You are on page 1of 33

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

net/publication/352361482

Effective predictive current control for a sensorless five-phase induction


motor drive

Article  in  International Journal of Power Electronics · January 2021


DOI: 10.1504/IJPELEC.2021.115583

CITATION READS

1 86

1 author:

Mahmoud Mossa
Minia University
98 PUBLICATIONS   576 CITATIONS   

SEE PROFILE

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

Model Predictive Control for AC Machine Drives View project

Multi-phase machine drives: Analysis and control View project

All content following this page was uploaded by Mahmoud Mossa on 28 February 2022.

The user has requested enhancement of the downloaded file.


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: 10.1504/IJPELEC.2021.10020183

Int. J. Power Electronics, Vol. X, No. Y, xxxx 1

Effective predictive current control for a sensorless


five-phase induction motor drive

Mahmoud A. Mossa
Electrical Engineering Department,
Faculty of Engineering,
Minia University, Egypt
Email: mahmoud_a_mossa@mu.edu.eg

Abstract: The paper aims to develop an effective predictive current control


(PCC) procedure for a sensorless five-phase induction motor (IM) drive. The
proposed PCC considers the discrete behaviour of the voltage source inverter
(VSI) and provides the voltage vectors utilising the finite control set (FCS)
principle, thus, there is no need for using the pulse width modulation (PWM)
techniques. A combined sliding mode-Luenberger observer is used to estimate
the motor speed, stator current, rotor flux, load torque, and stator and rotor
resistances. An effective pole placement procedure is proposed for selecting the
most appropriate gains of the observer to enhance the estimation process and to
improve the observer’s robustness against uncertainties in the system. The
drive performance has been tested for a wide range of speed changes. The
obtained results demonstrate the effectiveness of proposed PCC in obtaining
high dynamic performance from the drive. Moreover, the proposed sensorless
procedure has confirmed its validity to estimate precisely the required control
variables.

Keywords: five-phase IM drive; predictive control; current control; sensorless


control; sliding mode observer; Luenberger observer; pole placement.

Reference to this paper should be made as follows: Mossa, M.A. (xxxx)


‘Effective predictive current control for a sensorless five-phase induction motor
drive’, Int. J. Power Electronics, Vol. X, No. Y, pp.xxx–xxx.

Biographical notes: Mahmoud A. Mossa graduated and received his


Bachelor’s and Master’s in Electrical Engineering from the Faculty of
Engineering, Minia University – Egypt in 2008 and 2013, respectively. Since
January 2010, he was working as an Assistant Lecturer at the Electrical
Engineering Department at the same university. In November 2014, he joined
the Electric Drives Laboratory (EDLAB) at the University of Padova in Italy
for his PhD research activities. In April 2018, he was awarded the PhD in
Electrical Engineering (control of machine drives). Since May 2018, he is
working as an Assistant Professor at the Electrical Engineering Department at
Minia University in Egypt. His research interests are focused on developing
effective control techniques for different topologies of induction machine
drives based on model predictive control approaches.

Copyright © 20XX Inderscience Enterprises Ltd.


2 M.A. Mossa

1 Introduction

The ‘multi-phase’ expression used for the electrical machine drives denotes to a number
of phases greater than three. This type of machine drives exhibits several merits
compared with the three phase machine type. Some of these merits can be addressed
through the less torque ripples, the high torque density, the high fault tolerant capability,
and the reduction in the power rating of the inverter switches (Zhang et al., 2008; Heydari
et al., 2012; Rahman et al., 2017).
Nowadays, multi-phase machine drives are used extensively for several industrial
applications. The reasons for the preference of these machine types in such applications
are several; for example, when they are used in the applications which require large
power management (i.e., naval propulsion), the rating of each inverter phase is reduced,
and consequently the handled per-phase power is effectively reduced (Zhang et al., 2008;
Guzman et al., 2015; Bermudez et al., 2017). Another significant merit of a multi-phase
drive can be synthesised through the fault-tolerant capability, which is a vital requirement
in certain applications like more electric aircraft; as in this case the dynamics of the drive
will not be seriously affected by the loss of one or two phases (Abolhassani, 2005; Duran
et al., 2018; Huang et al., 2018). With the multi-phase drives, the rated value of the
current for the semiconductor switches is reduced when used in the electric vehicles and
the propulsions applications (Sadehgi and Parsa, 2010; Villani et al., 2010; Yan
et al., 2016), which as a result reduces the cost of required inverter.
The variable speed five-phase induction motor (IM) drive has been presented for the
first time at the year 1969 (Ward and Harer, 1969). A square wave mode is utilised to
operate the five-phase inverter of such drive, and then the pulse width modulation
(PWM) mode of operation was adopted. These PWM techniques are considered as
extension to the PWM technique used by the three-phase inverters (Prieto et al., 2011;
Payami et al., 2015; Xu et al., 2009). After that, the space vector PWM (SVPWM)
techniques were presented and adopted in multi-phase voltage source inverters (VSIs)
(Ryu et al., 2005; de Silva et al., 2004; Dujic et al., 2009; Dordevic et al., 2013; Iqbal and
Levi, 2006; Duran et al., 2007).
Up to the present time, the control of five-phase induction machine drives using
power converters has been based on the principle of mean value, using PWM and
SVPWM (Ojo et al., 2006; Hara et al., 2011; Darijevic et al., 2017). Using multiple PI
controllers besides the PWM increases the complexity of the control system which results
in increasing the computational time burdens and thus negatively affects the
microprocessor implementation time. Recent studies have showed that it is possible to
use the predictive control to control the multi-phase induction machine drives without
using modulators and linear controllers (i.e., PI controllers) (Cao et al., 2018; Cheng
et al., 2016; Riveros et al., 2010; Khan and Iqbal, 2009; Mossa, 2018). This new
approach will have a strong impact on the control of power electronics in the coming
decades. The advantages of predictive control are noticed through its ability to consider a
multi-objective case within the model, easy inclusion of nonlinearities within the model,
simple treatment of the system constraints, easy of digital implementation, and flexibility
of including modifications and extension of the control horizons according to the required
applications. Upon this, this paper concerns with developing an effective control
procedure for the five-phase IM drive based on the finite control set model predictive
control (FCS-MPC) principle, which actuates directly the switch states of the VSI and
thus, eliminates the usage of PWM modulators and PI controllers. A new formulation of
Effective predictive current control 3

