)(s + 110)
nism to identify unknown parameters and then adapt control
. (31)
 
parameters according to the physical characteristic of the 9900
0
unknown system. In fact, unlike offline methods, which rely on (s + 90)(s + 110)
existing datasets to form a control structure, this method can
Adaptation gains and the design filter are also chosen to
be applied to an IBR without the need for prior measured data
be 61 = 62 = 90 and q(s) = (s + 250)(s + 300), respec-
collection, so it uses posterior information for the adjustment.
tively. It should be noted that the predetermined system’s
Additionally, the online adaptation mechanism allows the
parameters (e.g., transfer function’s poles) can be arbitrar-
control system to adjust itself to possible changes in system
ily selected. Therefore, the closed-loop system’s speed and
parameters in real time during different contingencies.
performance can be tuned “without” changing the con-
troller’s gain, and hence increasing the closed-loop system’s
IV. S IMULATION R ESULTS
sensitivity.
This section verifies the proposed MAR algorithm through The simulation studies are carried out through two different
carrying out simulations of the test microgrid system (see scenarios. Besides, the performance of the proposed MAR
Fig. 3) in MATLAB/SimPowerSystems environment. The algorithm is compared with the robust control method in [18].
parameters of the system under study are presented in
Table I—which is just employed to simulate a test microgrid
system—otherwise, these parameters have “not” been used at A. Scenario 1: Step Load Change
any stage of the controller design process. The electrical loads The first scenario deals with the system’s robustness against
are constant impedance type. The loads’ power varies with a step load change. In this regard, the test microgrid system
the square of their terminal voltage magnitude. The control starts its operation in islanded mode at t = 0 s. At t = 0.4 s,
algorithm used in this study does not rely on the loads’ types Load 2 is increased by 50%, and at t = 0.6 s is decreased to
and their characteristics; its performance does not depend on its nominal value. Fig. 4 (up to t = 0.8 s) shows Scenario 1
them either. According to the dynamic model in (1) and (2), detailed above. Also, as Fig. 5 shows the voltage and current
IBRs’ output current is one of the state variables—which signals, IBR #2 responds well to this disturbance, and its
includes the power losses and the currents supplying loads— output voltage remains in a stable region without having any
but there is no constraint on the type of loads they are feeding. noticeable fluctuations.

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 2502 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 21, NO. 3, JULY 2024

Fig. 5. Scenario 1 (Step Load Change): IBR #2’s instantaneous three-phase

(a) voltages and (b) currents. Fig. 7. Scenario 3 (Nonlinear Load Connection): IBR #2’s instantaneous
three-phase (a) voltages and (b) currents.

