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

net/publication/291974479

Fast-Speed Drives for Permanent Magnet Synchronous Motor Based on Model

Predictive Control

Conference Paper · October 2015

DOI: 10.1109/VPPC.2015.7352946

CITATIONS READS
12 92

1 author:

Abdelsalam Ahmed
Tanta University
29 PUBLICATIONS   157 CITATIONS   

SEE PROFILE

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

Development of Electric Vehicle Powertrain View project

All content following this page was uploaded by Abdelsalam Ahmed on 05 April 2016.

The user has requested enhancement of the downloaded file.

 Fast-Speed Drives for Permanent Magnet
Synchronous Motor Based on Model Predictive
Control
Abdelsalam A. Ahmed
Dept. of Electrical Power Systems and Machines
Faculty of Engineering, Tanta University
Tanta, Egypt
[email protected]

Abstract—fast speed-tracking with limits commitment for and current controllers in a single MPC, instead of keeping the
currents and voltages of drivetrain is crucial for EV/HEV. This conventional cascade loop. In [8], a predictive function control
study proposes a fast-speed drive for permanent magnet was designed to control the speed of PMSM. The response was
synchronous motor (PMSM) based on Model Predictive Control improved using a second-order linear state observer. However,
(MPC). Fast speed tracking is achieved through optimization of the method required a lot of effort to tune the observer gains
the DC source utilization. With common torque and current under different speed conditions. Fast torque control system
limits, the proposed technique predicts the optimum voltage based on model predictive control was proposed in [10].
vector applied to the driver. The performance of the PMSM Experimental results showed good performance of proposed
under Field Oriented Control (FOC) and MPC is first analyzed
MPC torque control system compared with conventional torque
by a simulation study at different working circumstances. The
control system.
effectiveness of the proposed MPC has been validated by
simulation results. Then, an experimental setup has been In this paper a model-based predictive current control
established for the PMSM drive using a TMS320F2812 DSP. The (MPCC) method for nonlinear systems with inherent output
effectiveness of the proposed MPC has been validated by limitation is presented. The proposed technique is based on
experimental results. With simulation and experimental results, optimization of the voltage vector applied to the power
the presented MPC has been investigated by comparing to the converter. With the same torque and current limits, the
cascaded PI control based on SVPWM to highlight the proposed MPC-based FOC is compared to the SVPWM-based
improvements, especially regarding speed tracking response.
FOC. Analysis, simulation and experimental set-up are
introduced to verify the proposed system performances.
Keywords—model predictive current control; field oriented
control; PMSM; space vector PWM; digital signal processor; II. DESCRIPTION OF DRIVE SYSTEM
cascaded PI control; DC source utilization
The overall experimental system is depicted in Fig. 1.
I. INTRODUCTION
Field Oriented Control (FOC) with cascaded PI controllers
and modulators are having many constraints for producing fast
and smooth speed performances over large speed operating
range. The solution proposed here is the Model-based
predictive control (MPC). A MPC is one of the typical optimal
controllers used in power electronics and electric drive
systems and decides control inputs by an optimal computation
so as to track a system output to a reference [1].
Several predictive control techniques have been proposed in
the literature for power electronics and electrical drives [2-6].
MPC is the only one among the so-called advanced control
techniques which has been extremely successful in practical
applications in recent decades [1, 2, 3]. A comparative
assessment of Space Vector Pulse Width Modulation
(SVPWM) and MPC for VSI in terms of THD is presented in
[4]. MPC was found superior in terms of THD. Recently,
several authors have developed different predictive techniques
for PMSMs [7, 8, 9]. Authors in [7] introduce augmented state
vector that includes nonlinearity as cross-effect into a new state
variable: measured disturbance. This article combined speed Fig. 1. Experimental setup for real tests

