This article has been accepted for publication in IEEE Transactions on Energy Conversion.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

Robust Vector Current Control of Grid-interactive

Smart Packed E-Cell Multi-level Inverters under
Non-ideal Grid Conditions
Mahdieh S. Sadabadi, Senior Member, IEEE, Mohammad Sharifzadeh, Senior Member, IEEE, Eric Laurendeau,
Meysam Gheisarnejad, Mohammad-Hassan Khooban, Senior Member, IEEE, Ramon Portillo, Member, IEEE,
Leopoldo Garcia Franquelo, Life Fellow, IEEE, Kamal Al-Haddad, Life Fellow, IEEE.

Abstract—This paper proposes a convex optimization-based benefits in terms of better harmonic profiles, reduced dV /dt
strategy for the design of vector current controllers in a grid- stress, reduced voltage stress on semiconductor components,
interactive single-phase nine-level Packed E-Cell (PEC9) in- low switching power losses, and less distortion of input
verter under uncertain grid conditions, from stiff to weaker
conditions, and non-ideal grid voltage conditions. The non- currents [8]–[10]. In the past few decades, several topologies
ideal grid-interactive PEC inverter with the grid impedance for multi-level VSIs have been proposed. Some of these
uncertainty is modeled by a multivariable polytopic model subject technologies include Cascaded Connected H-bridge (CHB) or
to a disturance in a state space framework. By virtue of this Series Connected H-bridge (SCHB) [11], Flying Capacitor
novel modeling approach, a robust two-degree-of-freedom H∞ (FC) [12], Neutral Point Clamped (NPC) [13], and Packed
multivariable Proportional-Integral (PI) vector current control
strategy is then proposed. The proposed control mechanism U-Cells (PUCs) inverter topology [14], which is a mixture of
guarantees robust stability, employing a quadratic Lyapunov FC and CHB topologies. The researchers in [15] have recently
function, and robust performance against the uncertainty in developed a Packed E-Cell (PEC) inverter, which is a novel
grid impedance and non-ideal grid voltage conditions, also it single DC source compact multi-level inverter topology. In the
provides inverter currents with Total Harmonic Distortion and topology of the PEC inverter, two U-Cells are substituted by
harmonic components complying with IEEE Standard 1547
requirements. The effectiveness of the proposed grid-interactive one E-Cell; hence, the PEC inverter offers a more optimized
smart PEC9 inverter equipped with the proposed robust vector design than PUCs and other proposed multi-level inverters
control strategy is evaluated by simulation and hardware-in-the- [15]. In addition, in the PEC inverter topology, nine voltage
loop experimental case studies. levels are generated and the number of semiconductor devices
Index Terms—Grid-interactive inverters, Packed E-Cell is reduced [15].
(PEC9) inverter, Vector current control, Convex Optimization,
Polytopic model uncertainty. One of the relevant control problems in grid-interactive
multi-level PEC inverters is the control of the inverter injected
I. I NTRODUCTION current, which is a challenging problem due to the uncertainty
of the grid impedance at the Point of Common Coupling

V OLTAGE source inverters (VSIs) are the main inter-

face between distributed generation (DG) units (e.g.,
photovoltaics and vehicle-to-grid technologies) and the main
(PCC). The main source of causing this uncertainty is that
the grid impedance parameter can vary under a wide range,
encompassing grid operation from stiff to weaker conditions
grid [1]. Conventionally, two-level VSIs were used for the [1]. In addition to the uncertainty in the grid impedance, the
integration of DG units to the main AC grid, see [1]–[5] grid voltage might be subject to distortion due to unbalanced
and references therein. However, two-level VSIs suffer from conditions and harmonic distortion [16]–[18], which might
causing electromagnetic interference and high dV /dt stress lead to a non-ideal grid. A number of control methods for
across devices [6], [7]. On the contrary, multi-level VSIs offer multi-level PEC inverters among other topologies have been
proposed in the literature. For example, several model-based
M. S. Sadabadi is with the Department of Electrical and Electronic
Engineering, The University of Manchester, Manchester, United Kingdom and data-driven-based control approaches such as finite-set
(e-mail: [email protected]). Corresponding author: M. S. model predictive control (MPC) [19], sliding mode control
Sadabadi. (SMC) method [20], [21], passivity-based control [22], direct
M. Sharifzadeh, Eric Laurendeau, M. Gheisarnejad, and K. Al-
Haddad are with the Department of Electrical Engineering, Ecole nonlinear predictive control [23], loop-shaping-based control
de Technologie Superieure, Montreal, QC, Canada (e-mails: moham- [24], adaptive fuzzy control method [25], Q-learning-based
[email protected]; [email protected]; kamal.al- controller [26], feedback linearization, and artificial neural
[email protected]).
M. Gheisarnejad and M. H. Khooban are with the Department of Electrical network (ANNs)-based control [27] have been designed for
and Computer Engineering, Aarhus University, Aarhus, Denmark (e-mails: a single-phase grid-interactive PEC inverter. Although these
[email protected]; [email protected]). approaches focus on the design of current controllers for grid-
R. Portillo and L. G. Franquelo are with the Department of Electronic
Engineering, University of Seville, Seville, Spain (e-mails: [email protected]; connected PEC inverters under a specific grid condition with
[email protected]). a priori known grid impedance, the design of such controllers