Fig. 6. Scenario 2 (Short-Circuit Fault): IBR #2’s instantaneous three-phase

(a) voltages and (b) currents. Fig. 8. Scenario 3 (Nonlinear Load Connection): Nonlinear load’s instanta-
neous three-phase (a) voltages and (b) currents.
B. Scenario 2: Short-Circuit Fault
The objective of this case study is to evaluate the perfor- nonlinear load can be seen in Fig. 8. The nonlinear load draws
mance of the proposed control method with respect to a sudden a non-sinusoidal current that does not follow the characteristic
short-circuit fault. In this regard, a single-phase short-circuit of its voltage waveform. Although nonlinear loads generate
fault (phase a) with a resistance of Rslg = 0.01 strikes Bus 4 harmonics intensively, the voltage regulation is well achieved,
at t = 0.8 s and is next cleared at t = 1 s. This event causes and there is not any considerable fluctuation in response to the
some fluctuations in IBRs’ output currents. Nevertheless, the connection of the nonlinear load.
output voltages show minor, negligible oscillations and remain Consequently, the proposed MAR method is able to guar-
stable in response to this fault. Fig. 4 (from t = 0.8 s to antee a robust performance of the closed-loop system dur-
t = 1.2 s) shows active and reactive powers associated with ing the scenarios mentioned above. In this regard, Fig. 10
Scenario 2 detailed above. shows the dq-components of IBRs’ output voltages. Also,
Fig. 6 presents IBR #2’s responses. Its results show that after Figs. 4(a) and (b) shows the instantaneous active and reactive
a short transient, the output current of the unit converges to powers of IBRs #1 and #2, respectively. According to the
its steady-state value. As expected, the tracking performance IEEE recommendations, the maximum allowable voltage THD
of the output voltage is completely acceptable and shows the is 5%. In this regard, Fig. 9(a) depicts IBR #2’s voltage THD,
robustness of the closed-loop system. for example. The results confirm that the voltage THD of
IBR #2 does not exceed its maximum threshold except for
C. Scenario 3: Nonlinear Load Connection transients. The voltage THD results affirm the robustness of
In this scenario, the performance of the proposed control the designed control system against grid-side disturbances,
scheme is evaluated by the connection of a 3-phase nonlinear which is obtained through the online adaptation mechanism
harmonic load, a six-pulse diode rectifier, to the test system. to estimate the unmatched input disturbances. Also, the per-
To this end, a 3-phase nonlinear harmonic load with R = formance of the proposed method in tracking the reference
40  and L = 0.01 mH is connected to Bus 4 at t = 1.2 s signals can be investigated using the voltage tracking error
and then disconnected at t = 1.4 s. The responses of the plotted in Fig. 9(b). Moreover, to examine the performance of
output voltage and current of IBR #2 are plotted in Fig. 7. the proposed method from another perspective, the evaluation
Additionally, the instantaneous voltages and currents of the of the IBRs trajectories (Vod , Voq , and also output tracking

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 AFSHARI et al.: MAR, PRIMARY CONTROL ENFORCING PREDETERMINED DYNAMICS OF INTEREST IN ISLANDED MICROGRIDS 2503

Fig. 12. Output voltage trajectories of IBRs.

Fig. 9. Scenarios 1, 2, and 3 (Step Load Change, Short-Circuit Fault, and

Nonlinear Load Connection): (a) Voltage THD of IBR #2 and (b) the voltage
tracking errors of IBRs #1 and #2.

Fig. 13. Results produced by the controller in [18] for comparisons:

(a) Voltage tracking errors and (b) voltage THD of IBR #2—solid lines and
dash-dotted ones are associated with the proposed method and the method
in [18], respectively.

method in [18]. The authors in [18] have presented a robust

Fig. 10. Scenarios 1, 2, and 3 (Step Load Change, Short-Circuit Fault, and control structure for islanded ac microgrids using two cascaded
Nonlinear Load Connection): Performance of the (a) d and (b) q components control loops. The inner loop in this method is based on an
of the output voltages.
integral sliding mode control. The outer loop uses a mixed
H2 /H∞ controller. The design procedure of both control
loops requires knowledge of the system’s physical parameters.
In addition to being overly conservative, it is necessary to solve
LMIs to obtain the outer loop control parameters for different
operation conditions. Nevertheless, the proposed method does
not need any prior algebraic calculations. Indeed, it uses online
learning mechanisms to identify the physical parameters and
adjust the control system parameters.
In this regard, the performance of the algorithm selected
is compared with that of the proposed method using the
first scenario of the simulation studies. To this end, Load 2
encounters 50% and is then decreased at t = 0.3 s and
t = 0.5 s, respectively. Fig. 13 displays the tracking errors and
Fig. 11. Trajectories of voltage tracking errors. voltage THDs of both methods. As a result, it is concluded
that the proposed method has a solid tracking performance
errors) are displayed in Figs. 11 and 12 for all the simulation and disturbance rejection capability compared with the chosen
scenarios. They reveal that the output voltages’ dq-components robust method.
have practically reached their reference values (starting from
zero initial condition). As a result, the tracking errors have V. E XPERIMENTAL R ESULTS
converged to a bounded range around zero. The test rig depicted in Fig. 14 is deployed to conduct
experimental examinations related to IBR #2 in the Sim-
D. Comparisons With Results of Sliding Mode Control ulation Results section. It is built by SEMIKRON intelli-
For the last case of simulation studies, the performance of gent power modules using insulated gate bipolar transistors
the introduced method is compared with the presented control (IGBTs) (based on “SKM 50 GB 123 D” modules). Besides,