978-1-4673-7637-2/15/$31.00 ©2015 IEEE

 It includes a TMS320F2812 DSP control board with its B. Motor model discretization
different peripherals connectors, a voltage source IGBT-based To predict the current at sampling instant of 1 with
IPM drives with interface and measurement circuits, 3-ph the measured position, speed and currents at sampling instant
PMSM with an incremental encoder with integrated of , a discrete model of the motor and its drive are needed.
commutation signals that used as motor feedback, an Applying the forward Euler discretization theory at (1), the
electromagnetic brake device, single-phase adjustable output resulting discrete time model is given in form of 1
transformer, PC with Code Composer Studio (CCS) installed as:
[11]. The specifications of the drive system are described in
Table I. The specifications of the used PMSM are described in 1
Table II.
1
TABLE I. SPECIFICATIONS OF THE DRIVE SYSTEM
1
Device Specification, Value (unit) 1 0
Load Adjustable brake unit: 24 (V), 0.5 (A), 0~10 (Nm), 110 (W)
DSP TMS320F2812:150MHz, 32-bit fixed-point CPU, 12-bit ADC
1kVA, 50Hz, 220V input voltage, 0~250V output voltage,
0 1
Transformer
4 A rated current .
Power IPM power module with six IGBT power transistors: 0 1
board PS21867-AP (600V, 30 A); Hall-based current sensors, 20A.
Encoder Incremental encoder with 2500 pulses/rev 0 0

TABLE II. SPECIFICATIONS OF THE PMSM 0 0 (2)

Specification Unit Value 0 0
Rated/Max. speeds rpm 2000/2500
Standstill torque, Nm 3 Where, , , , are equivalent inertia of motor and
Standstill current A 2.5
Torque C. (KT) Nm/A 1.2
load, mechanical angular velocity, electrical angular velocity,
Pole pairs 3 and sampling time, respectively.
Voltage C. (Ke) V/Krpm 80
C. Inverter and switching states
Rotor moment of inertia 10-4 Kg.m2 4.4
Friction coefficient N.m.s/rad 0.00001 For 2-level three-phase inverter, the inverter switching
Stator resistance Ω 3 voltages directly in the d-q frame can be derived as:
Inductance mH 11.5
Flux linkage Web 0.178
(3)

III. DESIGN AND IMPLEMENTATION OF MPC FOR PMSM

A. Motor continuous state-space model and state variables The vector is the switching function for the three
for MPC phases with the value 1 when upper leg switch is ON and 0
when it is OFF. is the voltage level of DC link.
The first step in the design of a MPC consists in
determining a discrete-time model for the system. The linear D. Optimization of the DC voltage utilization
model has to be pursued by an appropriate choice of the state The objectives of the MPC scheme are torque current
variables. The state space representation for the PMSM can be reference tracking, torque by ampere optimization, and current
summarized as , . Where; the state magnitude limitation. These objectives are expressed in the
vector is and the input signal is the 2-D following cost function.
voltage vector . According to this choice, the
continuous state-space model for the motor is then: 1 1
1 , 1 . (4)
0
The first term represents the minimization of the reactive
power, allowing the torque by ampere optimization. The
. second term is defined for tracking the torque-producing
0 current. These terms are weighted according to the factors of
and . These factors are taken 1 and 2, respectively. The
0 0
third term is a nonlinear function for limiting the amplitude of
0 0 (1) the stator currents. This term, i.e., , is a highly nonlinear
function that takes into account restrictions on the maximum
0 0 stator current, imposing hard constraints on the maximum
values of the stator currents. This function is given by (5). The
final term is responsible of minimizing the switching frequency executing current loop is 100 µs and executing PI control
of inverter according to the value of its weighting factor, λ . algorithm of speed loop is 1ms.Those programs are coded with
C language through the CCS programming. For the purpose of
∞ ̂ ̂ , fairly comparison, all of these conditions are fixed for both
, (5)
0 ̂ ̂ . MPC and SVPWM techniques.

