You are on page 1of 6

2014 17th International Conference on Electrical Machines and Systems (ICEMS),Oct.

22-25, 2014, Hangzhou, China

The Use of Genetic Algorithms for


Speed and Current Controller Designs for
IPMSM Drive
Pakorn Chottiyanont1, Mongkol Konghirun2, Wanchak Lenwari3
1, 2
Dept. of Electrical Engineering, King Mongkuts University of Technology Thonburi, Thailand
3
Dept. of Control System and Instrumentation Engineering, King Mongkuts University of Technology Thonburi, Thailand
E-mail:pakorn213@gmail.com, mongkol.kon@kmutt.ac.th, wanchak.len@kmutt.ac.th

Abstract Recently, speed and current controller have been controller to improve the speed controller of the IPMSM.
developed for interior permanent magnet synchronous motor However, in the commercial drive systems, the controllers
(IPMSM). Each method has its own advantage and limitation. should be easy to set up and tune in particular for operator.
However, in the commercial drive system, the controller should
Therefore the classical PI controller has been widely used due
be easy to set up and tune in particular for operator, therefore the
classical proportional plus integral (PI) controller has widely used to its simple structure while achieving zero steady-state error.
due to its simple structure while achieving zero steady-state error. However, finding out the parameters of controller is not any
However finding out the parameters of the controller is not any easy task because of complex dynamics of IPMSM.
easy task because of complex dynamics of IPMSM. For this To obtain the controller parameters, the trial and error
reason, this paper presents a novel automated design employed method is very simple but does not guarantee optimal
genetic algorithm (GA) to optimize PI speed and current parameters. Therefore, an optimization based Genetic
controller of IPMSM. The drive control system is designed on the Algorithms (GA) has been developed to improve the controller
basis of vector control scheme incorporated with the maximum performance. [5] proposed the online optimization. The GA
torque per ampere (MTPA) control strategy to improve the drive was implemented and run real time while the controller
performance. Simulation and experimental results verify the hardware operated in the normal condition. Another approach,
effectiveness of the proposed method.
offline optimization [6], is carried out by the simulation.
Mathematical model is necessarily required for evaluating
I. INTRODUCTION potential controller parameters. Therefore, in this case, it was
The permanent magnet synchronous motor drive have been depend on the accuracy of the model. However, by using
widely used in high performance applications because of their offline algorithm, it is possible to avoid risk of ill-suited
advantages, such as wide operating speed range, high power control parameters on the system hardware due to over-tuned
density, high torque to inertia ratio and high efficiency. parameter that commonly met in online tuning.
The permanent magnet synchronous motors are also In this paper, the genetic algorithms optimization is used to
classified as surface-mounted permanent magnet (SPM) and optimize the PI controller in order to obtain precise speed and
interior permanent magnet (IPM). The magnets of the SPM current controls with the minimum error possible. The
proposed method combines the vector control strategy with
motor are attached on the surface of the rotor, whereas those of
the IPM motor are buried inside. The effective air gap in the MTPA control technique. As a result, the effectiveness of the
magnetic flux path of stator inductance component in d-axis proposed method is higher than the trial and error solution.
(Ld) and q-axis (Lq) are the same in SPM and unequal in IPM The paper is organized as follows. Section II and III
motors. From this reason, it is possible to use the reluctance explains the mathematical model of the IPMSM and control
torque of IPM to increase the total output torque and operate at algorithm of the proposed method. Then, Section IV describes
very high speed condition [1]. However, the conventional the principle and optimization procedure of the proposed
field-oriented control strategy (vector control) is used only on genetic algorithm method. To demonstrate the effectiveness of
the amplitude of the q-axis of the stator current and neglects the proposed method, the simulation are carried out by using
the reluctance torque. Therefore, the maximum torque per MATLAB/Simulink in Section V and the experimental results
ampere control (MTPA) is presented [2]-[3]. Results give the are illustrated in Section VI. Finally, the conclusion is clarified
highest torque output compared to the current by utilizing both in Section VII.
the electromagnetic and the reluctance torques.
Moreover, in order to improve the performance of the drive
system, [2] proposed the speed controller based on state- II. THE MATHEMATICAL MODEL
dependent Riccati equation (SDRE) for IPMSM drive. This Generally, the mathematical model of IPMSM drive is
method not only yielded fast dynamic response but also small defined in the rotating dq frame of reference. The stators
steady-state speed error and robustness under load torque voltage and the torque equations of the rotating dq reference
disturbance. Additionally, [3] and [4] utilized the fuzzy logic frame are given in (1) and (2), respectively.