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 2504 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 21, NO. 3, JULY 2024

Fig. 16. Experimental results (with the 40 ms/div horizontal axis) asso-
ciated with the simulation results related to IBR #2 shown in Fig. 5; load
connection test in which active power and reactive power are shown by
traces in magenta and lawn green with 84.85 V/div [(for the peak value)—or
equivalently 60 V/div (for the rms value)]—and 20 A/div (for the peak value),
respectively—their “actual” per unit (pu) values for voltage and current signals
are noted in the left-bottom corner of the figure.

Fig. 14. Test rig used in the experiments and details of IBR—housed in the the comparison between simulation and experiment results is
Laboratory for Advanced Power and Energy Systems (LAPES) at Georgia
Southern University—where experiments have been conducted.
feasible. In this regard, Fig. 15 replicates the time frame from
t = 0.3 s to t = 1.0 s in Fig. 4, and Fig. 16 reproduces
Fig. 5. Comparing Fig. 15 with Fig. 4 and Fig. 16 with Fig. 5
reveals that simulations and experiments match well with
each other. The agreement on time responses demonstrated
in simulations and experimental results, which are altogether
consistent, reveal the effectiveness of the proposed control
methodology employed in IBR #2.
VI. C ONCLUSION
A multivariable, adaptive, robust control method for the
islanded operation of ac microgrids has been proposed and
studied in this paper. The presented method provides a stable
and robust performance for autonomous microgrids and is
independent of microgrids’ load dynamics, size, and topology.
What is more, there is no need to know the nominal values
of the system’s physical parameters to design the controller.
Online adaptation and estimation mechanisms identify the
system’s physical parameters; the controller parameters are
tuned according to an arbitrary predetermined model of inter-
est chosen by the controller designer. Compared with other
control methods, the proposed strategy is less conservative;
Fig. 15. Experimental results (with the 100 ms/div horizontal axis) associated unlike conventional robust algorithms, there is no need to do
with simulation results related to IBR #2 shown in Figs. 4(a) and (b) (for any mathematical calculations or solve algebraic equations
the time frame from t = 0.3 s to t = 1.0 s): (a) Load connection test in
which active power and reactive power are shown by traces in black and cyan
to adjust the control parameters. The MAR controller has a
with 6.00 kW/div and 1.30 kvar/div (their “actual” pu values for active and simple structure compared with high-order robust methods,
reaction power signals are noted in the left-bottom corner of the figures), is less conservative, and, more importantly, does not need
respectively; and (b) load disconnection test with the signal information
detailed in Fig. 15(a).
to be re-tuned over time. Furthermore, it gives the system
operator a large degree of freedom to select an ideal dynamic
SEMIKRON “SKHI 21A (R)” gate drives and protection cir- for the closed-loop system to imitate. The proposed control
cuitry are employed to make the converter functional. Verivolt strategy can be applied to the inner voltage controller of
“IsoBlock I-ST-1c”/“IsoBlock V-1c” current/voltage sensors GFM inverters for the reliable operation of isolated microgrids.
are hooked to digital inputs to measure the currents and Finally, various comparative simulations and experiments have
the voltages, respectively. dSPACE “MicroLabBox (MLBX)” been employed to verify the accuracy and validity of the
using a real-time processor and field-programmable gate arrays proposed method in response to different scenarios and distur-
(commonly known as FPGAs) and benefiting from PWM bances. Future work includes employing an optimal criterion
signals (generated by digital inputs/outputs) connects the VSC for choosing adaptation gains to be addressed as the potential
under test to the printed circuit boards of the measurement and development.
drive circuits. R EFERENCES
Furthermore, all of the parameters of the setup deployed are [1] M. Farrokhabadi et al., “Microgrid stability definitions, analysis, and
similar to those of simulations, as stated in Table I. Therefore, examples,” IEEE Trans. Power Syst., vol. 35, no. 1, pp. 13–29, Jan. 2020.