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

becomes more difficult when grid conditions are uncertain or current control technique for a grid-connected two-level VSI
vary under a wide range. In other words, all these control with an LCL filter under grid impedance variation has been
strategies do not take into account robustness in the controller developed. Although each control loop in [29] includes a low-
synthesis for grid-interactive PEC inverters. As a result, these order controller, this approach requires an extra current sensor
methods do not guarantee satisfactory transient performance for measuring the current of the filter capacitor. Moreover,
against uncertainty in the grid impedance value or non-ideal the works in [28]–[30] have not considered the case of robust
grid voltage conditions. Consequently, the uncertainty in the performance under grid voltage distortion. A metaheuristics
grid impedance value or the grid voltage distortion might approach for optimizing the control gains of the current
adversely impact the design and performance of the proposed controller of grid-connected two-level VSI operating under
controllers. Hence, it is essential to develop a current control normal and abnormal grid conditions has been presented in
technique for grid-interactive PEC inverters that can systemat- [31]. The main drawback of this test-driven design approach
ically consider the parametric uncertainty such as uncertainty is the lack of the theoretical guarantee of desired performance
in the grid impedance as well as non-ideal grid conditions. under various grid conditions from stiff to weak. Similar
This paper develops a novel active current control mech- to the proposed current control strategy, all these control
anism for grid-interactive PEC9 inverters subject to grid design approaches are based on robust control methods and
impedance uncertainty non-ideal grid voltage conditions. To their variations. However, the main advantage of the proposed
incorporate the uncertainty in the dynamic model of the grid- multivariable robust current control scheme compared to these
interactive PEC9, we develop a multivariable linear time- approaches is the simplicity of the control structure and ease
invariant uncertain model whose region of uncertainty is a of implementation, which has great importance in practice.
polytope. In the proposed polytopic model, the uncertainty Furthermore, the proposed vector current control strategy in
in the grid impedance is modeled by a convex hull of two this paper guarantees robust current tracking performance in
given vertices. This novel modeling framework of the grid- the presence of both grid impedence uncertainty and non-ideal
interactive PEC9 enables us to reformulate the control design grid voltage conditions.
and performance specification of the grid-interactive PEC9 Several simulation, software-in-the-loop (SIL), and
inverter in terms of a convex optimization problem subject hardware-in-the-loop (HIL) experimental case studies are
to a set of Linear Matrix Inequality (LMI) constraints. The carried out on a grid-interactive PEC inverter setup to evaluate
optimal solution of the proposed optimization problem is a the performance of the smart PEC9 inverter, controlled by
robust multivariable Proportional-Integral (PI) vector current the proposed robust optimization-based control strategy.
controller in direct and quadrature (dq) rotating reference The experimental HIL results comply with IEEE Standard
frame with a two-degree-of-freedom (2DOF), i.e., feedback 1547 requirements [32] and demonstrate that the proposed
and feedforward terms. The proposed active current control robust H∞ vector current controller can ensure a satisfactory
scheme guarantees robust stability and H∞ performance for performance (in terms of fast dynamics and a zero steady-state
grid-interactive PEC9 inverters with uncertain grid impedance. current tracking error) if its synthesis process is based on the
In addition to robustness to grid impedance uncertainties and proposed systematic control design technique.
robust performance under non-ideal grid voltage conditions, The main contributions of this paper are twofold. First, a
the main advantage of the proposed control system compared multi-level PEC9 inverter interactive with a non-ideal grid
to MPC and SMC in [19]–[21] is the simplicity in the structure under a wide range of impedance, encompassing from stiff to
and implementation of the proposed controller. Furthermore, weaker grid conditions can be cast as a multivariable polytopic
compared to a conventional PI controller with only two model and non-ideal grid voltage conditions. Second, by virtue
design control parameters, the proposed vector current control of this novel dynamic modeling description and the design of a
framework has a higher degree of freedom with feedback and robust 2DOF H∞ multivariable PI vector current controller, the
feedforward control gain matrices, leading to better transient stability, desired performance, and offset-free current tracking
performance in the current trajectories of the grid-interactive in a non-ideal grid-interactive PEC9 inverter subject to the grid
PEC9 inverter. impedance uncertainty are guaranteed. Finally, the simplicity
Several robust control techniques for two-level VSIs have of the proposed control structure facilitates its implementation
been proposed in the literature, e.g., [28]–[31] and references in practice.
therein. In [28], a robust vector-controlled current controller The organization of the paper is as follows: Section II
has been proposed for a two-level VSI connected to a weak develops the architecture of a grid-interactive PEC9 inverter
grid. Despite the effectiveness of the proposed control scheme, and also develops a polytopic representation of such a system
the proposed current controller has a high order (7th-order under the grid inductance uncertainty. The details of the vector
dynamic controller), which limits its implementation in prac- current control structure and the LMI-based control design
tice. The study in [30] proposed an adaptive H∞ current procedure are given in Section III. The current controller,
control strategy for a two-level VSI interfaced with a weak simulation, SIL, and HIL experimental results are presented
grid. However, the proposed controller depends on the system in Section IV. Finally, Section V concludes the paper.
equilibrium point and includes a derivative term, which is Notation: The notation used in this paper is standard. In
sensitive to measurement noises. In [29], a robust multi-loop particular, In is an n × n Identity matrix, ∥X∥2 is the two-norm

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

TABLE I
S WITCHING STATES AND VOLTAGE LEVELS IN A PEC9 INVERTER - T HE
SYMBOLS ↑, ↓, AND − INDICATE CHARGING , DISCHARGING , AND
NEUTRAL CASES FOR CAPACITORS C1 AND C2 IN F IG . 1, RESPECTIVELY.

State (S1 , S2 , S3 , S4 , S5 , S6 , S7 ) C1 C2 Vinv

1 (1, 0, 0, 0, 1, 1, 0) −∗ − Vdc = E
2 (1, 0, 0, 0, 1, 0, 1) − ↑ Vdc −V2 = 43 E
3 (1, 0, 1, 0, 1, 0, 0) ↑ ↑ Vdc −V1 −V2 = 21 E
4 (1, 1, 0, 0, 0, 1, 0) ↓ ↓ V1 +V2 = 21 E
5 (1, 1, 0, 0, 0, 0, 1) ↓ − V1 = 41 E
6 (0, 0, 0, 1, 1, 1, 0) − − 0
7 (1, 1, 1, 0, 0, 0, 0) − − 0
8 (0, 0, 0, 1, 1, 0, 1) − ↓ −V2 = − 41 E
9 (0, 0, 1, 1, 1, 0, 0) ↓ ↓ −V1 −V2 = − 21 E
10 (0, 1, 0, 1, 0, 1, 0) ↑ ↑ −Vdc +V1 +V2 = − 21 E
11 (0, 1, 0, 1, 0, 0, 1) ↑ − −Vdc +V1 = − 43 E
12 (0, 1, 1, 1, 0, 0, 0) − − −Vdc = −E
Fig. 1. Schematic diagram of a single-phase grid-interactive PEC9 inverter.

of X, and X T denotes the transpose of X. For a symmetric reference frame as below (see [33] for dq transformation of
matrix X, the positive definite and negative definite operators single-phase signals):
are shown by X ≻ 0 and X ≺ 0, respectively.
dig,d (t)
L = −Rig,d (t) + Lω0 ig,q (t) +Vinv,d (t) −Vg,d (t),
dt (2)
II. M ODELING OF A G RID - INTERACTIVE PEC I NVERTER
dig,q (t)
L = −Rig,q (t) − Lω0 ig,d (t) +Vinv,q (t) −Vg,q (t),
The single-phase grid-interactive nine-level Packed E-Cell dt
(PEC9) inverter as the case study is depicted in Fig. 1. Table I where ω0 is the angular frequency, (ig,d (t), ig,q (t)),
shows the switching states of the PEC9 inverter as well as its (Vinv,d (t),Vinv,q (t)), and (Vg,d (t),Vg,q (t)) are the dq components
capacitors’ charge/discharge condition. It is worth mentioning of the inverter current, the terminal voltage of the PEC9
that indicated capacitors charging/discharging conditions in inverter, and the grid voltage, respectively. In (2), Lω0 ig,q (t)
Table I are for the grid current direction shown in Fig. 1. and Lω0 ig,d (t) indicate the coupling terms in d and q current
It is observed that there are two switching states in which signals, respectively.
the output voltage level of the inverter is 12 E and two others It is assumed that the multi-level PEC9 inverter is connected
in which this voltage level is − 21 E, with the difference that to a grid under non-ideal voltage conditions such as harmonic
for each of them, and a given direction of the grid current, distortion and unbalanced conditions [18]. The impact of the
one produces the charging of the capacitors while the other grid’s non-ideal conditions can be modeled as an external dis-
produces their discharge. The integration of such redundant turbance with twice the fundamental frequency for unbalanced
states with the PWM technique achieves to maintain a balance conditions and even harmonics, e.g., ±6th, for harmonic loads
of the voltages of the DC capacitors to a quarter of the DC [16]–[18]. To this end, by separating the dc components from
input voltage, i.e., V4dc . Therefore, no complicated or external Vg,d (t) and Vg,q (t), one can obtain that:
voltage control is required for PEC9 capacitors, which leads
to some level of simplification in the grid-interactive PEC9 Vg,d (t) = Vd + vd (t),
(3)
control strategy [15]. Vg,q (t) = Vq + vq (t),
where Vd and Vd are ideal d and q components of the grid
A. Polytopic Model of a Grid-interactive PEC9 Inverter under voltage and vd (t) and vq (t) model the non-ideal grid condition
Non-ideal Grid Conditions caused by the grid distortion.
Considering Fig. 1, the dynamic equation of grid-interactive The dynamic model in (2) under the non-ideal grid voltage
PEC9 inverter with an L filter can be written as follows [7]: modeled by (3) can be represented in the state-space frame-
work as follows:
dig (t) 
L = −Rig (t) +Vinv (t) −Vg (t), (1) ẋg (t) = Ag xg (t) + Bg u(t) + Bd (V (t) + d(t)),
dt G(s) : (4)
y(t) = Cg xg (t),
where Vg (t), ig (t), and Vinv (t) are the grid voltage, the inverter
where
current, and the terminal voltage of the inverter, respectively.
1
− RL
   
