Вы находитесь на странице: 1из 11

DocumentToPDF trial version, to remove this mark, please register this software.

Fuzzy Logic Speed Control of Vector Controlled Induction


Motor Drives
Anil Kumar Yaramasu and Naga venkata siva .k
Rao & Naidu Engineering College, Ongole
E-mail: anilkumar_666@yahoo.co.in

Abstract: This paper proposes the practical implementation of a fuzzy logic control
scheme applied to a two phase (d-q) current model of an induction motor. Most
conventional fuzzy controller uses fuzzy logic to tune a PI or PID. The proposed fuzzy logic
control scheme directly transforms the inputs, speed errors and rate of change in speed
errors, into the output quantity, the change in inverter command frequency. The proposed
approach is simpler in implementation and its trajectory tracking capability was
investigated. The simulation results show that it has a superior performance compared to a
fuzzy tuned PI controller or a simple PID controller.
Keywords: Fuzzy systems, nonlinear systems, Electric machines and drives, intelligent control,
Analysis and design.

(I). INTRODUCTION
The most important aspect of the vector control of induction motors is the
transformation of the stator currents in to a torque producing component and a flux producing
component. To enable the flux producing current component to align with the rotor magnetic
flux, the accurate estimation of a transformation parameter called the unit vector is required [1].
However, if this unit vector can be correctly determined, then the ac drive performance will
depend on the effectiveness in producing the appropriate torque command. Some researchers
have already successfully implemented fuzzy controllers to vector control drives [2]. Some use
an error driven discrete time set-point controller, other prefer to use the fuzzy logic to tune the PI
constants of a proportional-integral torque controller.
The de facto method in the conventional vector drive is to use the
Proportional-Integral (PI) controller to generate the torque command [3]. In recent years, fuzzy
logic has been successfully applied in many control applications [4]. Fuzzy logic controller has
been shown to be insensitive to external disturbance and small unknown or erroneous
information. A conventional PI controller requires accurate sensor inputs and appropriate values
of the PI constants to produce high performance drive. Therefore the unexpected change in load
conditions or environmental factors would deteriorate the drive performance. In contrast, fuzzy
logic controllers use heuristic input-output relations to deal with vague and complex situations.

1
DocumentToPDF trial version, to remove this mark, please register this software.

The sensor input needs not to be very precise and theoretically it is more robust to fluctuations in
operating conditions. Hence fuzzy controller offers the benefits of low cost and higher
reliability.

(II). VECTOR CONTROL OF INDUCTION MOTORS


Vector control in induction motor is about entirely orienting the rotor flux in the direct-axes [5].
This is carried out by using a unit vector to transform the command current in the rotating d-q
axis to the stationary d-q axis. Mathematically, the transformation is made using the following
equations (1).

(1)

Where, id, iq commands referred to the rotating d-q axis,


id s &iq s is stator currents referred to the stationary d-q axis, and ?´is the rotor flux position
angle. To achieve a dynamic performance equivalent to a separately excited dc motor, the flux
command ids (analogous to the flux current in the dc motor) is usually kept constant and the
torque current command iqs is varied according to the speed signal command. In addition, for
flux weakening Operations Purposes at higher speeds, the current limit of the motor have to be
taken into considerations. Taking the effective values of the currents,

(2)

where, Im is the maximum current per phase in the motor and Iqs is the torque current command
obtained from the fuzzy controller. The current command in the real three phase abc domain can
then be calculated using equations (3)

(3)

To drive the motor, a suitable inverter with the capability of tracking the command current can
then be used to bring the motor currents to the command values so as to produce the desired
: the direct vector control
results. There are two ways to determine the rotor flux position angle ?H
uses flux sensors or motor input currents/voltages to synthesize ?ú and the indirect field
orientation uses the rotor slip to calculate ?°. In this paper, the indirect vector control is adopted.
For a perfectly tuned vector control, the rotor flux in the q axes is zero and the rotor flux in the
d-axes can be expressed in a differential equation (4).