The term λSW . n is augmented to the cost function, which B. Simulation results
will make this specific switching pattern less favorable for 1) Transient response at change of load and speed
choosing. If there is no change in switching state from last Phase voltage, applied to the VSI, and phase current are
sampling instant, the function will be equal to zero, hence no simulated for both MPC and SVPWM techniques in Fig. 2. The
additional term is added to the cost function and is therefore effect of optimizing the DC voltage utilization by MPC is
more favorable for choosing. The factor of is taken 0.001. observed in this figure at transition from no load to full load
When increase, the switching frequency is reduced but torques at 0.4 s. In case of MPC, the voltage vectors are
this reduction comes with a price of higher current ripple and optimally chosen to keep the phase current very close to its
distortion. reference at no- and full-load conditions, as depicted in Fig.
| 1 | | 1 | 2(a). At the same load conditions, with SVPWM, the active
voltage vectors are applied to the VSI causing a relatively high
| 1 | (6) current flows during no-load operation and unmatched current
Inverter dead-time is considered to prevent the short-circuit response after loading, as shown in Fig. 2(b). Also, it can be
faults at commutation instants. The optimal value of the observed that the voltage vectors in MPC technique are
objective function is applied during the next sampling period. optimally chosen with fast response to accomplish the
commanded performance within predetermined ratings.
The objective function is evaluated for each of the eight Another notice is that the switching frequency with MPC is
possible voltage vectors in order to calculate the future optimal lower than that with SVPWM method.
value of the load current. For all possible switching states for Phase voltage (V) 150
three phases, the voltage vectors components are predicted 100
depending on the measured rotor position. Then, the values of 50
stator current components are assigned according to the 0
-50
predicted voltage vectors, measured stator currents , -100
machine’s parameters, and the used sampling time period . -150
The stator currents are hardly constrained for guaranteeing 0.395 0.4 0.405 0.41 0.415 0.42

the overcurrent precautions at different speed levels.

Phase current (A)

4
3
IV. ASSESSMENT OF MODEL PREDICTIVE CURRENT CONTROL 2
1
0
A. Conditions for comparision between MPC-based FOC and -1
-2
SVPWM--based FOC -3
-4
This paper makes a comparison mainly between a MPC-
0.395 0.4 0.405 0.41 0.415 0.42
based vector control with a SVPWM-based vector control. PID Time (sec)
controller with anti-windup is used for controlling speed in (a)
both control methods. For the output of speed regulator, the q- 150
axis current component and torque are limited at 3 A and 6 Nm
Phase voltage (V)

100
for both strategies. Sampling time of the simulated results for 50
speed loop is considered as 1000 . The DC–link 0
voltage is leveled at 200V. The proportional and integral gains -50
of the PI speed controller are taken as 0.04 and 2, respectively -100
for both techniques. Whereas, two anti-windup PI current -150
0.395 0.4 0.405 0.41 0.415 0.42
controllers are used for control the current components in d and
q axes in SVPWM technique. These controllers are limited at 4
3
Phase current (A)

20 . The proportional and integral gains of the current 2

Response

controller are tuned at 25 and 20, respectively. The output of 1

Reference
0
the controllers represents the required voltage components in d -1
and q. Both voltage components are restricted at 100V. -2
-3
In experimental implementations, the PWM switching -4
0.395 0.4 0.405 0.41 0.415 0.42
frequency of inverter is designed with 10 kHz, dead-band is Time (sec)
considered as 3µs to prevent three phase legs of inverter from
shooting through. The control sampling frequency of current (b)
and speed loops are determined as 10 kHz and 1 kHz, Fig. 2. Effect of voltage utilization techniques on phase voltage and phase
respectively. That means the computational time of the DSP for current: (a) phase voltage (upper) and phase current (lower) for PMC
technique, (b) phase voltage and phase current for SVPWM technique
 Another comparison side, transient torque responses for the load torque is shown in Fig. 5. The amplitude of the speed
both MPC and SVPWM at load torque and speed changes are is changed from 500 to 1000 RPM at time 0.2s and the
shown in Fig. 3 and Fig. 4, respectively. As shown in Fig. 3 (a, amplitude of the load torque is changed from 0 to 2Nm at time
and b) for MPC, when the load torque is increased, the 0.4s. The maneuver corresponds to controlled starting, a speed
developed torque is increased and the speed is decreased. The increase, and a load torque response. It is possible to see that
same is happened for SVPWM as shown in Fig. 3 (c, and d). It MPC achieves an excellent dynamic performance with no
can be seen that the torque response with MPC is faster than overshoots in speed.
that of SVPWM. Hence, the speed can be retrieved fast to its
command value. 1000
900
MPC SVPWM