the cost function is presented which is different from the classic form of quadratic type
(Khan and Iqbal, 2009).
The research point of sensorless control of multi-phase motor drive are still not
widely discussed, however few number of studies handled the sensorless control of this
type of machine drives as an extension to the sensorless procedures used for the
three-phase machine drives. One of these studies proposed the extended Kalman filter as
a solution for the speed estimation of series connected five phases two motor drive
system as presented in (Khan and Iqbal, 2009). Another sensorless approach is the
MRAS-based sensorless control of a vector controlled five-phase IM drive presented in
Abu-Rub et al. (2010). A sliding mode observer for the sensorless control of a five-phase
IM has been proposed by Kong et al. (2013), which utilised a sigmoid function instead of
the sign function to reduce the chattering effect. The sensorless operation of five-phase
IM drive is still attracting more ideas through which the shortages in the previous
techniques are avoided.
The contribution of the current paper can be addressed through utilising a new
formulation of the cost function that is used by the predictive controller compared with
the classic form which is based on minimising the absolute values of the torque and stator
flux errors. Moreover, the paper proposes a combination between the sliding mode
observer and an effective Luenberger observer for the sensorless operation. The gains of
the Luenberger observer are designed in an effective way so that the observer’s stability
is guaranteed for a wide range of speed operation. The outputs from the observer are the
stator current, the rotor flux, the rotational speed, the load torque, and the stator and rotor
resistances. The stator currents are estimated to limit the noise contents in the predicted
values of the currents which are present if the currents are used directly from the
measurements; this led to improving the predicted signals of the currents, fluxes and the
developed torque as well. In addition, the load torque is estimated to enhance the
estimation of the speed. As the stator and rotor resistances are exist in the derivatives of
the stator current and stator flux components as well, so they are estimated to prevent the
effect of any outer disturbance which can affect their values especially at the low-speed
operation. The feasibility of the sensorless procedure is achieved through the obtained
testing results, which shows high and precise transient dynamic performance of the
five-phase IM drive.
The paper is organised as following: firstly, the model of the five-phase IM is
introduced, then the proposed predictive current control (PCC) approach is explained in
details; after that the proposed Luenberger-sliding mode observer is described in a
systematic manner through which the observer’s gains selection criterion is deeply
investigated. The complete system configuration of the control system is then introduced
and explained; finally, the test results are introduced which confirmed the effectiveness
of the proposed sensorless estimation procedure during the high-speed and low-speed
operations of the five phase IM drive.

2 Modelling of five-phase IM

The equivalent model of the machine can be represented in a common reference frame,
which rotates with an arbitrary angular speed a,k as illustrated in Figure 1. me,k is the
electrical angular rotor position and it equals ( me,k = P m,k), where m,k is the mechanical
4 M.A. Mossa

angular position of the rotor. In addition, s,k refers to the angular position of the d-axis of
the common reference frame with respect to the phase ‘a’ magnetic axis of the stator.

Figure 1 D-Q (park) transformation of five-phase induction machine model (see online version
for colours)

a s
me

Va
s

b i ar V se e
i br

V sb
i cr i er

id
r
V d
s
V sc

c d q

The transformation from (abcde) to (dq) coordinates is performed as following:


s s
Vdq A s Vabcde , isdq A sisabcde , and s
dq As s
abcde (1)

r r r r r r
Vdq A r Vabcde , i dq A r i abcde , and dq Ar abcde (2)

cos s,k cos s,k cos s,k 2 cos s,k 2 cos s,k

sin s,k sin s,k sin s,k 2 sin s,k 2 sin s,k
2 1 cos2 cos4 cos4 cos2
As * (3)
5 0 sin2 sin4 sin4 sin2
1 1 1 1 1
2 2 2 2 2
In the same way, the rotor variables are transformed utilising the same transformation
procedure, except that s,k is replaced by k, where k = s,k me,k. Where k is the
instantaneous angular position of the (d-axis) of the common reference frame with
respect to the phase ‘a’ magnetic axis of the rotor.
Thus, the transformation matrix for the rotor will be as following:
Effective predictive current control 5

cos k cos k cos k 2 cos k 2 cos k

sin k sin k sin k 2 sin k 2 sin k

2 1 cos2 cos4 cos4 cos2


Ar * (4)
5 0 sin 2 sin 4 sin 4 sin 2
1 1 1 1 1
2 2 2 2 2
Then, from equations (1)–(4), the mathematical model of the five-phase induction
machine can be expressed in a discrete form in the d-q reference frame through the
following relations which outline the voltages balance and the fluxes-currents
relationships as:
d ds,k
Vds,k R s ids,k a,k qs,k (5a)
dt
d qs,k
Vqs,k R s iqs,k a,k ds,k (5b)
dt
d xs,k
Vxs,k R s i xs,k (5c)
dt
d ys,k
Vys,k R s i ys,k (5d)
dt
d 0s,k
V0s,k R s i0s,k (5e)
dt
d rs,k
Vdr,k 0 R r i dr,k a,k me,k qr,k (5f)
dt
d qr,k
Vqr,k 0 R r i qr,k a,k me,k dr,k (5g)
dt
d xr,k
Vxr,k 0 R r i xr,k (5h)
dt
d yr,k
Vyr,k 0 R r i yr,k (5i)
dt
d 0r,k
V0r,k 0 R r i 0r,k (5j)
dt
The flux linkage equations in the d-q reference frame are represented by:

ds,k Lls Lm i ds,k L mi dr,k (6a)

qs,k Lls Lm i qs,k L mi qr,k (6b)

xs,k L lsi xs,k (6c)


6 M.A. Mossa

ys,k Lls i ys,k (6d)

0s,k L ls i 0s,k (6e)

dr,k Llr L m idr,k L m ids,k (6f)

qr,k Llr L m iqr,k L m iqs,k (6g)

xr,k Llr i xr,k (6h)

yr,k L lr i yr,k (6i)

0r,k L lr i 0r,k (6j)

where Rs and Rr are the per phase stator and rotor resistances, while Lm, Lls, and Llr
denote to the magnetising inductance, stator leakage-inductance and rotor leakage-
inductance, respectively.
The developed torque of the motor can be expressed by:
Lm
mk P dr,k i qs,k qr,k i ds,k (7)
Lr

where Lr refers to the rotor self-inductance, and P is the pole pairs.


The torque-speed relationship can be represented by:
J d me,k
mk m l,k (8)
P dt
where ml,k is the load torque and J is the moment of inertia. It is obvious from the
previous equations that the five-phase IM model is similar to that of three-phase IM type,
except that the five-phase IM model has extra (x-y) set of component which are not
torque producing components, but they are involved in the total machine losses.

