Вы находитесь на странице: 1из 5

International Journal of Technical Research (IJTR)

Vol. 2, Issue 2, Jul.-Aug. 2013

DESIGN OF PID CONTROLLERS USING MULTIOBJECTIVE


OPTIMIZATION WITH GA AND WEIGHTED SUM OBJECTIVE
FUNCTION METHOD
Parvesh Kumar1, Jitesh Raheja2, Dr. Shiv Narayan3
1
EEE, 2ECE, 3EEE, 1.3PEC University of technology, 2OITM Hisar
1
raheja.parvesh@gmail.com, 2rahejajitesh@gmail.com, 3shivnarayan@pec.ac.in

AbstractThis paper treats a tuning of the PID controllers using The parallel controllers are mostly preferred for higher
the multi-objective optimization with GA with weighted sum order systems. The transfer function of PID controller in
approach. The design objective was to apply the GA in the aim of Laplace transform is defined for a continuous system as
tuning the optimum solution of the PID controllers (KP, KI and
KD) by minimizing the multi-objective function. The potential of KD s2 KPs KI
Gc (s)
using multi-objective GA is to identify the Pareto optimal s (1)
solution. The other methods are applied to make the comparisons
between a classical approach based on the Ziegler-Nichols The proportional controller response is proportional to the
method [1,2,3] and Ant Colony Optimization [4]. Simulation control error. The controller error is defined as the difference
results demonstrate that the new tuning method using multi- between the set point and the process output. The proportional
objective ant colony optimization has a better control system controller output is the multiplication of the system error
performance compared with the classic approach and the genetic signal and the proportional gain. Proportional term can be
algorithms. mathematically expressed as
Keywords PID, GA, ACO, Muti-objective optimization, Pterm K P e(t) (2)
Controllers.
The integral control applies a control signal to the system
I. INTRODUCTION which is proportional to the integral of the error. The offset
PID controllers are widely used in industrial applications in introduced by the proportional control is removed by the
different forms. It can be used as a stand-alone or as a part of integral action but a phase lag is added into the system.
direct digital control (DDC) and distributed control system Integral term can be mathematically expressed as
(DCS) Figure 1. is a simple block diagram shoeing the I term KI e(t)dt (3)
schematic of the PID controller and it is known as non- There is a proportion between the derivative controller output
interacting form or parallel form.
and the rate of change of the error. Derivative control is used
to decrease and eliminate overshoot of system response and
introduce a phase lead action that removes the phase lag
introduced by the integral action.
de(t) (4)
Dterm K D
dt

TABLE 1: PID characteristic parameters

Closed- Rise Overshoot Settling Steady


Loop Time Time State
Response Error
Increasing Fast Increase Small / Decrease
No
Figure 1. Block diagram of a PID controller KP Effect
Increasing Fast Increase Increase Decrease

ISSN 2278-5787 Page 52


International Journal of Technical Research (IJTR)
Vol. 2, Issue 2, Jul.-Aug. 2013

Ki H (s)
T (s) (8)
Increasing Small Decrease Decrease Small / 1 H (s)
/ No No
Kd Effect Effect III. WEIGHTED SUM OBJECTIVE FUNCTION METHOD
Since applying weighted objective functions is one of the
most popular approaches for solving multi-objective
optimization problems, and maybe the simplest one, it is
The control signal to the plant is given by
t
shortly described here. Usage of weighted objective functions
de(t) (5) means the following definition of the optimization problem
u(t) K P e(t) K I e( )d KD
0
dt [6,7]:
Each one of the three gains (KP, KI and KD) of the classical LA min( F ), (9)
PID control has an effect on the response of the closed loop T
system. Table 1 summarizes the effects of each of PID where F [ f 1 , f 2 , f 3, f 4 , f 5 , f 6 , f 7 , f 8 ] : vector of the
control parameters. It will be known that any changing of one objective functions, f 1 : settling time (ts) f 2 :overshoot (OS)
of the three gains will affect the characteristic of the system
response. f 3 :rise time f 4 : integral absolute error f 5 : integral square
error f 6 :sensitivity function f 7 :complementary sensitivity
function f 8 :undershoot. [ 1, 2 , 3 , 4 , 5 , 6 , 7 , 8 ]:
is the nonnegative weights.
i n


i 1
i 1 (10)

