Received November 9, 2020, accepted November 20, 2020, date of publication December 9, 2020,

date of current version December 21, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.3043604

An Optimal Model Predictive Control Method for

Five-Level Active NPC Inverter
ZHAN LIU 1 , (Member, IEEE), ZHENGLONG XIA 1, (Member, IEEE), DAN LI2 , YU WANG2 ,
AND FEI LI 1 , (Member, IEEE)
1 School of Electrical Engineering and Automation, Jiangsu Normal University, Xuzhou 221116, China
2 Xinyi Power Supply Branch, State Grid Jiangsu Electric Power Company Ltd., Xuzhou 210024, China
Corresponding author: Zhan Liu ([email protected])
This work was supported in part by the National Natural Science Foundation of China under Grant 51907083, and in part by the Natural
Science Foundation of the Jiangsu Higher Education Institutions of China under Grant 19KJB470003.

ABSTRACT Finite Control Set Model Predictive Control (FCS-MPC) can improve the control performance
of Five-level Active NPC (5L-ANPC) inverter effectively. However, with the increasing of the levels of
inverter, the traditional FCS-MPC method has many problems, such as large amount of rolling optimization
calculation and difficulty in weight coefficient design. In order to solve those problems, an Optimal Model
Predictive Control (O-MPC) method is proposed. First, an optimal control method based on level jump
limitation is proposed to reduce the number of switch states. Second, in order to solve the conflict problem
of multiple control objectives and reduce the design difficulty of weight factors, a satisfactory optimal control
method based on hierarchy is proposed. After the optimization, the calculation time of the controller is greatly
reduced and the control performance of the multi-objective control system will be significantly improved.
At last, the effectiveness and feasibility of the proposed O-MPC are validated by the experimental results.

INDEX TERMS Five-level active NPC (5L-ANPC) inverter, an optimal model predictive control (O-MPC)
method, switching states, satisfactory optimal control.

I. INTRODUCTION and three-level conversion. Compared with traditional con-

Traditional NPC three-level inverter cannot be applied to trol, it has outstanding advantages in reducing system loss,
6kV and above high-voltage system because of the limita- improving system efficiency and realizing optimal control
tion of power devices. Therefore, multilevel converters have of system [5], [6].The realization of predictive controller
attracted a great deal of interest in high-power medium- needs to consider the influence of all switch states on sys-
voltage applications due to the low voltage stresses, low dv/dt tem control variables [7]–[9]. In order to select the optimal
and high-power quality [1], [2]. Diode clamped, flying capac- switching state, the control system needs to calculate and
itor and cascaded H-bridge multilevel converters are the three compare all possible switching states, so the calculations of
classic multilevel topologies. However, with the increasing the whole systerm is directly proportional to the amount of
demands of higher voltage level and higher performance, all switching states existing in the converter. For multi-level and
of them have encountered some problems, such as the mass multi-phase converter systems, it is necessary to consider an
number of diodes, flying capacitors, isolated DC sources and appropriate optimization method to reduce the computational
the resulting complex control problems. Compared with clas- complexity of the predictive control. In [10], an improved
sic multilevel topologies, five-level ANPC converter has the five-level bidirectional converter based on FCS-MPC was
advantages of fewer switches, fewer capacitors and simpler presented, which consisted in using the discrete time nature
voltage balance method [3], [4]. of the improved five-level bidirectional converter to define
FCS-MPC technology, with the outstanding advantages in its state in each sampling interval. In response to the problem
reducing system loss, improving system efficiency and realiz- of the conventional five-phase FCS-MPC suffers from heavy
ing optimal control of system, is widely studied in two-level computational burden, a novel FCS Model Predictive Current
Control (MPCC) with continued modulation was proposed
The associate editor coordinating the review of this manuscript and in [11]. In [12], a simple Model Predictive Direct Power
approving it for publication was Emanuele Crisostomi . Control (MP-DPC) of single-phase pulse width-modulated

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
221414 VOLUME 8, 2020
Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

