Abstract: In this paper, a novel design method for self-tuning PID controller of in mechanisms system using the particle

swarm optimization (PSO) algorithm is presented. This paper demonstrated in detail how to employ the PSO to search efficiently the optimal PID controller parameters of in mechanisms system. The proposed approach had superior features, including easy implementation, stable convergence characteristic, and good computational efficiency. Fast tuning of optimum PID controller parameters yields high-quality solution. Using the PSO approach, the initial PID parameters under normal operating condition can be found out. By the same way, the best parameters of PID controller under full-load condition can be found, too. The proposed self-tuning PID controller will automatically tune its parameters under these ranges. Moreover, the PC-based controller is implemented to control the position of the motor mechanism coupling system. The simulation and experimental results will show the potential of the proposed controller. Indexing terms: mechanisms system, self-tuning PID

controller, particle swarm optimization, optimal control. I. Introduction

uring the past decades, the process control techniques in the industry have made great advances. Numerous control methods such as adaptive control, neural control, and fuzzy control have been studied [1-5]. Among them, the best known is the proportional integral derivative (PID) controller, which has been widely used in the industry because of its simple structure and roust performance in a wide range of operating conditions. Unfortunately, it has been quite difficult to tune properly the gains of PID controllers because many industrial plants are often burdened with problems such as high order, time delays, and nonlinearities [1-6]. Over the years, several heuristic methods have been proposed for the tuning of PID controllers. The first method used the classical tuning rules proposed by Ziegler and Nichols. In general, it is

often hard to determine optimal or near optimal PID parameters with the Ziegler-Nichols formula in many industrial plants [1-3].
For these reasons, it is highly desirable to increase the capabilities of PID controllers by adding new features. Many artificial intelligence (AI) techniques have been employed to improve the controller performances for a

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

Design of Self-Tuning PID Control in a Mechanisms System

Rong-Fong Fung and 2Chih-Cheng Kao 1 Department of Mechanical and Automation Engineering National Kaohsiung First University of Science and Technology 2 Department of Electrical Engineering Kao Yuan Institute of Technology
wide range of plants while retaining their basic characteristics. AI techniques such as neural network, fuzzy system, and neural-fuzzy logic have been widely applied to proper tuning of PID controller parameters [1-2]. Particle swarm optimization (PSO), first introduced by Kennedy and Eberhart, is one of the modern heuristic algorithms. The PSO technique can generate a high-quality solution within shorter calculation time and stable convergence characteristic than other stochastic methods [7-9]. Much research is still in progress for proving the potential of the PSO in solving complex power system operation problems. Because the PSO method is an excellent optimization methodology and a promising approach for solving the self-tuning PID controller parameters problem. This controller is called the PSO self-tuning PID controller. The PID method is the most popular controller up to now. Despite the progression of many control theories, the PID controller is still the majority of industrial processes [11-13]. Due to the easily understanding of the physical sense for parameters of PID controller, engineers used to apply it to practical objects. However, the PID controller is not robust to wide parameter varying and large external disturbance. Especially for the highly coupling nonlinear system, the PID controller is lack of adaptive capability. Usually, the parameters of PID controller are manually tuned under ideal condition, that is the operating point without load. However, these parameters are mostly not suitable for the condition with full load. To achieve practical requirement, engineers have to adjust the parameters under different operating conditions. However, the robustness is limited with a small range. A rule to overcome this disadvantage is called self-tuning rule. Many researches and reports of self-tuning PID have been published. The parameter tuning at any time instance is usually based on a structurally fixed mathematical model produced by on-line identification procedure [14-15]. Unfortunately, recent plants are mostly difficult to obtained their fixed mathematical models. This paper proposed an intelligent self-tuning PID controller. Engineers will easily accept the straightforward design procedure. At the same time, the robustness will be expended to large range. In this paper, a practical high-order mechanisms system with a PID controller is adopted to test the performance of the proposed PSO self-tuning PID controller. This paper proposes a self-tuning PID control method to the position control of slider-crank mechanism.

~ 227 ~

2.1

PM Synchronous A model of a PM synchronous motor can be simplified to the following block diagram.

Usually, the PM synchronous motor is coupled with a gear speed reducer with a gear ratio of n . Hence, the applied torque can be described as

where

and viscous damping ration, respectively. 2.2 Slider Crank Mechanism In this section, Hamilton's principle and Lagrange multiplier are used to derive the differential equation for the slider-crank mechanism. The slider crank mechanism system is shown in Fig. 2 [16].