3 Proposed PCC approach

To develop the proposed predictive current controller for the machine, the model of the
machine is presented in a discrete form as stated previously, thus all variables are
discretised and sampled at different sampling instants (i.e., kTs, (k + 1)Ts). Where k
denotes to the sampling instant and Ts denotes to the sampling time.
The control aims to maintain at each sampling instant the predicted values of the
stator current components at instant (k + 1)Ts [ids,k+1 (flux component), iqs,k+1 (torque
component)] and the two other components [ixs,k and iys,k] as close as possible to their
relative reference values. To realise this, the stator currents are estimated using an
improved Luenberger observer to limit the distortion in the current values that is present
in case that the currents being directly measured. The Luenberger observer is combined
with a sliding mode observer for estimating the speed, the angular position, and the stator
and rotor resistances. The resistances are estimated as they are involved in the calculation
of the stator current derivatives which are used for the prediction and for the speed
estimation as well. So to avoid any mismatch in the resistances values under certain
Effective predictive current control 7

operating conditions (specifically at low speed operation), then the proposed observer
estimate their values as well, this contributes effectively in enhancing the estimation
precision of the observer. The reference values of the stator current components are
obtained via applying the rotor field orientation principle, in which the rotor flux is
aligned with the d-axis of the rotating synchronous frame, and thus the following
expressions are derived:
dr,k r,k , and qr,k 0.0 (9)

Lm
mk P r,k i qs,k (10)
Lr
Assuming that the change in the rotor flux during steady state operation is very small and
thus can be neglected, then from the rotor voltage equation defined in the d-q reference
frame, it results:
d dr,k
0.0 R r i dr,k 0.0 (11)
dt
Thereafter, from equations (6f), (6g), (9), (10) and (11), the reference values of stator
current components are calculated by
*
r,k
i*ds,k (12)
Lm

1 L r m*k
i*qs,k (13)
P L m *r,k

While the references of the other two components i*xs,k , i*ys,k are set to zero.
The reference values in equations (12) and (13) are then transformed to the ( - )
quantities i* s,k , i* s,k to be used by the control system.
The error vector can be then evaluated in terms of the actual and reference values of
the stator current components at instant kTs as:
i* s,k i s,k i* s,k i s,k 0.0 i xs,k 0.0 i ys,k
ek j j
Isn Isn Isn Isn (14)
ei s,k jei s,k ei xs,k jei ys,k

where Isn is the rated value of stator current.


From equation (14), it can be realised that via selecting the appropriate voltage
vectors, the error tends to be zero. The selection of the optimal vectors is mainly based on
a criterion which relies on achieving a negative variation in the error direction at each
instant happened that its value exceeds the permissible limit by the controller.
According to that, the control has to achieve the following condition:
2 2 2 2
ek ei s,k ei s,k ei xs,k ei ys,k E max (15)

where Emax is a permissible error limit selected for reducing the switching frequency.
8 M.A. Mossa

From equation (15), it can be deduced that when the actual error amplitude | k|
oversteps the value Emax, then the control starts searching between the eight possible
voltage vectors and selects the one which causes a predicted negative variation in the
value of the error, and then applies it to the motor in the next control cycle.
Then, by differentiating equation (15), it results
2 2 2 2
d ek ei s,k ei s,k ei xs,k ei ys,k
0.0 (16)
dt dt
Equation (16) can be represented by
dei s,k dei s,k deixs,k deiys,k
ei ei s,k ei xs,k ei ys,k
d ek s,k
dt dt dt dt 0.0 (17)
dt ek

From equation (17), and based on the hypothesis which explains the way by which the
optimal voltage vectors are selected, the cost function that will be used by the control
system can be expressed in one of the following two forms:
dei s dei s de ixs dei ys
k ei s,k ei s,k ei xs,k ei ys,k 0 (18)
dt k dt k dt k dt k

di s di s di xs di xs
k ei s,k ei s,k ei xs,k ei ys,k 0 (19)
dt k dt k dt k dt k

Here, the second formulation is used. The derivatives of stator current components are
obtained from the equivalent mathematical model of the motor in the stationary reference
frame as:
di s 1 Rr Lm
V s,k Ri s,k s,k me,k s,k Lt i s,k (20)
dt k Lt Lr Lr

di s 1 Rr Lm
V s,k Ri s,k s,k me,k s,k Lti s,k (21)
dt k Lt Lr Lr

L2m R r Ls
where L t Ls , and R Rs .
Lr Lr

di xs 1
Vxs,k R s i xs,k (22)
dt k Lls

di ys 1
Vys,k R s i ys,k (23)
dt k Lls
All terms of the cost function given by equation (19) are now obtainable and given as
functions of the applied voltage vectors. Then the control can predict the value of
equation (19) for one prediction horizon (k + 1)Ts, and select the voltage vector which
minimises this value and apply it to the motor terminals.
Effective predictive current control 9

4 Implementation procedure of the proposed PCC approach

The sequence of operation for the proposed PCC procedure can be separated into two
phases.

4.1 Stator currents prediction phase


The values of the stator current components are evaluated at instant (k + 1)Ts using
equations (20), (21), (22) and (23) as:
di s
 s,k 1 i s,k Ts (24)
dt k

di s
 s,k 1 i s,k Ts (25)
dt k

di xs
xs,k 1 i xs,k Ts (26)
dt k

di ys
ys,k 1 i ys,k Ts (27)
dt k

Consequently, the current errors are computed by equation (14) and condition of
equation (15) can be then checked.

4.2 Voltage vectors selection phase


The selection of the optimal voltage vectors is carried out through checking the values of
equation (19). The control predicts the value of equation (19) at instant (k + 1)Ts using
the predicted values of the stator current errors and their derivatives as follows:

i di s di s
k 1 ei s,k 1 ei s,k 1
dt k 1 dt k 1
i
(28)
di xs di ys
eixs,k 1 eiys,k 1
dt k 1 dt k 1

The derivatives are given by equations (20), (21), (22) and (23). The stator flux
components are obtained at instant (k + 1)Ts through applying the same procedure to
equations (24) to (27) starting from equations (5a) to (5d).
Then at any instant at which the predicted value of the absolute error |ek+1| oversteps
the pre-determined value Emax, then the control will search for the voltage vector that
minimises the value of equation (19), and apply it to the stator terminals in the next cycle.
The proposed PCC enables to investigate the speed effects on overall dynamic
performance during the different implementation stages (the prediction and the voltage
selection stages), and this can be an addition which was not presented for PCC for the
five phase IM before. To explain clearly, the speed role and estimation suitability are
discussed in each stage in the following sections.
10 M.A. Mossa

4.2.1 Speed influence during the prediction phase


From equations (20) and (21), it is noticed that the derivatives of the stator current
components include a part dependent on the speed and consequently equations (24) and
(25) need the speed to perform the prediction, while the derivatives of the stator current
components (x-y) do not contain any variable related to the speed as given by equations
(22) and (23). The prediction step goal is to decide whether the predicted error value
exceed the error limit or not. Accuracy of the prediction is important but not crucial and
therefore the speed estimation and prediction by the previous proposed method is
appropriated.