rectifiers with constant switching frequency using modula-

tion function optimization was proposed.
In order to make the FCS-MPC can be better used in
other converter topologies, various optimization methods for
MPC have been widely studied [13], [14]. In [15], a fast
finite-level-state model predictive control (FFLS-MPC) strat-
egy was proposed in predictive control regulated modular
multilevel converter (MMC), which aims to overcome the
high computational complexity. However, this method has
certain limitations and is only suitable for multi-level con-
FIGURE 1. The topological structure of five-level ANPC inverter.
verters with multiple sub-modules in a similar MMC struc-
ture. In [16] a new algorithm, based on the control of the
switching period within the cost function was proposed, Therefore, an Optimal Model Predictive Control (O-MPC)
which was easy to design and implement, demanding very method is presented to solve the above problems. This pro-
low computation power, and enabled a combination of the posed method can effectively reduce the number of switching
advantages of FCS-MPC with the benefits of a PWM-like states involved in the rolling optimization, so the efficiency
power quality. An improved FCS-MPC algorithm with fast of the processor will be improved. In addition, because of
computation and fixed switching frequency is proposed for the difficulty at selecting weight factors in predictive control,
two-level three- phase inverters in [17]. Compared with the the design method of weight factor in the O-MPC controller
conventional fixed switching frequency FCS-MPC method, will be also described in detail.
the number of sectors involved in the FCS-MPC calculation This paper is organized as follows. the 5L-ANPC topol-
can be reduced from 6 to 1, which greatly improves the ogy and the control principle of floating-capacitor voltage
computation efficiency. In [18], A fixed switching frequency are summarized in Section II. In Section III, mathematical
scheme for finite-control-set model predictive control was models of each control objective are presented. In Section IV,
presented to achieve an output waveform quality that com- the optimization method of 5L-ANPC inverter switching state
pares well to that of a pulse-width-modulator-based linear is introduced in detail. Section V shows the design of O-MPC
controller, while retaining the benefits of model predictive controller for 5L-ANPC Inverter. The experimental results
control. The above schemes are all aimed at the optimization are presented in Section VI. Finally, in Section VII, the con-
of two-level or three-level converter. When the number of clusion is drawn.
levels in the converter increases, these methods cannot fully
adapt to the new structure. II. ANALYSIS ON THE STRUCTURE OF 5L-ANPC
Compared with MPC of two-and three-level con- INVERTER AND THE CONTROL PRINCIPLE OF
verter, the MPC of 5L-ANPC converter has the following FLOATING-CAPACITOR VOLTAGE
characteristics: A. THE STRUCTURE OF 5L-ANPC INVERTER
1) 5L-ANPC converter has more switching states, which The topological structure of five-level ANPC inverter is dis-
means that the number of rolling optimizations of pre- played in Figure 1, where Cup and Cdn are the capacitance
diction model increases and the amount of calculation of upper bus and lower bus respectively, while Uup and Udn
rises, which poses a challenge to the hardware design are the voltages on the two capacitors respectively; Cf is the
of control system [18]. Floating-Capacitance (FC) of the inverter; Ufx is the voltage
2) The stable operation of the 5L-ANPC converter needs of the Cfx , where x represents phase a, b, or c.
to meet two basic control objectives when predictive Tab. 1 lists the switching states of five-level ANPC inverter.
control is carried out. The two control objectives are The switches Sx1–Sx4 and Sx10 –Sx40 are operated comple-
the control of neutral point potential and the control of mentarily. It can be seen that each phase of the inverter can
floating capacitor voltage in each leg of 5L-ANPC con- output five levels, a total of eight switching states, as shown
verter. The difficulty of control is that the establishment in Table 1, where ifx and inx are the corresponding FC current
of predictive model is more complicated because of the and neutral-point current, iox is the phase current.
cross coupling between the two control objectives [19]. The voltage of FC in each phase should be stable at 1/4 of
3) 5L-ANPC converter has 512 switching states, where DC bus voltage, which is the premise for 5L-ANPC inverter
the switching must follow certain rules, which will to operate normally. Therefore, it is necessary to study the
greatly increase the complexity of predictive control voltage balance of FC.
algorithm [20].
Considering the above factors, the traditional MPC applied B. THE VOLTAGE CONTROL OF FC
in 5L-ANPC converter has some disadvantages, such as com- Figure 2 shows the flow chart of voltage control of FC. From
plex prediction model, many rolling optimization times and Table 1 and Figure 2, it can be seen that the balance of the volt-
complex control algorithm. ages across the FC can be achieved by a simple closed-loop

VOLUME 8, 2020 221415

 Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

TABLE 1. Switching states of the five-level ANPC inverter.

FIGURE 3. The space vector graph for five-level inverter.

where Va , Vb and Vc are the output voltage of phase a, b and

c respectively; E is 1/4 of the DC bus voltage, the switching
state corresponding to Sx = −2 is V0, the switching state
corresponding to Sx = −1 is V1 and V2, the switching
state corresponding to Sx = 0 is V3 and V4, the switch-
FIGURE 2. Flow diagram of FC voltage control.
ing state corresponding to Sx = 1 is V5 and V6, the switching
state corresponding to Sx = 2 is V7, where x represents phase
control. It can be seen from Table 1 that the FC voltage is a, b, or c.
affected by the switching states (V1, V2) and (V5, V6), when According to the analysis, there are 5∧ 3 = 125 voltage
the current flow paths are coincident, the switching states states that can be combined in three-phase five-level con-
V1 and V2 have opposite effects on FC and the switching verter, which correspond to 125 different switching states.
states V5 and V6 have opposite effects on FC too. Therefore, In the three-phase static coordinate system, the voltage space
when each phase output voltage has two levels E and −E, vector of the five-level converter can be expressed as:
the switching states which guarantee the stability of the FC
2
voltage can be easily chosen by detecting the direction of the V= E · (Sa + α · Sb + α 2 · Sc ) (2)
phase current io and the voltage of FC. 3
where α = ej 3 π . The voltage space vector diagram corre-
It should be pointed out that there is coupling between the 2

control of FC voltage and the control of neutral-point voltage sponding to 125 switching states of five level inverter is given
in some states. When the redundant switch state is selected to in Figure 3.
control the voltage of FC, the influence of the two optional The digital combination (such as 414) in the circle shown
switch states on the neutral-point voltage is different. The in Figure 3 represents the level state of phase a, b and c
traditional control methods put the control of the FC voltage respectively from left to right. In 5L-ANPC converter, digit
in the first place, which will inevitably affect the control 4 represents level state +2, digit 3 represents level state +1,
of the neutral-point voltage and lead to the increase of the digit 2 represents level state 0, digit 1 represents level state
fluctuation of the neutral-point voltage. −1, and digit 0 represents level state −2.

