ARCHIVES OF ELECTRICAL ENGINEERING VOL. 72(1), pp.

5 –22 (2023)

DOI 10.24425/aee.2023.143687

Current ripple reduction for finite control set model

predictive control strategy of grid-tied inverter
with reference current compensation

NAN JINo , WUCHUANG FAN, JIE FANG B , JIE WU, YONGPENG SHEN

Zhengzhou University of Light Industry

College of Electrical and Information Engineering
China
e-mail: B [email protected]

(Received: 07.06.2022, revised: 26.09.2022)

Abstract: In the finite control set model predictive control (FCS-MPC) strategy of the
grid-tied inverter, the current ripple (CR) affects the selection of optimal voltage vectors,
which leads to the increase of output current ripples. In order to solve this problem, this
paper proposes a CR reduction method based on reference current compensation (RCC)
for the FCS-MPC strategy of grid-tied inverters. Firstly, the influence of the CR on optimal
voltage vector selection is analyzed. The conventional CR prediction method is improved,
which uses inverter output voltage and grid voltage to calculate current ripples based on the
space state equation. It makes up for the shortcomings that the conventional CR prediction
method cannot predict in some switching states. The improved CR method is more suitable
for the FCS-MPC strategy. In addition, the differences between the two cost functions are
compared through visual analysis. It is found that the sensitivity of the square cost function
to small errors is better than that of the absolute value function. Finally, the predicted CR is
used to compensate the reference current. The compensated reference current is substituted
into the square cost function to reduce the CR. The experimental results show that the
proposed method reduces the CR by 47.3%. The total harmonic distortion (THD) of output
current is reduced from 3.86% to 2.96%.
Key words: cost function, current ripple reduction, finite control set model predictive
control (FCS-MPC), grid-tied inverter, reference current compensation (RCC)

© 2023. The Author(s). This is an open-access article distributed under the terms of the Creative Commons Attribution-
NonCommercial-NoDerivatives License (CC BY-NC-ND 4.0, https://creativecommons.org/licenses/by-nc-nd/4.0/), which per-
mits use, distribution, and reproduction in any medium, provided that the Article is properly cited, the use is non-commercial,
0 and no modifications or adaptations are made.
6 Nan Jin et al. Arch. Elect. Eng.

Nomenclature

Symbol Implication Symbol Implication

CR Current ripple 𝑔1 Type A cost function
RCC Reference current compensation 𝑔2 Type B cost function
VSI Voltage source inverter ∇𝑔1 Gradient of 𝑔1
𝑢 𝑎𝑏𝑐 Output voltage ∇𝑔2 Gradient of 𝑔2
𝑒 𝑎𝑏𝑐 Grid voltage 𝛿 Cost function error
𝑖 𝑎𝑏𝑐 Output current 𝑖 ∗𝛼𝛽_com Compensated reference current
𝑝
𝑈𝑑𝑐 DC voltage 𝑖 𝛼𝛽_rip Predictive current ripple
𝐿 Filter inductance 𝑑 Voltage vector duration
𝑅 Stray resistance 𝑢𝑚 Modulation amplitude
𝑝
𝑖 𝛼𝛽 Predictive current 𝑈tri Triangular carrier amplitude
𝑖 ∗𝛼𝛽 Reference current Δ𝑖 𝐿 Inverter output current error
Δ𝑢 Inductance fluctuating voltage 𝑋 (𝑘) Value of 𝑋 at (𝑘)-th instant
𝑇 Control period 𝑋 (𝑘 + 1) Value of 𝑋 at (𝑘 + 1)-th instant

1. Introduction

With the increasing scale of renewable energy represented by photovoltaic and wind energy
in the modern power system, the requirements for the efficiency and reliability of grid-tied
inverters are gradually improved [1]. Three-phase inverters were widely used in renewable energy
power generation [2–4]. As a connection between renewable energy and the power grid, the
output current of inverters controlled by conventional finite control set model predictive control
(FCS-MPC) causes a large ripple due to the limitation of the switching frequency. The current
ripple (CR) affects the selection of the optimal voltage vector at the (𝑘 + 1)-th instant so that the
optimal voltage vector is inaccurately selected. Eventually, it reduces the performance of grid-tied
inverters. With the high penetration of large-scale renewable energy into the grid, the stability of
the grid was reduced. Therefore, it is necessary to improve the FCS-MPC strategy.
The CR of grid-tied inverters is one of the main factors for the total harmonic distortion (THD)
of grid-tied inverter output current. The FCS-MPC strategy of voltage source inverters (VSI) [5–9]
has been studied. The optimal voltage vector is selected by comparing the predicted value and
the reference value without pulse width modulation (PWM), phase-locked loop (PLL) and other
controllers. For the FCS-MPC strategy, only one voltage vector is applied during a switching
period. The CR can be reduced by increasing the switching frequency. However, it will lead to
a large number of switching losses. If the switching loss is reduced by lowering the switching
frequency, the CR of the inverter output will increase.
Many studies have been devoted to eliminating the CR of grid-tied inverters. A current ripple
compensation technique has been proposed [10]. The error between the average control current
Vol. 72 (2023) Current ripple reduction for finite control set model 7