In (1), L is the summation of the grid inductance and the ω0 L 0
Ag = , Bg = ,
inductance of the filter while R represents the equivalent series −ω0 − RL 0 L1
(5)
resistance of the grid and filter. The dynamic equation of the − L1
   
0 1 0
grid-interactive PEC9 inverter is transformed to a dq rotating Bd = , Cg = .
0 − L1 0 1

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

 T
and xg (t) = ig,d (t), ig,q (t) is the state vector, 4) Axis decoupling: The decoupling between the d and q
T
current signals is maximized.

u(t) = Vinv,d (t),Vinv,q (t) is the input signals,

y(t) = ig,d (t), ig,q (t)
T
is the output signals, and 5) Minimization of the impact of the distorted grid volt-
V (t) = [Vd ,Vq ]T is the ideal grid voltage and age: The impact of the grid disturbances d(t) on the
current tracking performance of the grid-tied PEC9 is
d(t) = [vd (t), vq (t)]T represents the grid disturbance.
minimized.
Since the value of the grid inductance and grid resistance
are uncertain or unknown in practice, the parameters L and R For good current reference tracking and attenuation of the
are assumed to be uncertain but belong to intervals L ∈ [L L̄] grid disturbance (due to the grid distortion), a small sensitivity
and R ∈ [R R̄] with known (L, R) and (L̄, R̄), encompassing ¯ function S(s), representing the transfer function from the
¯ ¯ ¯
grid operation from stiff to weaker conditions. Taking into current reference i∗dq to the inverter current error i∗dq − ig,dq ,
account the grid inductance uncertainty, the state space model is required. The sensitivity function can mathematically be
in (4)-(5), particularly (Ag , Bg , Bd ), is subject to parameter described as follows:
uncertainty. To incorporate such uncertainty in the dynamic S(s) = (I2 + G(s)C(s))−1 . (8)
model of the grid-interactive PEC9 inverter, a polytopic rep-
resentation is adopted in which the uncertainty in the grid where C(s) is the transfer function of the current controller,
inductance is modeled via a convex hull of four known vertices and G(s) is the open-loop transfer function of the grid-
(q = 4) as stated below [34]: interactive PEC9 model. To achieve the above-mentioned
q q
control objectives, we will consider the minimization of a cost
Ag (λ ) = ∑ λi Agi , Bg (λ ) = ∑ λi Bgi , function J = ∥W (s)S(s)∥∞ in the frequency domain, where
i=1 i=1 W (s) is a stable weighting filter aimed to shape the singular
q (6)
values of sensitivity function such that they are small in the
∑ λi = 1, λi ≥ 0, i = 1, . . . , 4. low frequency [35].
i=1
We then propose the following 2DOF multivariable PI
The polytope vertices are computed according to the mini- current control policy to minimize the cost function J:
mum and maximum value of (L, R) as follows:
u(t) = k p y(t) + ki v(t) + kw w(t) + k f i∗dq ,

" R
# "
1
# C(s) : (9)
− L¯ ω0 L 0 v̇(t) = i∗dq − y(t),
Ag1 = ¯ R , Bg1 = ¯ 1 ,
−ω0 − L¯ 0 L
¯
where w(t) is the state of the weighting filter W (s) (see
" ¯ #
− R̄L̄ ω0
 1
0
 Subsection III-A). The proposed structural controller in (9)
Ag2 = , Bg2 = L̄ ,
0 L̄1 includes a feedback control gain K = [k p ki kw ] and a feedfor-
−ω0 − R̄L̄
" # " # (7) ward control gain k f .
1
− R̄L ω0 L 0 Remark. It is worth mentioning that the structure of the
Ag3 = ¯ , Bg3 = ¯ 1 , proposed active current control law is in the form of a modified
−ω0 − R̄L 0 L
" ¯ #  1 ¯  Proportional-Integral (PI) controller. In other words, in the
R
− L̄¯ ω0 0 control update rule u(t) in (9) (i.e., k p y(t) + k f i∗dq ) represents
Ag4 = R , Bg4 = L̄ 1
the proportional term with a setpoint weight, ki v(t) = ki (i∗dq −
R
−ω0 − ¯ 0 L̄
L̄
y(τ))dτ represents the integral part, and kw w(t) is a retrofitting
III. P ROPOSED ROBUST V ECTOR C URRENT C ONTROL
term designed to improve the transient performance of the grid
D ESIGN S CHEME
current trajectories.
A robust vector current control strategy for the single-phase The cost function J can be optimized over a robust 2DOF
grid-interactive PEC9 inverter depicted in Fig. 1 is proposed in control policy with the structure given in (9). Hence, the
this section. The main goal is to design a current controller that problem of the current control design is transferred to the
satisfies the following performance specifications for all values following H∞ optimization problem in which the control gain
of the grid impedance in the set L ∈ [L L̄] and R ∈ [R R̄], matrices are decision variables:
which covers grid operation from stiff ¯to weaker conditions:
¯
min ∥W (s)S(s)∥∞ (10)
1) A Theoretical Certificate of Robust Stability: The K,k f
closed-loop grid-interactive PEC9 inverter with the poly- In the following, we will discuss the design of the weighting
topic uncertainty in (6)-(7) is robustly stable. filter W (s) and also propose a systematic way for solving the
2) Current reference tracking performance: The closed-loop above optimization problem.
grid-tied PEC9 inverter augmented with the proposed
current controller can track all step dq current reference A. Design of Performance Weighting Filter
signals i∗dq with a zero steady-state error. Considering that fact that the performance weight W (s)
3) A Theoretical Certificate of Robust Performance: The forms the frequency response of the sensitivity function S(s),
response of the current trajectories of the closed-loop the minimization of the cost function J = ∥W (s)S(s)∥∞ in (10)
grid-interactive PEC9 inverter to current reference sig- provides the desired performance characteristics of the closed-
nals i∗dq is fast with no or small overshoot. loop grid-interactive PEC9 inverter in the frequency domain.