III. MODEL OF 5L-ANPC INVERTER SYSTEM B. MODELING OF 5L-ANPC INVERTER LOAD SYSTEM
A. SPACE VECTOR MODEL OF 5L-ANPC INVERTER
Referring to Figure 1, the load dynamic current equation for
If the output state of each bridge arm is represented by the each phase of 5L-ANPC inverter is obtained:
switching variables Sa, Sb and Sc, then the three-phase output
dia

voltage is: 
 uao = L + Ria + ea + uno
dt


 
dib

Va = Sa · E

ubo = L + Rib + eb + uno (3)
Vb = Sb · E (1) 
 dt
uco = L dic + Ric + ec + uno
 


Vc = Sc · E

dt
221416 VOLUME 8, 2020
Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

where R and L are the load resistance and inductance, respec- vector in the αβ coordinate system, eα,β is the EMF vector of
tively, uxo (x = a, b, c) is the voltage generated by the inverter, load in the αβ coordinate system.
ex (x = a, b, c) is the electromotive force (EMF) of the load, Applying a sampling period Ts , the derivative term of
and ix (x = a, b, c) is the load current, uno is the voltage current derivative can be approximately expressed as:
between load neutral point n and inverter neutral point o. diα,β iα,β (k + 1) − iα,β (k)
Substituting equation (3) into equation (2), the dynamic = (11)
dt Ts
vector equation of load current is as follows:
    By introducing equation (11) into equation (10), the pre-
d 2 2 dictive mathematical model of inverter output current can be
u=L (ia + αib + α ic ) + R (ia + αib + α ic )
2 2
dt 3 3 obtained:
2 2 RT s Ts
+ (ea + αeb + α 2 ec ) + (uno + αuno + α 2 uno ) (4) iα,β (k + 1) = (1 − )iα,β (k) + (uα,β (k) − eα,β (k))
3 3 L L
According to the definitions of voltage space vector, load (12)
current vector and back EMF space vector, and assuming that By discretizing equation (6) with the same method, a math-
the output three-phase voltage is strictly symmetrical, the last ematical model for predicting the neutral point voltage fluc-
part is zero, the simplified dynamic vector equation of load tuation of 5L-ANPC inverter can be obtained:
current can be obtained:
Ts
di uo (k + 1) = (Ha ia (k) + Hb ib (k) + Hc ic (k)) + uo (k)
u = L + Ri + e (5) 2C
dt (13)
where u is the voltage vector generated by the five-level
inverter, i is the load current vector, e is the EMF vector of By discretizing equation (8), a mathematical model for pre-
load. dicting the voltage fluctuation of FCs in each leg of 5L-ANPC
The fluctuation of the neutral point voltage uo in 5L-ANPC inverter is obtained:
inverter is determined by the output current io of the neutral Ts
ufx (k + 1) = | Hfx ix + ufx (k) − Uf∗ | (14)
point and the capacitance value of the upper and lower buses, Cf
as shown in: where: Uf∗ represent the setting voltage of FC.
duo io 1 The average number of switching actions of the system at
= = (Ha ia + Hb ib + Hc ic ) (6)
dt 2C 2C the next moment is fswitch :
where C is the DC side half bus capacitance, and C = Cup = 12
X
Cdn . Ha , Hb , Hc are the three-phase neutral point connection fswitch = (fsai + fsbi + fsci )/36 (15)
flag variables of inverter, which are defined as: i=1
where: fsa , fsb and fsc represent the action times of three-phase
(
1, (Statex = V2 or V3 or V4 or V5)
Hx = (7) switching devices respectively.
0, (Statex = V0 or V1 or V6 or V7) The control strategy also uses an estimation of the future
where x represents phase a, b, or c. reference current. Depending on the sampling time applied
The mathematical model of voltage fluctuation of FC Cfx and the computational constrains, the estimation can be
can be described by the following equation: obtained by a second-order extrapolation given by:
dufx ifx 1 i∗α,β (k + 1) = 3i∗α,β (k) − 3i∗α,β (k − 1) + i∗α,β (k − 2) (16)
= = Hfx ix (8)
dt Cf Cf
IV. OPTIMIZATION OF SWITCHING STATES OF
where Hfx is the FC access flag variable in each phase, x
5L-ANPC INVERTER
represents phase a, b, or c. which is defined as:
 In order to reduce the number of rolling optimizations,
1,
 (Statex = V1 or V5) it is necessary to optimize all switching states of 5L-ANPC
Hfx = 0, (Statex = V0 or V3 or V4 or V7) (9) inverter. The switching states that do not meet the output
 requirements of the inverter are deleted in advance, and the
−1, (Statex = V2 or V6)