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 AFSHARI et al.: MAR, PRIMARY CONTROL ENFORCING PREDETERMINED DYNAMICS OF INTEREST IN ISLANDED MICROGRIDS 2505

[2] R. Pérez-Ibacache, C. A. Silva, and A. Yazdani, “Linear state-feedback [22] M. Babazadeh and A. Nobakhti, “Robust decomposition and structured
primary control for enhanced dynamic response of AC microgrids,” control of an islanded multi-DG microgrid,” IEEE Trans. Smart Grid,
IEEE Trans. Smart Grid, vol. 10, no. 3, pp. 3149–3161, May 2019. vol. 10, no. 3, pp. 2463–2474, May 2019.
[3] Z. Wang, Y. Yu, W. Gao, M. Davari, and C. Deng, “Adaptive, optimal, [23] M. S. Sadabadi, Q. Shafiee, and A. Karimi, “Plug-and-play voltage
virtual synchronous generator control of three-phase grid-connected stabilization in inverter-interfaced microgrids via a robust control strat-
inverters under different grid conditions—An adaptive dynamic pro- egy,” IEEE Trans. Control Syst. Technol., vol. 25, no. 3, pp. 781–791,
gramming approach,” IEEE Trans. Ind. Informat., vol. 18, no. 11, May 2017.
pp. 7388–7399, Nov. 2022, doi: 10.1109/TII.2021.3138893. [24] T. Dragičević, “Model predictive control of power converters for robust
[4] Y. Huang, Q. Sun, Y. Li, H. Zhang, and Z. Chen, “Adaptive-discretization and fast operation of AC microgrids,” IEEE Trans. Power Electron.,
based dynamic optimal energy flow for the heat-electricity integrated vol. 33, no. 7, pp. 6304–6317, Jul. 2018.
energy systems with hybrid AC/DC power sources,” IEEE Trans. Autom. [25] M. Cucuzzella, G. P. Incremona, and A. Ferrara, “Decentralized sliding
Sci. Eng., early access, Jul. 12, 2022, doi: 10.1109/TASE.2022.3188277. mode control of islanded AC microgrids with arbitrary topology,” IEEE
[5] C. Zhai, H. D. Nguyen, and X. Zong, “Dynamic security assessment of Trans. Ind. Electron., vol. 64, no. 8, pp. 6706–6713, Aug. 2017.
small-signal stability for power grids using windowed online Gaussian
[26] M. B. Delghavi and A. Yazdani, “Sliding-mode control of AC voltages
process,” IEEE Trans. Autom. Sci. Eng., early access, May 12, 2022,
and currents of dispatchable distributed energy resources in master-slave-
doi: 10.1109/TASE.2022.3173368.
organized inverter-based microgrids,” IEEE Trans. Smart Grid, vol. 10,
[6] L.-N. Liu and G.-H. Yang, “Distributed fixed-time optimal resource
no. 1, pp. 980–991, Jan. 2019.
management for microgrids,” IEEE Trans. Automat. Sci. Eng., vol. 20,
no. 1, pp. 404–412, Jan. 2023, doi: 10.1109/TASE.2022.3155163. [27] J. R. Massing, M. Stefanello, H. A. Grundling, and H. Pinheiro,
[7] M. Davari, W. Gao, Z.-P. Jiang, and F. L. Lewis, “An optimal primary “Adaptive current control for grid-connected converters with LCL filter,”
frequency control based on adaptive dynamic programming for islanded IEEE Trans. Ind. Electron., vol. 59, no. 12, pp. 4681–4693, Dec. 2012.
modernized microgrids,” IEEE Trans. Autom. Sci. Eng., vol. 18, no. 3, [28] J.-W. Jung, N. T.-T. Vu, D. Q. Dang, T. D. Do, Y.-S. Choi, and
pp. 1109–1121, Jul. 2021. H. H. Choi, “A three-phase inverter for a standalone distributed gen-
[8] A. La Bella, S. R. Cominesi, C. Sandroni, and R. Scattolini, “Hierarchi- eration system: Adaptive voltage control design and stability analysis,”