2930

978-1-4799-5162-8/14/$31.00 2014 IEEE


Fig. 1. Block diagram of the proposed method.

d available stator current depended on the inverter capability is


R + L n pr Lq i written in (3), where Is is the magnitude of stator current.
vd s dt d d 0
= +
i n (1)
vq n L d
Rs + Lq q p r PM

p r d
dt id = I s 2 iq 2 (3)

3 Substituting (3) into (2), the torque equation can be


Te = n p (PM iq + ( Ld Lq )iq id ) (2) rewritten and then it is differentiated with respect to q-axis
2
current and setting the equation to zero in as shown in (4) and
Where: simplifies to (5).
vd , v q stator voltage components in dq-axis;
dTe 3 1
id , iq stator current components in dq-axis; = n p (PM + ( Ld Lq )id ( Ld Lq )iq 2 ) = 0 (4)
diq 2 I s iq 2
2

Ld , Lq stator inductance components in dq-axis;


PM permanent magnet flux linkage;
iq 2
r rotor speed in angular frequency; PM + ( Ld Lq )id ( Ld Lq ) =0 (5)
id
np number of pole pairs;
Rs stator resistance; Solving (5) for d-axis current in order to obtain the MTPA
Te electromagnetic torque; equation as shown in (6) [1].

PM PM 2
id = + iq 2 (6)
III. CONTROL ALGORITHM WITH MTPA 2( Lq Ld ) 4( Lq Ld ) 2
In conventional vector control system, the d-axis current
command is normally set to be zero in order to make the The d-axis current command is determined by using
control scheme simply. Therefore, there is only equation (6) as clearly shown in Fig.1. The q-axis current
electromagnetic torque regulated by setting only q-axis current
command is generated from this controller and simultaneously
command. However, when maximum torque is required in
below rated speed range, the reluctance and electromagnetic used to compute the d-axis current command for reaching
torques are simultaneously controlled through regulating both MTPA control. Then, both d- and q-axis currents are regulated
of d- and q- axis currents under the available inverter by classical PI current controllers. After that voltage
capability. commands in stationary reference frame are used as input of
In order to achieve maximum torque per ampere control, the space vector PWM to generate pattern switching signals for
relationship between d- and q- axis currents and maximum inverter as shown in Fig.1.

2931
IV. GENETIC ALGORITHM OPTIMIZATION APPROACH
A. Principle of the genetic algorithm method
A GA is a stochastic global search heuristic that is inspired
by the theories of evolution and natural selection [7]. GAs try
to imitate the properties of natural selection and natural
genetics. It begins with initial population that consists of coded
strings of binary numbers, called chromosomes that hold
possible solutions of the problem. Their performance is
evaluated by a fitness function. Strings with higher fitness
value are potentially survived to the next generation as parents
until it finds the optimal solution to the problem.
GA operates on a population of potential solutions, termed
individuals, applying the principle of evolution, simulated by
means of mathematical operations that mimic the process of
selection, crossover, mutation and replacement. A basic GA
routine is shown in Fig. 2. The routine starts with generating a
population of individuals randomly. In each iteration,
successive populations of strings were generated through
genetic operation. Individuals in the current population are
evaluated using a measure of their objective function values,
called fitness function. The GA will seek the solution that
maximizes or minimizes the fitness function. A fitness
function measures the fitness of an individual to survive in a
population of individuals. Individuals with a lower value of
Fig. 2. Basic GA routine.
fitness function have a higher probability of contributing one
or more offspring in the next generation. The selection of
fitness function used to evaluate fitness of each individual is current population. At the end of the optimization process, the
the crucial step and can be defined in many different ways GA returns a set of optimum parameter provided the least IAE
based on different target specifications. possible within the predefined search area of the parameters as
given in Table I. The plot of the objective function over the
number of iterations is shown in Fig. 4. The final s-domain of
B. Optimization procedure speed, d- and q- axis current controller employing the
In this paper, the optimizations are performed through 80 proposed optimization procedure are given in (8), (9) and (10),
generations with each generation having 150 individuals. This respectively.
is an offline optimization. The duration of the whole
optimization process depends on the time needed to complete t
each simulation for each individual. Fig. 3 shows the timing IAE = e(t ) dt (7)
diagram when GA optimization is performed in each 0
simulation. The total simulation time is 20 seconds.
Since the speed and current control are to minimize K i _ speed 16.004
magnitude error, in this work, an evaluate function is based on Gspeed ( s ) = K p _ speed + = 0.195 + (8)
s s
an Integral of Absolute Error (IAE) as given in (7), where e are
the combined error between the command and feedback of
speed and dq-axis current. K i _ id 298.002
Gid ( s ) = K p _ id + = 0.492 + (9)
To enhance the efficiency of the optimization, it should be s s
noticed from Fig. 3. In each simulation, the optimization also
takes two measurements of stator current (Is) at the different K i _ iq 38.271
times: just before applied the load torque variation and stopped Giq ( s ) = K p _ iq + = 0.091 + (10)
s s
the simulation as another input to GA procedure. The
measured Is will be compared with the maximum current of the
IPMSM (Ismax). If the measured Is is higher than the Ismax, the
optimization process will consider this as a bad individual. V. SIMULATION RESULTS
This is to ensure that the current control loop works efficiently To demonstrate the effectiveness of the proposed speed
and is stable during the load torque variation. The program controller, the simulations are carried out by using
recursively runs a simulation reproducing the whole motor Matlab/Simulink. The IPMSM and system parameters are
drive control system, testing each of the individuals in the presented in Table I.