Motor speed (RPM)

3 3 800
700
Motor torque (Nm)

2 Motor torque (Nm) 2

600
1 1 500
400
0 0
300
-1 -1 200
0.4 0.401 0.402 0.403 0.404 0.405 0.4 0.401 0.402 0.403 0.404 0.405
(a) (c) 100
1000 1000 0
0 0.1 0.2 0.3 0.4 0.5 0.6
Motor speed (RPM)

Motor speed (RPM)

980 980 Time (sec)

Fig.5. Simulation results of MPCC of the PMSM: speed response
960 960
Another comparison side, response of torque and stator
940 940
current components for both MPC and SVPWM at load torque
920
0.4 0.401 0.402 0.403 0.404 0.405
920
0.4 0.401 0.402 0.403 0.404 0.405
and speed changes are shown in Fig. 6 and Fig. 7, respectively.
(b) (d) A torque response is shown in Fig. 6(a). It is important to point
Time (sec) Time (sec)
out the fast dynamic behavior of the strategy. This can be
Fig. 3. Transient torque response at load torque change: (a) and (b) for torque explained by the fact that MPC is a direct strategy that does not
and speed with MPC, (c) and (d) for torque and speed with SVPWM.
require an inner PI control loops for the stator currents and
Also, as shown in Fig. 4 (a, and b) for MPC, when speed is modulators. Hence, there is no bandwidth limitation for the
increased, the developed torque is increased for the electromagnetic torque dynamics. The behavior of the stator
acceleration period. For the same action with SVPWM, the current components at steady state and transient operations are
torque response is shown in Fig. 4 (c, and d). It can be noticed presented in Fig. 6(b, c). It can be noticed that d-axis current
that the torque response with MPC is limited and strict more component is fully controlled to be zero and the q-axis current
than that of SVPWM. Whereas, the developed torque with component is controlled to give fast torque response within the
SVPWM overshoots above the limited value (3 Nm) and with prescribed limitations.
some oscillations. This results in fast speed tracking with MPC MPC
than that with SVPWM, as depicted in Fig. 4 (b) and (d),
Torque (Nm)

4
respectively. 2
MPC SVPWM 0
5 5
-2
4 4 0 0.1 0.2 0.3 0.4 0.5 0.6
Motor torque (Nm)

Motor torque (Nm)

(a)
Quadrature current (A)

3 3
2 2 6
1 1 4
2
0 0
0
-1 -1 -2
0.19 0.2 0.21 0.22 0.23 0.19 0.2 0.21 0.22 0.23 0 0.1 0.2 0.3 0.4 0.5 0.6
(a) (c) (b)
1200 1200
Direct current (A)

4
2
Motor speed (RPM)

Motor speed (RPM)