4.2.2 Speed influence during the voltage selection phase


According to the proposed PCC procedure, the selection of the optimum voltage vector is
based on minimising the cost function defined by equation (28), which can be
reformulated after separating the parts in the current derivatives that only depend on the
speed (apex( )) from the other one which are free of the speed sample but depend on the
applied voltages (apex(u)) as:
u u
i di s di s di s di s
k 1 ei s,k 1
ei s,k 1
dt dt k 1 dt dt k 1
(29)
u u i
di xs di ys
ei xs,k 1 ei ys,k 1 0.0
dt dt

where
u
di s 1 Rr
u s,k 1 Ri s,k 1 s,k 1 (30)
dt k 1 Lt Lr

di s 1 Lm
me,k 1 s,k 1 Lt i s,k 1 (31)
dt k 1 L t Lr
u
di s 1 Rr
u s,k 1 Ri s,k 1 s,k 1 (32)
dt k 1 Lt Lr

di s 1 Lm
me,k 1 s,k 1 Lt i s,k 1 (33)
dt k 1 Lt Lr

After the substitution of these terms in the cost function, the form of equation (28) will
tend to be as:
u u u u
di s di s di xs di ys
ei s,k 1
ei s,k 1 eixs,k 1 ei ys,k 1
dt k 1 dt k 1 dt dt
(34)
di s di s
ei s,k 1
ei s,k 1
dt k 1 dt k 1
Effective predictive current control 11

From equation (34), it can be concluded that this condition can be achieved using more
than one voltage vector, but the optimum voltage vector that will be selected is that one
which minimises equation (28) or maximise the left side of equation (34). Maximisation
of the left side does not depend on the speed but only on the applied voltage vector at this
instant, and thus the speed has no effect on the voltage selection procedure.

5 Proposed Luenberger-sliding mode observer

The operation methodology of the observer can be described through the equations which
are obtained from the motor model expressed in the stationary reference frame as
follows:
dî s Lm 1 1
Rî s,k j me , k ˆ r,k Vs,k G s,k î s,k (35)
dt k Ls Lr Tr Ls

d r Lm 1
î s,k j ˆ me,k ˆ r,k (36)
dt k Tr Tr

where îs,k, ˆ r,k are the estimated values of the stator current and the rotor flux
Lr
respectively, while G is the matrix of the observer gains, and Tr is the rotor time
Rr
L2m
constant, and 1 is the leakage coefficient.
Ls L r

5.1 Design of gains


The observer’s gains are selected using the traditional way via adjusting the relationship
between the observer’s poles and the IM poles to be proportionally varying, but this can
result in poles with a large imaginary part, which may negatively affect the stability of
the observer specifically at the high-speed operation. Moreover, the observer’s gains
contain speed dependent terms which can negatively affect the estimation process
especially at the low-speed operation as reported in (Kubota et al., 1993), in which the
gains matrix is defined by
T
g1 g2 g3 g4
G (37)
g2 g1 g4 g3

The eigenvalues as calculated by Kubota et al. (1993) are expressed by


(k 1)
g1 R s Lr R r Ls (38)
Ls L r L2m

g2 (k 1) ˆ me (39)

R r Ls kR s L r
g3 (k 1) (40)
Lm
12 M.A. Mossa

(k 1) Ls Lr L2m ˆ me
g3 (41)
Lm
As can be noticed from equations (39) and (41), the observer gains are defined in terms
of the machine parameters and the speed, which influences the stability of the observer.
For example; at high speed, with large imaginary part, the observer poles overlap the
stability limit. In addition, at high speed with large imaginary part of the poles, the speed
increases and will be corrupted with high level of noise content, and to avoid that, the
maximum value of the gains must be limited, and thus the maximum permissible speed is
restricted and this action is not preferred in numerous industry applications.
At low-speed operation, the poles which are relating to the poles which are near from
the imaginary axis of the complex s-plane, will move slowly resulting in a stiff observer
response.
Thus, to avoid these issues, an effective pole placement method has been adopted
here. The proposed pole placement procedure gives better performance through aligning
the observer’s poles to the left of the motor’s poles (to get a fast exponentially decreasing
response), with small change in the imaginary part of the poles. The gains are calculated
according to the following criteria:
The observer is a closed-loop system, which can be described in a state space form
through the following equations:
dî s
A11 G î s,k A12 ˆ r,k B1vs,k G s,k (42)
dt k

dˆ r
A 21î s,k A 22 ˆ r,k (43)
dt k

where
Rs (1 )
A11 *I a r11 * I (44)
Ls Tr

Lm 1
A12 *I me * a r12 * I a i12 * (45)
Ls L r Tr

Lm
A 21 *I a r21 * I (46)
Tr

1
A 22 *I me * a r22 * I a i22 * (47)
Tr
And
1 0 0 1
I= , (48)
0 1 1 0

From equation (48), the observer gains matrix can be expressed by


Effective predictive current control 13

a r11 g1 g2
A11 G (49)
g2 a r11 g1

Equation (49) has the following characteristic equations:


2
s2 2 a r11 g1 s a r11 g1 g 22 0.0 (50)

s2 2 ns
2
n 0.0 (51)
From equations (50) and (51), the natural frequency n and damping coefficient are
calculated as follows:

2 a r11 g1
n g 22 a r11 g1 , and (52)
2
g 22 a r11 g1

From equation (52), it can be realised that the observer gains have a clear effect on the
observer response. So that a criterion for selecting the most appropriate gains for the
observer has to be adopted in such a way that the system dynamic response becomes
stable and insensitive to the motor speed.
From equation (50), the observer poles are computed as follows:
s1 a r11 g1 jg2, and s 2 a r11 g1 jg2 (53)

The criterion based on which the eigenvalues of the proposed observer are selected can
be described through the following items:
Eigenvalues should be with negative real parts to ensure the system stability; and
thus (g1) should be negative.
Their position is on the left of the complex (s-plane) compared to the eigenvalues of
the observed system; this is to ensure a fast convergence for the observer’s state.
Selecting of g1 and g2 so that |g2 / (ar11 + g1)| is small, this is to get high damping
response from the observer [see equation (52)].
To limit the accompanied noise in the estimated variables, the observer gains are
selected with small values.
The observer poles should not be placed far from the left of the imaginary (j ) axis,
because if this happened, the elements of the observer gains matrix become large
which results in increasing the accompanied noise in the estimated signals.
The previous addressed items for the gains selection criterion are derived from the
analysis of the poles and the natural response of the control system described by a
transfer function which represents the differential equation that describes and outlines the
system behaviour. The poles and zeros of the transfer function outline the complete
system response. It is known that the natural response of a linear single input single
output (SISO) system to a given initial condition can be defined by the following
expression
j
y n (t) Ci e pi t (54)
i 1
14 M.A. Mossa

where the subscript n denotes to the natural response. From equation (54), poles location
in the complex (s-plane) defines the j (imaginary) components in the natural response of
the system as shown in Figure 2, which illustrates the system poles locations on the
complex (s-plane) for the natural response of the system.