2932
Fig. 3. Timing diagram for optimization process.

Fig. 5. Simulation results of speed control response.


Fig. 4. Fitness of the best solution against iteration.

The simulation consists of two scenarios to verify the drive


performance of the proposed method. The first scenario shows
the speed transient behavior when the speed command is
stepped back to 1500 rpm to 1000 rpm at t= 4 s and then
stepped back to 1500 rpm at t= 8 s, while the load torque is
maintained at 0.1 N.m. The next scenario shows the torque
transient behavior when the load torque is stepped up from 0.1
to 0.3 N.m at t= 12 s and then stepped back to 0.1 N.m at t= 16
s, while the speed command is kept at 1500 rpm.
(a)
The proposed GA and the trial error method are compared in
term of speed response as shown in Fig. 5. In first scenario,
when zoomed in Fig. 6 at t= 4 and t= 8 s as shown in Fig. 6(a)
and Fig. 6(b), respectively, it can be seen that motor speed of
GA method has small overshoot, less than trial and error
method. Moreover, in the next scenario, the speed response
when stepped the load torque at t= 12 and 16 s. are expanded
the speed response in the transient response. Fig. 7(a) and Fig.
7(b) enlarge the transient response at t= 12 and 16 s,
respectively. It is clearly seen from the simulation results, the
improvement in drive performances can be achieved when
(b)
compared with the trial and error method in term of speed Fig. 6. (a) Simulation results of speed control response at t= 4 s. (b) Simulation
response, overshoot, and settling time. results of speed control response at t= 8 s.

2933
(a)

Fig. 8. Experimental results of speed and phase-a stator current response using
trial and error controller when stepped the motor speed.

(b)
Fig. 7. (a) Simulation results of speed control response at t= 12 s. (b)
Simulation results of speed control response at t= 16 s.