2
DocumentToPDF trial version, to remove this mark, please register this software.

(4)

where T, is the rotor time constant, ?õ


dr is the rotor flux in the d axes, and M = (3/2) x stator
magnetizing inductance. From (2), the value of, ?P
dr can be calculated and the torque developed
in the motor can be derived from:

(5)

where, P is the number of poles in the machine. From the motor dynamics, the relation between
the rotor speed Wr, and the torque will be:

(6)

where J is the inertia of the rotor and TI is the load torque.

(III). THE FUZZY LOGIC CONTROLLER


The block diagram of the controller used in this paper is illustrated in Fig. 1.

Fig 1: The proposed Speed Controller.


A. The Fuzzification Process:
The proposed fuzzy logic controller is basically driven by the normalized speed error e (k) (the
difference between the desired speed and the rotor speed) and de (k) (the change of e (k)) [2] i.e.

(7)

(8)

where Wr *(k) is the desired speed and Wr (k) is the rotor speed The output of the fuzzy
controller decides the value of ?ºiqs which is the increase or decrease in the torque current

3
DocumentToPDF trial version, to remove this mark, please register this software.

command iqs. Therefore the fuzzy logic controller can be viewed as a nonlinear function
generator that maps e (k) and de (k) into ?niqs using fuzzy relation rules.

Fig. 2 The input fuzzy membership functions


The next step is to decide the appropriate shape of the membership functions for e (k) and de (k).
Since it is easier to see the influence of e (k) on the value of ?¾
iqs as compared to de (k), five
fuzzy sets are assigned to e (k) and only three fuzzy sets are assigned to de (k) in their respective
universe of discourse. More fuzzy sets in e (k) will lead to higher precision in this input space.
Consequently, if the number of fuzzy sets in a particular universe of discourse in increased to
infinity, then all fuzziness will be lost and it will be equivalent to a conventional input domain.
To simplify mathematical computations, the shape of the fuzzy sets on the two extreme ends of
the respective universe of discourse is taken as trapezoidal whereas all other intermediate fuzzy
sets are triangular. This is illustrated in Fig. 2. The most widely used fuzzifier, the singleton
fuzzifier is not used because the fuzzy singleton which has a membership value of 1 at a certain
input value and zero elsewhere may not be adequate to handle input uncertainties. Such
uncertainties occur when the input data is corrupted by measurement noise or when the input
sensors are time varying nonlinear devices. Note that the membership functions have
asymmetrical shape with more crowding near the origin. This permits higher precision at steady
state. The Zero membership function is a triangle with narrow base which will reduces fuzziness
and increases the precision near the zero error state. Mathematically, the grade of membership of
an input e (k) in the various input fuzzy set can be computed as follows:

(9)

4
DocumentToPDF trial version, to remove this mark, please register this software.

Similarly, for de@), the grades of memberships are as follows:

(10)

B. The Fuzzy Rules


The universe of discourse for the controller output contains 5 singleton fuzzy sets as shown in
Fig. 3. Singletons contributes to a degree of membership of unity at certain predetermined
values and zero otherwise is lack of overlapping between adjacent sets which reduces its degree
of fuzziness and thus is less capable of handling imprecise inputs. The fuzzy rules matches the
fuzzified inputs obtained from equations (9) and (1 0) to these output fuzzy sets to produce some
sort of fuzzy outputs. Based on the above concepts, the rule table [6], which contains 15 rules, is
as shown in table 1. Each cell from row 2 and column 2 onwards represents a fuzzy if-then rule.
Atypical rule reads as follows:
Rule 1: If e (k) is NB and de (k) is N then the controller output should be NB.

Fig. 3 The Singleton Output Fuzzy Sets


All the proposed rules can be combined in the following fuzzy associative matrix (FAM) table.

Table 1: The Proposed Fuzzy Associative Matrix


Mathematically, rule 1 can be computed as:
(Minimum operation) ---------- (11)

(Product operation) ---------- (12)


In this paper, the minimum operation and the product operation are used in separate cases to
compare the suitability of each type of operation in applying to the vector control of an
induction motor. The 15 rules indicated above can be further reduced to 9 rules by combining all

5
DocumentToPDF trial version, to remove this mark, please register this software.

the rules in column 2 to a single rule and all rules in the last column into another single rule as
shown below:
If e (k) is NB then the controller output is NB, i.e.

(13)
If e (k) is PB then the controller output is PB, i.e.