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

A common choice of the multivariable weighting filter W (s) C. Proposed LMI-based Robust Vector Control Design
is given as follows [35]: In this subsection, the theoretical results for the design of
 s +ω ∗  the proposed robust 2DOF vector control mechanism (9) are
Md d
∗ 0 presented.
W (s) =  s+ω εd Ms +ωq∗  , (11)
0 q Theorem 1: If there exist matrices P, H, k f , and positive
s+ω ∗ εq q scalars β , η, and γ so that the following inequality constraints
where Md and Mq are the maximum magnitudes of the fre- are hold:
quency response of the diagonal elements of S(s), ωd∗ and ωq∗ A j P + PATj + B j H + H T BTj (CP)T
 
Bi j
are the desired closed-loop bandwidth for d and q components,  BTij −γ 2 × I2 DT  ≺ 0 (17a)
and εd and εq are two positive scalars that are chosen to be CP D −I2
   
sufficiently small, i.e., εd << 1 and εq << 1, to certify an P I6 −ηI6 H T
≻ 0, ≺ 0, P ≻ 0 (17b)
approximate integral action in low frequencies for both d and I6 β I6 H −I2
q current channels [35]. for j = 1, . . . , q, then the state feedback controller gain K
The transfer function W (s) in (11) is a low-pass filter and is obtained as K = HP−1 , where H and P are a solution of
can be presented in the state-space framework as follows: the inequality conditions in (17). Moreover, the vector current
ẇ(t) = Aw w(t) + Bw (i∗dq − y(t)), controller in (9) guarantees:

W (s) : (12)
z(t) = Cw w(t) + Dw (i∗dq − y(t)), 1) The robust stability of the closed-loop dynamics for
all values of the grid impedance in given intervals
where w(t) is the vector of the states of W (s) in (11) and z(t) L ∈ [L L̄] and R ∈ [R R̄];
is a performance output vector. The state-space matrices in 2) The ¯robust performance ¯ of ∥W (s)S(s)∥∞ ≤ γ for the
(12) are obtained as follows: entire range of the grid impedance in given intervals
L ∈ [L L̄] and R ∈ [R R̄];
" #
 ∗ 1
0

ωd εd 0 Md ¯ ¯ current reference i∗ ;
Aw = − , Dw = , 3) Offset-free tracking of
0 ωq∗ εq 0 M1q √ dq
" # (13) 4) ∥K∥2 < ηβ , i.e., preventing large control actions.
ωd∗ (1 − Mεd ) 0 Proof 1: See Appendix in Section VI.
d
Bw = ε , Cw = I2 .
0 ωq∗ (1 − Mqq ) Remark 1: The inequality conditions proposed in (17) are
presented as a set of linear matrix inequalities (LMIs) with
B. Closed-loop Dynamics of Grid-interactive PEC9 Inverters respect to unknown decision variables P, H, k f , γ 2 , β , and η.
By augmenting the dynamics of the grid-interactive PEC9 The controller gains K = [k p ki kw ] and k f of feedback
inverter in (4) with the polytopic model uncertainty in (6)-(7) and feedforward terms are obtained through a solution of the
with the dynamics of the weighting filter in (12), as well as following convex optimization problem:
the dynamics of the proposed vector current control strategy
min α1 γ + α2 β + α3 η
in (9), the state-space polytopic model of the transfer matrix γ,β ,η,H,P,k f
(18)
function W (s)S(s) is presented as follows: subject to the set of LMIs in (17)
ẋ(t) = (A(λ ) + B(λ )K) x(t) + Bi i∗dq + Bv (V (t) + d(t)),

WS : where α1 , α2 , and α3 are positive scalars, which characterize a
z(t) = Cx(t) + Di∗dq trade-off between the robust H∞ performance ∥W (s)S(s)∥∞ ≤
 T (14) γ and large control actions. The feedback control gain is then
where x(t) = ig,d (t) ig,q (t) v(t) w(t) and obtained as K = HP−1 , and the feedforward gain k f is directly

Ag (λ ) 02 02
 
Bg (λ )
 obtained by solving the optimization problem in (18),
A(λ ) =  −I2 02 02 , B(λ ) =  02 , D. Proposed Algorithm for Robust Vector Current Controller
−Bw 02 Aw 02 Design
   
Bg (λ )k f Bd (15) The following algorithm summarizes the design steps of
Bi =  I2 , Bv =  02 , robust vector current control design for the grid-interactive
Bw 02 PEC9 inverter:
Inputs: Closed-loop dynamics and desired current tracking
 
C = −Dw 02 Cw , D = Dw .
performance specifications
It is worth mentioning that the dynamics in (14) are exposed
1) Polytopic modeling of the grid-interactive PEC9 inverter
to polytopic model uncertainty with the following vertices:
      and constructing q = 4 vertices in (7).
Ag j 02 02 Bg j Bg j k f 2) Weighting filters design (design of W (s) in (11)) based
A j = −I2 02 02 , B j = 02 , Bi j = I2 , (16) on the desired closed-loop bandwidth and current track-
−Bw 02 Aw 02 Bw
ing performance.
for j = 1, . . . , q, where Ag j and Bg j are given in (7). Fig. 2 3) Obtaining the state-space matrices of the closed-loop
shows the block diagram of a closed-loop grid-interactive PEC9 system in (15) and (16) with the polytopic model
PEC9 with its proposed vector current controller in (9). uncertainty.

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

Fig. 2. Block diagram of a closed-loop grid-interactive PEC9 inverter augmented with the proposed vector current control system.

TABLE II
PARAMETERS OF THE G RID - INTERACTIVE PEC9 INVERTER IN F IG . 1.
(a)
Parameter Value 0

Mag (dB)
Grid voltage (Vg ) 120 V (rms) -50
DC source (Vdc ) 180 V -100
Grid resistance (R) 0.1 Ω 10 0 10 1 10 2 10 3 10 4 10 5
DC-Link capacitance (C1 ,C2 ) 2500 µF, 2500 µF
(b)
Nominal frequency ( f0 ) 60 Hz
0
Mag (dB)

Sampling time (Ts ) 40 µs (rad/s)

Switching frequency ( fsw ) 2 KHz -50

-100
10 0 10 1 10 2 10 3 10 4 10 5
Output: A robust 2DOF H∞ active vector current controller. (c)
0
4) Solving the convex optimization problem in (18) by opti-
Mag (dB)

-50
mization toolboxes, e.g., YALMIP [36], and semidefinite (rad/s)
-100
programming (SDP) solvers (e.g., MOSEK [37]).
10 0 10 1 10 2 10 3 10 4 10 5
5) Finding the decision variables k f , P, and H as a solution (d)
of the optimization problem. 0
Mag (dB)

6) Forming the gain of the feedback control as K = HP−1 -50 (rad/s)