VI. EXPERIMENTAL RESULTS


In order to ensure the simulation results, the proposed drive Fig. 9. Experimental results of speed and phase-a stator current response using
system has been implemented by using 100 watts IPMSM GA-based controller when stepped the motor speed.
controlled by 32-bit fixed-point microcontroller
(STM32F103VE). The dc-bus voltage is 40 volt from the dc
power supply and then connected to three-phase inverter. The
PWM gate signals are generated by the space vector pulse
width modulation. The rotor position and stator currents are
measured by a quadratic encoder pulse (QEP) sensor and Hall-
effect current sensor, respectively.
To verify the effectiveness of the proposed method, Fig. 8
show the experiment results using trial and error controller
when stepped down the motor speed from 1500 to 1000 rpm
while the load torque is maintained at 0.1 N.m. It can be seen
that the motor speed (m) has precisely track the speed
command (m*) in steady-state and it has an undershoot
(14.19%). After that, when stepped back the motor speed from Fig. 10. Experimental results of speed and phase-a stator current response
1000 to 1500 rpm and load torque kept at 0.1 N.m, the using trial and error controller when stepped the load torque.
transient response of the motor speed has an overshoot
(14.61%) and accurately follow the motor speed command. In Figs. 10 and 11, the experiment results show the motor
However, to improve the performance of the drive system, the speed response when stepped load torque form 0.1 to 0.3 N.m
experimental results using GA-based controller is shown in and stepped back to 0.1 N.m while the motor speed kept at
Fig. 9, the motor speed has smaller undershoot (4.87%) when 1500 rpm. It is clearly seen that the motor speed has smaller
stepped down and lower overshoot (4.23%) when stepped up overshoot, undershoot and faster speed recovery (1 second) to
the speed command than trial and error controller. Moreover, the load variation than the trial and error controller. These
the settling time of the GA-based controller has much faster results confirm the drive performance from the GA-based
than the than trial and error controller. controller proposed method.

2934
VII. CONCLUSION
The use of GA for speed and current controller designs for
IPMSM drive is proposed. The optimization procedures have
been demonstrated thoroughly and an automated selection of
controller parameters is achieved with minimum control error
possible. The results obtained in the simulations and
experiments confirm the effectiveness of the design method
with accurate speed controls during both the transient and
steady-state conditions. With this technique, it can be
considered as a useful tool to automatically tune the controller
in particular for a high performance drive system.

Fig. 11. Experimental results of speed and phase-a stator current response ACKNOWLEDGMENT
using GA-based controller when stepped the load torque.
This work was supported by the Higher Education Research
Promotion and National Research University Project of
TABLE I Thailand, Office of the Higher Education Commission.
MOTOR AND SYSTEM PARAMETERS FOR SIMULATIONS

Parameter value
REFERENCES
Rating power (W) 100
[1] M. H. Rashid, Power electronics handbook, Academic Press, 2001
Rating torque (N.m) 0.318 [2] T. Do, S. Kwak, H. Choi, J. Jung, "Suboptimal Control Scheme Design
Rating voltage (V) 24 for Interior Permanent Magnet Synchronous Motors: An SDRE-Based
Approach," IEEE Trans. Power Electronics, pp. 3020 - 3031, July 2013
Number of pole pairs 4
[3] M. N. Uddin and R. S. Rebeiro "Online efficiency optimization of a
Direct inductance, Ld (mH) 1.17 fuzzy-logic-controller-based IPMSM drive", IEEE Trans. Ind.
Quadrature inductance, Lq (mH) 1.41 Appl., vol. 47, no. 2, pp.1043 -1050 2011
[4] R. S. Rebeiro and M. Nasir Uddin, "Performance of FLC Based Online
Permanent magnet flux linkage (V.s) 0.01442 Adaptation of Both Hysteresis and PI Controllers for IPMSM Drive",
Stator resistance, Rs () 0.79 IEEE Trans. on Ind. Appl., vol. 48, no. l, Jan'!Feb. 2012, pp. 12-19.
Motor inertia (g,cm2) 42 [5] F. Cupertino , E. Mininno , D. Naso , B. Turchiano and L.
Salvatore "On-line genetic design of anti-windup unstructured
GA generation gap 0.9 controllers for electric drives with variable load," IEEE Trans. Evol.
GA crossover rate 0.7 Comput., Vol. 8, no. 4, pp. 347 -364, 2004
GA mutation rate 0.5 [6] W. Lenwari, M. Sumner, and P. Zanchetta, The use of genetic
algorithms for the design of resonant compensators for active filters,
GA parameter ranges for PI speed Kp_speed 0 100 IEEE Trans. Ind. Electron., vol. 56, no. 8, pp. 2852 2861, August
controller optimization Ki_speed 0 400 2009
Kp_id 0 100 [7] Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution
GA parameter ranges for PI d-axis Programs, 1996 :Springer-Verlag
current controller optimization Ki_id 0 400
GA parameter ranges for PI q-axis Kp_iq 0 100
current controller optimization Ki_iq 0 400

2935

You might also like