effective switching states are involved in the rolling optimiza-
C. DISCRETE PREDICTION MODEL OF 5L-ANPC INVERTER tion control.
In view of the above objectives, it is necessary to put
According to formula (5), the voltage equation in the αβ
forward requirements for the voltage output of the inverter,
coordinate system is
which should meet the following two conditions:
diα,β 1) Each phase output voltage of the inverter shall not jump
uα,β = L + Riα,β + eα,β (10)
dt more than one level, and the output level of each phase
where uα,β is the voltage vector generated by the five-level voltage shall not exceed the output level range of the
inverter in the αβ coordinate system, iα,β is the load current inverter;

VOLUME 8, 2020 221417

 Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

FIGURE 4. The switching state diagram of five-level inverter.

2) The output line voltage cannot jump more than one

level too.
If the output voltage does not meet the two conditions
FIGURE 5. The switching state optimization diagram of five-level inverter.
given in this paper, there will be cross-level jump, which
will lead to the decrease of output level and the increase
of output harmonic. For 5l-anpc inverter, switching chaos 125 switching states in total. Through formula (17), we can
will occur when switching state, which will lead to system get 27 switching states that meet the first condition. By fur-
failure. ther optimizing the 27 switching states obtained by equation
According to the above two requirements, the effective (18), the number of switching states that can be output is
switching state of the inverter can be calculated at the next reduced to 15. Figure 5 is the switching state optimiza-
moment. If the current output switch state is [sa (k), sb (k), tion diagram of five-level inverter, in which the dotted line
sc (k)], and the first condition is satisfied, the output switching represents the switching state meeting condition 1 but not
state at the next moment is as follows: condition 2, and the solid line represents the switching state

sa (k + 1)
 
sa (k)
  
sja that can be output at the next moment after optimization
sb (k + 1) = sb (k) + sjb  (17) control.
sc (k + 1) sc (k) sjc
V. MODEL PREDICTIVE CONTROLLER FOR
where: sjx represents the level jump variable of x, where x 5L-ANPC INVERTER
represents phase a, b, or c. |sjx | ≤ 1, and 0 ≤ sx (k + 1) ≤ 4. A. DESIGN OF MODEL PREDICTIVE CONTROL SYSTEM
According to the above formula, the calculation formula of FOR 5L-ANPC INVERTER
line voltage at the next moment can be deduced:
      Figure 6 shows the block diagram of O-MPC system for
sab (k + 1) sab (k) sja − sjb 5L-ANPC inverter. In order to accurately predict and control
sbc (k + 1) = sbc (k) + sjb − sjc  (18) the neutral point potential, the voltage of floating capacitor
sca (k + 1) sca (k) sjc − sja and the expected output load current, the voltage and load
According to the calculation of five-level inverter, the number current of each capacitor should be sampled in real time.
of all switching states is 125, the maximum number of switch Then, according to the switch state of the inverter output at
states meeting the first condition is 27, and the maximum the current moment, the optimized switch state is selected
number of switching states satisfying both conditions is 15. by the switch state optimization controller and substituted
A conclusion is drawn that the traditional MPC method needs into the prediction model, i.e. (12), (13), (14), (15). Then
to calculate 125 times each time, while the proposed method the predicted value of the required control target is obtained.
only needs to calculate 15 times at most, which can save Finally, the switching state with the smallest value func-
88% of the calculation time and improve the computational tion is selected to output and the optimal control target is
efficiency of the processor. Figure 4 shows the switching state obtained.
diagram of five-level inverter, in which the state represented The design objectives of 5L-ANPC inverter predictive con-
by box represents all switching states of five-level inverter, troller are as follows:
and the state represented by solid dot is the optimized switch- 1) Load current reference trajectory tracking;
ing state satisfying two conditions. 2) Balance control of neutral point potential;
Taking the current switching states (2,2,2) as an exam- 3) Balance control of FCs;
ple, if not optimized, the five-level inverter can output 4) Reduce the switching frequency.

221418 VOLUME 8, 2020

Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

FIGURE 6. The block diagram of O-MPC system for 5L-ANPC inverter.

Using the value function, the above control objectives can

be described by formula:
J = Ki (|i∗α (k + 1) − iα (k)| + |i∗β (k + 1) − iβ (k)|)
+ Kf |ufa (k + 1) + ufb (k + 1) + ufc (k + 1)|
+ Ko |uo (k + 1)| + Kn fswitch (19)
where Ki , Ko , Kf and Kn are the weight factors of load current,
neutral point potential deviation, the voltage deviation of
three-phase FCs and switching frequency optimization.
The O-MPC method flow chart of 5L-ANPC inverter is FIGURE 7. The flow chart of O-MPC system for 5L-ANPC inverter.
given in Figure 7. Firstly, the voltage and load current on
each capacitor are measured and sampled, and the final opti- TABLE 2. Priorities of control objective.
mized switch state and its number are determined according
to formula (17) (18). Then, the rolling optimization stage is
started. The load current and capacitor voltage are obtained
through formula (12), (13), (14) and (15) respectively for each
optimized switching state. Finally, the optimal switching state
obtained by equation (19) is applied to the converter.