(14)
.
Each rule will produce an output fuzzy value. Through the defuzzification process all these
output fuzzy values combine all the output fuzzy values to produce a discrete single value
output.

C. The Defuzzification Process


Many defuzzifiers have been presented in fuzzy logic literature [7]; however, there are no
scientific or mathematical bases for the preference of any of them. Consequently, defuzzification
is considered as an art rather than simplicity. The most popular defuzzification method is the
centroid method where

(15)

The centroid defuzzifier can be interpreted as a conditional expectation in probability


distribution. However it since singleton output sets are used, a very simple defuzzification using
the computed average moment is used.
Mathematically, it can be represented as:

The deffuzzified output is used to control the torque current.

(17)

(IV). SIMULINK IMPLEMENTATION


A. Fuzzy Controller Implementation
The defuzzified output from the fuzzy algorithm calculations are used to determine the change
f of the inverter and the change of command torque current
of command switching frequency ?—
?út -Mathematically, it can be represented as:

6
DocumentToPDF trial version, to remove this mark, please register this software.

(18)

(19)
where ?ýf (t) = Kf * Ymean, Kf is a suitable defuzzfiing constant to convert the final deftizzified
output from the fuzzy controller Ymean to the change of frequency. This value is obtained by
trial and error in this simulation. And ?±iqs (t) = ki * Ymean, ki is a suitable defuzzifying constant
to convert the final defuzzified output from the fuzzy controller Ymean to the change of torque
current command. The torque current command determines the pulse width of the inverter
output pulses. The value of ki is also obtained by trial and error. As mentioned earlier, to achieve
a dynamic performance that is equivalent to a separately excited dc motor, the flux command ids
has to be kept constant and the torque current command iqs is varied according to the speed
signal command. Therefore, in our simulation, the flux command current ids are fixed at 40% of
the current limit for normal operations.
B. The Inverter Implementation
In this paper, a simple approach is adopted. The generation of the PWM switching waveforms is
implemented using a closed loop PWM with speed error feedback. The speed error determines
the command frequency to the inverter and the voltage output to the motor must be controlled in
synchronism with the command frequency. In addition, to maintain the optimum flux level in
the machine, the concept of constant volt/hertz has to be applied. Since the motor is rated at
200V and 50 Hz, the optimum volt/hertz ratio per phase is 200/ (50*v£3). To keep the PWM
scheme as simple as possible, a single pulse per half cycle is used. The pulse width of this pulse
in relation to the half cycle time interval (the duty cycle) then determines the voltage output.
C. Simulation Results &Analysis:
The command speed against time curve takes a trapezoidal shape. This is to ensure that the ramp
up (region I), the steady speed (region II) and the ramp down (region III) characteristics are
investigated and compared. The motor speed against time curves for the fuzzy tuned PI
controller, the error driven fuzzy logic controller and the simple PID controller (where the
Proportional constant, the Integral constant and the derivative constant are arbitrarily selected)
are obtained. The performance curves for a steady speed of 400 rpm and 1000 rpm are shown in
Fig. 5. From the simulation results shown in Fig. 5, for the steady speed of 400 rpm the fuzzy
tuned PI controlled motor could follow the command speed quite closely on ramp up, shows
slight overshoot when reaching the constant speed region and slight oscillations in region I1 and
did not follow the command speed well, especially when the command speed is almost zero in
region -III. The same controller demonstrated similar results when the constant speed is raised to
600 rpm (Fig. 6) except that the amplitude of the initial overshoot when reaching the desired