Figure 2 System pole positions on the pole-zero plot (see online version for colours)

From Figure 2, the idea behind adopting the gains selection criterion can be easily
understood, this can be described through the following items:
1 For a real pole that is located in the left-half of the s-plane (pi = ), this gives an
exponentially decaying signal (yn(t) = Ce t). The pole location in the s-plane
determines the rate of the decay; for example, poles which are located far from the
origin in the left-half of the s-plane decay rapidly, while that one’s which are located
near the zero are decaying slowly.
2 For a real pole that is located in the right-half of the (s-plane) (pi = ), this gives an
exponentially increasing signal (yn(t) = Ce t), which means that the system will be
unstable.
3 If the pole is located at the origin (pi = 0), this defines a component with a fixed
magnitude and determined by the initial conditions.
4 For a complex conjugate pole pair (pi = ± j ) that is located in the left-half of the
s-plane, this gives a decaying sinusoid signal of the form of [Ae t sin( t + )], this
can be shown in Figure 2 and takes the blue colour. The values A and are
determined using the initial conditions, and is the oscillation frequency of the
sinusoid. In case that the pole is located in the right-half of the s-plane, this will give
an increasing sinusoid signal of the form of [Ae t sin( t + )], this can be shown in
Figure 2 and takes the red colour.
After describing the criterion upon which the gains have to be selected, the gains g1 and
g2 are computed as follows:
Effective predictive current control 15

Rs (1 )
g1 k , and g2 kp (55)
Ls Tr

where k is an arbitrary positive constant, and kp is an arbitrary value ( –1).


The suitable values for k and kp are found by trial-and-error to be 0.70. The
Luenberger observer evaluates the speed utilising the Lyapunov’s stability criterion,
which can be expressed by
d ˆ me
K1 (56)
dt k

where K1 is a positive constant, and = ( s,k – îs,k) ˆ r,k is the correction element.
To improve the speed estimation process, the load torque (external disturbance) is
also evaluated as following.
d ˆ me p
ˆk
m ˆ l,k
m (57)
dt k J

dmˆl
K2 (58)
dt k

where m ˆ k is the estimated motor torque obtained from the observed currents and fluxes
according to equation (7), m ˆ l,k is the estimated load torque, J is the moment of inertia,
and K2 is a positive constant.
To testify the effectiveness of the gains calculation procedure, a root locus analysis
for the conventional and proposed gains is shown in Figure 3. It can be noticed that at
high-speed operation, the imaginary term of the proposed observer poles [Figure 3(b)] are
lower than their respective values which are obtained using the conventional gains
[Figure 3(a)] this action contributes effectively in enhancing the estimation robustness.

Figure 3 (a) Classic Luenberger observer (b) Proposed Luenberger observer (see online version
for colours)

(a) (b)
16 M.A. Mossa

5.2 Simultaneous estimation of stator and rotor resistances


In order to eliminate the effect of parameters variation under different operating
conditions, the stator and rotor resistances are estimated simultaneously with the speed
and load torque. A sliding mode observer is used for this purpose, which takes the
estimated values of the stator current and rotor flux obtained from the Luenberger
observer, and then estimates the resistances as shown in Figure 4.

Figure 4 Combined Luenberger and sliding mode observer

is, k Rs
i s, k
Sliding mode
ˆ r,k observer Rr
Luenberger
observer me, k

Vs,k
m l,k

For the synthesis of the sliding mode observer, the following state representation can be
adopted:
x Ax Bv, and, y Cx (59)
T
x i s,k i s,k r,k r,k (60)
T
1
0 0 0
Ls
B (61)
1
0 0 0
Ls

1 0 0 0
C (62)
0 1 0 0

Lm L m me,k
0
Ls L r Tr Ls L r
L m me,k Lm
0
Ls L r Ls L r Tr
A (63)
Lm 1
0 me,k
Tr Tr
Lm 1
0 me,k
Tr Tr

1 1 1
where .
Ts Tr
Effective predictive current control 17

The sliding mode observer can be written as following:

xˆ ˆ ˆ Bv k *sgn(e)
Ax (64)

yˆ Cxˆ (65)

The discontinuous function k sgn(e) maintains trajectory of the observed vector state on
the slide surface.
where e = c(x xˆ ), and thus

e c x xˆ cAcT e cg Axˆ ck *sgn(e) (66)

And,
A ˆ A
A (67)
By considering the rotor resistance as the first observed element, thus the error matrix for
the rotor resistance can be determined from equations (63) and (67) as:
(1 ) Lm
Rr 0 Rr 0
Lr Ls L2r
(1 ) Lm
0 Rr 0 Rr
Lr Ls L2r
A (68)
Lm Rr
Rr 0 0
Lr Lr
Lm Rr
0 Rr 0
Lr Lr

The positive definite candidate Lyapunov function can be defined as following:


1
L eT e R r2 (69)
a
where ‘a’ is an arbitrary value, and by taking the derivative of equation (69) respecting to
the time, then the derivative of ‘L’ along the solution trajectory of the observer on the
sliding mode will be as following:
2
L eT e eT e R r Rˆ r (70)
a
By substitution from equation (66) into equation (70), this results in
T
eT e eT cAcT c ˆ T e (ck)T e *sgn(e)
e (c Ax) (71)

eT e eT cAcT c eT c Axˆ eT ck *sgn(e) (72)

The gains of the vector k are selected such that the Lyapunov condition in the following
condition is satisfied:
L 0.0 (73)
18 M.A. Mossa

From equation (73) and from equations (71) and (72), the following sub-conditions are
obtained:

cT (cAc)T e eT cAcT c 0 (74)

(ck)T e*sgn(e) (75)

eT ck *sgn(e) 0 (76)