B. DESIGN METHOD OF WEIGHT FACTORS

The role of the weight factor is to adjust the proportion of the
control objective in the value function. When the value of the is realized according to the order of priority from high to low.
weight factor K is larger, it means that the corresponding con- The determination of weight factor takes the control objective
trol objective has a greater proportion in the value function, to reach the set control range as the goal, that is to achieve
which means that the target has a greater priority. When the the satisfactory control of the objective. The weight factors
value of K is smaller, it means that the proportion of the cor- of other control objectives are determined in turn.
responding control objective in the value function is smaller. Take 5L-ANPC inverter control system as an example.
In the actual system, the satisfactory control effect can be Firstly, three control objectives of the system are determined:
achieved by setting the weighting coefficient K flexibly. the output level does not jump across levels, the control
In order to alleviate the conflict between different indexes of the FC voltage, the control of the neutral point voltage
and realize the global optimization that that takes the interests and the control of the switching frequency. Secondly, the pri-
of all indexes into account, this paper designs the value ority of the three control objectives is determined, as shown
function based on the satisfactory optimization method. in Table 2. In the optimization process, objective satisfaction
According to the importance of control objectives, replaces the optimal one, in exchange for more control free-
we divide the priority of control objectives, and adopt the idea dom, so that lower priority control targets participate in the
of hierarchical optimization to design the weight coefficient optimization process.
of each control objective within the satisfaction range. Specif- When the control objectives conflict, the conflicting targets
ically, all control objectives are divided into different levels can be prioritized again. In this way, all targets can be divided
according to different priorities, and the control of each target according to different priorities, so there will be no conflict.

VOLUME 8, 2020 221419

 Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

FIGURE 8. The surface constrained by weight factors Ki , Kf and Kn .

FIGURE 10. The system delay compensation method.

FIGURE 11. The experimental prototype of 5L-ANPC inverter.

of calculation time when executing the code, resulting in the

mismatch between the pulse action time and the sampling
FIGURE 9. Taking switch state 400 as an example, the specific flow chart time, which affects the control effect and causes the control
of the O-MPC optimization. results to produce errors. Especially in the multilevel transfor-
mation, due to the number of rolling optimizations, the above
In this paper, because the control objectives of neutral point problems are particularly prominent.
potential voltage and FC voltage are both 5% of the set A simple solution to compensate for this delay is to con-
voltage, the priority of the two control objectives is set as the sider the calculation time and apply the selected switching
same. state after the next sampling instant. In this way, the control
Taking 5L-ANPC as an example, in order to simplify the algorithm is modified as follows: 1) Measurement of the load
calculation, the number of weight factors can be simplified currents. 2) Application of the switching state. 3) Estimation
because the control priority of floating capacitor voltage of the value of the currents at time tk−1 , considering the
and neutral point voltage is the same. Let K0 = Kf , then applied switching state. 4) Prediction of the load currents
the weight factors are only three Ki , Kf , Kn . The constraint for the next sampling instant tk+1 for all possible switching
relations of the three weight factors are as follows: Ki + states. 5) Evaluation of the cost function for each prediction.
Kf + Kn = 1. According to the priority division in Table 2, 6) Selection of the switching state that minimizes the cost
the priority of current control is higher than that of floating function.
capacitor voltage and neutral point voltage, while the priority The schematic diagram is shown in Figure 10. The specific
of switching frequency control is the lowest. It can be seen steps are as follows:
that Ki > Kf > Kn . Thus, the selection range of weight 1) Output the switch state calculated from the previous
coefficient can be greatly simplified, and the weight coef- beat;
ficient diagram meeting the above conditions is shown in 2) System AD sampling;
Figure 8. Then the algorithm can be further optimized in the 3) Delay compensation;
actual experiment to get the weight factors which can meet 4) Query optimization switch sequence;
the system requirements. 5) Online optimization.
If the current switch state is 400, the process of selecting
the optimal switch state is shown in Figure 9. VI. EXPERIMENTAL VERIFICATION
To verify the effectiveness of the proposed control strat-
C. DESIGN METHOD OF DELAY COMPENSATION egy, the 5L-ANPC inverter experimental prototype is built
Ideally, the controller sampling, calculation and output should for experimental verification, as shown in Figure 11. The
be completed at the same time point. However, the actual core controller of the experimental platform uses DSP
digital processing system needs to consume a certain amount TMS320C28346 and FPGA Spartan 6. FPGA is mainly used

221420 VOLUME 8, 2020

Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

FIGURE 13. Dynamic waveforms of 5L-ANPC inverter under the proposed

O-MPC method. (a) Waveform of line voltage uab . (b) the given and actual
load current waveforms of phase A.

