This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

A Constant Switching Frequency Finite-Control-Set Predictive Current

Control Scheme of Five-Phase Inverter with Duty Ratio Optimization

Cheng Xue, Wensheng Song, Member, IEEE, Xuesong Wu, and Xiaoyun Feng

Abstract- As is well known, the additional low-order harmonic converter system, multiphase inverters have several salient
currents and the huge computation amount of the cost function features：First, higher power and current can be delivered
value with all available control sets, are the two special issues of using power electronic devices in huge-power and
FCS-MPCC for multiphase inverters. Thus, this paper proposes low-voltage occasion, such as: electromagnetic launch system,
a finite-control-set model predictive current control (FCS-MPCC) railway locomotive traction drive, marine and aerospace
scheme with duty ratio optimization for five-phase inverters. propulsion. Second, there are more control variables that can
Firstly, a group of the reconstructed virtual voltage vectors (V3s) be manipulated to afford stronger fault tolerance [5]-[8], and
with low-order harmonic elimination is defined and employed as the enhanced reliability is subsequently coming. Third,
the control sets in FCS-MPCC scheme. Thus, it leads to no multiphase inverters can also provide an improved
requirement for harmonic items and weighting factors existing in magnetomotive force waveform, lower torque pulsation
the cost function. Then, a combination of zero vectors and the [9]-[10], and higher torque density for their applications
selected virtual voltage vector during each control horizon is [11]-[12].
adopted to rearrange the pulse sequence, which can realize Model predictive control (MPC) has been becoming a hot
constant switching frequency. In addition, the durations of the topic of academic and industry for more than three decades,
selected virtual voltage vector and zero vectors during each and the most classical form is the finite-control-set MPC
control horizon are also estimated to minimize the steady-state (FCS-MPC) [13]. Among various kinds of FCS-MPC
current ripples in different subspaces based on cost function methods, model predictive current control with the
optimization. Finally, a comparison of the proposed finite-control-set (FCS-MPCC) is one of the most important
V3-FCS-MPCC scheme and the other FCS-MPCCs is presented and classical solution in power electronics. MPC schemes
in simulation and experimental test. Simulation and have been widely reported for three-phase inverter
experimental results verify that the proposed V3-FCS-MPCC applications [14] and popularized to a five-phase inverter
scheme can remain the simplified structure, fast dynamic driving a star-connected R-L load [15], [16]. FCS-MPCC is a
performance, constant switching frequency, and achieve typical direct current control scheme that predicts the future
minimum current ripples in different subspaces. evolution of currents in discrete time using a predictive model.
Index Terms- Current control, finite-control-set(FCS), model It takes into account the discrete nature of power converters
predictive control(MPC), multi-phase inverters, constant
to solve the online optimization problem using a single cost
switching frequency, virtual voltage vector (V3), duty ratio
function and does not require an external modulator [17]. The
switching vector that minimizes cost function is selected as
the output of the controller. The FCS-MPCC scheme has an
I. INTRODUCTION
intuitive concept and straightforward implementation.
The interest in the area of multiphase inverters has Furthermore, the considered variables and control objectives
substantially increased during the last decade because of the of system can be easily presented by designing a specific cost
potential advantages that they offer for multiphase machine function with adjustable weighting factors. However,
drives [1]-[4]. Compared with conventional three-phase power FCS-MPCC needs to solve the optimization problem during
Manuscript received January 02, 2017; revised Febuary 26, 2017; accepted each control horizon, which leads to the heavy calculation
for publication May 06, 2017. This work was supported in part by the National amount. Especially, for long steps prediction and multilevel
Natural Science Foundation of China under Project 51577160. Recommended or irregular converter topologies application, the calculation
for publication by Associate Editor XXXXXX. (Corresponding author: complexity can rise exponentially [17]. As presented in [18]
Wensheng Song.). and [19], a synchronous current control scheme with a full set
The authors are with the School of Electrical Engineering, Southwest of switching states taken as the input control set of the
Jiaotong University, Chengdu 610031, China (e-mail: FCS-MPCC for five-phase drive system was addressed.
[email protected]; [email protected]; [email protected]; Usually, a heavy calculation amount means that a lower
[email protected]). sampling frequency should be employed for the sake of
executing solution procedure, which will increase current

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

ripple and deteriorate current waveform quality.  Employing the pre-defined virtual voltage vectors as the
In order to reduce the calculation effort due to the control set to reduce the low-order current harmonics. Hence,
increased number of iterations, various advanced techniques it is not necessary to design harmonic item and the weighting
have been applied for multi-phase system. A simplified factor in the cost function.
algorithm that employs adjacent voltage vector was proposed  Enhancing the output current quality, comparing with
to reduce the available voltage vectors of the control set in the several traditional FCS-MPCC schemes. Steady state
[20]. Similarly, [21] proposed a computation reduction performance will be better while the fast dynamic response is
solution that eliminates unfeasible switching states inherent in reserved.
reduced-switch-count inverters. These two methods can  Taking the discrete nature of power converters into
reduce computational burden effectively but at the expense of account and selects the best virtual voltage vector by
worse control optimality. Additionally, according to minimizing a predefined cost function. Hence, the simplified
predefined constraints, a predictive control technique was structure is reserved and no modulation stage is required.
adopted in [22] to reduce the available switching states of  Realizing constant switching frequency for inverters.
voltage source inverters. However, the shortcoming of the
adopted scheme is the weak algorithm portability. Therefore, Vdc Sa1 Sb1 Sc1 Sd1 Se1
how to reduce the computational cost in FCS-MPCC scheme 2 a
L R
b
for multi-phase drive system still needs to be studied further. c
d
Recently, to achieve excellent steady-state performance of e
three-phase inverter-fed motor drive systems, duty cycle Vdc Sa2 Sb2 Sc2 Sd2 Se2
2
control concept was proposed to be applied in FCS-MPC [23],
[24]. For example, in order to reduce torque ripple of
induction motor, a combination of the selected active voltage
vector and null vector in each control horizon is adopted in Fig. 1. The topology of DAB dc-dc converters
the proposed FCS-MPC reported in [23]. Different from
The paper is organized as follows. The discrete model of
traditional optimal duty cycle control, [23] reported that the
five-phase inverter system is described and the basic concept
way of selecting voltage vector firstly based on the principle
of the FCS-MPCC algorithm is presented in Section II.
of cost function minimization can no longer remain optimal
Section III explains the proposed scheme with duty ratio
in minimizing torque and flux errors. Thus it realizes vector
optimization. Section IV shows the comparative experimental
selection and duty ratio optimization simultaneously. The
results of the proposed method and traditional FCS-MPCC
duration of the active vector is obtained from deadbeat torque
schemes. Finally, Section V concludes this paper.
controller. Consequently, better torque performance under
lower sampling frequency condition can be observed. In
II. SYSTEM MODELING AND TRADITIONAL FCS-MPCC OF FIVE-PHASE
addition, better steady-state performance during wide speed
range can be achieved, compared with the traditional INVERTERS
duty-cycle optimal control schemes. As for multi-phase drive
The main topology of a two-level five-phase voltage
system, the work with an evaluation of the FCS-MPC quality
source inverter connected to a passive R-L load is depicted in
using duty ratio optimization is only discussed in [25], where
Fig. 1. The load is taken star-connected and the input dc-link
a five-phase inverter is considered for a two-motor
voltage Vdc is regarded as being constant. Five-phase bridge
three-phase drive. However, the work reported in [25] is not
legs are symbolled and expressed as a,b,c,d,e, respectively.
available for an independent multi-phase load system.
Symbol S with subscript represent the drive signal of each
This paper aims to deal with the aforementioned issues in
switch device. The upper and lower devices of each leg
the FCS-MPCC for five-phase drive systems. Virtual voltage
operate in a complementary manner.
vectors (V3s) are reconstructed firstly for releasing the
computational burden of digital signal processor. Then, a new A. Voltage vectors distribution
integrated FCS-MPCC method with the duty ratio estimation
of V3s in five-phase inverters is proposed, which can enhance The five-phase inverter has 32 voltage vectors, which can
torque waveform quality, reduce the low-order harmonic be defined as the binary value of Vi=Sa1Sb1Sc1Sd1Se1. “1”
components of phase currents, and keep constant switching indicates the upper switch of the corresponding leg is on,
frequency for the power switches. Compared with the while “0” indicates the lower switch is on. Since five-phase
existing FCS-MPCC schemes for multiphase inverters, the variables present four degrees of freedom with the exception
salient features of the proposed FCS-MPCC scheme can be of zero-sequence components, two space vectors should be
summarized as: used in order to define the equivalent transformation of
 Proposing an integrated FCS-MPCC method with duty five-phase variables [26]. The fundamental components of the
ratio optimization and making it expandable to five-phase inverter variables and the harmonics with order k=10m±1
inverters even other multi-phase system. (m=1,2,3…) are transformed into the α-β subspace, whereas