cal predictive control of microgrids in islanded operation,” IEEE Trans. IEEE Trans. Energy Convers., vol. 29, no. 1, pp. 46–56, Mar. 2014.
Autom. Sci. Eng., vol. 14, no. 2, pp. 536–546, Apr. 2017. [29] M. Rubagotti, A. Estrada, F. Castanos, A. Ferrara, and L. Fridman,
[9] S. M. Azimi and S. Lotfifard, “A nonlinear controller design for power “Integral sliding mode control for nonlinear systems with matched and
conversion units in islanded micro-grids using interconnection and unmatched perturbations,” IEEE Trans. Autom. Control, vol. 56, no. 11,
damping assignment tracking control,” IEEE Trans. Sustain. Energy, pp. 2699–2704, Nov. 2011.
vol. 12, no. 1, pp. 284–292, Jan. 2021. [30] G. Song and G. Tao, “A partial-state feedback model reference adaptive
[10] M. Raeispour, H. Atrianfar, M. Davari, and G. B. Gharehpetian, control scheme,” IEEE Trans. Autom. Control, vol. 65, no. 1, pp. 44–57,
“Fault-tolerant, distributed control for emerging, VSC-based, islanded Jan. 2020.
microgrids—An approach based on simultaneous passive fault detec- [31] L. Wen, G. Tao, and Y. Liu, “Multivariable adaptive output rejection of
tion,” IEEE Access, vol. 10, pp. 10995–11010, 2022. unmatched input disturbances,” Int. J. Adapt. Control Signal Process.,
[11] U. Bose, S. K. Chattopadhyay, C. Chakraborty, and B. Pal, “A novel vol. 30, nos. 8–10, pp. 1203–1227, Aug. 2016.
method of frequency regulation in microgrid,” IEEE Trans. Ind. Appl., [32] S. Sahoo, S. Mishra, S. Jha, and B. Singh, “A cooperative adaptive droop
vol. 55, no. 1, pp. 111–121, Jan. 2019. based energy management and optimal voltage regulation scheme for DC
[12] Y. Lin et al., “Research roadmap on grid-forming inverters,” Nat. Renew. microgrids,” IEEE Trans. Ind. Electron., vol. 67, no. 4, pp. 2894–2904,
Energy Lab. (NREL), Golden, CO, USA, Tech. Rep. NREL/TP-5D00- Apr. 2020.
73476, 2020. [33] A. Afshari, M. Karrari, H. R. Baghaee, G. B. Gharehpetian, and
[13] M. Raeispour, H. Atrianfar, H. R. Baghaee, and G. B. Gharehpetian, J. M. Guerrero, “Robust cooperative control of isolated AC microgrids
“Robust sliding mode and mixed H2 /H∞ output feedback primary subject to unreliable communications: A low-gain feedback approach,”
control of AC microgrids,” IEEE Syst. J., vol. 15, no. 2, pp. 2420–2431, IEEE Syst. J., vol. 16, no. 1, pp. 55–66, Mar. 2022.
Jun. 2021. [34] N. M. Dehkordi and S. Z. Moussavi, “Distributed resilient adaptive
[14] S. K. Sahoo, A. K. Sinha, and N. K. Kishore, “Control techniques in control of islanded microgrids under Sensor/Actuator faults,” IEEE
AC, DC, and hybrid AC–DC microgrid: A review,” IEEE J. Emerg. Sel. Trans. Smart Grid, vol. 11, no. 3, pp. 2699–2708, May 2020.
Topics Power Electron., vol. 6, no. 2, pp. 738–759, Jun. 2018.
[35] A. Afshari, M. Karrari, H. R. Baghaee, and G. B. Gharehpetian,
[15] H. Karimi, M. Karimi-Ghartemani, and K. Sheshyekani, “Robust control
“Resilient synchronization of voltage/frequency in AC microgrids under
of three-phase voltage source converters under unbalanced grid condi-
deception attacks,” IEEE Syst. J., vol. 15, no. 2, pp. 2125–2136,