7
DocumentToPDF trial version, to remove this mark, please register this software.

constant speed was reduced considerably. For the error driven fuzzy logic controller, the ramp
up stage and the ramp down stage shows that the motor speed is able to follow the command
speed more closely compared to the PI tuned fuzzy logic controller for both the 400 rpm and the
1000 rpm constant running speed. The oscillations at region II were also reduced except that the
motor speed showed a considerable lag near the transition point between region I and region II.
The remarkable lagging characteristic shown by the fuzzy PI tuned controlled motor at the end
of region III where the command speed dropped to almost zero speed was not present in the
error driven fuzzy logic controlled motor. The simulation results verifies that the error-driven
fuzzy logic controller generally produces less cycles of oscillations after reaching the steady
command speed and lower magnitude of overshoots and undershoots. The PID controlled drive
showed slightly inferior characteristics compared to the other two fuzzy drives if the steady state
command speed is low (400 rpm) except that the large amount of lag between the actual motor
speed and the command speed that was found in the fuzzy tuned PI controllers is not present
here. When the steady state command speed is high (1000 rpm), the ramp up and ramp down
performance of the PID controlled drive are superior to the other two fuzzy logic controllers.
However, the oscillations at region I1 are worse both in terms of the number of cycles and the
magnitude of deviations from the command speed. This is due to the fact that the P and I
constants adopted arbitrarily are close to the optimum values only for the ramp up and ramp
down regions at 1000 rpm steady speed. At different operating conditions; the optimum values
of the P and I constants are different and hence it needs continuous and meticulous tuning if the
optimum performance is to be maintained.

V. CONCLUSION
The Simulation results show that the error-driven fuzzy controller has shown the
best overall performance, surpassing the other controllers. This may be due to the fact that both
fuzzy-tuned PI controller and the simple PI controller depend on the proper choice of the PI
constants. Fuzzy tuned PI controller has to search for the optimum PI constants which are not
required in the case of a fuzzy controller driven by an output error. A simple PI or PID controller
with no proper choice of the PI constant will not be able to produce the desired results.
Furthermore, a good choice of the PI or PD) constants for one particular operating condition
most likely will not be optimum when the operating condition changes.

8
DocumentToPDF trial version, to remove this mark, please register this software.

SPEED Vs TIME CURVE


Fuzzy tuned PI controller for 400 rpm & 90 secs

SPEED Vs TIME CURVE


Fuzzy Error driven controller for 400 rpm & 90 secs

SPEED Vs TIME CURVE


PID controller for 400 rpm & 90 secs

Fig. 5 Performance curve for fuzzy controlled and conventional PID drives (I)

9
DocumentToPDF trial version, to remove this mark, please register this software.

SPEED Vs TIME CURVE


Fuzzy tuned PI controller for1000 rpm & 90 secs

SPEED Vs TIME CURVE


Fuzzy Error driven controller for 1000 rpm & 90 secs

SPEED Vs TIME CURVE


PID controller for 1000 rpm & 90 secs

Fig.6 Performance curve for fuzzy controlled and conventional PID drives (II)

10
DocumentToPDF trial version, to remove this mark, please register this software.

(VI). REFERENCES

[l] W. P. Hew, M.R. Tamjis and S. M. Saddique, “An adaptive neural network for the estimation

of torque and air gap flux signals in the direct field oriented control of an induction motor”,

Proc. of the seventh international conference on Electrical Machines and Drives, pp. 306-309,

Durham ,UK ,11-13 Sept. 1995.

[2] W. P. Hew, M.R. Tamjis and S. M. Saddique, “Application of Fuzzy Logic in the Speed

Control of Induction Motor Vector Control”, Proc. of the international conference on Robotics,

Vision and Parallel Processing for Industrial Automation, pp. 767-772, Ipoh, Malaysia, Nov.

28-30, 1996.

[3] W.P. Hew, MR. Tamjis and S.M. Saddique, “Application of fuzzy logic in the speed control

of induction motor vector drive”, Proc. of the international conference on Robotics, Vision and

Parallel Processing for Industrial Automation, pp. 767-772, Ipoh, Malaysia, Nov

[4] Lee C. C., “Fuzzy Logic in Control Systems: Fuzzy Controllers Part I, Part II”, IEEE Trans.

On Syst. Man. and Cyber. Vol2029, pp. 404-435, 1990.

[5] P. C. Sen, “Electric Motor Drives and Control: Past, Present and Future”, IEEE Trans. on

Industrial Electronics”, Vol. 37, No. 6, pp. 565-575, December 1990.

[6] Dubois D. and H. Prade, ‘Fuzzy Sets and Systems: Theory and Applications”, Academic

New York, 1980.

[7] Hellendoom H. and C. Thomas, “Defuzzifications in Fuzzy Controllers”, Intelligence and

Fuzzy Systems, Vol. 1, 28-30 1996. pp. 109-123, 1993.

11

Вам также может понравиться