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

the harmonics with k=10m±3 (m=1,2,3…) are mapped into large vectors, medium vectors, and small vectors can be
the x-y subspace. The voltage vectors can be expressed with expressed as
the switching functions as
VLarge Vmax =22 5Vdc cos36=0.6472Vdc
2 
V =V +jV = Vdc (Sa1 + Sb1 + 2 Sc1 + 3 Sd1 + 4 Se1 ) (1) VMiddle Vmid = 2 5Vdc =0.4Vdc (3)
5 V
2  Small Vmin =22 5Vdc cos 72=0.2472Vdc
Vxy =Vx +jVy = Vdc (Sa1 + Sc1 + 2 Se1 + 3 Sb1 + 4 Sd1 ) (2)
5 Thus, the active voltage vectors in the α-β subspace can be
divided into three decagons, from inner decagon to outer
Where =exp(j2π/5).
decagon: D1, D2, D3 [Fig.2 (a)]. The ratio between two
β y
consecutive magnitudes is a constant of 0.618-1. From Fig.2,
please note that the inner decagon D1 in the α-β subspace is
VL3 VS10
VL4 D3 VS7 D1
mapped as the outer decagon in the x-y subspace. The outer
VL5 VM4
VM3 D2 VL2
VS4 VM2
VM5 D2 VS3 decagon D3 in the α-β subspace is mapped as the inner
D1 VM2 D3 VM8
VM5
VS4 VS3
VS2 VM9 VL7
VL10
VL3 decagon in the x-y subspace. But, the middle decagon D2 of
the α-β subspace is still mapped as the middle decagon in the
VS5 VL4
V0 V31 VS1 VM1 VL1 V0 V31 VL6 VM1 VS6
VL6 VM6 VS6 α VS1 VM6 VL1 x

VS7
VS10
VL8
VL9
VM4
x-y subspace. The corresponding switching state of each
VS9 VM10 VL2
VM7 VS8
VM9
VL10
VM3 VL5
VM7 VS9
voltage vector is summarized and listed in Table I.
VL7 VM8 VS8 VM10

VL8
VL9
VS5
VS2 B. Prediction model of R-L load

The load current dynamics can be described by the

(a) (b) vector equation

Fig. 2. Space vector diagram of five-phase inverters in two subspaces. di

V =Ri +L (4)
(a) α-β subspace. (b) x-y subspace dt
where V, i are the space voltage and current vectors in both
TABLE I α-β and x-y subspace. R, L represent the resistive and inductive
SPACE VECTOR GROUPS AND CORRESPONDING SWITCHING loads. Assuming the small sampling interval Ts, and applying
STATES
the first-order discrete approximation to voltage and current
vector equations in (4), the future value of load current can be
predicted as
1
i (k +1)= [Li (k )+TsV (k +1)] (5)
RTs +L
Where V(k+1) represents the candidate voltage vectors in
the finite control set. The term RTs could be neglected if the
sampling interval is small enough.

C. Traditional FCS-MPCC Schemes

The overall control diagram of the traditional FCS-MPCC

for five-phase inverters is shown in Fig.3. The predictive
control scheme is performed in the following steps:
1) Measuring the load currents at kth sampling instant;
2) Predicting the load currents at (k+1)th sampling instant
for all candidate voltage vectors in the finite control set;
3) Evaluating the cost function value g for each prediction
The α-β and x-y subspace voltage vectors can be described result;
in Fig.2a and Fig.2b, respectively. According to the amplitude, 4) Selecting the corresponding switching state that can
the voltage vectors in α-β subspace are classified into four minimize the cost function g from all candidate voltage
types: zero vectors (V0,V31), small vectors (VS1-VS10), medium vectors;
vectors (VM1-VM10), and large vectors (VL1-VL10), as shown in Once when the prediction and control time horizons are set,
Fig.2a. The space plane pattern is divided as ten sectors. The the cost function remains as the only tuning tool that exists in
space angle range in each sector is π/5. Amplitudes of these the FCS-MPCC. Typically, two forms of cost function are used,
i.e., linear and quadratic [27]. For five-phase inverters, in order