and the reference current was compensated by a certain compensation signal. The problems of
inaccuracy inverter output current, average current tracking and current waveform distortion were
solved. To minimize line current ripple, each 60◦ sector was divided into six sub-sectors in [11].
Each sub-sector applied the latest three vectors to ensure the minimum line current ripple. A ripple
steering method in [12] was applied to reduce high-frequency CR of the inverter. This method
can provide soft switching for power semiconductors.
In addition, researchers have studied the multi-vector FCS-MPC modulation method in [13].
Two or three voltage vectors are applied in each control cycle. It can effectively reduce the CR
of inverter output current. Multiple voltage vectors are required for each control cycle in the
multi-vector FCS-MPC modulation strategy. The duration of each voltage vector needs to be
calculated. There are usually two methods to calculate the duration of each base vector in the
multi-vector FCS-MPC strategy. The first is based on the deadbeat control principle [14], while
the second is based on the principle that the basic voltage vector is inversely proportional to
the duration [15, 16]. Multi-vector MPC modulation can effectively reduce the inverter output
CR. However, due to the multi-vector MPC modulation strategy, more than one voltage vector
is applied in each control cycle. Compared with the conventional MPC modulation strategy, the
inverter switches the switching state more times at the same time. The switching loss is increased.
The ripple of inverter output current is caused by many factors, such as input low-frequency
ripples, high-frequency ripples, parasitic parameters and ultra-high frequency resonance noise
during power equipment switching. The CR of grid-tied inverters has been regarded as an im-
portant research object [17–19]. Although there are many related studies on CR characteristics
in [20, 21], an accurate prediction method for real-time CR has not been proposed yet. In [22],
a real-time CR prediction method based on a three-level T-type converter was proposed. In [23],
a dead-time effect compensation method based on CR prediction for voltage-source inverters was
proposed. The CR prediction method in [22, 23] calculates the CR slope through the Thevenin
equivalent circuit diagram corresponding to each voltage vector. Finally, the real-time CR is cal-
culated. However, since only one voltage vector is applied in each control cycle in the FCS-MPC
strategy, the method by Thevenin’s theorem cannot accurately predict the CR in the FCS-MPC
strategy. This CR prediction method in [22, 23] is not suitable for the FCS-MPC strategy. In
order to solve this problem, this paper improves the conventional CR prediction method for the
FCS-MPC strategy.
This paper proposes a current ripple reduction method based on reference current compen-
sation (RCC). In order to realize the reference current compensation, the conventional current
ripple prediction method is improved. The improved method uses inverter output voltage and grid
voltage to predict current ripples. It makes up for the shortcomings that the conventional current
prediction method cannot predict in some switching states. In addition, the differences between
two cost functions are compared through visual analysis. It is found that the sensitivity of the
square cost function to small errors is better than that of the absolute value function. Finally, the
compensated reference current is substituted into the square cost function to reduce the current
ripple.
The influence of current ripples on a two-level voltage source inverter is analyzed in section 2.
In section 3, the improved CR prediction method is introduced, and the difference between the
two cost functions is analyzed by visualization. Section 4 introduces the principle of RCC. Finally,
section 5 shows the experimental results.
8 Nan Jin et al. Arch. Elect. Eng.

2. Analysis of current ripple using FCS-MPC

This section introduces the structure of two-level voltage source grid-tied inverter and the
control principle of the conventional FCS-MPC strategy. In the FCS-MPC strategy, the influence
of current ripple on optimal voltage vector selection of two-level voltage source inverter is
analyzed.

2.1. FCS-MPC strategy for grid-tied inverter

The structure of a three-phase two-level VSI is shown in Fig. 1. 𝑢 𝑎 , 𝑢 𝑏 and 𝑢 𝑐 are the
output voltages of the VSI. 𝑒 𝑎𝑏𝑐 and 𝑖 𝑎𝑏𝑐 are the grid voltages and the inverter output currents,
respectively. 𝑈𝑑𝑐 is the voltage on the DC side. 𝐿 is the filter inductance, and 𝑅 is its stray
resistance.

Fig. 1. Structure of two-level voltage source grid-tied inverter

The mathematical model of the two-level VSI in the 𝛼𝛽 coordinate system can be expressed as
 d𝑖 𝛼   𝑅  " # 1
 d𝑡  − 𝐿 0  𝑖 0 𝑢 −𝑒
  " #

𝛼 𝐿  𝛼 𝛼
= + (1)
 

 d𝑖 𝛽   ,
1  𝑢𝛽 − 𝑒𝛽
 
𝑅  𝑖𝛽 
   0 −   0 
 d𝑡  
  𝐿  𝐿
where: 𝑢 𝛼 and 𝑢 𝛽 are the output voltages of the two-level VSI, 𝑖 𝛼 and 𝑖 𝛽 are the output currents
of the two-level VSI, 𝑒 𝛼 and 𝑒 𝛽 are the grid voltages, 𝐿 is the filter inductance and 𝑅 is the stray
resistance.
As the sampling frequency is much higher than the grid voltage frequency, the discretization
of (1) by Euler’s forward approximation can approximate the predicted current at the (𝑘 + 1)-th
instant. The 𝛼𝛽 components of the predictive current 𝑖 𝛼𝑝 and 𝑖 𝛽𝑝 can be expressed as
𝑅  " # 𝑇
1 − 0  𝑖 0 𝑢 −𝑒
" 𝑝#  " #
𝑖𝛼 𝐿  𝛼