-100
and constructing the control signal u(t) in (9).
10 0 10 1 10 2 10 3 10 4 10 5
7) Transformation of the control signal u(t) from the dq (rad/s)
framework to a single-phase sinusoidal signal.
IV. S IMULATION C ASE S TUDIES , SIL AND HIL R ESULTS , Fig. 3. Frequency response of the inverse of performance weight W −1 ( jω)
A NALYSES , AND D ISCUSSION (black dashed line) and sensitivity function S( jω) (black solid line): (a)
transfer function from the first input to the first output, (b) transfer function
To evaluate the performance and effectiveness of the pro- from the second input to the second output, (c) transfer function from the
posed robust vector control strategy in the current reference second input to the first output, and (d) transfer function from the first input
to the second output.
tracking and assess its robustness against the uncertainty in
the grid impedance value, two simulation case studies along performance specification, the performance weight W (s) in
with SIL and HIL analyses are carried out. In the simulation, (11) is designed as follows:
SIL, and HIL case studies, the grid-interactive PEC9 inverter " s +500 #
2
in Fig. 1 with parameters given in Table II is used. The grid- s+0.1 0
W (s) = s , (19)
interactive PEC9 inverter is controlled by the proposed robust 2 +500
0 s+0.1
vector control scheme.
The solution to the optimization problem given in (18) is
A. Robust H∞ Vector Current Controller a robust vector current controller. The frequency response of
Following the proposed control design procedure in Subsec- the inverse of the performance weight W (s) and the resulting
tion III-D, a robust H∞ vector current controller is designed sensitivity function for L = 2.5 mH are shown in Fig. 3. Fig. 4
for the PEC9 inverter of Fig. 1. To this end, we follow the shows the frequency response of the grid-interactive PEC9
design algorithm assuming that the grid inductance value is inverter with and without the proposed current controller for
unknown but belongs to L ∈ [2.5 5] mH. For the desired L = 2.5 mH. The low-frequency gain, particularly the zero DC

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

for j = 1, . . . , q. To determine the interval [L L̄], the values

of L and L̄ are initially chosen close to the ¯ nominal value
(a)
Mag (dB)
20 ¯
of L. Subject to the feasibility of the LMIs in (21), the
0
-20 uncertainty interval [L L̄] is iteratively stretched so that the
-40 ¯
largest possible uncertainty interval for the grid inductance
10 0 10 1 10 2 10 3 10 4 10 5
(b)
L is found. Following the aforementioned procedure, the
20
(rad/s)
uncertainty interval L ∈ [0.1 15] mH is obtained. It is worth
Mag (dB)

0
-20
mentioning that for the robust control design, the range of the
-40 grid inductance uncertainty is considered to be L ∈ [2.5 5] mH.
10 0 10 1 10 2 10 3 10 4 10 5 Since the uncertainty interval L ∈ [0.1 15] mH found via
(c) the robust stability analysis approach in (21) encompasses
Mag (dB)

0
the design range L ∈ [2.5 5] mH, the robust stability of the
-50 (rad/s)
proposed vector current control scheme for L ∈ [2.5 5] mH is
-100
10 1 10 2 10 3 10 4 10 5 fully demonstrated.
(d)
Mag (dB)

0
C. Simulation Results, Comparision, and Analyses
-50 (rad/s)
-100 The proposed vector current control approach for the grid-
10 0 10 1 10 2 10 3 10 4 10 5
(rad/s) interactive PEC9 inverter with parameters given in Table II is
evaluated by simulations in MATLAB/Simulink. The first case
study evaluates the current reference tracking performance of
Fig. 4. Frequency response of open-loop (black dashed line) and closed-loop
grid-interactive PEC9 inverter (black solid line) with the proposed robust H∞ the proposed controller for L = 2.5 mH. To this end, the refer-
vector current controller: (a) transfer function from the first input to the first ence current signal is initially considered as i∗g (t) = 5sin(ω0t).
output, (b) transfer function from the second input to the second output, (c) It is then assumed that the amplitude of the current reference
transfer function from the second input to the first output, and (d) transfer
function from the first input to the second output. is stepped up from 5 A to 7 A and stepped down to 5 A.
The voltage and current waveforms are shown in Fig. 5. As
gain, of the frequency response of the diagonal elements of one can observe from this figure, the grid voltage and the
the closed-loop system in Fig. 4 (also small values of S( jω) in PEC9 inverter current are in phase, so it is injecting pure
low frequencies in Fig. 3) indicates the offset-free (i.e., zero- active power into the grid. Moreover, the capacitors’ voltages
error) current tracking performance of the proposed controller. are well-balanced to their desired DC voltage level, ensuring
Furthermore, the magnitude of the frequency response of the a well-formed nine-level voltage waveform. The results of
off-diagonal elements of the closed-loop PEC9 system is very this case study demonstrate that the proposed vector current
small, which provides decoupling between d and q compo- controller is able to track the reference set point fast and with
nents of the inverter current. In addition, Fig. 4 demonstrates zero offsets.
that the closed-loop system has a much higher bandwidth than
As one can observe from these figures, the proposed con-
the open-loop system, leading to the fast transient response of
troller tracks the current reference set point variations with a
the grid-interactive PEC9 inverter.
zero steady-state error and provides a satisfactory transient
B. Robust Stability Analysis response, as the synchronization between grid voltage and
To analyze the robust stability of the closed-loop grid- inverter current is well obtained. Furthermore, the results
tied PEC9 system with respect to uncertainties in the grid indicate that the proposed vector current controller shows
inductance, we consider the closed-loop state matrix Acl (λ ) = robust performance under both stiff and weak grid conditions.
A(λ ) + B(λ )K in (14). This matrix is subject to the polytopic The second case study evaluates the robustness feature of
uncertainty with the following q = 2 vertices, as here only grid the proposed vector current controller against the uncertainty
inductance uncertainty is considered: of grid inductance. To this end, the grid inductance is assumed
to have a step change from L = 5 mH to L = 2.5 mH. It is im-
Acl, j = A j + B j K, (20) portant to emphasize that in practice the grid inductance does
where A j and B j ; j = 1, . . . , q, are defined in (16) and obtained not change suddenly; however, the main purpose of this case
by substituting the minimum and maximum values of L, i.e., study is to demonstrate the robust stability and performance of
L and L̄. The closed-loop state matrix Acl (λ ) is robustly stable the proposed vector current strategy for different values of the
¯ all values of L ∈ [L L̄] if there exist P ≻ 0, S , and S
for grid inductance. Fig. 6 shows the results of this case study. The
j 1 2
so that the following set¯ of LMIs are satisfied [38]: results show that the variation of L (uncertainty in the value of
     L) does not compromise the current tracking performance of
0 Pj S1   the PEC9 inverter with the proposed vector current controller,
≺ Acl, j −I6
Pj 0 S2 since the grid connectivity requirement of the PEC9 inverter is
(21)
 T acquired, that is, the phase synchronization of the grid voltage
  
S1 
+ Acl, j −I6 , and inverter current; furthermore, the voltage signals across
S2

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

(a) (a)
100 100

V(V)
V(V)

50 50

4.95 5 5.05 5.1

0 (b)
3.95 4 4.05 4.1 4.15 4.2 4.25 4.3
200
(b)
0
Current change from 5A to 7A Current change from 7A to 5A
200 -200
4.95 5 5.05 5.1
0 (c)
8

ig,dq (A)
6
-200 4
3.95 4 4.05 4.1 4.15 4.2 4.25 4.3 2
0
-2
(c)
200 4.95 5 5.05 5.1
(d)
Vinv(V)

200

Vinv(V)
0
0

-200 -200
3.95 4 4.05 4.1 4.15 4.2 4.25 4.3 4.95 5 5.05 5.1
Time (s) Time (s)

Fig. 5. Dynamic responses of the grid-interactive PEC9 inverter to current Fig. 7. Performance of the conventional PI vector controller against a step
reference changes occurred at t = 4 s and t = 4.25 s: (a) voltages of the DC change in L at t = 5 s: (a) Voltages of the DC link capacitors (V1 (t) and V2 (t))
link capacitors (V1 (t) and V2 (t)) and voltage of the DC link Vdclink (t), (b) and voltage of the DC link Vdclink (t), (b) grid voltage and inverter current –
grid voltage and inverter current – for better visibility, the inverter current is for better visibility, the current is multiplied by 10 –, (c) dq components of
multiplied by 10 –, and (c) nine-level voltage waveform Vinv (t). the inverter current ig,dq (t), and (d) nine-level voltage Vinv (t).

