Professional Documents
Culture Documents
14, 2011
Abstract
According to the high performance demand of speed control of an induction
motor, Fuzzy Logic Controller (FLC) gives superior behavior over wide range of
speed variation. Fuzzy logic is a robust controller for linear and non-linear system,
even if good mathematical representation of the system is not available. But, adapting
fuzzy controller parameters is very complex and depends on operator experience.
Particle Swarm Optimization (PSO) algorithm is proposed in this paper as an
optimization technique for adapting centers and width of triangle inputs membership
functions. The ordinary adapting method of FLC is represented too. Meanwhile,
based on the concept of optimization, ways of defining the fitness function of the
PSO including different performance criteria are also illustrated. The complete
mathematical model and simulation of an induction motor and inverter are
represented in this paper. The simulation results demonstrate that the proposed PSO-
FL speed controller realizes a good dynamic behavior of the I.M with very good
speed tracking.
.
.
.
.
.
.
.
1. Introduction
The modern heuristics algorithms; it
T
he design of a conventional
was developed through simulation of a
control system is normally based
simplified social system, and has been
on the mathematical model of a
found to be robust in solving
plant. If an accurate mathematical model
continuous non-linear optimization
is available with known parameters, it
problems [4, 5].
can be analyzed, for example, by a Bode
In this paper, a tradition method for
or Nyquist plot, and a controller can be
adapting fuzzy logic controller of non-
designed for the specified performance.
linear three phase induction motor
Such procedures are tedious and time-
speed control is represented. Then, the
consuming, and in addition of that for
PSO based method for adapting fuzzy
complex or non-linear system are
controller parameters are proposed as a
difficult to obtain good mathematical
modern intelligent optimization
model [1].
algorithm.
Fuzzy logic (FL) or Fuzzy
Also, this paper represents the
Inference System (FIS) was invented in
Sinusoidal Pulse Width Modulation
1970s, which was powerful tool in the
(SPWM) inverter as a voltage fed
field of control applications. The fuzzy
inverter. Which is used as volt/Hz
controller gives superior performance
control drive.
over the conventional control methods
2. Sinusoidal PWM Inverter
such as PID controller, even if the
accurate mathematical model is not The pulse width modulation
available [1, 2]. techniques have wide simple and
Unfortunately, adapting and choosing complex types, the Sinusoidal Pulse
of the fuzzy controller parameters such Width Modulation (SPWM) is the most
as: number of membership functions, popular used method of A.C. drives,
type of membership functions, number but it's not the efficient method [1, 6].
of rules, and formulating rules, are Voltage source inverter (VSI) should
strongly depend on the experience of the have a stiff source at the input [1], that
operator and designer. In spite of is, its Thevenin impedance ideally is
experience, adapting position and width zero. Thus, a large capacitor can be
of the input and output memberships are connected at the input if the voltage
very difficult and dependant parameters source is not stiff. A practical (VSI)
[2, 3]. consist of power bridge devices with
Nowadays, several new intelligent three output legs, each consist of two
optimization techniques have been power switches and two freewheeling
emerged like: Genetic Algorithm (GA), diodes, the inverter is supplied from
Ant Colony Optimization (ACO), D.C. voltage source via LC or C filter.
Simulated Annealing (SA), Bacterial In sinusoidal PWM the three output
Foraging (BF), and Particle Swarm legs considered as three independent
Optimization (PSO) [3, 4]. Due to its push-pull amplifiers [6] as shown in
high potential for global optimization, figure (1). The gating signals of each
PSO has received great attention in push-pull stage generated by
control system such as the search of comparing a constant level triangle
optimal PID controller and fuzzy signal of frequency ( fc ) called "carrier
controller parameters. PSO is one of signal", with 3-phase sinusoidal signals
2913
2914
current and flux) to twoaxis frame It's obviously that in squirrel cage
called (dq-axis stationary frame or park I.M Vqdr=0, then the pervious equation
transformation). Park transformation is can be rewritten:
applied to refer the stator variables to a d qs
= Vqs Rs i qs we ds ....( 8 )
e e
synchronously rotating reference frame
fixed in the rotor, by such
dt
transformation the stator and rotor
dds
e
parameters rotate in synchronous speed
= Vds Rs i ds + we qs ....( 9)
e e
and all simulated variables in the dt
stationary frame appear as D.C.
quantities in the synchronously rotating dqr
= Rr iqr ( we wr ) dr ....(10)
e
reference frame [1, 6].
The per-phase equivalent circuit dt
diagrams of an I.M. in two- axis
synchronously rotating reference frame ddr
= Rr i dr + ( we wr ) qr ....(11)
e
are illustrated in figure (5). From the dt
circuit diagram the following equations
can be written [1]: The development torque by interaction
Stator equation: of air gap flux and rotor current can be
dqs found as:
Vqs = Rs iqs + + we ds ...............(4)
e e
dt
Te= (3/2)(P/2) ..(12)
dds Where: Te is the electromagnetic
Vds e = Rs ids e + we qs ..............(5) torque.
dt By resolving the variables into de-qe
components:
Rotor equation:
Te= (3/2)(P/2) (dsiqs qsids )
e e
dqr
Vqr = Rr iqr + + (we wr )dr ...(6)
e e
dt
d ds qm
i ds =
e
= Rr idr + dr (we wr )qr ...(7) .....( 15 )
e e
Vdr Ls
dt
.....(13)
2915
2916
(MFs) for the input and output , Rr= 3.4 , rotor inertia= 4.5*10-3
variables and the determination of kg/m2.
fuzzy rules are not always easy. There 8. Particle Swarm Optimization
is no formal framework for the choice PSO is one of the optimization
of the parameters of FLC and hence the techniques first proposed by Eberhart
means of tuning them and learning and Colleagues [8, 9]. This method has
models in general have become an been found to be robust in solving
important subject of fuzzy control. problems featuring non-linearity and
7. Design of FL Controller: non-differentiability, which is derived
from the social-psychological theory.