𝐿  𝛼 𝛼
= + (2)

,
𝑖 𝛽𝑝
 
𝑅 𝑖 𝑇  𝑢𝛽 − 𝑒𝛽
1−  𝛽
 
 0 0 
 𝐿  𝐿
where 𝑇 is the control period.
Then the cost function 𝑔 of conventional FCS-MPC can be given as

𝑔 = 𝑖 ∗𝛼 − 𝑖 𝛼𝑝 + 𝑖 ∗𝛽 − 𝑖 𝛽𝑝 , (3)
where 𝑖 ∗𝛼 and 𝑖 ∗𝛽 are the reference currents.
Vol. 72 (2023) Current ripple reduction for finite control set model 9

The eight predicted currents are substituted into (3). The voltage vector that minimizes the
cost function is selected as the optimal vector and applied to the next period.

2.2. Influence analysis of current ripple

For constant-switching frequency FCS-MPC, each switch state corresponds to a current ripple.
The difference between the DC-link voltage and instantaneous AC-voltage determines the CR.
According to (1), ignoring the effect of the resistance 𝑅, it can be deduced that

𝑖 ∗𝛼𝛽 (𝑘 + 1) ≈ 𝑖 𝑃𝛼𝛽 (𝑘 + 1) = 𝑖 𝛼𝛽 (𝑘) + Δ𝑖, (4)

where Δ𝑖 = 𝑇 [u 𝛼𝛽 (𝑘) − e 𝛼𝛽 (𝑘)]/𝐿, u 𝛼𝛽 = [𝑢 𝛼 , 𝑢 𝛽 ] 𝑇 and e 𝛼𝛽 = [𝑒 𝛼 , 𝑒 𝛽 ] 𝑇 . i∗𝛼𝛽 (𝑘 + 1) is the

reference current at the (𝑘 + 1)-th instant.
According to (4), the current error Δ𝑖 can be expressed as

Δ𝑖 ≈ 𝑖 𝑃𝛼𝛽 (𝑘 + 1) − 𝑖 𝛼𝛽 (𝑘) = Δ𝑖 ∗ . (5)

When Δ𝑖 depends on u 𝛼𝛽 (𝑘) and e 𝛼𝛽 (𝑘) in the 𝛼𝛽 coordinate system, the amplitude of Δ𝑖 is
constant and the phase is consistent with i 𝑃𝛼𝛽 (𝑘 + 1). The amplitude and phase of Δ𝑖 are consistent
with Δ𝑖 by selecting 𝑢 𝛼𝛽 (𝑘).
Due to the influence of switching characteristics, the actual output voltage of the inverter
cannot reach the theoretical value. As shown in Fig. 2, 𝑢 out represents the actual output voltage
of the inverter, and Δ𝑢 represents the fluctuating voltage across the inductor. The time 𝑑, during
which the optimal voltage vector acts in each cycle, can be expressed as
 
𝑇 𝑢𝑚
𝑑= 1+ , (6)
2 𝑈tri

where 𝑢 𝑚 is the modulation amplitude and 𝑈tri represents the triangular carrier amplitude.

Fig. 2. Analysis of current ripple influence

10 Nan Jin et al. Arch. Elect. Eng.

The fluctuating current Δ𝑖 𝐿 through the inductance of the filter in the time 𝑑 is expressed as
𝑈𝑑𝑐
Δ𝑢 − 𝑢𝑜 𝑇 𝑈 + 𝑢
𝑑= 2
tri 𝑚
Δ𝑖 𝐿 = × ×
𝐿 𝐿 2 𝑈tri
𝑈𝑑𝑐 𝑢 𝑚 𝑈𝑑𝑐
− · 2 2
2 𝑈tri 2 𝑇 𝑈tri + 𝑢 𝑚 𝑈𝑑𝑐 𝑇 (𝑈tri − 𝑢 𝑚 )
= × × = 2
. (7)
𝐿 2 𝑈tri 4𝐿𝑈tri

According to (7), when the optimal voltage vector is applied at the (𝑘 + 1)-th instant, there is
an error Δ𝑖 𝐿 between the actual output current and ideal output current. The FCS-MPC strategy
selects the optimal vector by comparing each voltage vector with the reference value. The cost
function includes not only the difference between the predicted value and the reference value, but
also the error between actual output value and ideal output value.
As only one voltage vector is applied in every control period in the conventional FCS-MPC
strategy, the CR is large. The CR affects the selection of the optimal voltage vector. In order to
suppress the CR, it is necessary to predict the real-time CR. First, the influence of current ripples
on the optimal voltage vector selection in the conventional MPC strategy is analyzed. Then, the
conventional current ripple prediction method is improved to make it more suitable for the MPC
strategy.

