Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/355985225

Finite Control Set-Model Predictive Control Synthesized with SVPWM for quasi
Z-Source Inverter

Conference Paper · November 2021


DOI: 10.1109/COMPEL52922.2021.9646049

CITATIONS READS
0 45

2 authors, including:

Mohammed Adel Alhasheem


Arab Academy for Science, Technology & Maritime Transport
22 PUBLICATIONS   84 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Reliable Power Electronic Based Power System. View project

pOwer electroNics baSEd power sysTem (ONSET) View project

All content following this page was uploaded by Mohammed Adel Alhasheem on 05 January 2022.

The user has requested enhancement of the downloaded file.


Finite Control Set-Model Predictive Control
Synthesized with SVPWM for quasi Z-Source
Inverter
Abualkasim Bakeer∗† and Mohammed Alhasheem‡
∗ Dept. of Electrical Engineering, Faculty of Engineering, Aswan University, Aswan 81542, Egypt
† Dept. of Electrical Power Engineering and Mechatronics, Tallinn University of Technology, 19086 Tallinn, Estonia
‡ Dept. of Electrical and Control, Arab Academy for Science, Technology, and Maritime Transport, 2033 Cairo, Egypt

Abstract—This paper presents a fixed-switching operation capacitor voltage, and load current in case of a standalone RL
for the quasi Z-source inverter (qZSI) controlled by a finite load [5], grid-tied applications [6], and drives systems [7].
control set-model predictive control (FCS-MPC). This goal is These algorithms have the drawback of high computational
accomplished through integrating the FCS-MPC controller and
space vector pulse width modulation (SVPWM). The FCS-MPC burdens, as the cost function should be evaluated for the whole
algorithm defines the optimal switching sector and calculates the possible switching vectors. A powerful FCS-MPC algorithm
dwell time for the selected two active, zero, and shoot-through is proposed to reduce the computational power and simplify
vectors. Additionally, the optimal time for the shoot-through the cost function by using only one weighting factor for the
vector is determined using the corresponding cost function with capacitor voltage instead of two [8]. This powerful methodol-
a tuning parameter to adjust its value throughout the switching
period. The qZSI controlled variables consist of the inductor ogy depends on using a separate cost function for the inductor
current, capacitor voltage, and RL load current. In this way, a current that will be evaluated only twice at shoot-through case
fixed-switching frequency behavior is obtained while providing and non shoot-through case, then the decision can be taken
high-power quality. The theoretical analysis of the proposed to continue checking the whole non shoot-through vectors.
approach will be provided and confirmed using simulation results Later, the FCS-MPC based logical operation is proposed to
based on MATLAB/Simulink software.
Index Terms—Quasi Z-source inverter (qZSI), model predictive eliminate the weighting factors inside the cost function [9].
control (MPC), space vector pulse width modulation (SVPWM), In all these above-mentioned methods, the optimal switching
dwell time calculations. vector is applied during the whole sampling period, resulting
in a low performance of qZSI.
I. I NTRODUCTION To achieve the fixed-switching frequency operation with
qZSI controlled by FCS-MPC, the modulation stage is re-
FCS-MPC has several distinguishing characteristics, in-
quired. In [10], the modulated MPC (M2 PC) is proposed to
cluding quick response, low complexity, ease of constraint
compensate the delay during the FCS-MPC implementation
inclusion, and nonlinearity are obtained [1]. But due to the
with qZSI by combining both algorithms in [8], [11]. The
action of the FCS-MPC, which is applied directly to the
shoot-through duty-cycle is predicted as a continuous value
converter switches without any modulation stage, it has the
to balance the dc-link voltage, and the modulation index is
disadvantage of variable switching frequency [2]. Moreover,
predicted to regulate the load current in [12]. This can lead
the average switching frequency with FCS-MPC is difficult
to a fast tracking for the target variables and achieve the
to predict, as it depends on the converter operating point as
fixed-switching operation of qZSI. The same procedure has
well as the control design [3]. Several challenges may develop
been followed for the permanent magnet synchronous machine
when the converter operates at a variable switching frequency,
applications, but with the basic ZSI topology in [13].
including circuit resonance, magnetic component and filter
This work discusses the application of FCS-MPC to the
complexity, and a lack of power quality for critical applications
qZSI while achieving the fixed-switching operation. Following
as the harmonics spectrum spread in a wide range [4].
the same procedure as in the SVPWM, the time of each vector
Quasi Z-source inverter is one of the promising topolo-
should be calculated and optimized to achieve a certain target
gies in the single-stage configuration that has the buck and
[14]. These are referred to as dwell times. This modification
boost functionality with continuous input current and lower
is proposed to address the shoot-through state and provide
capacitor voltage stress compared to the basic circuit of ZSI
high power quality for the qZSI. This is accomplished by
[5]. Employing FCS-MPC with qZSI is accomplished with a
combining the FCS-MPC and SVPWM to select the optimal
simple way to improve the dynamic of the inductor current,
switching sector and calculate the dwell times for each applied
Corresponding author: Abualkasim Bakeer (e-mail: abualka- vector in the next sampling interval. The remaining of this
sim.bakeer@aswu.edu.eg). work is laid out as follows: the discrete model for qZSI
control objectives is introduced in Section II. While Section III The value of the future load current io ((k+1)T ) by which
addresses the proposed fixed-switching FCS-MPC algorithm the trajectory of the controlled load current can be predicted,
with the qZSI. Section IV displays the simulation results to is given by:
validate the algorithm and test the qZSI performance using
various test conditions, while the conclusion is reported in Ts Vx ((k + 1)T ) + Lio (kT )
io ((k + 1)T ) = (1)
Section V. L + RTs