The slider-crank mechanism consists of three parts: crank, rod and slider. This section establishes dynamic model of the slider-crank mechanisms system using Largange method. Through reduction and incorporation, we can derive the dynamic equation is [16]:

where
1 2 m r 2 2 n2 m1 R m 2 3 sin J m 2 M 1 m2 3 l sin sin m r 2 1 m 2 3 l sin sin m r 2 1 m2 l 2 3 l 2 sin 2 m 3

1 2 2 r2 m rl n2 m 2+m3 sin cos m2 3 sin cos Bm 2 N , 1 m cos sin m l 2 sin 1 m gl cos 2 rl 2 cos 3 2 m3 2 2 2

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

II. Slider Crank Actuated by a PM Synchronous

F r sin D BE F BE l sin where m1 , m2 and m3 are the mass of crank, rod and
slider, respectively, and r and l are length of crank and rod, and and are angle of crank and rod. The translation position transforming

Fig 1 Block Diagram of a PM synchronous motor

, that is X B cos cos l ' r l

X B can be obtained by

t iq m m nK nJ nB r r

cosl sin ' r r l

2 2 2

1 2

(3)

(1) III. Particle Swarm Optimization In 1995, Kennedy and Eberhart first introduced the particle swarm optimization (PSO) method. The method has been found to be robust in complex system, which is derived from the social-psychological theory. Instead of using evolutionary operators to manipulate the particle (individual). Each particle is treated as a volume less particle in g-dimensional search space and keeps track of its coordinates in the problem space, which are associated with the best solution (evaluating value) it has achieved so far [7-10]. This value is called pbest. Another best value that is tracked by the global version of the particle swarm optimizer is the overall best value, and its location, obtained so far by any particle in the group, is called gbest. The PSO concept consists of, at each time step, changing the velocity of each particle toward its Pbest and gbest locations. For example [9-10], the jth particle is represented as xj = (xj,1, xj,2, xj,g) in the g-dimensional , space. The best previous position of the jth particle is recorder and represented as pbestj = (pbestj,1, pbestj,2, , pbestj,g). The index of best particle among all of the particles in the group is represented by the gbestg. The rate of the position change (velocity) for particle j is represented as vj = (vj,1, vj,2, vj,g). The modified velocity , and position of each particle can be calculated using the current velocity and distance from pbestj,g to gbestj,g as shown in the following formulas:
) v (jt, g1) w (jt, g 1 rand () * ( pbest j , g x (jt, g ) v ) c* ) 2 Rand () * ( gbest g x (jt,g ) c*

is the torque applying in the direction of , r K t is the torque constant, J m and Bm are the inertia

Fig 2 A Slider Crank Mechanism System

M + N , BU 0 (2) D+ T

(4)

1 ) x (jt, ) x (jt, g (jt, g , j ,2, , n; g ,2, , m v ) 1 1 g

where n m t
) v (jt, g

number of particles in a group; number of members in a particle; pointer of iterations (generations); velocity of the particle j at iteration t,
) Vgmin v (jt, g gmax ; V

nK B t 0 U iq

w inertia weight factor; c 1, c 2 acceleration constant; rand(), Rand() random number between 0 and 1;

~ 228 ~

t; pbeatj pbest of particle j; gbest gbest of the group. In the above procedures, the parameter Vmax determined the resolution, or fitness, with which regions be searched between the present position and the target position. If Vmax is too high, particles might fly past good solution.. If Vmax is too small, particles may not explore sufficiently beyond local solution. The PSO algorithm was mainly utilized to determine three optimal controller parameters kp, ki, and kd. We defined three controller parameters to compose an individual K by K:=[kp, ki, kd]; its dimension is n3. The searching procedures were shown as below:[10] Step1) Specify the lower and upper bounds of the three controller parameters and initialize randomly the individuals of the population including searching points, velocities, pbests, and gbest. Step2) Calculate the evaluation value of each individual in the population. Se3 C m a ec i i dasea ao vl t ) o pr ah n v ul vl t n a e p e di ui u with its pbest. The best evaluation value among the pbest is denoted as gbest. Step4) Modify the member velocity v of each individual K according to
1 v (jt, ) w (jt ) 1 rand ()* ( pbest j , g (j t, g ) v c* k ) g

The structure of the proposed fuzzy PID controller is shown in figure 3. The detailed design procedures are

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

) x (jt, g

current position of particle j at iteration

described in the following sections.