3. Proposed current ripple prediction method for FCS-MPC strategy

3.1. Proposed current ripples prediction
Taking a two-level VSI as an example, the structure is shown in Fig. 1. According to the inverter
output voltage and grid voltage, the inverter output current in one control cycle is calculated, as
shown in (8).
∫𝑡
A(𝑡−𝑡0 ) B𝑢  AΔ𝑡 
i(𝑡) = i (𝑡 0 ) · 𝑒 + 𝑒 A(𝑡−𝜏) B𝑢(𝜏) d𝜏 = i (𝑡0 ) · 𝑒 A(𝑡−𝑡0 ) + 𝑒 −1 , (8)
A
𝑡0

 𝑇
where: A = diag (−𝑅/𝐿, −𝑅/𝐿), B = diag (1/𝐿, 1/𝐿), 𝑡 − 𝑡 0 = 𝑇, i(t) = 𝑖 𝛼 (𝑡) 𝑖 𝛽 (𝑡) , 𝑖 𝛼 (𝑡) and
 𝑇
𝑖 𝛼 (𝑡) are the output currents of the inverter at the 𝑡 instant in the 𝛼𝛽-axis, i(t0 ) = 𝑖 𝛼 (𝑡0 ) 𝑖 𝛽 (𝑡0 ) ,
𝑖 𝛼 (𝑡 0 ) and 𝑖 𝛽 (𝑡 0 ) are the output currents of the inverter at the 𝑡0 instant in the 𝛼𝛽-axis.
According to (8), the output current of the inverter at the 𝑘−th instant can be represented by
" # " # 𝑅
" #
𝑖 𝛼 (𝑘 + 1) 𝑖 𝛼 (𝑘) − 𝑅 𝑇 1 − 𝑒 − 𝐿 𝑇 𝑢 𝛼 (𝑘) − 𝑒 𝛼 (𝑘)
= 𝑒 𝐿 + . (9)
𝑖 𝛽 (𝑘 + 1) 𝑖 𝛽 (𝑘) 𝑅 𝑢 𝛽 (𝑘) − 𝑒 𝛽 (𝑘)

According to the above analysis, the CR is expressed as

"𝑝 # " # " # " # " #
𝑖 𝛼_rip 𝑖 𝛼 (𝑘 + 1) 𝑖 𝛼 (𝑘) 𝑖 𝛼 (𝑘)  − 𝑅 𝑇  1 − 𝑒 − 𝑅𝐿 𝑇 𝑢 (𝑘) − 𝑒 (𝑘)
𝛼 𝛼
𝑝 = − = 𝑒 𝐿 −1 + . (10)
𝑖 𝛽_rip 𝑖 𝛽 (𝑘 + 1) 𝑖 𝛽 (𝑘) 𝑖 𝛽 (𝑘) 𝑅 𝑢 𝛽 (𝑘) − 𝑒 𝛽 (𝑘)
Vol. 72 (2023) Current ripple reduction for finite control set model 11

The proposed CR prediction method for the FCS-MPC strategy of grid-tied inverters makes
up for the shortcomings of the conventional CR prediction method that the CR cannot predict
in some switching states. The proposed method calculates the real-time CR by i 𝛼𝛽 (𝑘), u 𝛼𝛽 (𝑘)
and e 𝛼𝛽 (𝑘). It is not necessary to calculate the CR slopes under each voltage vector through
Thevenin’s equivalent circuit.

3.2. Visualization of cost function

In the conventional FCS-MPC modulation strategy, the cost function is mainly divided into
two types. One is the absolute value of the difference between the reference value and the predicted
current value. This type of cost function is collectively referred to as type A in this paper, and it
is given by
𝑔1 = 𝑖 ∗𝛼 − 𝑖 𝛼𝑝 + 𝑖 ∗𝛽 − 𝑖 𝛽𝑝 . (11)
The other is the sum of the difference square between the reference value and the predicted
current value. In this paper, this kind of value function is called type B, which leads to
h ih i𝑇
𝑔2 = 𝑖 ∗𝛼 − 𝑖 𝛼 𝑖 ∗𝛽 − 𝑖 𝛽𝑝 𝑖 ∗𝛼 − 𝑖 𝛼𝑝 𝑖 ∗𝛽 − 𝑖 𝛽𝑝
𝑝
. (12)

The cost function 𝑔1 is less stable than the cost function 𝑔2 [24]. According to [25] and [26],
the output current error of the cost function 𝑔1 is greater than 𝑔2 .
In order to display the change trend of the two cost functions, the partial derivatives of 𝑢 𝛼
and 𝑢 𝛽 are calculated for the two cost functions, respectively. The gradient of 𝑔1 (named ∇𝑔1 ) is
written as
!𝑇
𝜕𝑔1 𝜕𝑔1 𝑇 𝜕𝑖 𝛽𝑝
  𝑝
𝜕𝑖
∇𝑔1 (𝑢 𝛼 , 𝑢 𝛽 ) = = − 𝑖 ∗𝛼 − 𝑖 𝛼
𝑝 𝛼
− 𝑖 ∗𝛽 − 𝑖 𝛽
𝑝
. (13)
𝜕𝑢 𝛼 𝜕𝑢 𝛽 𝜕𝑢 𝛼 𝜕𝑢 𝛽

The gradient of the 𝑔1 (named ∇𝑔2 ) cost function is expressed as

!𝑇

𝜕𝑔2 𝜕𝑔2
𝑇   𝜕𝑖 𝑝   𝜕𝑖 𝛽𝑝
∇𝑔2 (𝑢 𝛼 , 𝑢 𝛽 ) = = −2 𝑖 ∗𝛽 − 𝑖 𝛼𝑝 𝛼
−2 𝑖 ∗𝛽 − 𝑖 𝛽 𝑝
. (14)
𝜕𝑢 𝛼 𝜕𝑢 𝛽 𝜕𝑢 𝛼 𝜕𝑢 𝛽