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

to achieve the sinusoidal current load waveform, not only the predefined cost functions (6) and (7). In addition, in these
currents in α-β subspace need to be considered into the cost three traditional FCS-MPCCs, the heavier computational
function, but also the harmonic currents in x-y subspace need efforts of the control scheme, the smaller low-order current
to be contained. Therefore, the cost function must be relative to harmonics. For example, computational burden in MPCC31 is
four current error terms, and usually defined as the following heaviest, while the low-order current harmonic values are
structures: smallest in these three schemes.
g = i* (k +1)-i (k +1) + i* (k +1)-i (k +1)
(6) III. PROPOSED FCS-MPCC WITH DUTY RATIO OPTIMIZATION
+1 ix* (k +1)-ix (k +1) +2 i*y (k +1)-i y (k +1)
A. Definition of Virtual voltage vectors
g =[i* (k +1)-i (k +1)]2 +[i* (k +1)-i (k +1)]2
(7) As aforementioned, heavy computational burden and large
+1[ix* (k +1)-ix (k +1)]2 +2 [i*y (k +1)-i y (k +1)]2 low-order current harmonics are two existing key issues in
where i* with the subscript represents the desired value of the the traditional FCS-MPCC of the five-phase drive system.
corresponding currents in α-β and x-y subspace, respectively. According to (4), the harmonic currents ix, iy are produced by
λ1 and λ2 represent the weighting factors of cost function. In the harmonic voltages existing in x-y subspace. Hence, if the
this paper, the quadratic form is employed here as it can harmonic average voltages are approximately zero in x-y
produce a higher penalization of bigger errors compared with subspace during each control interval, the harmonic currents
smaller ones and better insight into the cost values. In ix, iy can be eliminated effectively.
addition, the differential process of cost function in (6) is very From Fig.2a and Fig.2b, it can be noticed that the two
difficult, due to the expression type of the absolute value inverter states with aligned large and medium voltage vectors
function. The expression of cost function in (7) is widely in the α-β subspace, e.g., inverter states VL1 and VM1, become
adopted in the FCS-MPCC schemes, especially in small and medium voltage vectors pointing to the opposite
FCS-MPCC with duty-ratio optimization or PWM stage [28]. directions in the x-y subspace. On the basis of this principle,
Thus, quadratic form of cost function in (7) is also used in as the amplitude ratio between the medium and small voltage
this paper. vectors is 0.618-1 in x-y subspace, the duration ratio of
medium vector VM1 and large vector VL1 in α-β subspace can
ik* be set as 0.618 to synthesize a zero-mean volt-second product
Minimization SIGBT in the x-y subspace. The process is presented in Fig.4. This
of Load
Predictive model g function combination mode of large and medium vectors and the
ik +1 corresponding dwell times design can eliminate low-order
ik Finite Control Set Sk current distortion, and form ten active virtual voltage vectors
(V3s) [29], named as Vv1–VV10, in α-β subspace, as shown in
Fig. 5.
Fig. 3. Block diagram of traditional MPCC schemes VM1 VL1 VM1
VL1 zero-mean
Vv1
In general, there are three classical FCS-MPCC schemes
for five-phase inverter according to the categories of the (a) (b)
selected FCS. The FCS-MPCC that uses all the available
Fig. 4. Schematic diagram of synthetic vectors V v1 in different
switching states is termed as FCS-MPCC31. All the ten largest
voltage vectors and zero voltage vectors are taken as the subspaces. (a) α-β subspace. (b) x-y subspace
control set, is named as the FCS-MPCC11. Similarly,
FCS-MPCC21 method takes ten large vectors, ten medium β
vectors and zero vectors as the input control set [27]. In Vv4 Vv3
essence, [27] has demonstrated that small vectors is not
Vv2
necessary to be included as elements in the input control set Vv5
for the reason that small vectors in α-β subspace will become
higher voltage vectors in the x-y harmonic subspace, which V0 V31 Vv1
leads to large low-order current harmonics. In addition, the Vv6 α
0.5527Vdc
calculation burden of FCS-MPCC21 and FCS-MPCC31
Vv10
schemes is very heavy.
Vv7
Anyway, all of these traditional FCS-MPCCs applying a
Vv8 Vv9
single switching state during each control interval, inevitably
lead to an additional harmonic excitation in the x-y subspace,
despite there are the harmonic component constraints in the Fig. 5. Virtual voltage vectors in α-β subspace

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

Following the above defined combination rule, the amplitudes of control interval.
virtual vectors in α-β and x-y subspaces are expressed as The prediction can be realized through one step forward
 Vv   (1-kv )  0.4Vdc  kv  0.6472Vdc shift of the load model. And cost function with the predicted
 (8) currents during (k+2)th control interval is shown as
 Vv xy  kv  0.2472Vdc  (1-kv )  0.4Vdc g =[i* (k +2)-i (k +2)]2 +[i* (k +2)-i (k +2)]2 (11)
Where kv represents duration ratio of the large voltage vector
during every control interval Ts. Substituting kv=0.618 into (8) yields D. Duty ratio determination
 Vv   0.5527Vdc The optimization control object of MPCC is to force the
 (9) predictive currents to reach the reference values. In traditional
 Vv xy  0 FCS-MPCC schemes, a single switching state is widely
Please note that the amplitudes of these 10 virtual voltage vectors adopted during the entire control interval Ts, and it leads to
are all equal to 0.5527Vdc. large current ripples, especially in the low sampling
frequency condition. Although the virtual voltage vectors are
B. Cost function optimization used as the input control set, they still cannot eliminate the
tracking error completely. In most cases, it is not necessary
The goal of cost function is to adjust the system behavior for the virtual voltage vector being applied in the entire
through comparing the predicted values with reference values. control interval to meet the desired control requirements. The
It plays a great role in the FCS-MPCC scheme. For corresponding duty ratio of the selected virtual voltage vector
five-phase drive system, in order to achieve sinusoidal current during each control interval should be estimated, in order to
waveform and avoid current distortion in the x-y subspace, force the actual currents to track the reference values
the harmonic current references ix* and iy* in cost function accurately. Thus, this paper proposes a combination solution
shown in (6) are both set to zero. In the proposed FCS-MPCC of virtual voltage vector and zero voltage vectors to further
scheme, these 10 virtual voltage vectors and zero voltage reduce tracking error. In this condition, if the dwell time of
vectors are set as the all available elements in the input virtual voltage vector Vvj during the entire control interval is
control set. Thus, harmonic currents ix and iy can be defined as djTs, zero voltage vector is applied in the rest
eliminated automatically by the pre-defined virtual voltage interval (1-dj)Ts, where dj represents the duty ratio of virtual
vectors. Therefore, the issue of low-order harmonic currents vector VVj during the entire control interval.
is unnecessary to be considered into cost function. In other For the predictive current in α-β subspace, the
words, the controlled quantities in the cost function are only computational expressions at (k+2)th sampling instant can be
current components in α-β subspace. The proposed deduced from (4) as
FCS-MPCC scheme can reduce computational burden further.
 1
The modified cost function is shown as i (k +2)= RT +L [Li (k +1)+d jTsVvj (k +2)]

j  1, 2...10 (12)
s
g =[i* (k +1)-i (k +1)]2 +[i* (k +1)-i (k +1)]2 (10) 
i (k +2)= 1 [Li (k +1)+d T V (k +2)]
Please note that there is no weighting factor. It is easy to be   RTs +L
 j s vj 

implemented and applied in the practical application. Where Vvjα, Vvjβ are α-axis and β-axis components of
virtual voltage vector VVj, respectively. The optimization
C. Digital delay compensation
control object is to minimize cost function g in (7) by
As is well known, in real-time implementation, one-step selecting the appropriate duty ratio dj, which yields
delay between the desired vector and the actual vector caused g
=0 (13)
by digital processing will deteriorate the performance of d j
FCS-MPCC, particularly when the calculation period is
significant compared with the sampling period [30]. To Substituting (11) and (12) into (13), dj can be derived from
compensate this delay, the digital delay compensation scheme (14). dj is in the range from 0 to 1. And fortunately, the best
in [30] is employed in this paper. solution dj is irrelative to the weighting factor λ for the reason
The measured currents and the applied switching state at that harmonic currents are not contained into the adopted cost
the kth sampling instant are used in (5) to estimate the load function, the computation process of dj can be simplified.
current values at the (k+1)th sampling instant. Then, these
estimated currents are used as the initial prediction values for
all available vectors in the defined control set during (k+2)th
[Vvj (k +2)i* +Vvj (k +2)i* ](RTs +L)-L[Vvj (k +2)i (k +1)+Vvj (k +2)i (k +1)]
dj= (14)
Ts [Vv2j (k +2)+Vv2j  (k +2)]