2
R r Rˆ r
T
c Axˆ e e T c Axˆ 0 (77)
a
The underlined term in equation (77) can be evaluated as following
(1 ) Rr Lm R r
0 0
Lr Ls L2r
ˆ
(1 ) Rr Lm R r i s,k
0 0
T 1 0 0 0 Lr L L2 iˆ e s,k
c Axˆ e
s r s,k
* (78)
0 1 0 0 Lm R r Rr ˆ r,k e s,k
0 0
Lr Lr ˆ r,k

Lm R r Rr
0 0
Lr Lr

(1 ) Rr Lm R r
î s,k r,k
T Lr Ls L2r e s,k
c Ax e * (79)
(1 ) R r Lm R r e s,k
î s,k r,k
Lr Ls L2r

where
e s,k i s,k î s,k
(80)
e s,k i s,k î s,k

T (1 ) Lm
c Ax e î s,k ˆ r,k R re s,k
Lr Ls L2r
(81)
(1 ) Lm
î s,k ˆ r,k Rre s,k
Lr L s L2r

Then by substituting from equation (81) into equation (77), it results:


2 (1 ) Lm
R r Rˆ r 2 î s,k ˆ r,k Rre s,k
a Lr Ls L2r
(82)
(1 ) Lm
2 î s,k ˆ r,k R re s,k 0.0
Lr Ls L2r

Then from equation (82), the rotor resistance can be estimated as following:
Effective predictive current control 19

aL m
Rˆ r ˆ r,k Lm î s,k e s,k ˆ r,k Lm î s,k e s,k (83)
Ls L2r
In the same manner, the stator resistance can be estimated. The error matrix can be
obtained from equations (63) and (67) as following:
1
Rs 0 0 0
Ls
1
A 0 Rs 0 0 (84)
Ls
0 0 0 0
0 0 0 0
T
From equation (77), the underlined term { c Axˆ e eT c Axˆ } can be evaluated as
following:

Rs
0 0 0 ˆ
Ls i s,k
1 0 0 0 Rs ˆ e
0 0 i s,k *
T s,k
c Axˆ e eT c Axˆ 2 0 (85)
0 1 0 0 Ls ˆ r,k e s,k
0 0 0 0 ˆ
r,k
0 0 0 0

Then from equation (85) and by substitution into equation (67), it results:
2 2
R s Rˆ s î s,k e s,k î s,k Rs 0.0 (86)
a Ls
Thus from equation (86), the stator resistance can be estimated as following:
a
Rˆ s î s,k e s,k î s,k e s,k (87)
Ls

6 Complete system configuration

The overall layout of the proposed PCC drive can be constructed collecting the parts of
the five phase IM drive, the proposed PCC controller and the Luenberger-sliding mode
observer as shown in Figure 5. The control procedure starts with measuring of the voltage
and current of the stator, and then the measured values are sampled and used as input to
the proposed Luenberger-sliding mode observer. No existence for the PWM voltage
modulator is observed in this configuration. The reference values of the stator current
components are obtained as described in Section 3 based on the rotor field orientation
principle in the d-q reference frame, and then they are transformed to the - reference
frame using the estimated rotor flux angle ˆ r,k 1 obtained from the Luenberger observer
as follows:
20 M.A. Mossa

ˆ r,k Im ˆ r,k 1
1 a tan (88)
Re ˆ r,k 1

The estimated values of the stator and rotor resistances are used for computing the
predicted values of the stator current components and the fluxes as well; this is to prevent
any expected mismatch in their values especially at the low-speed operation. It can be
also noticed that the controller uses the estimated currents and the estimated fluxes from
the Luenberger observer instead of using the measured quantities which are usually suffer
from remarkable noise content, and thus the prediction quality is improved. The predicted
value of the rotor speed at instant (k + 1)Ts is obtained starting from the estimated speed
samples at instants (k – 1)Ts and kTs, and then fed to a PI speed controller to get the
reference torque m*k 1 which is used to get the references of the stator current d-q
components.

Figure 5 System layout for the proposed PCC for five phase IM drive (see online version
for colours)

Vs,k
1
Z

Vs,k & i s,k

ˆ me,k
*
me,k 1 me,k 1
Ts me, k me, k 1
Ts me,k 1
1
Z

m*k 1
*
r k1 1 Lr
P Lm
i *qs,k 1
ˆ
1 r, k 1 1 r, k 1
tan
Lm r, k 1

i * s,k 1 i *s,k 1 ˆ r,k 1 r, k 1


u (i) , i 0,1,...7
~ Rˆ s
i s,k 1
~ Rˆ r
i s,k 1
ˆi
s,k
~
ixs,k 1
Vs, k
Vs,k 1 ~
iys,k 1 ˆ me,k

i*xs,k 1 i*ys,k 1

Figure 6 shows the sequence of implementation for the proposed PCC procedure and how
the voltage vectors are selected. The mechanism of voltage selection starts from the
prediction of stator current components and comparing the predicted values with their
Effective predictive current control 21

corresponding references, then the obtained predicted errors will be used with the cost
function defined by equation (28). The optimal voltage vector to be selected will be
identified by the minimum value of the cost function, and then this voltage vector is
delayed by a one-step sampling time to be applied at instant kTs.

Figure 6 Sequence of implementation (see online version for colours)

us,k & is,k ˆi , ˆ , u s, k ui


s, k r, k me, k , m l, k

Rs , R r

~i ~
u s,k u s,k ~
e ki E max i s, k 1 & i is,k 1
1 1
~i ~i
ixs,k 1 & iys, k 1

i i
k 1 k 1

7 Testing results

The proposed sensorless PCC approach for the five-phase IM drive is tested using the
MATLAB/Simulink software, the parameters of the motor is addressed in Table 1.
Table 1 Parameters and data specifications for the five-phase IM drive

Parameter Value Parameter Value


Rs 2.8 J 0.008
Rr 2.4 P (pole pairs) 2
Ls 0.2388 DC link voltage 150 Vdc
Lr 0.2388 Rated Toque 4 Nm
Lls 0.0088 K1, and K2 [high speed] 65 and 35
Llr 0.0088 K1, and K2 [low speed] 1,000 and 1,000
Lm 0.23

The tests are carried out for a wide speed range for checking the feasibility of the
proposed sensorless technique. For high-speed operation, the reference is set to
150 rad/sec, a load torque of 3 Nm is applied and removed at times of (1.5 and 3 sec),
respectively. The reference value of the rotor flux is set to 0.35 Vs based on the
calculation of the machine ratings. The maximum error limit Emax is set to 0.03.
22 M.A. Mossa

Figure 7 Rotor speed (rad/sec) (see online version for colours)

Figure 8 Developed torque (Nm) (see online version for colours)

Figure 9 Rotor flux components (Vs) (see online version for colours)

Through the obtained results in Figures 7, 8 and 9 which show the speed, torque and rotor
flux profiles, respectively; it can be said that the proposed controller exhibits high
dynamic performance via tracking precisely the speed reference, while the sensorless
approach proves its validity through the accurate estimation of the rotor speed. In
addition, a decoupled control is achieved to the rotor flux components dr,k and qr,k while
Effective predictive current control 23

tracking the imposed flux reference. Moreover, the estimated values for the rotor flux
components are matching their corresponding actual values, which confirm the validity of
the proposed observer.
Figure 13 shows the five-phase stator currents, which are precisely tracking their
references; this is also confirming the feasibility of the proposed PCC approach.
Moreover, the estimated values of the stator currents are also giving a very good
matching with the measured ones and this reconfirm the validity of the proposed
observer.