The difference between the two types of cost functions shown in Fig. 3 and Fig. 4 is obtained
by the method of visualized analysis.
Figure 3(a) and Fig. 4(a) represent the cost function values of type A and B under different
reference currents, respectively. Figure 3(b) and Fig. 4(b) represent the trend and isoline of the
corresponding cost function changes, respectively. It can be seen that the type A cost function
(named 𝑔1 ) changes relatively smoothly and is relatively sensitive to different errors. The type B
cost function (named 𝑔2 ) is more sensitive to large errors. When the error is less than 0.4, 𝑔2
changes slowly.
In order to analyze the effects of the two cost functions more specifically, a cost function error
𝛿 is defined as
𝑝 𝑇
h ih i
𝛿 = min 𝑖 ∗𝛼 − 𝑖 𝛼𝑝 + 𝑖 ∗𝛽 − 𝑖 𝛽𝑝 − min 𝑖 ∗𝛼 − 𝑖 𝛼 𝑖 ∗𝛽 − 𝑖 𝛽 𝑖 ∗𝛼 − 𝑖 𝛼 𝑖 ∗𝛽 − 𝑖 𝛽 .
𝑝 𝑝 𝑝
(15)
12 Nan Jin et al. Arch. Elect. Eng.

(a)

(b)

Fig. 3. Type A visual analysis graphs of cost function and gradient map: (a) visual analysis graphs;
(b) gradient map

(a)

Fig. 4. Type B visual analysis graphs of cost function and gradient map: (a) visual analysis graphs
Vol. 72 (2023) Current ripple reduction for finite control set model 13

(b)

Fig. 4. Type B visual analysis graphs of cost function and gradient map: (a) visual analysis graphs;
(b) gradient map

The three-dimensional graph of the error 𝛿 is shown in Fig. 5. 𝛿 is greater than or equal
to 0 throughout the three-dimensional graph, which means that 𝑔2 is always smaller than 𝑔1 for
different basic voltage vectors. Therefore, it can be concluded that the type B cost function is
more effective in reducing the control error than the type A cost function. For small errors, the
sensitivity can be improved by the RCC method proposed in this paper. The optimal voltage
vector is accurately selected without the influence of the CR.

Fig. 5. Three-dimensional graph of two cost functions error 𝛿

According to the above analysis, the difference between type A and B cost functions are
summarized.
1. According to Fig. 3(b) and Fig. 4(b), the type A cost function is more sensitive to small
errors. The type B cost function is more sensitive to large errors.
2. According to Fig. 5, the error of the type B cost function is smaller than the errors of type A.
The type B cost function is selected to reduce the control error. The RCC method is used to
improve its sensitivity to small errors.
 14 Nan Jin et al. Arch. Elect. Eng.

4. Reference current compensation method of FCS-MPC strategy

grid-tied inverter

According to the above analysis, the CR in certain areas of the sector has a great influence on
vector selection. The inverter output CR causes the inaccurate selection of some optimal voltage
vectors. As a result, the THD of the output current waveform is larger. In order to eliminate
the influence of the CR on the selection of voltage vectors, the CR suppression method based
on RCC is proposed in this paper. By predicting the real-time CR at the (𝑘 + 1)-th instant, the
reference current is corrected in advance to compensate for the impact of the CR. The voltage
vector predicted based on the reference current after compensation can better fit the reference
current before compensation. The schematic diagram of RCC is depicted in Fig. 6. The influence
caused by the CR can be effectively compensated. The compensated formula for the reference
current is given in (16). " # " # "𝑝 #
𝑖 ∗𝛼_com 𝑖 ∗𝛼 𝑖
∗
= ∗
− 𝛼_rip
𝑝 , (16)
𝑖 𝛽_com 𝑖𝛽 𝑖 𝛽_rip
h i𝑇
where: 𝑖 ∗𝛼𝛽_com = 𝑖 ∗𝛼_com , 𝑖 ∗𝛽_com , 𝑖 ∗𝛼_com and 𝑖 ∗𝛽_com stand for the component of the compensated
reference current in the 𝛼𝛽-axis.

Fig. 6. Reference current compensation diagram

The cost function is obtained by substituting the compensated reference current into the type B
cost function. The cost function is
𝑝 𝑇
h ih i
𝑔 = 𝑖 ∗𝛼_com − 𝑖 𝛼 𝑖 ∗𝛽_com − 𝑖 𝛽 𝑖 ∗𝛼_com − 𝑖 𝛼 𝑖 ∗𝛽_com − 𝑖 𝛽 .
𝑝 𝑝 𝑝
(17)

The control schematic of the proposed RCC method is shown in Fig. 7. The flow chart of the
proposed RCC method is shown in Fig. 8.
Step 1: Collect the grid side voltage 𝑒(𝑘) and three-phase output currents of the two-level VSI
𝑖(𝑘).
𝑝 𝑝
Step 2: Predict the output current 𝑖 𝛼𝛽 and the real-time CR 𝑖 𝛼𝛽_rip of the inverter under each voltage
vector, respectively.
Step 3: Compensate the reference current based on the predicted current ripple.
Step 4: The compensated reference current 𝑖 ∗𝛼𝛽_com is substituted for the original reference current
𝑖 ∗𝛼𝛽 to construct a cost function.
Step 5: Calculate the cost function 𝑔 and select the optimal voltage vector.
Vol. 72 (2023) Current ripple reduction for finite control set model 15