100
(a) where KP and KI are the proportional and integral parameters.
The term ud (t) is added to cancel out the interaction compo-
V(V)

50 nents in the PEC9 current dynamics. Although, PI controllers

4.95 5 5.05 5.1 are largely used for grid-tied inverters to ensure dq current
(b) reference tracking, the design of such controllers becomes
200
harder when it is required to guarantee desired performance for
0
-200
grid-tied inverters operating under uncertain grid conditions,
4.95 5 5.05 5.1 from stiff to weaker.
(c)
8 For the comparative case study, the previous case study is
ig,dq (A)

6
4 repeated. To this end, the parameters of the conventional PI
2
0 controller in (22) are chosen to be KP = 5 and KI = 10. The
4.95 5 5.05 5.1
parameters are designed based on a design guideline provided
(d)
200 by [39]. The results of this case study are depicted in Fig. 7. As
Vinv(V)

0 one can observe from this figure, although the PI control can
-200 track the reference current signals with zero offsets, its tran-
4.95 5 5.05 5.1
Time (s)
sient performance under the grid inductance change occurred
at t = 5 s is poor. In other words, the conventional PI vector
Fig. 6. Robustness feature of the proposed H∞ vector controller against a step
current controllers are not inherently robust against the grid
change in L at t = 5 s: (a) Voltages of the DC link capacitors (V1 (t) and V2 (t)) inductance uncertainty. The comparative case study confirms
and voltage of the DC link Vdclink (t), (b) grid voltage and inverter current – the superiority of the proposed H∞ vector control approach
for better visibility, the current is multiplied by 10 –, (c) dq components of
the inverter current ig,dq (t), and (d) nine-level voltage Vinv (t).
over its conventional counterpart in terms of robustness to
parameters uncertainty and smooth transient responses.
the capacitors are self-balanced to a quarter of the DC input
D. Software-In-the-Loop Validation
voltage amplitude.
For a comparative case study, we consider the conventional The feasibility and performance validation of the proposed
Proportional-Integral (PI) vector current controller [39], struc- robust optimized vector current controller for a grid-interactive
tured as follows: PEC9 inverter in Fig. 1 is evaluated by a sofware-in-the-
  Z   
0 −1

loop (SIL) platform by using an OPAL-RT real-time digital
u(t) = KP i∗dq − y(t) + KI i∗dq − y(τ) dτ + ω0 L y(t) simulator. In the platform, the subsystems including PEC9 and
1 0
| {z }
ud (t)
the designed controller are modeled in the OPAL-RT.
+ (V (t) + d(t)),
To highlight the effectiveness of the proposed control
(22) scheme in current tracking, in the first test, we consider the

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

(a) Full experiment time window. (a) Full experiment time window.

(b) Zoomed version of results for magnitude changes from 2 A to 10 A. (b) Zoomed version of results for magnitude changes from 2 A to 10 A.

(c) Zoomed version of results for magnitude changes from 10 A to 6 A. (c) Zoomed version of results for magnitude changes from 10 A to 6 A.
Fig. 8. Dynamic response of the stiff-grid-interactive PEC9 inverter with Fig. 9. Dynamic response of the weak-grid-interactive PEC9 inverter with
equivalent DC-link capacitors voltage Vdclink (t), nine-level voltage Vinv (t), equivalent DC-link capacitors voltage Vdclink (t), nine-level voltage Vinv (t),
the grid voltage Vg (t), and the inverter current ig (t) under changes in the the grid voltage Vg (t), and the inverter current ig (t) under changes in the
magnitude of the reference current 2 → 10 → 6 A- Note that for better magnitude of the reference current 2 → 10 → 6 A- Note that for better
visibility, the grid current is multiplied by 10. visibility, the grid current is multiplied by 10.

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

(a) Full experiment time window. (a) Stiff grid conditions.

(b) Zoomed version of results. (b) Weak grid conditions.

Fig. 10. Dynamic response of the stiff-grid-interactive PEC9 inverter with Fig. 11. Dynamic response of the PEC9 inverter under non-ideal grid voltage
equivalent DC-link capacitors voltage Vdclink (t), nine-level voltage Vinv (t), the conditions with equivalent DC-link capacitors voltage Vdclink (t), nine-level
grid voltage Vg (t), and the inverter current ig (t) under changes in the phase voltage Vinv (t), the grid voltage Vg (t), and the inverter current ig (t)- Note that
of the reference current 0 → π4 → 0 rad. for better visibility, the grid current is multiplied by 10.

current tracking under both stiff and weak grid conditions with
Lg = 0 mH (Lg is only the grid inductance and L ̸= Lg ) and
Lg = 5 mH. In this test, the current reference is initially set
to i∗g (t) = 2 sin(ω0t) where the grid voltage is in phase with
the inverter current to achieve a unity power factor on the
grid side. The current reference setpoint is then changed to
i∗g (t) = 10 sin(ω0t) and returns to i∗g (t) = 6 sin(ω0t). Fig. 8 and
Fig. 9 depict the dynamic response of the grid-tied PEC9
including the DC-link voltage Vdclink (t), nine-level voltage
waveform Vinv (t), the grid voltage Vg (t) and the inverter current
ig (t) for both stiff and weak grid conditions. Note that for
better visibility, the current is multiplied by 10 in Fig. 8, Fig. 9,
and Fig. 10.
The second test evaluates the current reference tracking
performance of the proposed vector control strategy. To this
end, the phase of the reference current i∗g (t) = 2 sin(ω0t + θ ) Fig. 12. Experimental setup of grid-connected PEC9 inverter.
is subject to changes from 0 to π4 to 0 rad. The results of this