0 K p , p , Ki0 , i , KD , D K K 0 K

Obtained By PSO

e(t )
_

Command

Fuzzy Rule
d dt

) (t

PID Self-Tuning

Slider Crank

Xb (t)

e(t)

Fig 3 Structure of Fuzzy PID Controller 4.1 Self-Tuning PID In most researches, the parameter self-tuning is based on known mathematical models. Even the models unknown, the on-line identification or parameter estimation is proposed to establish an estimated model. However, the modeling error usually causes the unexpected conditions. Fortunately, for any complex systems, the PSO approach can easily obtain the most proper parameters under no-load and full-load conditions. Take the proportional controller for example. Under operating conditions between no-load and full load, the proportional gain should be in the range of the gains of full-load and no-load conditions. Let the proportional gains are Pmin and Pmax under no-load and full-load conditions respectively. The proportional gain should be in the range of Pmax to Pmin . According the common sense [17], if the proportional (P) controller gain s increases, then the rising time and steady state error will be reduced. Too large gain will make the great overshoot and extreme oscillation. Too small gain will make steady state error existed. If the absolutions of error and error derivation are large, then the proportional gain should be working on largest number to achieve fast rising time, that is Pmax . If the error and error derivation are small enough, then the proportional gain should be Pmin to maintain the minimum steady state error. The relationship can be shown as figure 4. Hence, let the tuning rule is defined as

2 Rand () * ( gbest g (j t, g ) c* k ) 1 1 j ,2, , n; g ,2,3

(5)

When g is 1, vj,1 represents the change in velocity of kp controller parameter. When g is 2, vj,2 represents the change in velocity of ki controller parameter. When g is 3, vj,3 represents the change in velocity of kd controller parameter. Step5) If If Step6)
v (jt, g1) gmax , then v (jt,g1) gmax . V V ( t ) 1 j ,g

min g

, then

( t ) 1 j ,g

min g

Modify the member position of each individual K according to

k (j t,g 1) (j t,g (jt,) k ) v g1 min max k g k (j t, g1) k g

K P K P K P 1 E , E t 0 t t (6) 0 where K P min , P max min and is the P K P P adjusting factor. The adjusting factor will be decided
on-line by fuzzy rule.

where

min max k g and k g represent the lower

Step7)

Step 8)

and upper bounds, respectively, of member g of the individual K. If the number of iterations reaches the maximum, then go to Step 8. Otherwise, go to Step 2. The individual that generates the latest gbest is an optimal controller parameter..

IV. Fuzzy Self-Tuning PID Controller Fig 4 Tuning Rule of Proportional Gain

~ 229 ~

K I K I0 K I 1 E , E t t t (7) 0 Let K I min , I max min and I max and I K I I I min are the gains under no-load and full-load conditions
respectively.

By the same conception, the integral controller is helpful for steady state and hurtful for transient state. Hence, the integral (I) controller gain should be s increased along with the error and error derivation decreasing. The curve is shown as figure 5. Let the tuning rule is defined as

K D D K D E , E t K 0 1 t t (8) 0 Let K D min , D max min and Dmax D K D D and Dmin are the gains under full-load and no-load
conditions respectively

Large differential controller can increase the speed of response. At the same time, large differential controller will cause large steady state error. Therefore, the differential (D) controller gain should be decreased s along with the error and error derivation decreasing. The tuning rule can be shown as following figure and equation.

The most significant of the proposed PID tuning method is that the initial parameters and tuning ranges of PID controller are supported from PSO approach. It will

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

bring the confidence in the controller for operators. After this, the tuning work will be finished by the fuzzy rule. 4.2 Adjusting Factor In order to intelligently and automatically tune the contro e parameters, the adjusting factor is lr ls obtained by fuzzy theory. By fuzzy control theory, the first work is to defuzzify the reference variables. The reference variables in designing procedure are position error and position error difference. There are seven linguistic variables used in this paper for each reference variables, that are large negative (LN), medium negative (MN), small negative (SN), zero (ZO), small positive (SP), medium positive (MP) and large positive (LP). The membership functions representing these linguistic expressions are defined as triangular type function shown in figure 7. In order to process the reference variable more efficiently, the position error is deal with linear convert from 0.05m~+0.05m into 0~1. At the same time, the position error difference is converted from 100m/s ~+100m/s to 0~1. The output factor is restricted in 0~1, shown in figure 7, too.