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

The first step of calculating dj in the proposed method is to V0 VM1 VL1 V0 V0 VM1 VL1 V31 VL1 VM1 V0
Sa1 Sa1
select an optimal virtual voltage vector or zero vector at the Sb1 Sb1
maximum duty ratio value (dj =1) from the cost function Sc1 Sc1
Sd1 Sd1
estimation for 11 times, and then the best duty cycle of the Se1 Se1
selected virtual voltage vector can be deduced from (13), as d1(1-kv)Ts d1kvTs d1(1-kv)Ts d1kvTs d1kvTs d1(1-kv)Ts
2 2 2 2
shown in Fig. 6. The last step is to assemble pulse sequence Ts Ts
units with virtual voltage vector and zero-vector, which will
(a) (b)
be translated into the drive signals to five-phase inverters.

β 3
Fig. 7. Switching sequence of virtual voltage vector (V v1 ). (a) Initial
Step1: select the best V
at the maximum duty ratio circle mode. (b) Symmetrical mode.
Vvj
Step2: determine the best
3 Start
duty ratio circle of V
VV(k+2)j Min{gj}
α Sampling
Phase-currents
Predicting VVV
iα(k+2)j iβ(k+2)j selction
Delay
Compensation
Calculating
Calculating
gj(k+2)
Corresponding
Coordinate
Fig. 6. The determination steps d j in the proposed FCS-MPCC duty ratio dj
Transformation j++
iα(k+1) iβ(k+1)
scheme.
Generating
N
j>11 switching sequence
j=1
The first step of calculating dj in the proposed method is to Y
End
select an optimal virtual voltage vector or zero vector at the
maximum duty ratio value (dj =1) from the cost function
Fig. 8. The determination steps d j in the proposed FCS-MPCC
estimation for 11 times, and then the best duty cycle of the
selected virtual voltage vector can be deduced from (13), as scheme.
shown in Fig. 6. The last step is to assemble pulse sequence
units with virtual voltage vector and zero-vector, which will From the above analysis, the proposed FCS-MPCC scheme
be translated into the drive signals to five-phase inverters. employing the virtual voltage vectors as input control set is
easy to be implemented. Moreover, the computational burden
E. Switching sequence and implementation in the proposed FCS-MPCC is dramatically lower than that in
FCS-MPCC21 method, because there are only 11 vectors for
Take Vv1 as an example, Fig.7a shows the original cost function evaluation during every control interval. Fig.8
combination switching sequence of the selected virtual shows the flowchat diagram of the proposed V3-FCS-MPCC11
voltage vector Vv1 and zero voltage vectors, where the with duty-ratio optimization scheme. The block diagram for
duration ratio of VM1 and VL1 are fixed at 0.618. Due to the the proposed FCS-MPCC scheme is shown in Fig. 9, where
asymmetry of this pulse sequence, there exists some the dashed box indicates the duty ratio estimator. Compared
additional harmonics around the switching frequency. And it to traditional FCS-MPCC, the simplicity structure is reserved.
is not easy to be realized in digital processors such as DSP
TMS320F2000 series. For the most commonly used IV. SIMULATION RESULTS
pulse-width-modulation (PWM) generation technique, the
inverted and non-inverted signals are symmetrical to the To verify the effectiveness of the proposed FCS-MPCC
middle of the PWM period. This type of PWM waves can be with duty-ratio optimization scheme, a simulation comparison
considered as the standard switching sequence for the of the proposed scheme and four traditional FCS-MPCC
real-time systems [26][31]. Therefore, according to the schemes is shown in the MATLAB/Simulink environment.
specific PWM generation technique, symmetrical Five kinds of current control schemes are named as
arrangement of switching sequence is firstly required, as FCS-MPCC11, FCS-MPCC21, FCS-MPCC31, FCS-MPCC11-V3,
presented in Fig.7b. Thus, the actual applied voltage vector and the proposed FCS-MPCC11-V3-DRO, respectively. Where
sequences are changed from V0–VM1–VL1–V0 to FCS-MPCC11-V3 represents a form of traditional
V0–VM1–VL1–V31–VL1–VM1–V0. The corresponding duration FCS-MPCC11 scheme employing 10 virtual vectors and zero
of each switching state is shown in Fig.7b. The switching vectors as the control set; FCS-MPCC11-V3-DRO represents the
sequence in Fig.7b is absolutely same with that of the proposed FCS-MPCC11 with duty ratio optimization of virtual
traditional space vector PWM (SVPWM), which can realize voltage vectors.
constant switching frequency.

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

The Selected
g Cost Function Minimizer V3
Eq.(10) gj Min (g)
Drive
Signal Vabcde
The Selected V3
i (k +2) j i (k +2) j

Predictive Model Duty Ratio Calculation

Eq.(4) Eq.(13) d

Constant Switching Frequency

VVj (j =1...11) Sequence

i (k +1) Coordinate
Delay
Transformation
i (k +1) i (k +1) Compensation i (k )
abcde/αβ

Current Sensor
FCS Contained
Ten V3+zero vectors

Fig. 9. The proposed FCS-MPCC11- V3-DRO scheme of five-phase inverters.

Fig. 10 shows steady-state simulation results of the 2 2 Reference current

output phase current, current FFT analysis, and current 1 1

iβ iy (A)
trajectories in two bi-dimensional subspaces for the

ia(A)
0 0
traditional FCS-MPCC11, FCS-MPCC21, FCS-MPCC31, -1 -1
Harmonic current

FCS-MPCC11-V3, and the proposed FCS-MPCC11-V3-DRO -2 -2

0 0.04 0.08 0.12 0.16 0.2 -2 -1 0 1 2
scheme at same sampling frequency. In this figure, it is Time(s) iα ix(A)
clear that the phase current waveform and the current
(e)
ripples on two bi-dimensional subspaces in the proposed
method are superior to the other four FCS-MPCC schemes. Fig. 10. Simulation comparison of steady-state performance for five
FCS-MPCC schemes at same sampling frequency (10kHz) (a)
In addition, it should be noted that FCS-MPCC21 and FCS-MPCC11. (b) FCS-MPCC21. (c) FCS-MPCC31. (d) FCS-MPCC11-V3. (e)
FCS-MPCC31 method can reduce the low-order current Proposed FCS-MPCC11-V3-DRO method.
harmonics to a certain extent compared to FCS-MPCC11
scheme. Furthermore, Fig. 10b and c have verified that it is Fig. 11 shows steady-state simulation results of the
not necessary to set small vectors as elements in the control output phase current, current FFT analysis, and current
set. trajectories in two bi-dimensional subspaces for five control
2
schemes at same switching frequency. The average
2 Reference current
switching frequency in FCS-MPCC31 and FCS-MPCC11-V3
1 1
iβ iy (A)

is about 5kHz. Meanwhile, the switching frequency in the

ia(A)