Figure 10 Load torque (Nm) (see online version for colours)

Figure 11 Estimated stator resistance ( ) (see online version for colours)

Figure 12 Estimated rotor resistance ( ) (see online version for colours)


24 M.A. Mossa

Figure 13 Stator currents (A) (see online version for colours)


Effective predictive current control 25

Figure 14 Stator flux loci (see online version for colours)

Figure 15 Response of the control to a change in the absolute error (see online version
for colours)

Figure 10 illustrates the estimated load torque profile, while Figures 11 and 12 illustrate
the estimated stator and rotor resistances which are used by the controller, from which it
can be said that the observer managed in estimating properly these values, and hence the
robustness of the controller against external disturbance is enhanced. Figure 14 shows the
stator flux loci (iso flux), from which it can be confirmed that the flux inside the machine
is well managed and controlled. Figure 15 gives a focused view on the action taken by
26 M.A. Mossa

the control system at each instant the absolute value of the error exceed the hysteresis
limit Emax through switching from one voltage index to another.
For the low speed operation, the speed reference is set to 5 rad/sec, a load torque of 1
Nm is applied and removed at times of (1.5 and 3 sec), respectively.

Figure 16 Rotor speed (rad/sec) (see online version for colours)

Figure 17 Developed torque (Nm) (see online version for colours)

Figure 18 Rotor flux components (Vs) (see online version for colours)

Figure 19 Load torque (Nm) (see online version for colours)


Effective predictive current control 27

Figure 20 Estimated stator resistance ( ) (see online version for colours)

Figure 21 Estimated rotor resistance ( ) (see online version for colours)

Figure 22 Stator currents (A) (see online version for colours)


28 M.A. Mossa

Figure 22 Stator currents (A) (see online version for colours) (continued)

Figure 23 Stator flux loci (see online version for colours)


Effective predictive current control 29

Through the obtained results shown in Figures 16, 17 and 18, it can be recognised that the
proposed controller presents high dynamic performance by tracking precisely the
reference speed, while the sensorless approach proves its validity through the accurate
estimation of the rotor speed at the very low-speed range. In addition, a decoupled control
is achieved between the rotor flux components dr and qr. Furthermore, the estimated
values of the rotor flux components are highly coincidence with their actual values,
which confirm the validity of the proposed observer.
Figure 22 shows the five-phase stator currents, which are precisely tracking their
commands; this is also confirming the feasibility of the proposed PCC approach.
Figures 19, 20 and 21 show the estimated load torque profile and the estimated stator and
rotor resistances. The precise estimation of these quantities contributes in enhancing the
control system’s robustness and improved the estimation and the prediction of the
controlled variables during the implementation. Figure 23 shows the stator flux loci (iso
flux), while Figure 24 gives a detailed view of the control response expressed in terms of
the change in the voltage index according to the predicted error.

Figure 24 Detailed control action (see online version for colours)

8 Conclusions

The paper has presented an effective sensorless PCC approach for a five-phase IM drive.
A new formulation of the cost function is utilised by the controller, which enables the
investigation of the speed influence during the implementation steps. The proposed PCC
approach adopted the finite control set (FCS) principle for selecting the voltage vectors to
be applied to the motor, and thus the usage of a PWM became not necessary. An effective
formulation of the cost function used by the controller is derived through analysing the
behaviour of the absolute error value. An effective sensorless procedure is proposed for
30 M.A. Mossa

the purpose of estimating the rotor speed, the stator current, the rotor flux, the load
torque, and the stator and rotor resistances. The proposed observer exhibits high
performance in estimating these variables and demonstrating a robust behaviour under
external load changes. The contribution of the paper can be also illustrated through the
proposed selection criterion of the observer pole gains, which contributes effectively in
enhancing the precision of the observer’s estimation process and this is confirmed
through the obtained results for both high-speed and low-speed operations.

References
Abolhassani, T. (2005) ‘A novel multiphase fault tolerant high torque density PM motor drive for
traction application’, Proc. IEEE Int. Conf. Elect. Mach., pp.728–734.
Abu-Rub, H., Khan, R., Iqbal, A. and Ahmed, M. (2010) ‘MRAS-based sensorless control of a
five-phase induction motor drive with a predictive adaptive model’, IEEE International
Symposium on Industrial Electronics, Bari, pp.3089–3094.
Bermudez, M., Gonzalez-Prieto, I., Barrero, F., Guzman, H., Duran, M.J. and Kestelyn, X. (2017)
‘Open-phase fault-tolerant direct torque control technique for five-phase induction motor
drives’, IEEE Transactions on Industrial Electronics, Vol. 64, No. 2, pp.902–911.
Cao, B., Grainger, M., Wan, X., Zou, Y. and Mao, H. (2018) ‘Direct torque model predictive
control of a polyphase permanent magnet synchronous motor with current harmonic
suppression and loss reduction’, IEEE Applied Power Electronics Conference and Exposition
(APEC), San Antonio, TX, pp.2460–2464.
Cheng, X., Wensheng, S. and Xiaoyun, F. (2016) ‘Model predictive current control schemes for
five-phase permanent-magnet synchronous machine based on SVPWM’, IEEE 8th
International Power Electronics and Motion Control Conference (IPEMC-ECCE Asia), Hefei,
pp.648–653.
Darijevic, M., Jones, M., Dordevic, O. and Levi, E. (2017) ‘Decoupled PWM control of a
dual-inverter four-level five-phase drive’, IEEE Transactions on Power Electronics, Vol. 32,
No. 5, pp.3719–3730.
de Silva, N., Fletcher, E. and Williams, W. (2004) ‘Development of space vector modulation
strategies for five-phase voltage source inverters’, Proc. IEE Power Elect, Mach. Drives
Conf., PEMD, Edinburgh, pp.650–655.
Dordevic, O., Levi, E. and Jones, M. (2013) ‘A vector space decomposition based space vector
PWM algorithm for a three-level seven-phase voltage source inverter’, IEEE Transaction on
Power Electronics, Vol. 28, No. 2, pp.637–649.
Dujic, D., Jones, M. and Levi, E. (2009) ‘Generalized space vector PWM for sinusoidal output
voltage generation with multiphase voltage source inverters’, Int. J. Ind. Elect. Drives, Vol. 1,
No. 1, pp.1–13.
Duran, J., Toral, S., Barrero, F. and Levi, E. (2007) ‘Real time implementation of multidimensional
five-phase space vector pulse width modulation’, Elect. Lett., Vol. 43, No. 17, pp.949–950.
Duran, M.J., Gonzalez-Prieto, I., Rios-Garcia, N. and Barrero, F. (2018) ‘A simple, fast, and robust
open-phase fault detection technique for six-phase induction motor drives’, IEEE Transactions
on Power Electronics, Vol. 33, No. 1, pp.547–557.
Guzman, H., Barrero, F. and Duran, M.J. (2015) ‘IGBT-gating failure effect on a fault-tolerant
predictive current-controlled five-phase induction motor drive’, IEEE Transactions on
Industrial Electronics, Vol. 62, No. 1, pp.15–20.
Hara, A., Enokijima, H. and Matsuse, K. (2011) ‘Independent vector control of two induction
motors fed by a five-leg inverter with space vector modulation’, IEEE Industry Applications
Society Annual Meeting, Orlando, FL, pp.1–8.
Heydari, M., Varjani, A.Y., Mohamadian, M. and Fatemi, A. (2012) ‘Three-phase dual-output six-
switch inverter’, IET Power Electronics, Vol. 5, No. 9, pp.1634–1650.
Effective predictive current control 31