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

test are shown in Fig. 10. Simulink and a hardware-in-the-loop setup. The simulation
In the third case study, the performance of the proposed and hardware-in-the-loop results demonstrate the effectiveness
robust vector current control strategy is assessed against non- and robust performance of the proposed H∞ robust vector
ideal grid voltage conditions. It is assumed that i∗g (t) = control scheme for the grid-interactive PEC9 inverter under
6 sin(ω0t). The results of this case study for a stiff and weak uncertain grid inductance. The future scope of this work
distorted grid are shown in Fig. 11. Under the non-ideal includes the control design for multi-grid-tied inverters and
grid voltage, the proposed robust vector current controller analysis of their interactions.
attenuates the impact of distorted grid voltage on the inverter
current. Hence, the voltage across series-connected capacitors VI. A PPENDIX
is maintained at 90 V , and the grid current is sinusoidal, Proof of Theorem 1. According to [40], the inequal-
tracks the reference current i∗g (t), and is in phase with the grid ity conditions in (17a) for j = 1, . . . , q guarantee the ro-
voltage. Moreover, as one can see from Fig. 11, the inverter bust stability of the closed-loop system and robust perfor-
voltage produces a satisfactory nine levels under non-ideal grid mance ∥W (s)S(s)∥∞ ≤ γ with a Lyapunov function V (t) =
voltage conditions. xT (t)P−1 x(t) for all range of L ∈ [L L̄] and R ∈ [R R̄].
Furthermore, as the proposed control¯ law in (9) contains ¯ an
E. Hardware-In-the-Loop Experimental Validation integral term, the offset-free tracking of the constant reference
In the following, the performance and effectiveness of the current signals i∗dq is ensured. Finally, the inequality conditions
√
proposed robust vector current control scheme for a grid- in (17b) ensure that P−1 2 < β and ∥H∥2 < η. As a result,
interactive PEC9 inverter is evaluated by a hardware-in-the- √
∥K∥2 < ηβ .
loop experimental setup, depicted in Fig. 12.
To demonstrate the effectiveness of the proposed vector R EFERENCES
current control scheme under both stiff (Lg = 0 mH) and weak [1] M. S. Sadabadi, A. Haddadi, H. Karimi, and A. Karimi, “A robust active
(Lg = 5 mH) grid conditions, the case of current tracking is damping control strategy for an LCL-based grid-connected DG unit,”
IEEE Transactions on Industrial Electronics, vol. 64, no. 10, pp. 8055–
considered. To this end, the reference current is under changes 8065, 2017.
from i∗g (t) = 4 sin(ω0t) to i∗g (t) = 9 sin(ω0t). The results of this [2] J. He and Y. W. Li, “Analysis, design, and implementation of virtual
case study are shown in Fig. 13. The THD for the grid voltage impedance for power electronics interfaced distributed generation,” IEEE
Transactions on Industry Applications, vol. 47, no. 6, pp. 2525–2538,
and current under both grid conditions is analyzed in Fig. 14. 2011.
As seen from this figure, the THD of the grid side current is [3] X. Wang, Y. W. Li, F. Blaabjerg, and P. C. Loh, “Virtual-impedance-
below the standard limit of 5% [32]. based control for voltage-source and current-source converters,” IEEE
Transactions on Power Electronics, vol. 30, no. 12, pp. 7019–7037,
In the second case study, the reference current changes from 2015.
i∗g (t) = 4 sin(ω0t) to i∗g (t) = 4 sin(ω0t + π4 ). The results of this [4] H. Karimi, A. Haddadi, M. Karimi-Ghartemani, and M. Sadabadi, “A
case study, depicted in Fig. 15, confirm the effectiveness of the robust vector current controller with negative-sequence current capability
for grid-connected inverters,” energies, vol. 14, pp. 1–22, 2021.
proposed vector current control framework in current tracking. [5] S. Yazdani, M. Ferdowsi, and P. Shamsi, “Internal model based smooth
In the last case study, the performance and robustness transition of a three-phase inverter between islanded and grid-connected
of the proposed vector current framework are tested under modes,” IEEE Transactions on Energy Conversion, vol. 35, no. 1, pp.
405–415, 2020.
grid voltage disturbances. To this purpose, it is assumed [6] R. A. Rana, S. A. Patel, A. Muthusamy, C. W. Lee, and H.-J. Kim,
that the amplitude of the grid voltage is subject to changes “Review of multilevel voltage source inverter topologies and analysis of
from 170 V to 155 V. Fig. 16 shows the PEC current and harmonics distortions in FC-MLI,” Electronics, vol. 8, no. 1, pp. 1–37,
2019.
voltage trajectories. The results of this case study demonstrate [7] M. S. Sadabadi, M. Sharifzadeh, M. Mehrasa, S. Bacha, and K. Al-
that the proposed control scheme provides robustness to grid Haddad, “Convex optimization-based vector current control design for
disturbance and current tracking is performed regardless of grid-connected packed E-Cell inverters,” in 2021 IEEE Energy Conver-
sion Congress and Exposition (ECCE), Vancouver, BC, Canada, 2021,
changes in the grid voltage magnitude. pp. 2607–2612.
[8] S. Kouro, M. Malinowski, K. Gopakumar, J. Pou, L. G. Franquelo,
V. C ONCLUDING R EMARKS B. Wu, J. Rodriguez, M. A. Pérez, and J. I. Leon, “Recent advances
This paper proposes a modeling and robust control frame- and industrial applications of multilevel converters,” IEEE Trans. Ind.
Electron., vol. 57, no. 8, pp. 2553–2580, Aug. 2010.
work for grid-interactive PEC9 inverters under grid impedance [9] K. K. Gupta, A. Ranjan, P. Bhatnagar, L. K. Sahu, and S. Jain,
uncertainty, encompassing stiff and weaker grid conditions “Multilevel inverter topologies with reduced device count: A review,”
non-ideal grid conditions due to unbalanced conditions and IEEE Trans. Power Electron., vol. 31, no. 1, pp. 135–151, Jan. 2016.
[10] N. Vosoughi, S. H. Hosseini, and M. Sabahi, “A new transformer-
harmonic distortion. To address the uncertainty in the grid less five-level grid-tied inverter for photovoltaic applications,” IEEE
impedance and non-ideal grid conditions, a multivariable Transactions on Energy Conversion, vol. 35, no. 1, pp. 106–118, 2020.
dynamical polytopic model, a convex hull of two given [11] R. H. Baker and L. H. Bannister, “Electric power converter,” 1975, US
Patent 386743A.
models, is developed where the non-ideal grid is modeled [12] T. A. Meynard and H. Foch, “Multi-level choppers for high voltage
as a disturbance. The control problem for such a polytopic applications,” European Power Electronics and Drives Journal, vol. 2,
model is then formulated as a convex optimization problem no. 1, pp. 45–50, 1992.
[13] A. Nabae, I. Takahashi, and H. Akagi, “A new Neutral-Point-Clamped
whose solution delivers a robust 2DOF multivariable PI active PWM inverter,” IEEE Transactions on Industry Applications, vol. IA-17,
vector controller. Different case studies are carried out in no. 5, pp. 518–523, Sept. 1981.

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

(a) Stiff grid conditions.

(a) Stiff grid conditions.

(b) Weak grid conditions.

Fig. 14. THD analysis of the grid voltage and current when the proposed
(b) Weak grid conditions. active current controller is in place.
Fig. 13. Dynamic response of the PEC9 inverter under both stiff and weak
grid conditions with equivalent DC-link capacitors voltage Vdclink (t), nine-
level voltage Vinv (t), the grid voltage Vg (t), and the inverter current ig (t)
under changes in the magnitude of the reference current 4 → 9 A- Note that
for better visibility, the grid current is multiplied by 10.