where Vx ((k+1)T ) is the space vector of the qZSI output


II. F INITE C ONTROL S ET-M ODEL P REDICTIVE C ONTROL voltage at the load terminals (V0 : V7 ) in (α,β) coordinate
OF Q ZSI system as in (2), x is the number of the state signals vector
x ∈ [0:7], io (kT ) is the actual load current that includes two
The ZSI class has a unique switching state, called the shoot- components in (α,β), Ts is the sampling time, kT and (k+1)T
through state, when the boosting action is required. The aim of are the sampling interval, R and L are the resistance and
the control stage on the dc-side is to keep the dc-link voltage inductance of the standalone RL load, respectively.
fixed at a specified level to avoid stress across the switching
devices and to ensure normal operation of the qZSI. The 2
Vdc Q1 + aQ2 + a2 Q3

Vx ((k + 1)T ) = (2)
capacitor voltage, vC1 , is considered a simple way to achieve 3
this goal, where detecting the peak dc-link voltage is complex
due to the requirement of extra hardware circuits. As a result, where Vdc is the peak value for the dc-link voltage, which
the control complexity and cost may rise. Due to the presence equals (2vC1 (kT ) – vin (kT )) and vin (kT ) is the input voltage.
of a zero on the right half-plane in qZSI modeling, the dc- The reference value of the load current is computed according
side of qZSI has the non-minimum phase criteria [15]. Thus, to the desired output power (Po,ref ) as follows [17]:
the control design is difficult and constrained resulting in that r
the inductor current must be adjusted to control the capacitor 2Po,ref
i∗o ((k + 1)T ) = . (3)
voltage. The fundamental requirement on the ac side of the 3R
converter is the current or voltage control according to the
loading type. Therefore, the use of model predictive control 2) Capacitor Voltage and Inductor Current: These control
can solve this issue related to ZSI family. variables have different discrete models according to the
The algorithm of model predictive control is considered as switching state of the converter.
an optimization problem, which takes into account the accurate Non Shoot-through Case: The equivalent circuit of qZSI
model of the converter and its associated load. Furthermore, during this state is depicted in Fig. 3a. It includes the states
in the case of the qZSI, the model should include the inductor of x ∈ [0:6]. The future inductor current iL1 ((k+1)T ) and the
current and the capacitor voltage as controlled objectives to future capacitor voltage vC1 ((k+1)T ) during this case can be
ensure the stability of the converter [16]. In the following, given as:
the complete discrete model of qZSI control objectives is
discussed. Ts (vin (kT ) − vC1 (kT )) + L1 iL1 (kT )
iL1 ((k +1)T ) = (4)
L1 + RL1 Ts
A. Discrete Model of FCS-MPC for qZSI