Huang, W., Hua, W., Chen, F., Yin, F. and Qi, J. (2018) ‘Model predictive current control of open-
circuit fault-tolerant five-phase flux-switching permanent magnet motor drives’, IEEE Journal
of Emerging and Selected Topics in Power Electronics, Vol. 6, No. 4, pp.1840–1849.
Iqbal, A. and Levi, E. (2006) ‘Space vector PWM techniques for sinusoidal output voltage
generation with a five-phase voltage source inverter’, Elect. Power Comp. Syst., Vol. 34,
No. 2, pp.119–140.
Khan, R. and Iqbal, A. (2009) ‘Extended Kalman filter based speeds estimation of series-connected
five-phase two-motor drive system’, Simulation Modelling Practice and Theory, Vol. 17,
No. 17, pp.1346–1360.
Kong, W., Huang, J., Li, B., Kang, M. and Zhao, L. (2013) ‘improved sliding-mode observer for
sensorless control of five-phase induction motor’, International Conference on Electrical
Machines and Systems ICEMS, pp.2024–2027.
Kubota, H., Matsuse, K. and Nakan, O. (1993) ‘DSP-based speed adaptive flux observer of
induction motor’, IEEE Transaction on Industry Applications, Vol. 29, No. 2, pp.344–348.
Mossa, M. (2018) ‘Effective predictive flux control for a five phase induction motor drive with
inverter output filter’, International Review of Electrical Engineering (IREE), Vol. 13, No. 5,
pp.373–384.
Ojo, O., Dong, G. and Wu, Z. (2006) ‘Pulse width modulation for five-phase converters based on
device turn-on times’, Proc. IEEE Ind. Appl. Soc. Ann. Mtg IAS, Tampa, FL, CD-ROM, paper
IAS15, p.7.
Payami, S., Behera, R.K., Iqbal, A. and Al-Ammari, R. (2015) ‘Common-mode voltage and
vibration mitigation of a five-phase three-level NPC inverter-fed induction motor drive
system’, IEEE Journal of Emerging and Selected Topics in Power Electronics, Vol. 3, No. 2,
pp.349–361.
Prieto, J., Barrero, F., Toral, S., Jones, M. and Levi, E. (2011) ‘Analytical evaluation of switching
characteristics in five-phase drives with discontinuous space vector pulse width modulation
techniques’, Proceedings of the14th European Conference on Power Electronics and
Applications, Birmingham, pp.1–10.
Rahman, K., Iqbal, A., Al-Emadi, N. and Ben-Brahim, L. (2017) ‘Common mode voltage reduction
in a three-to-five phase matrix converter fed induction motor drive’, IET Power Electronics,
Vol. 10, No. 7, pp.817–825.
Riveros, A., Prieto, J., Barrero, F., Toral, S., Jones, M. and Levi, E. (2010) ‘Predictive torque
control for five-phase induction motor drives’, IECON – 36th Annual Conference on IEEE
Industrial Electronics Society, Glendale, AZ, pp.2467–2472.
Ryu, M., Kim, H. and Sul, K. (2005) ‘Analysis of multi-phase space vector pulse width modulation
based on multiple d-q space concept’, IEEE Transaction on Power Electronics, Vol. 20, No. 6,
pp.1364–1371.
Sadehgi, S. and Parsa, L. (2010) ‘Design and dynamic simulation of five-phase IPM machine for
series hybrid electric vehicles’, Proc. Green Tech. Conf., pp.1–6.
Villani, M., Tursini, M., Fabri, G. and Castellini, L. (2010) ‘Multi-phase fault tolerant drives for
aircraft applications’, Electrical Systems for Aircraft, Railway and Ship Propulsion, Bologna,
pp.1–6.
Ward, E. and Harer, H. (1969) ‘Preliminary investigation of an inverter-fed 5-phase induction
motor’, Proc. Int. Elect. Eng., Vol. 116, No. 6, pp.980–984.
Xu, H., Toliyat, A. and Petersen, J. (2009) ‘Five-phase induction motor drives with DSP-based
control system’, Proc. IEEE Int. Elec. Mach. And Drives Conf. IEMDC2001, Cambridge, MA,
pp.304–309.
Yan, H., Xu, Y., Zou, J., Zeng, D. and Zeng, F.(2016) ‘Phase current reconstruction for dual
three-phase permanent magnet synchronous motor drive in electric vehicles using two DC link
current sensors’, IEEE Vehicle Power and Propulsion Conference (VPPC), Hangzhou, pp.1–6.
View publication stats

32 M.A. Mossa

Zhang, X., Zhnag, C., Qiao, M. and Yu, F. (2008) ‘Analysis and experiment of multi-phase
induction motor drives for electrical propulsion’, Proc. Int. Conf. Elect. Mach., ICEM,
pp.1251–1254.

View publication stats

You might also like