0 0
Harmonic current proposed FCS-MPCC11-V3-DRO scheme is also set as 5kHz.
-1 -1
From Fig. 11, it is clear that, the phase current waveform
-2
0 0.04 0.08 0.12 0.16 0.2
-2
-2 -1 0 1 2 and the current ripples in the proposed FCS-MPCC11-V3-DRO
Time(s) iα ix(A)
scheme under the same switching frequency is still superior
(a) to those in the adopted FCS-MPCC31 and FCS-MPCC11-V3,
2 2 Reference current
which has verified the effectiveness of the proposed
1 1
scheme.
iβ iy (A)

2 2
ia(A)

Reference current
0 0
Harmonic current 1 1
iβ iy (A)

-1 -1
ia(A)

0 0
-2 -2 Harmonic current
0 0.04 0.08 0.12 0.16 0.2 -2 -1 0 1 2
iα ix(A) -1 -1
Time(s)
-2 -2
(b) 0 0.04 0.08 0.12 0.16 0.2 -2 -1 0 1 2
Time(s) iα ix(A)
2 2 Reference current
(a)
1 1
iβ iy (A)

2 2
ia(A)

Reference current
0 0
Harmonic current 1 1
iβ iy (A)

-1 -1
ia(A)

0 0
-2 -2
0 0.04 0.08 0.12 0.16 0.2 -2 -1 0 1 2 Harmonic current
iα ix(A) -1 -1
Time(s)
-2
0 0.04 0.08 0.12 0.16 0.2 -2-2
(c) -1 0 1 2
Time(s) iα ix(A)
2 2 Reference current
(b)
1 1
iβ iy (A)

2 2
ia(A)

0 Reference current
0
Harmonic current 1 1
iβ iy (A)

-1 -1
ia(A)

0 0
-2 -2
0 0.04 0.08 0.12 0.16 0.2 -2 -1 0 1 2 Harmonic current
iα ix(A) -1 -1
Time(s)
-2 -2
(d) 0 0.04 0.08 0.12 0.16 0.2 -2 -1 0 1 2
Time(s) iα ix(A)

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

(c) in α-β subspace cannot be eliminated completely. Fig. 13c
Fig. 11. Simulation comparison of steady-state performance for five and Fig. 13d show experimental results of the proposed
FCS-MPCC schemes at same switching frequency (5kHz) (a) FCS-MPCC31. FCS-MPCC11-V3-DRO scheme with asymmetric and
(b) FCS-MPCC11-V3. (c) Proposed FCS-MPCC11-V3-DRO method. symmetrical switching sequences, respectively. It is clear
that the proposed FCS-MPCC11-V3-DRO can further reduce
V. EXPERIMENTAL RESULTS more current ripples both in α-β and x-y subspaces
Fig. 12 shows a scale-down experimental test setup of compared with FCS-MPCC11-V3 in Fig. 13b. In addition,
five-phase inverter configuring with Infineon comparing Fig. 13c and Fig. 13d, the symmetrical
F4-30R06W1E3 insulated-gate bipolar transistor (IGBT, switching sequence scheme can achieve the better
30A/600V) modules. A dc power supply (Vdc=40V) is fed sinusoidal current quality and lower current THDi value
for the adopted inverter with five-phase star-connected than the asymmetric solution.
resistive-inductance loads (R=10Ω, L=4.5mH). The control 2
THDi=16.78% Reference current
scheme is programmed and implemented in TI DSP id
1
TMS320F28335. The output phase currents are measured

iβ iy (A)
ia 0
with LA 35-NP current Hall sensors.
Analog signal ua -1
Harmonic current
Oscilloscope input port
Host computer Signal -2
-2 -1 0 1 2
modulate iα ix(A)
DSP28335
Controller
(a)
Optical fiber
signal emission
2
THDi=8.19% Reference current
DC
source
id
DSP Emluator 1

iβ iy (A)
Current clamp ia 0
Harmonic current
Current sensor
Auxiliary
ua -1
Five-phase
power supply
inverter
-2
-2 -1 0 1 2
Optical fiber iα ix(A)
signal reception

Inductive load Resistive load

Driving module (b)
of IGBT

2
THDi=6.74% Reference current
Fig. 12. An experimental configuration of five-phase inverter test id
platform. 1
ia

iβ iy (A)
0
A. Steady-State Performance of Different FCS-MPCC ua Harmonic current
-1
Methods (Same Sampling Frequency)
-2
-2 -1 0 1 2
iα ix(A)
In order to verify the feasibility of the proposed
FCS-MPCC, at the same sampling frequency (10kHz), (c)
Fig.13 shows experimental results of current, phase-voltage,
current FFT analysis, and current trajectories in two 2
THDi=4.63% Reference current
id
bi-dimensional subspaces for the traditional FCS-MPCC11, 1
FCS-MPCC11-V3, and the proposed FCS-MPCC11-V3-DRO
iβ iy (A)

ia 0
schemes. It is clear that the phase current waveform and the Harmonic current

current ripples on two bi-dimensional subspaces in the ua -1

proposed method are superior to the other two FCS-MPCC -2

-2 -1 0 1 2
schemes. Current total harmonics distortion (THDi) value iα ix(A)

in the proposed scheme is smallest (THDi=4.63%). In Fig. (d)

13a, it can be found that, although FCS-MPCC11 scheme
Fig. 13. Comparison of steady-state performance of various FCS-MPCC
can suppress the harmonic currents in x-y subspace due to methods at same sampling frequency (10kHz) (a) Traditional FCS-MPCC11,
the cost function contained the harmonic items (7), there (b) FCS-MPCC11-V3, (c) Proposed FCS-MPCC11-V3-DRO method, (c)
inevitably exists 3th and 7th order harmonic excitations in Proposed FCS-MPCC11-V3-DRO with asymmetrical switching sequence, (d)
Proposed FCS-MPCC11-V3-DRO method with symmetrical switching
the x-y subspace because only a single switching vector is sequence.
applied during each control interval. In Fig. 13b,
FCS-MPCC11-V3 can reduce the harmonic currents in x-y Fig.14 shows current THDi values in three FCS-MPCCs
subspace partly. It should be mentioned that, the average with respect to different sampling frequency and reference
volt-second can be kept at zero when the instantaneous current value. In this figure, it is clear that the proposed
volt-second is still larger than that in the proposed FCS-MPCC11-V3-DRO can always achieve the smallest THDi
FCS-MPCC11-V3-DRO in each control interval, because value among the three FCS-MPCC schemes at different
FCS-MPCC11-V3 is a special case of FCS-MPCC11-V3-DRO sampling frequency and reference current. The same
with dj=1. Moreover, the current steady-state tracking error sampling frequency means the same computing time for the

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

execution of predictive control algorithm. For the traditional
FCS-MPCC11 and FCS-MPCC11-V3, although the current 1 1