tions,” IEEE Trans. Power Electron., vol. 34, no. 11, pp. 11278–11289,
Jun. 2021.
Nov. 2019.
[36] G. Tao, “Multivariable adaptive control: A survey,” Automatica, vol. 50,
[16] S. D’Arco, J. A. Suul, and O. B. Fosso, “Automatic tuning of cascaded
no. 11, pp. 2737–2764, 2014.
controllers for power converters using eigenvalue parametric sensitivi-
ties,” IEEE Trans. Ind. Appl., vol. 51, no. 2, pp. 1743–1753, Mar. 2015. [37] A. Tayyebi, A. Anta, and F. Dorfler, “Grid-forming hybrid angle
[17] M. Hamzeh, S. Emamian, H. Karimi, and J. Mahseredjian, “Robust control and almost global stability of the DC–AC power con-
control of an islanded microgrid under unbalanced and nonlinear load verter,” IEEE Trans. Autom. Control, early access, Jul. 26, 2022, doi:
conditions,” IEEE Trans. Emerg. Sel. Topics Power Electron., vol. 4, 10.1109/TAC.2022.3193953.
no. 2, pp. 512–520, Jun. 2016. [38] M. Davari, A. Aghazadeh, W. Gao, and F. Blaabjerg, “Detailed dynamic
[18] M. Raeispour, H. Atrianfar, H. R. Baghaee, and G. B. Gharehpetian, DC models of VSC considering controls for DC-fault simulations in
“Robust hierarchical control of VSC-based off-grid AC microgrids modernized microgrid protection,” IEEE J. Emerg. Sel. Topics Power
to enhancing stability and FRT capability considering time-varying Electron., vol. 9, no. 4, pp. 4514–4532, Aug. 2021.
delays,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 9, no. 6, [39] I. Subotic and D. Gros, “Power-balancing dual-port grid-forming power
pp. 7159–7172, Dec. 2021. converter control for renewable integration and hybrid AC/DC power
[19] S. Derakhshan, M. Shafiee-Rad, Q. Shafiee, and systems,” IEEE Trans. Control Netw. Syst., vol. 9, no. 4, pp. 1949–1961,
M. R. Jahed-Motlagh, “Decentralized robust LMI-based voltage Dec. 2022.
control strategy for autonomous inverter-interfaced multi-DG [40] F. Degioanni, I. G. Zurbriggen, and M. Ordonez, “Enhanced DC-
microgrids,” IEEE Trans. Power Syst., early access, Sep. 6, 2022, link voltage dynamics for grid-connected converters,” IEEE Trans. Ind.
doi: 10.1109/TPWRS.2022.3204625. Electron., vol. 69, no. 11, pp. 10787–10796, Nov. 2022.
[20] M. Shafiee-Rad, Q. Shafiee, M. S. Sadabadi, and M. R. Jahed-Motlagh, [41] Y. Gui, F. Blaabjerg, X. Wang, J. D. Bendtsen, D. Yang, and J. Stous-
“Decentralized voltage stabilization and robust performance satisfaction trup, “Improved DC-link voltage regulation strategy for grid-connected
of islanded inverter-interfaced microgrids,” IEEE Syst. J., vol. 15, no. 2, converters,” IEEE Trans. Ind. Electron., vol. 68, no. 6, pp. 4977–4987,
pp. 1893–1904, Jun. 2021. Jun. 2021.
[21] H. R. Baghaee, M. Mirsalim, G. B. Gharehpetian, and H. A. Talebi, [42] M. Davari and Y. A.-R. I. Mohamed, “Dynamics and robust control of a
“Decentralized sliding mode control of WG/PV/FC microgrids under grid-connected VSC in multiterminal DC grids considering the instanta-
unbalanced and nonlinear load conditions for on-and off-grid modes,” neous power of DC- and AC-side filters and DC grid uncertainty,” IEEE
IEEE Syst. J., vol. 12, no. 4, pp. 3108–3119, Dec. 2018. Trans. Power Electron., vol. 31, no. 3, pp. 1942–1958, Mar. 2016.