Fig. 7. Control schematic of the proposed reference current compensation

Fig. 8. Algorithm flow chart of the proposed method

The proposed method reduces the CR by compensating the reference current for the FCS-
MPC strategy of grid-tied inverters. First, the real-time CR is predicted by inverter output current,
inverter output voltage and grid voltage. Then, the reference current is compensated by the
predicted real-time CR. The basic control principle of the proposed method is similar to the
conventional MPC strategy. So, the proposed control method belongs to open-loop optimal
control. The stability of MPC has been verified in detail in [27]. This paper does not give detailed
proof.
16 Nan Jin et al. Arch. Elect. Eng.

Table 1 shows the effect comparison of different control methods. It can be seen from Table 1
that the double-vector in [16] and triple-vector MPC in [28] are multiple voltage vectors in
a control period. The CR is reduced by improving the accuracy of predicted value. But its
switching frequency also increases. The proposed method reduces the CR from the perspective
of reference value. Switching frequency will not increase.

Table 1. Comparison between the proposed method and multi-vector MPC strategy

Conventional Double-vector Triple-vector Proposed

MPC MPC MPC method
Current ripple large Smaller Small Smaller
Voltage vector in
1 2 3 1
each control period
Switching frequency Low Medium High Low
Algorithm
Small Medium Large Medium
calculation burden

According to the above analysis, the conclusions for CR reduction for the FCS-MPC strategy
of grid-tied inverters with RCC can be drawn as follows:
1. The CR will affect the selection of the optimal voltage vector at the next instant in the
FCS-MPC strategy.
2. The influence of the CR on the selection of the optimal vector can be eliminated by
compensating the reference current.

5. Experimental results
In order to verify the effectiveness of the proposed method, the experimental waveforms are
illustrated in this section. The experiment platform for a two-level voltage source grid-tied inverter
is shown in Fig. 9. The main control chip of the inverter is DSP28335. The parameters are shown
in Table 2.

Table 2. Experimental parameters

Parameter name Symbol Value Unit

DC voltage 𝑈𝑑𝑐 250 V
Peak of grid line voltage 𝑒 150 V
Filter inductance 𝐿 10 mH
Parasitic resistance 𝑅 50 mΩ
Grid frequency 𝑓 50 Hz
Sampling frequency 𝑓𝑠 10 kHz
Vol. 72 (2023) Current ripple reduction for finite control set model 17

Experimental waveforms include static experimental waveforms and dynamic experimental

waveforms. The steady-state experiment includes the comparison of inverter output current and
parameters mismatch experiments. The dynamic experiment includes inverter output current and
the current CR when the reference current changes from 6 A to 10 A.

Fig. 9. Experiment platform of two-level voltage source grid-tied inverter

5.1. Steady state experimental results

In Fig. 10, the CR of the proposed method is compared with that of the conventional FCS-MPC
strategy. When the reference current is 10 A, the maximum CR of the conventional FCS-MPC
strategy is 1.9 A. The maximum CR is reduced to 1 A. The amplitude and effective value of the
real-time CR are reduced. The variation of output current before and after the ripple at the peak
is more obvious. The THD of the conventional FCS-MPC is 3.86%. After adopting the proposed
method, the THD is reduced to 2.96%. According to the data of the analysis experiment, the
higher harmonic content of the inverter output current is suppressed after using the proposed
method.

Fig. 10. Inverter output current and current error

Figure 11 shows the THDs under different reference currents. It can be concluded that the
proposed method can reduce the current THD under different reference currents.
18 Nan Jin et al. Arch. Elect. Eng.

Fig. 11. The current THDs with different currents

In order to verify the robustness of the proposed method to parameter variations, the ex-
perimental study is carried out with inaccurate model parameters. The actual inductance in the
inverter is 10 mH. Two experiments are carried out according to the parameters set in [14]. The
inductance value in the controller is greater (20 mH) or less (5 mH) than the actual inductance
value. The experimental results are shown in Fig. 12 and Fig. 13.
It can be seen from Fig. 12 that when the inductance in the controller is greater than the
actual inductance, the CR becomes larger. The proposed method has an obvious effect on ripple
suppression.

Fig. 12. Inverter output current when the inductance in the controller is 20 mH

It can be seen from Fig. 13 that when the inductance in the controller is smaller than the
actual inductance, the current waveform is distorted. The effect of the proposed method on CR
suppression is weakened. But the proposed method is still effective when the parameters are
mismatched.
Vol. 72 (2023) Current ripple reduction for finite control set model 19

Fig. 13. Inverter output current when the inductance in the controller is 5 mH

5.2. Dynamic state experimental results

The performance of the proposed method is further verified by dynamic experiments. Figure 14
shows the dynamic three-phase current waveform and CR of phase a when the reference current
changes from 6 A to 10 A.
When the reference current changes from 6 A to 10 A, the inverter output current can quickly
track the change of the reference current. The inverter output current error does not increase with
the change of the reference current. When the reference current is 6 A, the THD is 4.25%. When
the reference current increases to 10 A, THD decreases to 2.96%. The amplitude of the inverter
output CR remains at 1 A. The proposed method does not show the overshoot phenomenon, which
further verifies the effectiveness of the proposed method.