The goal of multi-objective optimization problems is to find


the best compromise between multiple and conflicting
objectives. Considering all objectives in these problems, there
will be more than one solution that optimizes simultaneously
all the objectives and there is no distinct superiority between
these solutions. Usually there is not a single best solution
Figure 2. Characteristic response
being better than the remainder with respect to every objective.
In figure 2 characteristic response of a plant is shown with the Therefore, we face with a set of solutions which are better
different parameters rise time, settling time, overshoot and than remainder solutions called the Pareto front. Among the
peak time. feasible solutions, solutions belonging to the Pareto front are
known as non-dominated solutions, while the remainder
II. DESIGN PROCEDURE AND OBJECTIVES solutions are known as dominated. Since none of the Pareto
There are many criteria in time domain and frequency set solutions is absolutely better than the other non-dominated
domain for designing and evaluating controllers. Currently, in solutions, all of them are equally acceptable as regards the
the first stage, many controllers are designed according to the satisfaction of all the objectives [6,7,8,9,10].
frequency criteria since they are usually worked out fast.
IV. S IMULATION AND RESULTS
In this section, we presented the numerical results to
improve the performance of the proposed solution algorithm.
All the computation is implemented with Matlab
programing and Simulink. In this study, we utilised three
examples in order to illustrate the efficiency of the proposed
Figure 3. the diagram of closed loop system algorithm. The design problems are posed as multi-objective
problems which are solved using weighted sum approach.
Transfer functions and performance objectives used in this
system are expressed as follows. [5] Example 1. This example is presented to show the designing
Open-loop transfer function method described above. The plant G1(s), G2(s) and G3(s)
H (s) C(s)G(s) (6) multi-objective is controlled by a PID controller C(s) [11].
Sensitivity function
1 G1 (s)
4.228 (11)
S(s) (7) (s 0.5)(s 2 1.64s 8.456)
1 H (s)
Complementary sensitivity function

ISSN 2278-5787 Page 53


International Journal of Technical Research (IJTR)
Vol. 2, Issue 2, Jul.-Aug. 2013
27
G2 (s) (12) 2

(s 1)(s 3) 3
1 1.5

G 3 (s) 2
e 0.5 s (13)
s 2s 1 1

Kd
Controller
0.5
KD s 2 KP s KI
Gc (s) (14)
s 0
2

where 1.5
2.5
3

1 2
K P [0, ] 0.5 1
1.5

0.5

K I [0, ] (15) Ki
0 0
Kp

K D [0, ] Figure 5. The convergence of controller parameter values of G1(s)


The tuning parameter of the controllers are coded into a
feasible solution using the proposed algorithm [12-16]. The Bode plots of loop transfer with the designed PID
We chose the following objective function: controller are shown in Figure 6. This provides gain margin
and phase margin of 9.86 and 101.05, respectively.
LA min( f
1 1 2 f2 3 f3 5 f5 8 f8 ) (16) Bode Diagram

The objective function here is f 1 : settling time (ts) 20


Magnitude Plot of example 1 case 1

f 2 :overshoot (OS) f 3 :rise time f 5 : integral square error Magnitude ( dB)


0

and f 8 :undershoot. -20

Therefore the vector of the weights is -40

[0.2 0.2 0 0.2 0 0 0 0 0.2] (17) -60

-80
RESULTS: Comparison of step responses using present -45
Phase Plot of example 1 case 1
design with Ziegler Nichols and Ant Colony Optimization as -90

in multi-objective and the convergence of the controller


Phase (deg)

parameters are shown for three examples given above. -135

-180
Step Response

1.4 -225
-1 0 1 2
Multi-objective optimization 10 10 Frequency (rad/s) 10 10

Ziegler Nichols
Ant Colony optimization
1.2
Figure 6. Bode plots of G1(s)
Step Response
1
1.4
Multi-objective optimization with GA
0.8 Ziegler Nichol's
Amplitude

1.2
Ant Colony Optimization

0.6
1

0.4
0.8
Amplitude

0.2
0.6

0 Time (sec6onds)
0 2 4 8 10 1 0.4

Figure 4. Comparison of step responses using present design, 0.2


Ziegler Nichols and Ant Colony Optimization as in multi-
objective of G1(s)
0
0 1 2 3 me (seconds) 5
Ti4 6 7 8 9

Figure 7. Comparison of step responses using present design,


Ziegler Nichols and Ant Colony Optimization as in multi-
objective of G2(s)

ISSN 2278-5787 Page 54


International Journal of Technical Research (IJTR)
Vol. 2, Issue 2, Jul.-Aug. 2013
2
2.5

2
1.5

1.5
1
Kd

Kd
1

0.5
0.5

0 0
2.5 2.5

2 2
3.5 3
1.5 3 1.5 2.5
2.5 2
1 2 1 1.5
1.5
0.5 1
0.5 1
0.5 0.5
0 0 0 0
Ki Ki
Kp Kp

Figure 8. The convergence of controller parameter values of G2(s) Figure 11. The convergence of controller parameter values of
G3(s)
The Bode plots of loop transfer with designed PID controller
are shown in Figure 9. This provides gain margin and phase The Bode plots of loop transfer with the designed PID
margin of 4.23 and 68.45, respectively. controller are shown in Figure 12. This provides gain margin
Bode Diagram
and phase margin of 2.61 and 67.56, respectively.
50
Magnitude Plot of example 2 case 1 20
Madnitude and Phase Plot of example 1 case 1
10

0 0
Magnitude (dB)