0.8 0.8
quality can be improved in higher sampling frequency

Duty ratio d
Duty ratio d
0.6 0.6
condition, the algorithm execution process period will be
0.4 0.4
limited during shorter sampling interval. Fortunately, the 0.2 0.2
proposed FCS-MPCC11-V3-DRO can obtain better performance 0 0
0 0.1 0.2 0 0.1 0.2
under the limitation of low sampling frequency, due to Time(s) Time(s)

symmetrical switching sequence design with constant

switching frequency. FCS-MPCC11-V3 FCS-MPCC11-V3-DRO

(c)
40 FCS-MPCC11
THD of Phase Current(%)

Fig. 15. Comparison of steady-state performance of two FCS-MPCC

30
methods at same switching frequency (2kHz) (a) Current waveform. (b)
20 FCS-MPCC11-V3 Current trajectories in x-y subspace. (c) Duty ratio.
10
C. Evaluation of Dynamic Performance for the Proposed
0
2 FCS-MPCC11-V3-DRO
Sam
plin 5
2 FCS-MPCC11-V3-DRO Method
g fr
equ 1.5
enc t (A)
y(k
Hz) 10 1.2 curren
ence
Refer
As aforementioned, the proposed FCS-MPCC11-V3-DRO
Fig. 14. Current THDi values in three FCS-MPCCs with respect to
different sampling frequency and reference current
method can both improve the steady-state current
performance in α-β and x-y subspaces. In order to verify the
B. Steady-State Performance of Different FCS-MPCC dynamic performance of the current control in the proposed
Methods (Same Switching Frequency) method, Fig. 16 shows the dynamic results for a step in the
reference current at 2kHz sampling frequency. It can be
To verify the effectiveness of the proposed scheme fairly, observed that the current in α-β subspace can always track
Fig. 15 shows a comparison of steady-state performance of the reference value precisely and achieve fast dynamic
FCS-MPCC11-V3 and the proposed FCS-MPCC11-V3-DRO response. Meanwhile, from step change process of duty
methods at same switching frequency. The approximate ratio, the current tracking ability for reference step-change
switching frequency in FCS-MPCC11-V3 is about 2.51kHz is very fast (settling time is approximately 0.04ms),
when the sampling/control frequency is set to 5kHz. Moreover, the harmonic currents in x-y subspace are always
Meanwhile, the switching frequency in the proposed kept at zero, thus sinusoidal phase current with low
FCS-MPCC11-V3-DRO scheme keeps at 2kHz even if the harmonic distortion is obtained.
sampling/control frequency is set to 2kHz. From Fig. 15, it
is clear that, the phase current waveform and the current
2ms/div 2ms/div
ripples on two bi-dimensional subspaces of the proposed
method under lower sampling and switching frequencies, is
ia
still superior to those in the adopted FCS-MPCC11-V3.
)

Current THDi in the proposed scheme is 9.23% lower than

ua
12.04% in FCS-MPCC11-V3. In addition, the equivalent duty
ratio of virtual voltage vector in the FCS-MPCC11-V3 is
always 1 while it is accurately estimated around 0.6 in the (a)
A

proposed method. The effectiveness of the proposed 2 R e f e r e

1
scheme is proved obviously from the other perspective. 1 0.8
Duty ratio γ

id id
iβ iy (

0 0.6
ia ia H a
0.4
r m 0.04ms o n
- 1
ua ua 0.2
- 2 D y 0 n a m
- 2 0- 1 0.01 0
0.02 1
iα ix( A ) Time(s)
3 3
FCS-MPCC11-V FCS-MPCC11-V -DRO
(b) (c)
(a)
Fig. 16. Dynamic performances for a step in the reference current (2kHz
2 2
Reference current Reference current sampling frequency) (a) Current waveform. (b) Current trajectories. (d)
1 1 Duty ratio.
iβ iy (A)

iβ iy (A)

0 0 D. Parameter Sensitivity Analysis of the Proposed

Harmonic current Harmonic current

-1 -1 FCS-MPCC11-V3-DRO Method
-2 -2
-2 -1 0 1 2 -2 -1 0 1 2 The above experimental tests are carried on with the
iα ix(A) iα ix(A)
specific control parameters, i.e., R=10Ω, L=4.5mH. The
FCS-MPCC11-V3 FCS-MPCC11-V3-DRO load resistance has very small effect on the prediction
(b) process. In fact, it can be neglected for the reason that the

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

sampling interval Ts is short enough. Therefore, the presents the same trends. It can be noticed that from Fig.19,

o
influence of inductance parameters variation on control the larger the absolute value of the inductance mismatch
performance should be mainly analyzed. First, the ratio δ is, the larger the current ripple and THD value are.
inductance mismatch ratio δ is defined as And the negative ratio δ can result higher current ripple and
Lm THD values than the positive ratio δ in case of the same
 =1- (15) absolute value. Fortunately, current can still track the
L
reference value in Fig.17 no matter how the inductance

i
Where Lm is estimating inductance value and L is actual changes. It has proved the strong robustness of the
inductance value. L remains constant during the proposed FCS-MPCC scheme with duty ratio optimization.
experiment. 10 0.14 0.1358

current in subspace
αβ subspace

THD of Phase

Ripple value of
0.12
Fig. 17 shows the steady-state performance of the

Current(%)
8 7.51%

proposed FCS-MPCC11-V3-DRO method at 10kHz sampling 6

6.33% 7.28% 0.1
0.0936
t

0.08
frequency, when the mismatch ratio δ is +50% and -50%, 4
5.50%
4.63% 0.06
0.0643
0.0545 0.0504
xy subspace
0.0601
respectively. And the corresponding duty ratio is shown in 2 0.04
0.0547 0.0521 0.0483 0.0560
+50% +20% 0% -20% -50% +50% +20% 0% -20% -50%
Fig.18. Compared with Fig. 13d, it can be observed that Mismatch ratio δ Mismatch ratio δ

there is no significant difference between these three sets of (a) (b)

results. It confirms that the proposed FCS-MPCC11-V3-DRO Fig. 19. (a)The variation curve of phase current THD. and (b) Current
method is robust against the variation of control parameters. ripples in α-β and x-y subspaces
a

In order to clearly see the influence of inductance variation

on the system, Fig. 19 shows the variation curve of phase E. Cost Time Evaluation of the Proposed
current THDi and ripples in α-β and x-y subspaces with FCS-MPCC11-V3-DRO Method
respect to inductance mismatch. It should be mentioned that, Fig. 20 shows the real-time implementation details of the
the average steady-state ripples of currents in α-β and x-y proposed FCS-MPCC11-V3-DRO scheme. It can be seen that
subspaces in Fig.13d are 0.0504 and 0.0483, respectively, the proposed FCS-MPCC11-V3-DRO needs a sampling time of
r

which are very close to those in Fig. 17a. However, the at least 40μs. It means that the proposed scheme can still be
current ripples become slightly greater when the inductance executed below 20kHz sampling frequency. The main time
mismatch ratio equals -20% and -50% . consuming is to iterate the predictive model and calculate
2
the minimum cost function value while the duty ratio
id Reference current
calculation only occupies a tiny part during the entire
1
execution period of the proposed scheme.
iβ iy (A)

ia 0
Harmonic current
ua -1 k 10 20 30 40 50 60 70 80 90 k+1

Time(μs)
-2 A/D Convertion Delay Compensation
-2 -1 0 1 2
iα ix(A)
Current Transformation Predictive Model and Cost Function Minimization
y