Fig 5 Tuning Rule of Integral Gain

(a)

(b)

(C )
Fig 7 Membership Functions of Error, Error Difference and Adjusting Factor Fig 6 Tuning Rule of Differential Gain Secondly, the fuzzy relation and fuzzy rule are defined as 49 if-then rules for fuzzy inference. Due to the common sense, if the absolutions of error and error difference are large, then the adjusting factor needs to be large. Respectively, if the absolutions of error and error difference are small, then the adjusting factor is small, too.

~ 230 ~

The if-then rules can be established as followings.

The if-then rules can be tabulated as follow table. Table 1 If-Then Rules Table

Finally, by the centroid defuzzification method, the adjusting factor is calculated from

where x means the each central value of the linguistic variables of adjusting factor. Finally, applying the adjusting factor to eq (6)~(8), it will form the intelligent self-tuning PID controller. V. Simulation Results In this section, numerical simulation results are used to demonstrate the potential of the proposed control rule. To demonstrate the performance, the PID self-tuning method is compared with a fixed PID control. The actual slider crank mechanism dimensions are m1 =0.05454, m2 =0.2795, m3 =0.16, r =0.056, R=0.086, l =0.174, l ' =0.055,

Xb (m)

periodically translation position from 0.2107m to 0.1858m. The desired specifications are settling time t s =0.5sec, rising time

t r =0.25 sec, maximum overshoot M P <5% and steady state error ess <1%.
Figures 8(a) show the response of fixed PID. Using the PSO approach, we can get the parameters for optimal performance under no-load condition, where
no no K P =3.5695, K Ino =2.9267 and K D =0.7727. The

parameter variation is appeared by the mass of slider crank m3 changed from 0.16Kg to 0.8Kg. The external force is changed from 0Nt to 5Nt. While the load and parameter variation not existed, the performance can satisfy the

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

is Rule3 : If e is SN and e is SN, then t SP t t
(9)

is Rule1 : If e is LN and e is LN, then t LP t t is Rule2 : If e is MN and e is MN, then t MP t t

requirement. However, compared with these curves the response will not match the desired specification when the load or parameter varying existed. Obviously, the fixed PID obtained under no-load condition cannot achieve the robustness with parameter varying and load existed. By the same way, the PID parameters for full-load s condition, the optimal parameters for PID controller with full-load are

K Pfull =4.1942,

K Ifull =1.2997 and

full K D =1.1313. Theoretically the dynamic response under

t
LN MN SN ZO SP MP LP

e t

e t

LN LP LP LP LP LP LP LP

MN LP MP MP MP MP MP LP

SN ZO SP LP LP LP MP MP MP SP SP SP SP ZO SP SP SP SP MP MP MP LP LP LP

MP LP MP MP MP MP MP LP

LP LP LP LP LP LP LP LP

full-load will match the requirements. It is also proven in figure 9(a). However, once the parameter varying and external load are removed, the responses will cause the large steady-state error and unexpected transient state shown in figure 9. The large steady-state error and overshoot appeared to show the bad robustness of fixed PID. The proposed self-tuning PID controller is based on these two optimal parameters. Let the normal parameters are based on no-load condition, that is

K I0 K Ino
tuning

0 no K P K P =3.5695, 0 =2.9267, and K D D =0.7727. The K no

ranges
full P full I full D no P no I no D

are

selected

as

x max min , dt t max min , dt

1 0 i et i et 1 0 i et i et

P K K I K K D K K

K =0.6247, K =1.6270 and =0.3586. According to the position K

(10)

error and position error difference, the fuzzy rule will automatically find the optimal adjusting factor. Applying eqs (6)~(8), the responses of the proposed control rule are shown in figure 10(a). Obviously, the proposed self-tuning PID controller has great robustness. Under different operating conditions, it can still maintain the desired performance.
0.215 0.21 0.205 0.2 0.195 0.19
N orma l(m3=0.16 kg Fe =0Nt)

K t =0.6732, J m =0.00062 and

Bm =0.000153. The objective is to control the desired

0.185 0.18

C hang ed(m 3=0.8 kg Fe =0Nt) C hang ed(m 3=0.1 6kg Fe=5Nt) C hang ed(m 3=0.8 kg Fe =5Nt)

0.5

1.5

2 2.5 Time (sec)

3.5

(a) Simulation results

~ 231 ~

Xb (m)

Xb (m)

Xb (m)

Xb (m)

Xb (m0

Xb (m)