1000 1000 0
-2
800 800 -4
0 0.1 0.2 0.3 0.4 0.5 0.6
600 600 (c)
Time (sec)
400 400
0.19 0.2 0.21 0.22 0.23 0.19 0.2 0.21 0.22 0.23
Fig.6. Simulation results of MPC: (a) developed torque response, (b) q-axis
(b) (d)
Time(sec) Time(sec) current, (c) d-axis current.
Fig. 4. Transient torque response at speed change: (a) and (b) for torque and
speed with MPC, (c) and (d) for torque and speed with SVPWM.
Simulation results of phase voltage and phase current for
the PMSM under MPC and SVPWM vector control techniques
are depicted in Fig. 8 and Fig. 9, respectively. Fig. 8(a) shows
2) Dynamic behavior of MPC
the phase voltage for MPC. It can be noticed that the phase
The behavior of the MPC control scheme for a step in the current in MPC is fully matched its reference with fast
amplitude of the speed reference and step in the amplitude of response, as seen in Fig. 8(b). Whereas, using SVPWM, the
phase current doesn’t match the reference, as depicted in Fig. C. Experimental results
9(b). This is because the linearity nature of the PI current An experimental study is carried out with the aim of
controllers and modulator effect. corroborating the effectiveness of the proposed MPCC in
SVPWM comparison with a FOC based on SVPWM modulator. MPC
and SVPWM control algorithms have been implemented with a
Torque (Nm)

4
2 sampling time of 100 and the dead-time of 3 ,
0 and tested with a load of 3N.m. The DC link voltage is set to
-2
0 0.1 0.2 0.3 0.4 0.5 0.6 140V for both techniques. The speed performances are
(a) experimentally investigated for both techniques as in the
Qadrature current (A)

6 following figures.
4
2
0
Figure 10 shows the speed performance when it is
-2 referenced at 0, 500, and 1000 RPM for SVPWM as in Fig.
0 0.1 0.2 0.3 0.4 0.5 0.6
(b) 10(a) and for MPC as in Fig. 10(b). It can be noticed that with
MPC, the speed matches its reference very fast more than that
Direct current (A)

4
2
0
of SVPWM control method.
-2
-4
0 0.1 0.2 0.3 0.4 0.5 0.6
(c)
Time (sec)
Fig.7. Simulation results of SVPWM: (a) developed torque response, (b) q-axis
current, (c) d-axis current.

150
100
Phase voltage (V)

50
0
-50
-100 (a)
-150
0 0.1 0.2 0.3 0.4 0.5 0.6
(a)
4
Phase current (A)

-2

-4
0 0.1 0.2 0.3 0.4 0.5 0.6
(b)
Time (sec)

Fig. 8. Simulation results of phase voltage and current for the PMSM under
MPC (b)
150 Fig. 10. Experimental performance of PMSM at step-up speed: (a) SVPWM-
100 based VC, (b) MPC-based VC
Phase voltage (V)

50
0 The behavior of the predictive and SVPWM vector control
-50 techniques for a speed reversal is shown in Fig. 11. It shows the
-100 speed performance when it is reversed from 1000 RPM to -
-150 1000 RPM through stop point for SVPWM as shown in Fig.
0 0.1 0.2 0.3
(a)
0.4 0.5 0.6
11(a) and for MPC as in Fig. 11(b). It is possible to notice the
better dynamic performances achievable by the proposed MPC.
4 Response
3 Figure 12 shows the speed performance when it is loaded
Phase current (A)