As mentioned above, the first step in building the FCS- Ts


MPC is to identify the discrete model of the control objectives vC1 ((k + 1)T ) = vC1 (kT ) + (iL ((k + 1)T )
C1 1 (5)
during the possible switching states of the converter. The entire −iinv ((k + 1)T ))
system of qZSI with the proposed controller is illustrated in
Fig. 1. where RL1 is the equivalent series resistance (ESR) of the
1) Three-Phase Load Current: The actual three-phase cur- qZS inductor, L1 and C1 are the inductance and capacitance
rents will be measured as feedback signals to the control values of the qZSI network, respectively.
block, as shown in Fig. 1. These signals are transformed into Shoot-through Case: The equivalent circuit of qZSI during
stationary frame (α, β), resulting in two components instead of this state is shown in Fig. 3b. The discrete model of inductor
three. The three-phase inverter has six switched devices (Q1 : current and capacitor voltage can be determined as:
Q6 ) with possible eight switching states. The space vectors
of the inverter output voltage across the load terminals during Ts vC1 (kT ) + L1 iL1 (kT )
the eight switching states are depicted in Fig. 2. In the ZSI iL1 ((k + 1)T ) = (6)
L1 + RL1 Ts
family, the eight switching states could be divided into six
states as active vectors (V1 : V6 ), zero vector (V0 ), and shoot-
through vector (V7 ). When applying the shoot-through vector, Ts
all converter switches will be turned on at the same time. vC1 ((k + 1)T ) = vC1 (kT ) − iL ((k + 1)T ). (7)
C1 1
C2

+ −
vC2
L1 iL1 RL1 Dz L2 iL2 RL2

Q1 D1 Q3 D3 Q5 D5 R ia L
+
+ ib
vin vC1 C1 ic


Q2 D2 Q4 D4 Q6 D6

Q1 Q2 Q3 Q4 Q5 Q6

SVPWM

Sn d1 d2 d0 dsh
vC1 (kT )
io (kT )
FCS-MPC Optimization
iL1 (kT )

Po,ref VC1

Fig. 1: Configuration of qZSI controlled by the proposed FCS-MPC combined with SVPWM.

C2
β
+ −
V3 V2 vC2
S2
L1 iL1 RL1 L2 iL2 RL2

S3 S1
+
+
vin vC1 C1 iinv


V4 V0 ,V7 V1 α

S4 S6 (a) Non shoot-through case


C2

S5 + −
vC2
V5 V6 L1 iL1 RL1 L2 iL2 RL2

Fig. 2: Space vector diagram of qZSI output voltage (Vx ).