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

0.215 0.21 0.205 0.2 0.195 0.19 0.185 0.18
0.215 0.21 0.205 0.2 0.195 0.19 0.185 0.18

0.5

1.5

2 2.5 Time (sec)

3.5

0.5

1.5

2 2.5 Time (sec)

3.5

(b) Experimental Result with No-Load

0.22 0.215 0.21

(b) Experimental Result with 0.8 kg External Load

0.215 0.21 0.205

0.205 0.2 0.195 0.19 0.185

0.185 0.2 0.195 0.19

0.18 0.175
0.18

0.5

1.5

2 2.5 Time (sec)

3.5

0.5

1.5

2 2.5 Time (sec)

3.5

Experimental Result with 0.8 kg External Load Fig 8 Response of Translation Position with Fixed PID Control (Parameters Obtained Under No-Load Condition)
0.215 0.21 0.205 0.2 0.195 0.19 0.185 0.18

Experimental Result with No-Load Fig 9 Response of Translation Position with Fixed PID Control (Parameters Obtained Under Full-Load Condition)
0.215 0.21 0.205 0.2 0.195 0.19

Norm al(m3 =0.16 kg Fe=0Nt) Changed(m 3=0.8kg Fe=0N t) Changed(m 3=0.16kg Fe=5N t) Changed(m 3=0.8kg Fe=5N t)

N ormal (m3= 0.16kg Fe=0Nt) C hange d(m3 =0.8kg Fe=0Nt) C hange d(m3 =0.16 kg Fe =5Nt) C hange d(m3 =0.8kg Fe=5Nt)

0.185 0.18

0.5

1.5

2 2.5 Time (sec)

3.5

0.5

1.5

2 2.5 Time (sec0

3.5

(a) Simulation results

(a) Simulation Results

~ 232 ~

Xb (m)

Xb (m)

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

0.215 0.21 0.205 0.2 0.195 0.19 0.185 0.18

Fig 11 Experimental Instrument of Slider Crank Based on the same requirement of simulation, the experimental results by the HP 54601B 100 MHz 4 channel oscilloscope. Firstly, the fixed PID setup under no-load is experimented. Under no-load condition, the experimental result is shown in figure 8(b). However, the response shown in fig 8(c) has large overshoot when the load added. It shows the bad robust of fixed PID controller. The second experiment is fixed PID controller whose parameters are obtained under full-load condition. Figure 9(b) shows the experimental result. It shows the manual chosen parameters can achieve desired requirement under full-load condition. However, figure 9(c) shows the same controller applied when load removed. Obviously, the fixed PID controller cannot overcome the large change in operating situation. Finally, applying the proposed self-tuning PID controller to the same experimental instrument. The responses of translation position with no-load existed are shown in figure 10(b). It can achieve the same performance with the fixed PID controller under no-load condition. The response with load existed is shown in figure 10(c). The dynamic response is almost same with the result of no-load. It also approves that the experienced self-tuning PID controller has great robustness to against parameter variation and external load. VII. Conclusion This paper proposed a simple scheme of fuzzy PID method. Based on the PSO approach, the nominal values and tuning ranges of PID parameters can be accuracy get. Applying the intelligent fuzzy rule, the adjusting factor can be tuned on-line. A PC-based controller is implemented to apply to the translation position control of a slider crank mechanism. Simulation and experimental results show the proposed controller is more robust to fixed PID controller. Reference [1] A Vs l ui o PD cn o e wt fzy . io , T n g f I ot lr i uz ii n r ls h l i Proc. Inst.Elect. Eng. Contr. Theory oc g, Applicat., vol. 148, no. 1, pp. 1 Jan. 2001. 8, [2] T L Sn, B K ad adR Y sf ui o . . eg M. . hl ,n . uo,T n g f i n a neuro-fzy ot lry eec l rh ,IEEE uz cn o e gnt a oi m r lb i g t Trans. Syst., Man, Cybern. B, vol.29, pp. 226 236, Apr. 1999. [3] R A K ol gadJ P R y ei o ot a . . rh n n . . e, D s n f p m l i g i disturbance rejectionPID controllers using genetic a oi mIEEE Trans. Evol. Comput.,vol. 5, pp. l rh , g t 78 Feb. 2001. 82,

0.5

1.5

2 2.5 Time (sec)

3.5

(b) Experimental Result with No-Load

0.215 0.21 0.205 0.2 0.195 0.19 0.185 0.18

0.5

1.5

2 2.5 Time (sec)