FIGURE 12. Waveforms of phase voltage uan , current amplitude i voltage

of floating capacitor ufa and neutral point potential un of 5L- ANPC
inverter under the proposed O-MPC method.

TABLE 3. Experimental parameters.

FIGURE 14. The average switching frequency when Kn is changed. (a)

When Kn = 0.1. (b) When Kn = 0.2.

FIGURE 15. Waveforms of neutral point voltage and floating capacitor

in the calculation of rolling optimization in O-MPC algo- of 5L-ANPC inverter under the proposed O-MPC method. (a) Waveform of
rithm. The IGBT adopts Infineon FF300R12MS4, and uses neutral point voltage. (b) Waveform of floating capacitor in Phase A.
three-phase resistive load to replace the motor. The specific
parameters of the experiment are shown in Table 3. Figure 13 shows the dynamic waveforms of 5L-ANPC
In order to improve the utilization rate of DSP, the whole inverter under the proposed O-MPC method. It can be seen
rolling optimization process of predictive control is com- from Figure 13 (a) that there is no cross-level jump in
pleted in FPGA. Referring to the flow chart of the O-MPC line voltage level jump, which further verifies the control
in Figure 7, the specific implementation steps in FPGA are effect of switch state optimization described in Section IV.
as follows: 1) FPGA obtains the real-time data of actual cur- Figure 13 (b) shows the given and actual load current wave-
rent, neutral point voltage and three-phase floating capacitor forms of phase a, which can keep good sinusoidal degree
from DSP; 2) simplify the switch state involved in rolling before and after the load current mutation. At the same
according to the method mentioned in the Section IV; 3) cal- time, the load current can quickly track the given value after
culate the value function J corresponding to each switch state; the given current mutation, which shows that the proposed
4) select the switch state corresponding to the minimum value O-MPC method of 5L-ANPC inverter has good robustness
function J and send it back to DSP. ability and fast dynamic response ability.
Figure 12 shows the dynamic experimental waveforms Figure 14 shows the average switching frequency of
of 5L-ANPC inverter under O-MPC method. It can be seen 5L-ANPC inverter model predictive control when the weight
from the figure that the level jump of phase voltage is one factor Kn changes. Compared with Figure 14 (a) and (b), it can
level and there is no cross-level jump. Furthermore, it is be seen that when the weight factor Kn changes, the average
explained that the switch state optimization described in switching frequency of the system changes. At the same time,
Section IV has achieved the expected control effect; it can with the increase of the weight coefficient, the proportion
also be seen from the figure that the deviation of floating of the optimization of the average switching frequency in
capacitor voltage and neutral point potential of phase A the whole value function is increasing, and then the average
is small, and both can be stably controlled when the load switching frequency is reduced.
changes suddenly, which proves that the control theory in Figure 15 shows the experimental waveforms of neutral
Section V has good dynamic and steady-state control effect. point voltage and floating capacitor of 5L-ANPC inverter

VOLUME 8, 2020 221421

 Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

FIGURE 16. Dynamic experimental waveforms of upper and lower bus

voltages under the proposed O-MPC method. FIGURE 18. Experimental waveforms of ubn and iabc when optimizing
weight coefficients Kn .

the high-voltage and high-power driving system to reduce the

switching frequency of the equipment.