The function of the fuzzy controller
The technique is derived from research
is to observe the pattern of the speed
on swarm such as fish schooling and
loop error signal and correspondingly
bird flocking. In the PSO algorithm,
updates the control signal so that the
instead of using evolutionary operators
actual speed (wr ) matches the such as mutation and crossover to
*
command speed ( wr ) . There are two manipulate algorithms, the population
input signals to the fuzzy controller: the dynamics simulates a "bird flocks"
* behavior, where social sharing of
error E = wr - wr , and the change of information takes place and individuals
error CE (also it known as the future of can profit from the discoveries and
error), which is related to the derivative previous experience of all the other
of error dE/dt. A simple fuzzy logic companions during the search for food.
controller of two inputs and one output Thus, each companion, called particle,
can be used in this application, for each in the population, which is called
input three of triangle memberships swarm, is assumed to " fly " in many
was used, and five triangle direction over the search space in order
memberships for the output. For such to meet the demand fitness function [4,
FLC nine (3*3) of "If" statement rules 5, 9, 10].
was used. Adapting of the input/output
memberships positions are very For n-variables optimization
problem a flock of particles are put into
difficult, especially for one don't have
the n-dimensional search space with
experience of the system behavior.
Trial and error method can be used in randomly chosen velocities and
positions knowing their best values, so
such situations, in this work the final
far (P best) and the best position in the n-
suitable distribution of the input/output
memberships after several trial and dimensional space [8, 10]. The velocity
error iterations can be shown in figure of each particle, adjusted accordingly
to its own flying experience and the
(7). The overall system simulation is
other particles flying experience. For
illustrated in figure (8). System
performance of different command example, the ith particle is represented,
as:,
speed, with fuzzy logic controller,
.. (21)
under full-load condition is shown in
figure (9).
Where, the used squirrel cage In n-dimensional space, the best
induction motor name plate is: previous position of the ith particle is
3-ph I.M, 380 v., 2.2 kw, 2 poles, 50 recorded as:
Hz, Ls= 13.6 mH, Lr= 11.4 mH, Rs=2.3
2917
2918
10. Conclusions
A comparison performance between
..(25) the proposed PSO method and the
ordinary adapting method of the fuzzy
..(26) controller membership functions is
illustrated in figure (13), from which
..(27) the following tips can be concluded:
i. The proposed PSO adapting method
In this paper, the integrated of time can be considered as: very
weight square error ITSE is used for computationally efficient in
evaluating the accuracy performance of eliminate computing time, easy to
the fuzzy controller. A set of good implement, and simple concept.
control parameters can yield a good Unlike, the ordinary adapting
step response that will result in method which needs long adapting
performance criteria minimization in time and complex procedure.
the time domain, this performance ii. In the ordinary adapting method, the
criterion is called Fitness Function (FF) fuzzy controller gives a satisfactory
which can be formulated as follows [5, system performance: max. overshot
8]: = 4% in low seed & 0% in high
speed, rise time = 6 sec, settling
time = 6.5 sec, and steady state error
..(28) = 0%.
iii. The proposed PSO adapting method
Where: gives a superior performance, in
MP is maximum overshot. which: max. overshot = 0% in low
and high speed variation, rise time =
Ess is steady state error. 3 sec, settling time = 3 sec, and
is the weight factor can set to steady state error = 0%.
be larger than 0.7 to reduce the iv. The proposed PSO adapting method
overshot and steady state error, also gives very good speed tracking with
can be smaller than 0.7 to reduce the non-sluggish performance and high
rise time and settling time. robust controller than those obtained
by the ordinary adapting of fuzzy
The objective function is to logic controllers.
minimize the fitness function FF. the 11. References
PSO algorithm process can be
summarized in the flowchart shown in [1] Bimal K. Bose, Modern Power
figure (10). And the final obtained Electronic and AC Drives , Prentice
positions of the membership functions Hall, 2002.
from the particle swarm optimization [2] K. M. Passino and S. Yurkovich,
algorithm are illustrated in figure (11). "Fuzzy control", Addison
System performance of PSO-based for Wesley Longman, 1998.
different command speed is shown in [3] T. Senjyu, A. Miyazato and K.
figure (12). Uezato, " Enhancement of
Transient Stability of Multi-
Machine Power Systems by Using
Fuzzy-Genetic Controller",
2919
2920
2921
0.5
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
0.5
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Output MFs
1
0.5
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
2922
Command Speed
1
0.8
Speed (p.u)
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18 20
Time (sec)
2923
Start
If
No Max iteration
number reached?
Yes
End
2924
0.5
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
0.5
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Output MFs
1
0.5
0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
1
Command Speed
0.8
Speed (p.u)
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18 20
Time (sec)
Comparison Performance
1.2
0.8
Speed(p.u)
0.4
0.2
0
0 2 4 6 8 10 12 14 16 18 20
Time (sec)
2925