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.
 2506 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 21, NO. 3, JULY 2024

[43] G. Tao, Adaptive Control Design and Analysis, 1st ed. Hoboken, NJ, Mehdi Karrari (Senior Member, IEEE) received the
USA: Wiley, Jul. 2003, doi: 10.1002/0471459100. Ph.D. degree in control engineering from Sheffield
University, Sheffield, U.K., in 1991. Since 1991,
he has been with the Amirkabir University of Tech-
nology, Tehran, Iran. He has authored or coauthored
more than 150 technical articles and three books.
His main research interests include power system
modeling, modeling and identification of nonlinear
Amir Afshari received the B.Sc. degree in elec- dynamic systems, and large-scale and distributed
trical engineering from the Qom University of systems.
Technology, Qom, Iran, in 2016, and the M.Sc.
degree in electrical engineering from the Amirkabir
University of Technology, Tehran, Iran, in 2020.
He is currently pursuing the Ph.D. degree in elec-
trical and computer engineering with the Univer-
sity of Wisconsin–Madison, USA. His research Weinan Gao (Senior Member, IEEE) received
interests include controls of modern power grids the B.Sc. degree in automation and the M.Sc.
with high-penetration of power-electronics-based degree in control theory and control engineering
systems. from Northeastern University, Shenyang, China, in
2011 and 2013, respectively, and the Ph.D. degree
in electrical engineering from New York University,
Brooklyn, NY, USA in 2017.
He is currently a Professor with the State Key Lab-
oratory of Synthetical Automation for Process Indus-
tries, Northeastern University. Previously, he was
Masoud Davari (Senior Member, IEEE) was born
an Assistant Professor of Mechanical and Civil
in Isfahan, Iran, in September 1985. He received the
Engineering with the Florida Institute of Technology, Melbourne, FL, USA;
B.Sc. degree (summa cum laude) in electrical engi-
an Assistant Professor of Electrical and Computer Engineering with Georgia
neering (power) from the Isfahan University of Tech-
Southern University, Statesboro, GA, USA; and a Visiting Professor with
nology, Isfahan, in 2007, the M.Sc. degree (summa
the Mitsubishi Electric Research Laboratory (MERL), Cambridge, MA,
cum laude) in electrical engineering (power) from
USA. His research interests include reinforcement learning, adaptive dynamic
the Amirkabir University of Technology (Tehran
programming (ADP), optimal control, cooperative adaptive cruise control
Polytechnic), Tehran, Iran, in 2010, and the Ph.D.
(CACC), intelligent transportation systems, sampled-data control systems,
degree (Hons.) in electrical engineering (power elec-
and output regulation theory. He was a recipient of the Best Paper Award
tronics in energy systems) from the University of
in IEEE International Conference on Real-time Computing and Robotics
Alberta, Edmonton, AB, Canada, in 2016.
(RCAR) in 2018 and the David Goodman Research Award at New York
He was with the Grid Secure Operation Research Center, Iran, and the
University in 2019. He is an Associate Editor of IEEE T RANSACTIONS ON
Electric Power Research Institute (EPRI), Tehran, from January 2010 to
N EURAL N ETWORKS AND L EARNING S YSTEMS, IEEE/CAA J OURNAL OF
December 2011. From April 2015 to June 2017, he was a Senior Research
AUTOMATICA S INICA, Control Engineering Practice, Neurocomputing, and
and Development Specialist and a Senior Consultant with Quanta-Technology
IEEE T RANSACTIONS ON C IRCUITS AND S YSTEMS —II: E XPRESS B RIEFS.
Company, Markham, ON, Canada, in the field of the dynamic interaction of
He is an Editorial Board Member of Neural Computing and Applications and a
renewables with smart ac/dc grids and control, protection, and the automa-
Technical Committee Member of IEEE Control Systems Society on Nonlinear
tion of microgrids. In July 2017, he joined as a tenure-track Assistant
Systems and Control and IFAC TC 1.2 Adaptive and Learning Systems.
Professor with the Allen E. Paulson College of Engineering and Comput-
ing, Department of Electrical and Computer Engineering, Georgia Southern
University (GSU), Statesboro, GA, USA, where he was recommended for