Fig. 14. Dynamic experimental results of the proposed method when three-phase current rises
from 6 A to 10 A
20 Nan Jin et al. Arch. Elect. Eng.

6. Conclusions

The conventional FCS-MPC strategy of the voltage source grid-tied inverter can be affected
by the CR when selecting the optimal voltage vector and lead to an inaccurate selection of the
optimal voltage vector. This paper analyzes the influence of current ripples on the optimal voltage
vector selection of a two-level voltage source inverter. The current ripple prediction method is
improved. The improved method uses inverter output voltage and grid voltage to calculate current
ripples based on the space state equation. It makes up for the shortcomings that the conventional
current prediction method cannot predict in some switching states. The improved current ripple
predictive method is more suitable for the FCS-MPC strategy. In addition, the differences between
the two cost functions are compared through visual analysis. It is found that the sensitivity of the
square cost function to small errors is better than that of the absolute value function. Finally, the
predicted current ripple is used to compensate the reference current. The current ripple is reduced
by compensating the reference current.
The experimental results show that the proposed method is effective for two-level voltage
source inverters. However, the feasibility of the proposed reference current compensation method
in other inverter structures can also be studied, for example, in a T-type inverter and NPC (Neutral
Point Clamped) three-level inverter. In the future, we will study the application of the proposed
reference current compensation method in other structural inverters.

Acknowledgements
This research was supported in part by the National Nature Science Foundation of China under Grant
U2004166, in part by the Province Key Research and Development and Promotion of Special Projects
(Science and Technology) Grant 222102210059, and in part by Science and Technology Innovation Team
in Universities of Henan Province under Grant 22IRTSTHN017.

References
[1] Jayalakshmi N.S., Gaonkar D.N., Karthik R.P., Prasanna P., Intermittent power smoothing control for
grid connected hybrid wind/PV system using battery-EDLC storage devices, Archives of Electrical
Engineering, vol. 69, no. 2, pp. 433–453 (2020), DOI: 10.24425/aee.2020.133036.
[2] Sun K., Wang X.S., Li Y.W., Nejabatkhah F., Yang M., Lu X., Parallel operation of bidirectional inter-
facing converters in a hybrid AC/DC microgrid under unbalanced grid voltage conditions, IEEE
Transactions on Power Electronics, vol. 32, no. 3, pp. 1872–1884 (2017), DOI: 10.1109/TPEL.
2016.2555140.
[3] Vekhande V., Kanakesh V.K., Fernandes B.G., Control of three-phase bidirectional current-source con-
verter to inject balanced three-phase currents under unbalanced grid voltage condition, IEEE Transac-
tions on Power Electronics, vol. 31, no. 9, pp. 6719–6737 (2016), DOI: 10.1109/TPEL.2015.2503352.
[4] Li H., Xiao H., Yang G., Reconstructed current model predictive control of NPC three-level grid-tied
converter with current sensor fault, IEEE Access, vol. 9, pp. 141098–141106 (2021), DOI: 10.1109/AC-
CESS.2021.3119566.
[5] Guo L., Jin N., Gan C., Luo K., Hybrid voltage vector preselection-based model predictive control
for two-level voltage source inverters to reduce the common-mode voltage, IEEE Transactions on
Industrial Electronics, vol. 67, no. 6, pp. 4680–4691 (2020), DOI: 10.1109/TIE.2019.2931257.
Vol. 72 (2023) Current ripple reduction for finite control set model 21