3.5

Experimental Result with 0.8 kg External Load Fig 10 Response of Translation Position with self-tuning PID Control
VI. Experimental Results In order to demonstrate the proposed control rule, a PC-based experimental equipment is setup in this paper. The experimental instrument of slider crank is divided into three parts: actuator, slider crank and controller. The photographic is shown in figure 11. The first part consists of a PM Synchronous motor, driver.. The driver is worked on 3-phase, 220 V and 60 Hz. The slider crank is coupled with the PM motor. The translation position is measured by a photometer. The output of photometer scalar is 20000 pulse/m, which mapped to real translation position is 0~0.11m. To carry out the parameter varying and external load is to add an external mass (0.8Kg) on slider. The controller is based of a PC with Pentium-586 CPU. The PC plays the role of software development and data process. The data acquisition interface card (Advantech CO., PCL-1800) is installed in the ISA bus to handle the A/D and D/A process. The graphical software of Simulink is used to implement the proposed control rule. At the same time, the linear converts between the physical scale and voltage from sensors are also worked in this software.

~ 233 ~

slangiS & smetsyS nO ecnerefnoC lanoitanretnI 5002SSCI EEEI

[4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] Y Miuua T Y m m t adM. aea . t kr . a a o , n s , o K nd, A design of self-tuningPID controllers using a genetic a oi mn l rh ,i Proc. Amer. Contr. Conf., San Diego, g t CA, June 1999, pp. 1361 1365. T K w b adT T gm , A r l oe gnt . a ae n . aa i e cdd eec a i algorithm for matrixinequality design approach of rbsPD cn o e wt todges fedm out I ot lr i w er of eo , rl h e r in Proc. 12th IEEE Int. Symp. Intell.. Contr. Istanbul, , Turkey,July 1997, pp. 119 124. R. A. Krohling, H. Jaschek, and J. P. Rey, ei i P/I cn o e o am t n cn o D s n g I D ot lr r gn P r lf o o ot l i r ss m bsd o gnt a oi m i Proc. yt e ae n eec l rh , n i g t 12thIEEE Int. Symp. Intell. Contr.. Istanbul, Turkey, , July 1997, pp. 125 130. R C E e a adY S i o pro bteen . . br r n . h C m a sn e ht , i w gnt a oi m ada ie w r ot i t n eec l rh s nprc s a p m z i , i g t tl m i ao in Proc. IEEE Int. Conf. Evol. Comput.,Anchorage, AK, May 1998, pp.. 611 616. P JA gl e s gsl t nt i poe a ie .. nen,U i e co o m rv prc i n ei tl s a ot i t n i Proc. IEEE Int. Conf. Evol. w r p m z i ,n m i ao Comput., Anchorage, AK, May1998, pp. 84 89. H Y si ,K K w t ad Y F kym , . oh a . a a , n . uua a A d a particle swarm optimizationfor reactive power and voltage control considering voltage scrys s et IEEE Trans. Power Syst., vol. eui as s n t em , 15, pp. 1232 1239, Nov.2000. Z L G i , Prc Swarm Optimization . . a g A ai n tle Approach for Optimum Design of PID Controller in AVR System, IEEE Transactions on Energy Conversion, vol. 19, No. 2, June 2004, pp. 384 391. H. P. Huang, M. L. Roan, and J. C. Jeng, n O -line adaptive tuning for PID controllers, IEE Proceedings Control Theory and Applications, 2002, pp.60-67. P. Cominos, and N. Munro, controllers: recent PID tuning methods and design to specification,IEE Proceedings Control Theory and Applications, 2002, pp.46-53. B. Kristiansson, and B. Lennartson, Robust and optimal tuning of PI and PID controllers, IEE Proceedings Control Theory and Applications, 2002, pp.46-53. G. K. I. Mann, B. G. Hu, and R. G. Gosine, Time-domain based design and analysis of new PID tuning rules, IEE Proceedings Control Theory and Applications, 2001, pp.251-261. A. Visioli, Optimal tuning of PID controllers for integral and unstable processes, IEE Proceedings Control Theory and Applications, 2001, pp.180-184. R. F. Fung, K. W. Chen, and J. Y. Yen, Fuzzy Sliding Mode Controlled Slider-Crank Mechanism Using PM Synchronous Motor Drive, International Journal of Mechanical Science, Vol. 41, 1999, pp. 337-355. B. C. Kuo, Automatic Control, 6th Ed., Prentice-Hall, 1999.

~ 234 ~