VII. CONCLUSION
In this paper, an O-MPC method was proposed to shorten
the computation time of the predictive control algorithm
FIGURE 17. The time of one calculation cycle. (a) under the traditional and improve the multi-objective performances for five-level
MPC method. (b) under the proposed O-MPC method. ANPC inverter. By establishing the switching state of the
inverter and the set constraints, the number of switching
under the proposed O-MPC method. It can be seen from states participating in dynamic optimization is successfully
the waveform that the neutral point potential fluctuation reduced. Taking 5L-ANPC inverter as an example, the num-
and the voltage fluctuation of the floating capacitor are ber of switch states participating in rolling optimization is
all within 5% through the optimization of the no weight reduced from 125 to 15, which saves more than 80% of the
factor, and the satisfactory optimization control effect is calculation time compared with the traditional MPC. In order
achieved. to solve the conflict problem of multiple control objectives,
Figure 16 shows the dynamic experimental waveforms of all control objectives are divided into different levels accord-
upper and lower bus voltages under the proposed O-MPC ing to different priorities, and the control of each target is
method. It can be seen from the figure that before the con- realized based on the order of priority from high to low. The
trol is applied, the voltage of the neutral point potential value of the weight factor is not based on the optimal value
is extremely unbalanced, and after the control is applied, of a single target, but the optimal value when multiple control
the neutral point voltage tends to balance quickly. targets are satisfied. This method has improved the multi-
The time of one calculation cycle under the traditional objective performance of the control system effectively.
MPC method and the O-MPC method are shown in Figure 17.
From the figure, we can see that the traditional MPC needs REFERENCES
about 8 µs, while the proposed control strategy only needs [1] N. D. Dao and D.-C. Lee, ‘‘Operation and control scheme of a five-
2 µs, saving 6 µs, saving about 80% of the time. It can be seen level hybrid inverter for medium-voltage motor drives,’’ IEEE Trans.
Power Electron., vol. 33, no. 12, pp. 10178–10187, Dec. 2018, doi:
that the proposed control strategy can significantly improve 10.1109/TPEL.2018.2811182.
the computing efficiency of the processor. [2] E. Burguete, J. Lopez, and M. Zabaleta, ‘‘New five-level active neutral-
Figure 18 shows the experimental waveforms of phase point-clamped converter,’’ IEEE Trans. Ind. Appl., vol. 51, no. 1,
pp. 440–447, Jan. 2015, doi: 10.1109/TIA.2014.2334737.
voltage and load current of 5L-ANPC inverter when the [3] P. Barbosa, P. Steimer, J. Steinke, M. Winkelnkemper, and N. Celanovic,
weight coefficient Kn is optimized by model predictive con- ‘‘Active-neutral-point-clamped (ANPC) multilevel converter technology,’’
trol. It can be seen from the figure that the switching fre- in Proc. Eur. Conf. Power Electron. Appl., 2005, pp. 1–10.
[4] F. Kieferndorf, M. Basler, L. A. Serpa, J.-H. Fabian, A. Coccia, and
quency of phase B voltage is significantly reduced, and the G. A. Scheuer, ‘‘A new medium voltage drive system based on ANPC-5L
switching state of phase B does not act when the load current technology,’’ in Proc. IEEE Int. Conf. Ind. Technol., Viña del Mar, Chile,
amplitude of phase B is greater than that of other two phases. Mar. 2010, pp. 643–649.
In this way, not only the switching frequency of the phase [5] M. Narimani, B. Wu, V. Yaramasu, and N. R. Zargari, ‘‘Finite control-
set model predictive control (FCS-MPC) of nested neutral point-clamped
is reduced, but also the switching loss of the phase is min- (NNPC) converter,’’ IEEE Trans. Power Electron., vol. 30, no. 12,
imized. In addition, the O-MPC method is used to flexibly pp. 7262–7269, Dec. 2015, doi: 10.1109/TPEL.2015.2396033.
allocate the weighting factors corresponding to the switching [6] B. Stellato, T. Geyer, and P. J. Goulart, ‘‘High-speed finite con-
trol set model predictive control for power electronics,’’ IEEE Trans.
frequency to achieve the purpose of controlling the switching Power Electron., vol. 32, no. 5, pp. 4007–4020, May 2017, doi:
frequency of the equipment, which provides a new idea for 10.1109/TPEL.2016.2584678.