Duty Ratio Calculation Transmission Pulse

(a)
Other Process

2
id Reference current

1 Fig. 20. Real-time implementation details of proposed scheme during a

iβ iy (A)

ia sampling period.
0
Harmonic current
t

ua -1 F. A Comprehensive Comparison Performance of Three

-2
FCS-MPCC Schemes
-2 -1 0 1 2
iα ix(A)
Finally, a comprehensive comparison of the proposed
(b) FCS-MPCC scheme and two other methods is summarized
and listed in Table II. The proposed scheme can achieve the
Fig. 17. Inductance mismatch Effect on the load current (a) δ=+50%, (b)
u

best steady-state performance, and retain the fast transient

δ=-50%
performance as a potential feature in traditional
1 1
FCS-MPCC scheme. The proposed FCS-MPCC method is
0.8 0 .
robust
8
against the parameter mismatch. In addition, the
Duty ratio d

0.6 0 .
computational
6
consuming time for a whole control interval
0.4
in 4the proposed FCS-MPCC is about 40μs. Only ten virtual
D

0 .

0.2 0 .
voltage
2
vectors and zero vectors need to be evaluated in
0 0
every control interval. Moreover, it should be mentioned
0 0.1 0.2 0 0 0 . . 1 2
Time(s) T i that the maximum
m available ac evoltage of proposed
( scheme s

is 0.5527Vdc as shown in (9), which is smaller than that of

(a) (b)
conventional FCS-MPCC scheme.
Fig. 18. Inductance mismatch effect on duty ratio(a) δ=+50%, (b) δ=-50% TABLE II
Moreover, current THDi curve shown in Fig.19 also

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

EXPERIMENTAL COMPARISON OF THE PROPOSED CONTROL SCHEME AND [1] E. Levi, "Multiphase electric machines for variable-speed
applications," IEEE Trans. Ind. Electron., vol. 55, no. 5, pp.
OTHER SCHEMES 1893-1909, May 2008.
[2] E. Levi, "Advances in converter control and innovative exploitation
Control Schemes
FCS-MPCC11 FCS-MPCC21 FCS-MPCC11-V3 FCS-MPCC11-V3-DRO of additional degrees of freedom for multiphase machines," IEEE
Performances
Trans. Ind. Electron., vol. 63, no. 1, pp. 433-448, Jan. 2016.
Steady-state current General Good Good Excellent
[3] F. Barrero and M. J. Duran, "Recent Advances in the Design,
Dynamic performance Fast Fast Fast Fast Modeling, and Control of Multiphase Machines—Part I," IEEE
Trans. Ind. Electron., vol. 63, no. 1, pp. 449-458, Jan. 2016.
Cost effort Easy Heavy Easy Easy
[4] E. Jung, H. Yoo, S. K. Sul, H. S. Choi and Y. Y. Choi, "A
The maximum available
0.6472 Vdc 0.6472 Vdc 0.5527Vdc 0.5527Vdc nine-phase permanent-magnet motor drive system for an
output voltage
ultrahigh-speed Elevator," IEEE Trans. Ind. Appl., vol. 48, no. 3, pp.
987-995, May./June 2012.
[5] M. Darijevic, M. Jones and E. Levi, "An open-end winding
VI. CONCLUSION four-level five-phase drive," IEEE Trans. Ind. Electron., vol. 63, no.
1, pp. 538-549, Jan. 2016.
This paper proposes an integrated FCS-MPCC method [6] I. Gonzalez-Prieto, M. J. Duran, H. S. Che, E. Levi, M. Bermúdez
using virtual voltage vectors as control set with duty ratio and F. Barrero, "Fault-tolerant operation of six-phase energy
optimization for five-phase inverters. In order to reduce the conversion systems with parallel machine-side converters," IEEE
harmonic current and the calculation amount in the Trans. Power Electron., vol. 31, no. 4, pp. 3068-3079, April 2016.
predictive control algorithm, ten virtual voltage vectors are [7] L. de Lillo et al., "Multiphase power converter drive for
fault-tolerant machine development in aerospace applications," IEEE
re-defined in combination form of large and medium voltage Trans. Ind. Electron., vol. 57, no. 2, pp. 575-583, Feb. 2010.
vectors in the same direction in α-β subspace with constant [8] N. K. Nguyen, F. Meinguet, E. Semail and X. Kestelyn,
duration ratio 0.618-1. Thus, cost function can be simplified "Fault-tolerant operation of an open-end winding five-phase PMSM
without weighting factor in the proposed FCS-MPCC. drive with short-circuit inverter fault," IEEE Trans. Ind. Electron.,
Actually, if the cost function does not contain the x-y current vol. 63, no. 1, pp. 595-605, Jan. 2016.
[9] S. Williamson and S. Smith, "Pulsating torque and losses in
items, the controller could be less effective in dealing with multiphase induction machines," IEEE Trans. Ind. Appl., vol. 39, no.
non-ideal effect, mainly caused by the dead-time effect. 4, pp. 986-993, July./Aug. 2003.
However, the effect is not quite significant. Therefore, in [10] L. Parsa, H. A. Toliyat and A. Goodarzi, "Five-phase interior
practical applications, the x-y current items can be added into permanent-magnet motors with low torque pulsation," IEEE Trans.
cost function depends on the property of the controller. Ind. Appl., vol. 43, no. 1, pp. 40-46, Jan./Feb. 2007.
[11] H. A. Toliyat, T. A. Lipo and J. C. White, "Analysis of a
From a comprehensive experimental comparison of concentrated winding induction machine for adjustable speed drive
traditional FCS-MPCC, FCS-MPCC11-V3 and the proposed applications. II. Motor design and performance," IEEE Trans.
FCS-MPCC11-V3-DRO methods, the salient features of the Energy Convers., vol. 6, no. 4, pp. 684-692, Dec 1991.
proposed FCS-MPCC can be summarized as: [12] R. O. C. Lyra and T. A. Lipo, "Torque density improvement in a
six-phase induction motor with third harmonic current injection,"
1) The steady-state phase current waveform and the IEEE Trans. Ind. Appl., vol. 38, no. 5, pp. 1351-1360, Sep./Oct 2002.
current ripples on two bi-dimensional subspaces of proposed [13] S. Kouro, P. Cortes, R. Vargas, U. Ammann and J. Rodriguez,
method are superior to the other FCS-MPCC schemes. "Model predictive control—a simple and powerful method to control
2) The proposed scheme can realize constant switching power converters," IEEE Trans. Ind. Electron., vol. 56, no. 6, pp.
1826-1838, June 2009.
frequency.
[14] J. Rodriguez et al., "Predictive Current Control of a Voltage Source
3) The proposed scheme can retain the simplified Inverter," IEEE Trans. Ind. Electron., vol. 54, no. 1, pp. 495-503,
structure and fast dynamic performance of traditional Feb. 2007.
FSC-MPCC, which is easy to implement. [15] A. Iqbal, H. Abu-Rub, P. Cortes and J. Rodriguez, "Finite control set
model predictive current control of a five-phase voltage source
4) The proposed FCS-MPCC method is robust against the
inverter," in Proc. IEEE ICIT, Viña del Mar, Chile, 2010, pp.
inductance mismatch. 1787-1792.
5) The computational burden in the proposed FCS-MPCC [16] P. Cortés, L. Vattuone, J. Rodriguez and M. Duran, "A method of
is about 40μs. The algorithm can be still executed under predictive current control with reduced number of calculations for
20kHz sampling frequency. five-phase voltage source inverters," in Proc. IEEE IECON, Porto,
Portugal, 2009, pp. 53-58.
6) The proposed method takes the discrete feature of [17] W. Xie et al., "Finite-control-set model predictive torque control
inverter into account with no modulation stage, avoided the with a deadbeat solution for PMSM drives," IEEE Trans. Ind.
complex computational process of desired voltage in Electron., vol. 62, no. 9, pp. 5402-5410, Sept. 2015.
generalized predictive control with PWM stage at the same [18] C. S. Lim, N. A. Rahim, W. P. Hew, M. Jones and E. Levi, "Model
predictive current control of a five-phase induction motor," in Proc.
time.
IEEE IECON, Melbourne, Vic, Australia, 2011, pp. 1934–1940.
In this paper, a five-phase inverter is used as a practical [19] C. S. Lim, E. Levi, M. Jones, N. Abdul Rahim and W. P. Hew,
example. Moreover, the proposed algorithm with duty ratio "Experimental evaluation of model predictive current control of a
optimization can be extended into n-phase inverter five-phase induction motor using all switching states," in Proc. Int.
application. Conf. EPE-PEMC, Novi Sad, Serbia,, 2012, pp. LS1c.4-1-LS1c.4-7.
[20] P. Cortes, A. Wilson, S. Kouro, J. Rodriguez and H. Abu-Rub,
"Model Predictive Control of Multilevel Cascaded H-Bridge
Inverters," IEEE Trans. Ind. Electron., vol. 57, no. 8, pp. 2691-2699,
REFERENCES Aug. 2010.
[21] C. S. Lim, N. A. Rahim, W. P. Hew and E. Levi, "Model predictive
control of a two-motor drive with five-Leg-inverter supply," IEEE
Trans. Ind. Electron., vol. 60, no. 1, pp. 54-65, Jan. 2013.