[6] Estévez-Bén Adyr A., Alfredo Alvarez-Diazcomas, Juvenal Rodruguez Reséndiz, Transformerless
Multilevel Voltage-Source Inverter Topology Comparative Study for PV Systems, Energies, vol. 13,
no. 12 (2020), DOI: 10.3390/en13123261.
[7] Héctor López, Nimrod Vázquez, Juvenal Rodriguez et al., Analysis and implementation of a finite-
control-set by using model solution-based control for three-phase VSI, IET Power Electronics, vol. 10,
no. 14, pp. 1832–1840 (2017), DOI: 10.1049/iet-pel.2016.0819.
[8] Estévez-Bén Adyr A. et al., A new predictive control strategy for multilevel current-source inverter
grid-connected, Electronics, vol. 8, no. 8 (2019), DOI: 10.3390/electronics8080902.
[9] Tang M., Yang S., Zhang K. et al., Model predictive direct power control of energy storage quasi-Z-
source grid-connected inverter, Archives of Electrical Engineering, vol. 71, no. 1, pp. 21–35 (2022),
DOI: 10.24425/aee.2022.140195.
[10] Lu W., Li S., Chen W., Current-ripple compensation control technique for switching power
converters, IEEE Transactions on Industrial Electronics, vol. 65, no. 5, pp. 4197–4206 (2018),
DOI: 10.1109/TIE.2017.2762622.
[11] Zeng Z., Li Z., Goetz S.M., Line current ripple minimization PWM strategy with reduced zero-sequence
circulating current for two parallel interleaved three-phase converters, IEEE Transactions on Power
Electronics, vol. 35, no. 7, pp. 6931–6943 (2020), DOI: 10.1109/TPEL.2019.2958878.
[12] Samani R., Beyragh D.S., Pahlevani M., A new grid-connected DC/AC inverter with soft switching and
low current ripple, IEEE Transactions on Power Electronics, vol. 34, no. 5, pp. 4480–4496 (2019),
DOI: 10.1109/TPEL.2018.2863183.
[13] Zhang Y., Jiang H., Yang H., Model predictive control of PMSM drives based on a general discrete
space vector modulation, IEEE Transactions on Energy Conversion, vol. 36, no. 2, pp. 1300–1307
(2021), DOI: 10.1109/TEC.2020.3036082.
[14] Liu T., Chen A., Qin C., Chen J., Li X., Double vector model predictive control to reduce common-mode
voltage without weighting factors for three-level inverters, IEEE Transactions on Industrial Electronics,
vol. 67, no. 10, pp. 8980–8990 (2020), DOI: 10.1109/TIE.2020.2994876.
[15] Cao L.Z., Li Y.Y., Li X.Y., Guo L.L., Jin N., Cao H., A dual-vector modulated model predictive control
method for voltage source inverters with a new duty cycle calculation method, Energies, vol. 13, no. 16,
pp. 9204–9214 (2020), DOI: 10.3390/en13164200.
[16] Jin N., Chen M., Guo L., Li Y., Chen Y., Double-vector model-free predictive control method for voltage
source inverter with visualization analysis, IEEE Transactions on Industrial Electronics, vol. 69, no. 10,
pp. 10066–10078 (2022), DOI: 10.1109/TIE.2021.3128905.
[17] Kim S., Won I.J., Kim J., Lee K., DC-link ripple current reduction method for three-level inverters with
optimal switching pattern, IEEE Transactions on Industrial Electronics, vol. 65, no. 12, pp. 9204–9214
(2018), DOI: 10.1109/TIE.2018.2823662.
[18] Lu W., Li S., Chen W., Current-ripple compensation control technique for switching power
converters, IEEE Transactions on Industrial Electronics, vol. 65, no. 5, pp. 4197–4206 (2018),
DOI: 10.1109/TIE.2017.2762622.
[19] Zeng Z., Li Z., Goetz S.M., Line current ripple minimization PWM strategy with reduced zero-sequence
circulating current for two parallel interleaved three-Phase converters, IEEE Transactions on Power
Electronics, vol. 35, no. 7, pp. 6931–6943 (2020), DOI: 10.1109/TPEL.2019.2958878.
[20] Cervellini P., Antoszczuk P., Retegui R.G., Funes M., Current ripple amplitude measurement in
multiphase power converters, IEEE Transactions on Power Electronics, vol. 32, no. 9, pp. 6684–6688
(2017), DOI: 10.1109/TPEL.2017.2686784.
22 Nan Jin et al. Arch. Elect. Eng.

[21] Chang L., Jahns T.M., Prediction and evaluation of PWM-induced current ripple in IPM machines
incorporating slotting, saturation, and cross-coupling effects, 2017 20th International Conference on
Electrical Machines and Systems (ICEMS), Sydney, Australia, pp. 1–6 (2017).
[22] Wang Z., Zhao Z., Hammad Uddin M., Zhao Y., Current ripple analysis and prediction for three-level
T-type converters, 2018 IEEE Energy Conversion Congress and Exposition (ECCE), Portland, USA,
pp. 7251–7257 (2018).
[23] Shen Z., Jiang D., Dead-time effect compensation method based on current ripple prediction for
voltage-source inverters, IEEE Transactions on Power Electronics, vol. 34, no. 1, pp. 971–983 (2019),
DOI: 10.1109/TPEL.2018.2820727.
[24] Karamanakos P., Geyer T., Guidelines for the Design of Finite Control Set Model Predictive Con-
trollers, IEEE Transactions on Power Electronics, vol. 35, no. 7, pp. 7434–7450 (2020), DOI:
10.1109/TPEL.2019.2954357.
[25] Mirzaeva G., Goodwin G., Townsend C., A simple and effective strategy to reduce switching losses
under FS-MPC based on dynamically changing voronoi diagrams, 2017 12th IEEE Conference on
Industrial Electronics and Applications (ICIEA), Siem Reap, Cambodia, pp. 1516–1521 (2017).
[26] Mirzaeva G., Goodwin G.C., McGrath B., A new understanding and improvements of finite set model
predictive control in inverter applications, 2015 17th European Conference on Power Electronics and
Applications (EPE’15 ECCE-Europe), Geneva, Switzerland, pp. 1–10 (2015).
[27] Andrew E.T., Ahmed K.H., Holliday D., A new model predictive current controller for grid connected
converters in unbalanced grids, IEEE Transactions on Power Electronics, vol. 37, no. 8, pp. 9175–9186
(2022), DOI: 10.1109/TPEL.2022.3158016.
[28] Guo L., Chen M., Li Y., Wang P., Jin N., Wu J., Hybrid multi-vector modulated model predictive control
strategy for voltage source inverters based on a new visualization analysis method, IEEE Transactions
on Transportation Electrification (2022), DOI: 10.1109/TTE.2022.3161583.