Magnitude (dB) -10

-20

-50
-30

-40

-50
-100
4
-45 -60 x 10
Phase Plot of example 2 case 1 0

-90
-0.576
Phase (deg)
Phase (deg)

-135
-1.152

-180
-1.728

-225
-2.304

-270
-1 0
Frequency (rad/s) 1 2 -2.88
10 10 10 10 -1 0 1 2 3
10 10 10 10 10

Figure 9. Bode plots of G2(s) Frequency (rad/s)

Comparison of Step Response of example 3 case 1


1.4 Figure 12. The convergence of controller parameter values of
Multi0objective optimization with GA
Ziegler Nichol's
G3(s)
Ant Colony Optimization
1.2
V. CONCLUSIONS
1
In this thesis, design methodology has been developed for
the design of PID controller using multi-objective
0.8
optimization. The controller design problems have been posed
Amplitude

as multi-objective optimization problems, which have been


solved using weighted sum objective function approaches
0.6
with genetic algorithm (GA). Three examples have been
worked out to show the efficacy of the design methodology.
0.4
The design methodology is developed is general in nature and
permits to include any number of objective functions. The GA
0.2
has been found to solve such multi-objective optimization
problems efficiently.
0
0 2 4 6 8 10 12 14
Time (seconds)
REFERENCES
Figure 10. Comparison of step responses using present design,
[1] K. Astrom and T. Hagglund, PID Controller: Theory, Design and
Ziegler Nichols and Ant Colony Optimization as in multi- Tuning, 2nd ed, Library of Congress Cataloging-in-Publication Data,
objective of G3(s) 1994, pp.120-134.
[2] B. C. Kuo, Automatic control system, 7th ed., Prentice-Hall of India
PVT, 2003.

ISSN 2278-5787 Page 55


International Journal of Technical Research (IJTR)
Vol. 2, Issue 2, Jul.-Aug. 2013
[3] C. C. Hang, K. J. Astrom, and W. K. Ho, "Refinements of the [10] A. Afshar, A. Kaveh, and O. R. Shoghli, Multi-objective optimization
Ziegler&ndash, Nichols tuning formula", Proc. Inst. Elect. of time-cost-quality using multi-colony ant algorithm, Asian Journal
Eng.&mdash, Part D: Control Theory and Applications, vol. 138, no. Of Civil Engineering, vol. 8, no. 2, pp. 113124, 2007.
2, pp.111 -118 1991 [11] I. Chiha, N. Liouane and P. Borne, Multi-objective Ant Colony
[4] W. J. Gutjahr, "ACO algorithms with guaranteed convergence to the Optimization to tuning PID controller, Applied Computational
optimal solution", Information Processing Letters, vol. 82, no. Intelligence and Soft Computing 2012, Article ID 536326, 7 pages.
3, pp.145 -153 2002 (2002)
[5] H. Liang, W. Liang and Z. Jin-Hui, Multi-objective optimization for [12] A. Narang, S.L. Shah and T. Chen, Tuning of fractional PI controllers
controller design, Acta Automatic Sinica, vol.34, No.4 April, 2008. for fractional order system models with and without time delays,
[6] M. Nasri, H. Nezanabadi-Pour and M. Maghfoori, A PSO-Based American Control Conference, pp. 6674-6679, 2010.
optimum design of PID controller for a linear brushless DC motor, [13] H.T Toivonen and S. Totterman, Design of fixed-structure controllers
Proc. of World Academy of Science, Engineering and Technology 2007, with frequency domain criteria: a multiobjective optimization
vol. 26, pp. 211-215. approach, IEE Proc. on Control Theory and Applications, 153(1), pp.
[7] C.M. Fonseca and P.J. Fleming, Multiobjective optimization and 4652, 2006.
multiple constraint handling with evolutionary algorithmspart I: A [14] T. Ota, and S. Omatu, Tuning of the PID gains by GA, IEEE, Osaka
unified formulation, IEEE Trans. Syst., Man, Cybern, pp. 2637. Japan, pp. 272-274, 1996.
[8] I.K. Kookos, K.G. Arvanitis and G. Kalogeropoulos, PI Controller [15] A. G. Dupuis, M. Ghribi, and A. Kaddouri, Genetic Tuning of PID
Tuning via Multiobjective Optimization, Proc. of the 7th MED99, pp. Controllers, Proceeding of the IASTED International Conference:
408-419 Artificial Intelligence & Applications, Marbella, Spain, pp. 202-207,
[9] K. Deb, A. Paratap, S. Agarwal, and T. Meyarivan, A Fast and Elitist 2001.
Multi-objective Genetic Algorithm: NSGA-II, IEEE Trans. [16] C.M. Fonseca and P.J. Fleming, Genetic Algorithm for multi-
Evolutionary Computation, no. 2, pp. 182-197, 2002. objective Optimization: Formulation, Discussion and Generalization,
Automat. Contr. Sys. Eng.

ISSN 2278-5787 Page 56

Вам также может понравиться