+
+
vin vC1 C1
B. Cost Function Definition −

The following cost function is adopted to achieve the qZSI
controlled objectives [15]
(b) Shoot-through case
J = (i∗oα − ioα ((k + 1)T ))2 + (i∗oβ − ioβ ((k + 1)T ))2 +
Fig. 3: Cases of qZSI operation.
∗ ∗
λC (VC1 − vC1 ((k + 1)T )) + λL (IL1 − iL1 ((k + 1)T ))
(8)
∗ ∗
where VC1 is the voltage reference of C1 , IL1 is the current power, λC and λL are the weighting factors of capacitor C1
reference of inductor L1 , which is defined from the required voltage and inductor L1 current, respectively.
III. P ROPOSED F IXED - SWITCHING F REQUENCY FCS-MPC the required switching pattern. The implemented SVPWM
FOR Q ZSI WITH SVPWM strategy here is that utilizes 6 intervals for the shoot-through
In space vector modulation, it is possible to define each time throughout the switching period, as reported in [18]-
available vector in the qZSI in the (αβ) plane, as shown in [19]. The complete instructions of the proposed FCS-MPC
Fig. 2. The control technique calculates the prediction of the algorithm are given in Algorithm 1.
two active vectors that compose each sector at every sampling
IV. S IMULATION R ESULTS AND D ISCUSSION
time (Ts ) and evaluates the cost function. The cost function,
defined by (8), is evaluated for each sector. Each prediction is Here, in this section, we will show the performance of the
evaluated, where the duty-cycle for the two active vectors and proposed fixed-switching FCS-MPC combined with SVPWM
zero vector are calculated by solving the following equation: for qZSI with the system parameters listed in Table I. It is
worth noting that the input measurements from the qZSI topol-
δ ogy are obtained at each sample period, which corresponds
di = (9)
Ji to the SVPWM pattern’s switching frequency. The value of
where δ is the constant of proportionality, the subscript i weighting factor λC and λL are 3.75 and 5.20, respectively.
denotes the adjacent vectors, in this case (i = 1; 2), i = 0 The factor σ is set at 0.85 to increase the occupied time of the
corresponds to the duty-cycle of a zero vector, and i= 7 cor- shoot-through vector compared to the other vectors (i.e., two
responds to shoot-through. The sum of duty-cycle associated active vectors and the zero vector) during the switching period.
with the two active vectors, zero vector, and shoot-through All the inputs of the SVPWM block, which are generated from
vector equals unity as:
d1 + d2 + d0 + dsh = 1. (10) Algorithm 1: Pseudocode of the proposed fixed-
switching FCS-MPC for qZSI.
By solving the previous equations (9-10), it is possible to
obtain the value of δ and the duty-cycle for each vector as: 1 Measure iL1 (kT ), vC1 (kT ), io (kT );
2 Set Jopt = ∞;
3 Predict iL1 ((k + 1)T ), vC1 ((k + 1)T ), io ((k + 1)T ) at
σJ2 J0 Jsh
d1 = (11) the zero vector;
J1 J2 J0 + σJ1 J2 Jsh + σJ1 J0 Jsh + σJ2 J0 Jsh
4 Calculate J0 based on Vx = 0 using (8);
σJ1 J0 Jsh 5 Predict iL1 ((k + 1)T ), vC1 ((k + 1)T ), io ((k + 1)T ) at