being granted “early” promotion to an Associate Professor and the Award
of “early” Tenure in December 2021, and officially approved for both
in February 2022. He is currently the Founder and the Director of the Frede Blaabjerg (Fellow, IEEE) received the Ph.D.
Laboratory for Advanced Power and Energy Systems [LAPES (watch it on degree in electrical engineering from Aalborg Uni-
https://www.youtube.com/watch?v=mhVHp7uMNKo)], State-of-the-Art Cen- versity, Aalborg, Denmark, in 1995.
ter for Engineering and Research (CEaR), GSU, in 2021. He has developed From 1987 to 1988, he was with ABBScandia,
and implemented several experimental test rigs for research universities and Randers, Denmark. He became an Assistant Pro-
the power and energy industry. He has authored several papers published fessor, an Associate Professor, and a Full Profes-
in IEEE T RANSACTIONS and journals. His research interests include the sor of Power Electronics and Drives with Aalborg
dynamics, controls, the protections of different power electronic converters University, in 1992, 1996, and 1998, respectively.
utilized in the hybrid ac/dc smart grids, and hardware-in-the-loop (HIL) In 2017, he became a Villum Investigator. He is
simulation-based testing of modernized power systems. also a Honoris Causa with University Politehnica
Dr. Davari has been an active member and the Chapter Lead with the IEEE Timisoara (UPT), Timisoara, Romania, and Tallinn
Power and Energy Society Task Force on “Innovative Teaching Methods Technical University, Tallin, Estonia. He has authored or coauthored four
for Modern Power and Energy Systems” since July 2020. He has been an monographs, published more than 600 journal articles in the fields of power
active member and the Chapter Lead (for Chapter 3) with the IEEE Working electronics and its applications, and was an editor of ten books in power elec-
Group P2004—a newly established IEEE working group titled “Hardware- tronics and its applications. His current research interests include power
in-the-Loop (HIL) Simulation Based Testing of Electric Power Apparatus electronics and its applications, such as in wind turbines, PV systems,
and Controls” for IEEE Standards Association since June 2017. He is an reliability, harmonics, and adjustable speed drives. He was a recipient of the 33
invited member of the Golden Key International Honour Society. He was IEEE Prize Paper Awards, the IEEE PELS Distinguished Service Award in
a recipient of the 2019–2020 Allen E. Paulson College of Engineering and 2009, the EPE-PEMC Council Award in 2010, the IEEE William E. Newell
Computing (CEC) Faculty Award for Outstanding Scholarly Activity in the Power Electronics Award 2014, the Villum Kann Rasmussen Research Award
Allen E. Paulson CEC at GSU, the Discovery and Innovation Award from 2014, the Global Energy Prize in 2019, and the 2020 IEEE Edison Medal.
the 2020–2021 University Awards of Excellence at GSU, and one of the From 2006 to 2012, he was the Editor-in-Chief for the IEEE T RANSACTIONS
awardees of the 2021–2022 Impact Area Accelerator Grants (partially funded) ON P OWER E LECTRONICS . From 2005 to 2007, he was a Distinguished
at GSU. He was the Chair of the Literature Review Subgroup of DC@Home Lecturer of the IEEE Power Electronics Society and the IEEE Industry Appli-
Standards for the IEEE Standards Association from April 2014 to October cations Society from 2010 to 2011 and from 2017 to 2018. From 2019 to 2020,
2015. He is an invited Reviewer of several IEEE T RANSACTIONS /J OURNALS, he was the President of IEEE Power Electronics Society. He was the Vice
IET journals, Energies journal, and various IEEE conferences, the invited President of the Danish Academy of Technical Sciences, Lyngby, Denmark.
speaker at different universities and in diverse societies, and the best Reviewer From 2014 to 2020, he was nominated by Thomson Reuters, Toronto, Canada,
of the IEEE T RANSACTIONS ON P OWER S YSTEMS in 2018 and 2020. to be between the most 250 cited researchers in engineering in the world.

Authorized licensed use limited to: NUST School of Electrical Engineering and Computer Science (SEECS). Downloaded on January 12,2025 at 17:59:37 UTC from IEEE Xplore. Restrictions apply.