221422 VOLUME 8, 2020

Z. Liu et al.: O-MPC Method for Five-Level Active NPC Inverter

[7] P. Karamanakos and T. Geyer, ‘‘Guidelines for the design of finite control ZHENGLONG XIA (Member, IEEE) received the
set model predictive controllers,’’ IEEE Trans. Power Electron., vol. 35, B.S. degree in electrical engineering and automa-
no. 7, pp. 7434–7450, Jul. 2020, doi: 10.1109/TPEL.2019.2954357. tion, the M.S. degree in power system and automa-
[8] X. Zhang, L. Tan, J. Xian, H. Zhang, Z. Ma, and J. Kang, ‘‘Direct grid- tion, and the Ph.D. degree in power electronics
side current model predictive control for grid-connected inverter with LCL and power drives from the China University of
filter,’’ IET Power Electron., vol. 11, no. 15, pp. 2450–2460, Dec. 2018, Mining and Technology, in 2005, 2008, and 2014,
doi: 10.1049/iet-pel.2018.5338. respectively. Since 2014, he has been an Associate
[9] L. M. A. Caseiro, A. M. S. Mendes, and S. M. A. Cruz, ‘‘Dynamically
Professor with the Department of Electrical Engi-
weighted optimal switching vector model predictive control of power
neering and Automation, Jiangsu Normal Univer-
converters,’’ IEEE Trans. Ind. Electron., vol. 66, no. 2, pp. 1235–1245,
Feb. 2019, doi: 10.1109/TIE.2018.2829689. sity, China, where he is responsible for the Youth
[10] V. Monteiro, J. C. Ferreira, A. A. N. Melendez, and J. L. Afonso, Fund of the Foundation Research Project of Jiangsu. Since 2019, he has been
‘‘Model predictive control applied to an improved five-level bidirectional a Visiting Scholar with Aalborg University, Denmark. He has published more
converter,’’ IEEE Trans. Ind. Electron., vol. 63, no. 9, pp. 5879–5890, than 40 journal articles in the fields of power quality management and fault
Sep. 2016, doi: 10.1109/TIE.2016.2558141. diagnosis. His research interests include reactive compensation of power
[11] W. Zhao, T. Tao, J. Zhu, H. Tan, and Y. Du, ‘‘A novel finite-control-set systems, fault diagnosis, and circuit theory and systems.
model predictive current control for five-phase PM motor with continued
modulation,’’ IEEE Trans. Power Electron., vol. 35, no. 7, pp. 7261–7270,
Jul. 2020, doi: 10.1109/TPEL.2019.2954285.
[12] W. Song, Z. Deng, S. Wang, and X. Feng, ‘‘A simple model predictive
power control strategy for single-phase PWM converters with modulation
function optimization,’’ IEEE Trans. Power Electron., vol. 31, no. 7,
pp. 5279–5289, Jul. 2016, doi: 10.1109/TPEL.2015.2481323.
[13] Y. Li, Z. Zhang, K. Li, P. Zhang, and F. Gao, ‘‘Predictive current control
for voltage source inverters considering dead-time effect,’’ CES Trans. DAN LI was born in Xuzhou, Jiangsu, China,
Electr. Mach. Syst., vol. 4, no. 1, pp. 35–42, Mar. 2020, doi: 10.30941/CES- in 1986. He received the M.S. degree in elec-
TEMS.2020.00006. tronic engineering from the China University of
[14] T. Chen, O. Abdel-Rahim, F. Peng, and H. Wang, ‘‘An improved Mining and Technology, Xuzhou, China, in 2019.
finite control set-MPC-based power sharing control strategy for islanded His current research interests include automation
AC microgrids,’’ IEEE Access, vol. 8, pp. 52676–52686, 2020, doi: control system, modulation strategy, and system
10.1109/ACCESS.2020.2980860. improvement.
[15] X. Liu, L. Qiu, J. Ma, Y. Fang, W. Wu, Z. Peng, and D. Wang, ‘‘A fast
finite-level-state model predictive control strategy for sensorless modular
multilevel converter,’’ IEEE J. Emerg. Sel. Topics Power Electron., early
access, Jul. 9, 2020, doi: 10.1109/JESTPE.2020.3008186.
[16] Y. Yang, H. Wen, and D. Li, ‘‘A fast and fixed switching fre-
quency model predictive control with delay compensation for three-
phase inverters,’’ IEEE Access, vol. 5, pp. 17904–17913, 2017, doi:
10.1109/ACCESS.2017.2751619.
[17] M. Aguirre, S. Kouro, C. A. Rojas, J. Rodriguez, and J. I. Leon, ‘‘Switching
frequency regulation for FCS-MPC based on a period control approach,’’
IEEE Trans. Ind. Electron., vol. 65, no. 7, pp. 5764–5773, Jul. 2018, doi: YU WANG was born in Xuzhou, Jiangsu, China,
10.1109/TIE.2017.2777385. in 1989. He received the M.S. degree in elec-
[18] M. Tomlinson, H. D. T. Mouton, R. Kennel, and P. Stolze, ‘‘A fixed switch- tronic engineering from the China University of
ing frequency scheme for finite-control-set model predictive control— Mining and Technology, Xuzhou, China, in 2016.
Concept and algorithm,’’ IEEE Trans. Ind. Electron., vol. 63, no. 12, His current research interests include automation
pp. 7662–7670, Dec. 2016, doi: 10.1109/TIE.2016.2593997. control systems, modulation strategy, and system
[19] K. Wang, L. Xu, Z. Zheng, and Y. Li, ‘‘Capacitor voltage balanc- improvement.
ing of a five-level ANPC converter using phase-shifted PWM,’’ IEEE
Trans. Power Electron., vol. 30, no. 3, pp. 1147–1156, Mar. 2015, doi:
10.1109/TPEL.2014.2320985.
[20] Z. Liu, Y. Wang, G. Tan, H. Li, and Y. Zhang, ‘‘A novel SVPWM
algorithm for five-level active neutral-point-clamped converter,’’ IEEE
Trans. Power Electron., vol. 31, no. 5, pp. 3859–3866, May 2016, doi:
10.1109/TPEL.2015.2462124.

FEI LI (Member, IEEE) was born in Xuzhou,

Jiangsu, China, in 1982. He received the B.S.
ZHAN LIU (Member, IEEE) was born in Xiaoxian, degree in electrical engineering and automation
Anhui, China, in 1989. He received the B.S. degree from the China University of Mining and Tech-
in electrical engineering and automation, the M.S. nology, Xuzhou, China, in 2005, the M.S. degree
degree in power electronics and power drives, and in electronic engineering from the University of
the Ph.D. degree in electrical engineering from Duisburg–Essen, Germany, in 2009, and the Ph.D.
the China University of Mining and Technology, degree in power electronics and electrical drive
Xuzhou, China, in 2011, 2013, and 2016, respec- from the China University of Mining and Technol-
tively. Since 2017, he has been a University Lec- ogy, in 2017.
turer with the School of Electrical Engineering and He is currently working with the School of Electrical Engineering and
Automation, Jiangsu Normal University, China, Automation, Jiangsu Normal University, Xuzhou, as a University Lecturer.
where he is responsible for the National Natural Science Foundation of His current research interest includes power electronic converters for electri-
China. His research interests include power electronics, modern control cal drives and power quality.
theory, model predictive control, and multilevel converter.

VOLUME 8, 2020 221423