d2 = (12)
J1 J2 J0 + σJ1 J2 Jsh + σJ1 J0 Jsh + σJ2 J0 Jsh the shoot-through vector;
σJ1 J2 Jsh 6 Calculate Jsh based on Vx = 0 using (8);
d0 = (13) 7 Predict iL1 ((k + 1)T ) at the active vector;
J1 J2 J0 + σJ1 J2 Jsh + σJ1 J0 Jsh + σJ2 J0 Jsh
8 for x = 1 : 6 do
J1 J2 J0 9 Predict vC1 ((k + 1)T ), io ((k + 1)T ) using
dsh = (14)
J1 J2 J0 + σJ1 J2 Jsh + σJ1 J0 Jsh + σJ2 J0 Jsh 10 Vx = S(x)Vdc ;
where σ is a tuning parameter to adjust the shoot-through 11 Calculate J1 using (8);
time for better performance, and it is associated with the cost 12 if x < 6 then
function during the shoot-through vector (i.e., Jsh ). 13 Predict vC1 ((k + 1)T ), io ((k + 1)T ) using
Considering these expressions, as the new cost function, 14 Vx = S(x + 1)Vdc ;
which is evaluated at every Ts , and defined as: 15 Calculate J2 using (8);
16 else
g(k + 1) = d1 J1 + d2 J2 . (15) 17 Predict vC1 ((k + 1)T ), io ((k + 1)T ) using
The two vectors that minimize the cost function and the 18 Vx = S(1)Vdc ;
zero vector are selected and applied in the next sampling 19 Calculate J2 using (8);
interval. After selecting the optimal two vectors and defining 20 Calculate d1 (x), d2 (x), d0 (x), dsh (x) using
the corresponding duty-cycle of each vector, the following (11-14);
equations are defining the time for each applied vector in one 21 Evaluate J(x) = d1 (x)J1 + d2 (x)J2 ;
sampling period: 22 if J(x) < Jopt then
23 Set Jopt = J(x);
T1 = d1 Ts 24 Set d1 = d1 (x);
25 Set d2 = d2 (x);
T2 = d2 Ts
(16) 26 Set d0 = d0 (x);
T0 = d0 Ts 27 Set dsh = dsh (x);
Tsh = dsh Ts . 28 Set Sn = x;
The generated optimal switching sector and dwell times, 29 Calculate dwell times T1 , T2 , T0 , Tsh using (16);
from the FCS-MPC stage, fed the SVPWM stage to generate
TABLE I: Simulation parameters of qZSI.
tor current waveform. Furthermore, Fig. 6c shows the output
Parameter Symbol Value three-phase currents, which indicate the high performance of
Input voltage vin 100 V the control scheme in the case of the converter feeding a linear
qZSI inductance L1 = L2 2.5 mH RL load.
qZSI capacitance C1 = C2 470 µF
ESR of qZSI inductor RL1 = RL2 0.1 mΩ V. C ONCLUSION
Load inductance L 3 mH
Load resistance R 5Ω In this paper, the FCS-MPC of qZSI is modified to allow
Switching frequency Fsw 20 kHz the operation at a fixed-switching frequency while providing
Sampling time Ts 50 µs
high-power quality. The modification is based on optimally
calculate the time for the shoot-through vector using the
corresponding cost function and then use the SVPWM. This
enables the control system to enhance the performance and
reduce the THD of RL load. The system continues to have
a fixed-switching frequency and a fast dynamic response as
well. In this regard, the system’s steady-state and transient
performance become equivalent to those of the conventional
FCS-MPC.
Fig. 4: FFT spectrum analysis of the qZSI phase current at the
switching frequency of 20 kHz.