[14] Y. Ounejjar, K. Al-Haddad, and L. Gregoire, “Packed U Cells multilevel

converter topology: Theoretical study and experimental validation,”
IEEE Trans. on Ind. Electron., vol. 58, no. 4, pp. 1294–1306, Apr. 2011.
[15] M. Sharifzadeh and K. Al-Haddad, “Packed e-cell (PEC) converter
topology operation and experimental validation,” IEEE Access, vol. 7,
pp. 93 049–93 061, Jun. 2019.
[16] M. Eskandari and A. V. Savkin, “Robust PLL synchronization unit for
grid-feeding converters in micro/weak grids,” IEEE Transactions on
Industrial Informatics, vol. 19, no. 4, pp. 5400–5411, 2022.
[17] Y. Liao, X. Wang, X. Yue, and L. Harnefors, “Complex-valued multi-
frequency admittance model of three-phase VSCs in unbalanced grids,”
IEEE Journal of Emerging and Selected Topics in Power Electronics,
vol. 8, no. 2, pp. 1934–1946, 2019.
[18] “An enhanced phase-locked loop for non-ideal grids combining linear
active disturbance controller with moving average filter,” International
Journal of Electrical Power & Energy Systems, vol. 149, p. 109021,
2023.
[19] F. Sebaaly, M. Sharifzadeh, H. Y. Kanaan, and K. Al-Haddad, “Multi- Fig. 15. Dynamic response of the stiff-grid-interactive PEC9 inverter with
level switching-mode operation of finite-set model predictive control for equivalent DC-link capacitors voltage Vdclink (t), nine-level voltage Vinv (t), the
grid-connected packed E-cell inverter,” IEEE Transactions on Industrial grid voltage Vg (t), and the inverter current ig (t) under changes in the phase
Electronics, vol. 68, no. 8, pp. 6992–7001, 2021. of the reference current 0 → π4 rad.
[20] M. Babaie, M. Sharifzadeh, M. Mehrasa, and K. Al-Haddad, “Optimized

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.
 This article has been accepted for publication in IEEE Transactions on Energy Conversion. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/TEC.2024.3471191

[31] L. C. Borin, G. V. Hollweg, C. R. Osório, F. M. Carnielutti, R. C.

Oliveira, and V. F. Montagner, “Automated test-driven design for robust
optimized current control of GTIs operating under normal and abnormal
grid conditions,” Control Engineering Practice, vol. 143, p. 105790,
2024.
[32] “IEEE standard for interconnection and interoperability of distributed
energy resources with associated electric power systems interfaces,”
IEEE Std 1547-2018 (Revision of IEEE Std 1547-2003), pp. 1–138,
2018.
[33] M. Karimi-Ghartemani, “A unifying approach to single-phase syn-
chronous reference frame PLLs,” IEEE Trans. Power Electron., vol. 28,
no. 10, pp. 4550–4556, Oct. 2013.
[34] S. Boyd, L. E. Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix
Inequalities in System and Control Theory. Philadelphia: SIAM, 1994.
[35] S. Skogestad and I. Postlethwaite, Multivariable Feedback Control
Analysis and design. Wiley, Second Edition, 2001.
[36] J. Löfberg, “YALMIP: A toolbox for modeling and optimization in
MATLAB,” in Proc. IEEE Int. Symp. Comp. Cont. Syst. Design
(CACSD), 2004. [Online]. Available: http://control.ee.ethz.ch/∼joloef/
yalmip.php
[37] MOSEK ApS, The MOSEK optimization software, 2011. [Online].
Fig. 16. Dynamic response of the grid-interactive PEC9 inverter with Available: http://www.mosek.com
equivalent DC-link capacitors voltage Vdclink (t), nine-level voltage Vinv (t), [38] M. S. Sadabadi and D. Peaucelle, “From static output feedback to
the grid voltage Vg (t), and the inverter current ig (t) under grid voltage structured robust static output feedback: A survey,” Annual Reviews in
disturbances. Control, vol. 42, pp. 11–26, 2016.
[39] M. Kazmierkowski and L. Malesani, “Current control techniques for
three-phase voltage-source pwm converters: a survey,” IEEE Transac-
tions on Industrial Electronics, vol. 45, no. 5, pp. 691–703, 1998.
[40] A. Isidori, Lectures in Feedback Design for Multivariable Systems.
based algorithm first order sliding mode control for grid-connected Springer, 2016.
packed E-Cell (PEC) inverter,” in IEEE Energy Conversion Congress
and Exposition (ECCE), Baltimore, MD, USA, Sept.-Oct. 2019, pp.
2269–2273.
[21] J. Liu, X. Shen, A. M. Alcaide, Y. Yin, J. I. Leon, S. Vazquez, L. Wu,
and L. G. Franquelo, “Sliding mode control of grid-connected neutral-
point-clamped converters via high-gain observer,” IEEE Transactions on
Industrial Electronics, vol. 69, no. 4, pp. 4010–4021, 2022.
[22] M. Mehrasa, M. Babaie, A. Zafari, and K. Al-Haddad, “Passivity
ANFIS-based control for an intelligent compact multilevel converter,”
IEEE Trans. Ind. Inform., vol. 17, no. 8, pp. 5141–5151, 2021.
[23] F. Sebaaly, H. Y. Kanaan, J. Rodriguez, and K. Al-Haddad, “Direct
predictive control for a nine-level packed E-Cell (PEC9) converter based
shunt active power filter (SAPF),” in IECON 2021 – 47th Annual
Conference of the IEEE Industrial Electronics Society, Toronto, ON,
Canada, Oct. 2021, pp. 1–7.
[24] M. S. Sadabadi, M. Sharifzadeh, M. Mehrasa, H. Karimi, and K. Al-
Haddad, “Decoupled dq current control of grid-tied packed E-Cell
inverters in vehicle-to-grid technologies,” IEEE Trans. Ind. Electron.,
vol. 70, no. 2, pp. 1356–1366, Feb. 2023.
[25] M. Mehrasa, M. Babaie, M. Sharifzadeh, S. Bacha, and K. Al-Haddad,
“An adaptive fuzzy passivity-based control strategy for grid-tied packed
E-Cell converter,” in IECON 2021 – 47th Annual Conference of the
IEEE Industrial Electronics Society, Toronto, ON, Canada, Oct. 2021,
pp. 1–6.
[26] M. Gheisarnejad, M. Sharifzadeh, M. Khooban, and K. Al-Haddad,
“Adaptive fuzzy Q-learning control design and application to grid-
tied nine-level packed E-Cell (PEC9) inverter,” IEEE Transactions on
Industrial Elecetronics, vol. 70, no. 1, pp. 1071–1076, 2023.
[27] M. Mehrasa, M. Babaie, M. Sharifzadeh, and K. Al-Haddad, “An
input–output feedback linearization control method synthesized by arti-
ficial neural network for grid-tied packed E-Cell inverter,” IEEE Trans.
Ind. Appl., vol. 57, no. 3, pp. 3131–3142, 2021.
[28] M. Davari and Y. A.-R. I. Mohamed, “Robust vector control of
a very weak-grid-connected voltage-source converter considering the
phase-locked loop dynamics,” IEEE Transactions on Power Electronics,
vol. 32, no. 2, pp. 977–994, 2016.
[29] J. Wang, I. Tyuryukanov, and A. Monti, “Design of a novel robust current
controller for grid-connected inverter against grid impedance variations,”
International Journal of Electrical Power & Energy Systems, vol. 110,
pp. 454–466, 2019.
[30] Z. Zhang, P. Wang, P. Jiang, F. Gao, L. Fu, and Z. Liu, “Robust
control method of grid-connected inverters with enhanced current quality
while connected to a weak power grid,” IEEE Transactions on Power
Electronics, vol. 37, no. 6, pp. 7263–7274, 2022.

Authorized licensed use limited to: Deakin University. Downloaded on November 04,2024 at 10:48:23 UTC from IEEE Xplore. Restrictions apply.
© 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See https://www.ieee.org/publications/rights/index.html for more information.