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2017.2707440, IEEE
Transactions on Power Electronics

> TPEL-Reg-2016-01-0006 _Final_Version<

[22] M. J. Duran, J. Prieto, F. Barrero, and S. Toral, “Predictive current Converters with Modulation Function Optimization. IEEE Trans.
control of dual three-phase drives using restrained search Power Electron., vol. 31, no. 7, pp. 5279 – 5289, Jul. 2016.
techniques,” IEEE Trans. Ind. Electron., vol. 58, no. 8, pp. [29] L. Zheng, J. E. Fletcher, B. W. Williams and X. He, "A novel direct
3253–3263, Aug. 2011. torque control scheme for a sensorless five-phase induction motor
[23] Y. Zhang and H. Yang, "Model predictive torque control of drive," IEEE Trans. Ind. Electron., vol. 58, no. 2, pp. 503-513, Feb.
induction motor drives with optimal duty cycle control," IEEE Trans. 2011.
Power Electron., vol. 29, no. 12, pp. 6593-6603, Dec. 2014. [30] P. Cortes, J. Rodriguez, C. Silva and A. Flores, "Delay compensation
[24] S. A. Davari, D. A. Khaburi and R. Kennel, "An improved in model predictive current control of a three-phase inverter," IEEE
FCS–MPC algorithm for an induction motor with an imposed Trans. Ind. Electron., vol. 59, no. 2, pp. 1323-1325, Feb. 2012.
optimized weighting factor," IEEE Trans. Power Electron., vol. 27, [31] Y. Ren and Z. Q. Zhu, "Reduction of both harmonic current and
no. 3, pp. 1540-1551, March 2012. torque ripple for dual three-phase permanent-magnet synchronous
[25] C. S. Lim, E. Levi, M. Jones, N. A. Rahim and W. P. Hew, "A machine using modified switching-table-based direct torque control,"
comparative study of synchronous current control schemes based on IEEE Trans. Ind. Electron., vol. 62, no. 11, pp. 6671-6683, Nov.
FCS-MPC and PI-PWM for a two-motor three-phase drive," IEEE 2015.
Trans. Ind. Electron., vol. 61, no. 8, pp. 3867-3878, Aug. 2014.
[26] H.-M. Ryu, J.-H. Kim and S.-K. Sul, "Analysis of multiphase space
vector pulse-width modulation based on multiple d-q spaces
concept," IEEE Trans. Power Electron., vol. 20, no. 6, pp.
1364-1371, Nov. 2005.
[27] C. S. Lim, E. Levi, M. Jones, N. A. Rahim and W. P. Hew,
"FCS-MPC-Based current control of a five-phase induction motor
and its comparison with PI-PWM control," IEEE Trans. Ind.
Electron., vol. 61, no. 1, pp. 149-163, Jan. 2014.
[28] W. Song, Z. Deng, S. Wang, and X. Feng. A Simple Model
Predictive Power Control Strategy for Single-phase PWM

Xuesong Wu was born in Bengbu, Anhui, China, in

Cheng Xue was born in Bengbu, Anhui, China, in 1994, 1994, received the B.S. degree in electrical engineering
received the B.S. degree in electrical engineering from from Southwest Jiaotong University, Chengdu, China,
Southwest Jiaotong University, Chengdu, China, in 2015, in 2016, where he is currently working toward his M.S
where he is currently working toward his M.S degree in degree in electrical engineering.
electrical engineering. His current research interest is power electronics
and motor drives.
His current research interests include multiphase
machine control and predictive control.

Wensheng Song (M’13) received the B.S. degree in

electronic and information engineering, the Ph.D. degree Xiaoyun Feng received the B.S.,M.S., and Ph.D.
in electrical engineering from Southwest Jiaotong degrees in electrical engineering from Southwest
University, Chengdu, China, in 2006 and 2011, Jiaotong University, Chengdu, China, in 1983,1988,
respectively, where he is currently an associate professor and 2001.
in the School of Electrical Engineering. Since 1983, she has been with the school of
From Sep. 2009 to Sep. 2010, he was a visiting electrical engineering, Southwest Jiaotong University,
scholar with the department of electrical engineering and where she is currently a full professor. She was a
computer science, University of California, Irvine, USA. From Jul. 2015 to visiting professor at the University of Tokyo, between
Dec. 2015, he was a visiting scholar at University of Alberta, Edmonton, October 1998 and October 1999. She is the author or coauthor of more
Canada. His current research interests include digital control and than 70 papers.
modulation methods of electrical AC-DC-AC railway traction drive Her major research interests include the analysis and control of
systems, and multilevel converters. He is the corresponding author of this electrical traction converter and motor drive system, the design of railway
paper. traction characteristics and train optimizing operation.

0885-8993 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.