(a) Inductor current

Fig. 5: Zoom view of the qZSI dc-link voltage.

the FCS-MPC algorithm, are updated once at each sampling


interval by calling the MATLAB script of the FCS-MPC. By
utilizing the proposed control method, the sampling period
is 50 µs and therefore the switching frequency equals 20
kHz. Fig. 4 validates that the switching frequency operation is
fixed at 20 kHz, and the harmonics spectrum appears on the (b) Capacitor voltage
multiplication of the switching frequency. As a result, the total
harmonic distortion (THD) is calculated around at 3.46%. This
is also achieved as a result of efficiently calculating the dwell
times for the qZSI vectors, considering the optimal duration
of the shoot-through vector. The dc-link voltage of the qZSI is
shown in Fig. 5. It has a pulsating shape, where its value equals
zero six times when the shoot-through vector is applied. This
complies with the utilized SVPWM. Moreover, its maximum
value equals 250 V, corresponding to (2vC1 − vin ) when one
of the two active vectors is applied. (c) Load current
Figure 6 shows the dynamic of the qZSI controlled objec-
tives when the reference output power is increased from 300 Fig. 6: Performance of qZSI controlled objectives.
W to 600 W (i.e., 100%) at the instant 0.25 s. The capacitor
voltage vC1 is maintained fixed at its reference (i.e., at 175 V) R EFERENCES
after the power change, see Fig. 6b. Furthermore, due to the
[1] I. S. Mohamed, S. Rovetta, T. D. Do, T. Dragicević, and A. A. Z. Diab,
nature of FCS-MPC, the performance of the proposed control “A neural-network-based model predictive control of three-phase inverter
method preserved the fast response characteristic in the induc- with an output lc filter,” IEEE Access, vol. 7, pp. 124 737–124 749, 2019.
[2] C. Jiang, G. Du, F. Du, and Y. Lei, “A fast model predictive control with [11] A. Olloqui, J. Elizondo, M. Rivera, M. Macías, O. Micheloud, R. Pena,
fixed switching frequency based on virtual space vector for three-phase and P. Wheeler, “Modulated model predictive rotor current control (m 2
inverters,” in IEEE International Power Electronics and Application pc) of a dfig driven by an indirect matrix converter with fixed switching
Conference and Exposition (PEAC), 2018, pp. 1–7. frequency,” in IEEE 2nd Annual Southern Power Electronics Conference
[3] N. Panten, N. Hoffmann, and F. W. Fuchs, “Finite control set model (SPEC), 2016, pp. 1–6.
predictive current control for grid-connected voltage-source converters [12] Y. Guo, H. Sun, Y. Zhang, Y. Liu, X. Li, and Y. Xue, “Duty-cycle
with lcl filters: A study based on different state feedbacks,” IEEE predictive control of quasi-z-source modular cascaded converter based
Transactions on Power Electronics, vol. 31, no. 7, pp. 5189–5200, 2015. photovoltaic power system,” IEEE Access, vol. 8, pp. 172 734–172 746,
[4] Y. Yang, H. Wen, and D. Li, “A fast and fixed switching frequency model 2020.
predictive control with delay compensation for three-phase inverters,” [13] H. Mahmoudi, M. Aleenejad, and R. Ahmadi, “Modulated model predic-
IEEE Access, vol. 5, pp. 17 904–17 913, 2017. tive control for a z-source-based permanent magnet synchronous motor
[5] A. Bakeer, M. A. Ismeil, A. Kouzou, and M. Orabi, “Development of drive system,” IEEE Transactions on Industrial Electronics, vol. 65,
mpc algorithm for quasi z-source inverter (qzsi),” in 3rd International no. 10, pp. 8307–8319, 2017.
Conference on Control, Engineering & Information Technology (CEIT), [14] M. Rivera, “A new predictive control scheme for a vsi with reduced
2015, pp. 1–6. common mode voltage operating at fixed switching frequency,” in
[6] O. Ellabban, J. Van Mierlo, and P. Lataire, “Experimental study of the IEEE 5th International Conference on Power Engineering, Energy and
shoot-through boost control methods for the z-source inverter,” EPE Electrical Drives (POWERENG), 2015, pp. 617–622.
journal, vol. 21, no. 2, pp. 18–29, 2011. [15] A. Ayad, P. Karamanakos, and R. Kennel, “Direct model predictive
[7] A. Bakeer and A. A. Ahmed, “Performance evaluation of pmsm based current control strategy of quasi-z-source inverters,” IEEE Transactions
on model predictive control with field weakening operation and bidirec- on Power Electronics, vol. 32, no. 7, pp. 5786–5801, 2016.
tional quasi z-source inverter,” in IEEE Nineteenth International Middle [16] A. Bakeer, G. Magdy, A. Chub, and D. Vinnikov, “Predictive control
East Power Systems Conference (MEPCON), 2017, pp. 741–746. based on ranking multi-objective optimization approach for quasi z-
[8] A. Bakeer, M. A. Ismeil, and M. Orabi, “A powerful finite control set- source inverter,” CSEE Journal of Power and Energy Systems, 2020.
model predictive control algorithm for quasi z-source inverter,” IEEE [17] A. Bakeer, M. A. Ismeil, M. Orabi, and R. Kennel, “Control of switched-
Transactions on Industrial Informatics, vol. 12, no. 4, pp. 1371–1379, inductor quasi z-source inverter (sl-qzsi) based on model predictive con-
2016. trol technique (mpc),” in IEEE International Conference on Industrial
[9] Y. Xu, Y. He, and S. Li, “Logical operation-based model predictive Technology (ICIT), 2015, pp. 2248–2253.
control for quasi-z-source inverter without weighting factor,” IEEE [18] S. Singh and S. Sonar, “A new svpwm technique to reduce the inductor
Journal of Emerging and Selected Topics in Power Electronics, vol. 9, current ripple of three-phase z-source inverter,” IEEE Transactions on
no. 1, pp. 1039–1051, 2020. Industrial Electronics, vol. 67, no. 5, pp. 3540–3550, 2019.
[10] O. Romero, A. Olloqui, M. Macias, F. Martell, O. Micheloud, J. Eli- [19] W. Liu, Y. Yang, T. Kerekes, and F. Blaabjerg, “Generalized space vector
zondo, and M. Rivera, “Finite states-modulated model predictive control modulation for ripple current reduction in quasi-z-source inverters,”
of a quasi-z-source inverter with lcl filter,” in IEEE Southern Power IEEE Transactions on Power Electronics, vol. 36, no. 2, pp. 1730–1741,
Electronics Conference (SPEC), 2017, pp. 1–6. 2020.

View publication stats

You might also like