2
1
suddenly. The motor starts with no load and then is loaded at
0 1000 RPM with 2Nm for SVPWM as in Fig. 12(a) and for
-1 MPC as in Fig. 12(b). It is can be noticed the better dynamic
-2
-3 Reference
performances achievable by the proposed MPC compared to
-4 PI.
0 0.1 0.2 0.3 0.4 0.5 0.6
(b)
Time (sec)
(b)
Fig. 9. Simulation results of phase voltage and current for the PMSM under
SVPWM.
 developed MPC technique, the liner PI controllers and
modulators are omitted from the drive system. The influences
of replacement of SVPWM by MPC are analyzed with
simulation results at transient times. Then the performance of
the developed controllers has been studied at different
operating conditions. Experimental results demonstrate the
effectiveness of the proposed MPC. From experimental results,
compared to SVPWM, the presented MPC offers improved
speed tracking response and better dynamic performance for
PMSMs overall operation widespread under the same working
limitations. Finally, we can conclude that the MPC presents
(a)
robust speed control with fast speed tracking, no overshoots,
guarantee stator currents within limits, robust for widespread
speed levels, and it can be considered a non-linear controller
for non-linear systems.
REFERENCES
[1] P. Cortes, M. P. Kazmier kowski, R. M. Kennel, D. E. Quevedo, and J.
Rodriguez, “Predictive control in power electronics and drives,” IEEE
Transaction on Industrial Electronics, Vol. 55, No. 12, pp. 4312–4324,
Dec. 2008.
[2] J. Rodriguez and P. Cortes, Predictive Control of Power Converters and
Electrical Drives, Hoboken, NJ, USA: Wiley, 2012.
[3] Alfonso Damiano, Gianluca Gatto, Ignazio Marongiu, Aldo Perfetto, and
(b) Alessandro Serpi, “Operating constraints management of a surface-
mounted PM synchronous machine by means of an FPGA-based model
Fig. 11. Experimental performance of PMSM at reversing speed: (a) predictive control algorithm,” IEEE Trans. on Indus. Informatics, Vol.
SVPWM-based VC, (b) MPC-based VC 10, No. 1, Feb. 2014.
[4] Irtaza M. Syed, Kaamran Raahemifar, “Space Vector PWM and Model
Predictive Control for Voltage Source Inverter Control”, World
Academy of Science, Engineering and Technology International Journal
of Electrical, Computer, Electronics and Communication Engineering
Vol.8 No:11, pp. 1562-1568, 2014.
[5] S. Kouro, P. Cortes, R. Vargas, U. Ammann, and J. Rodriguez, “Model
predictive control-a simple and powerful method to control power
converters,” IEEE Trans. Ind. Electron., vol. 56, no. 6, pp. 1826–1838,
Jun. 2009.
[6] F. Morel, X. Lin-Shi, J.-M. Retif, B. Allard, and C. Buttay, “A
comparative study of predictive current control schemes for a permanent
magnet synchronous machine drive,” IEEE Trans. Ind. Electron., vol. 56,
no. 7, pp. 2715–2728, Jul. 2009.
(a) [7] S. Bolognani, L. Peretti, and M. Zigliotto, “Design and implementation
of model predictive control for electrical motor drives,” IEEE
Transactions on Industrial Electronics, vol. 56, no. 6, pp. 1925 –1936,
2009.
[8] S. D. Cairano, D. Yanakiev, A. Bemporad, I. V. Kolmanovsky, and D.
Hrovat, “Model predictive idle speed control: Design, analysis, and
experimental evaluation,” IEEE Trans. Control Syst. Technol., vol. 20,
no. 1, pp. 84–97, Jan. 2012.
[9] H. Liu and S. Li, “Speed control for PMSM servo system using
predictive functional control and extended state observer,” IEEE Trans.
Ind. Electron., vol. 59, no. 2, pp. 1171–1183, Feb. 2012.
[10] Yuya Hozumi, Shinji Doki and Shigeru Okuma, “Fast Torque Control
System of PMSM based on Model Predictive Control”, IECON 2009
Proceedings, p.p. 1147-1151, 2009.
(b)
[11] A. Ahmed, A. Quntao and S. Li, “DSP-based implementation of
Fig. 12. Experimental performance of PMSM at loading: (a) SVPWM-based permanent magnet synchronous motor drives for EV/HEV applications”,
VC, (b) MPC-based VC 16th International middle east power systems conference (MEPCON'14),
Cairo, Egypt, December 23-25, 2014.
V. CONCLUSION
This paper addressed the simulation and experimental
implementations of MPC and FOC with SVPWM applied for
PMSM. The goal was to obtain an alternative to the classical
FOC method with improved dynamics. The developed MPC is
based on optimization of the DC source utilization. With the

View publication stats