SAMS, 2002, Vol. 42, pp.

16231637

A NEW APPROACH TO FUZZY CONTROL OF INTERCONNECTED SYSTEMS

MAGDI S. MAHMOUDa,*, MANAR M. SABRYb,y and SALAH G. FODAc,z
a

Faculty of Engineering, Arab Academy for Sciences and Technology, P.O. Box 2033 Al-Horriya, Cairo, Egypt b Projects and Design Division, Saudi Arabian Texaco (J.O.), P.O. Box 9720, Ahmadi 61008, Kuwait c Electrical Engineering Department, King Saud University, P.O. Box 800, Riyadh 11421, Saudi Arabia
(Received 29 May 2000)

This paper develops a new approach to the control of interconnected system using fuzzy system theory. The approach is based on incorporating a group of local estimators on the system level to generate the inputoutput database. An array of feedback fuzzy controllers is then designed to ensure the asymptotic stability of the closed loop system. The developed technique is applied to an unstable large-scale system, and extensive simulation studies are carried out to illustrate the potential of the new approach. Keywords:

1. INTRODUCTION In control engineering research, problems of decentralized control and stabilization of interconnected systems are receiving considerable interest in recent years [1,2] where most of the effort is focused on
*Corresponding author. E-mail: magdi@eng.kuniv.edu.kw y E-mail: mmanar@ncc.moc.kw z E-mail: sfoda@ksu.edu.sa 1623

1624

M.S. MAHMOUD et al.

dealing with the interaction patterns. It is concluded that a systematic approach to deal with the problems of interconnected systems is twofold: First is to base the analysis and design effort on the subsystem level using conventional control methods and second is to deal with interactions effectively. These methods are facilitated, in general, by virtue of several mathematical tools including linearization, delay approximation, decomposition and model reduction. This constitutes the so-called model-based control system approach for which we have seen numerous techniques [3]. Most of the available results have so far overlooked the operational knowledge of the interconnected system under consideration. In [4], a knowledge-based control system approach has been suggested to deal with the analysis and design problems of interconnected systems by incorporating both the simplest available model as well as the best available knowledge about the system. For single physical systems, one of the earlier efforts along this direction has been based on the development of an expert learning system [56]. An alternative approach has been to integrate elements of discrete event systems with differential equations [7]. A practically supported third approach has been the use of fuzzy logic control by successfully applying fuzzy sets and systems theory [9]. For interconnected systems, the foregoing approach motivates the research into intelligent control by combining techniques of control and systems theory with those from artificial intelligence. The main focus should be on integrating a knowledge base, an approximate (humanlike) reasoning and/or a learning process within a hierarchical structure. Fuzzy logic controllers [10] are generally considered applicable to plants that are mathematically poorly understood (there is no acceptable mathematical model for the plant) and where experienced human operators are available for satisfactorily controlling the plant and providing qualitative rules of thumb (qualitative control rules in terms of vague and fuzzy sentences). 1. Hierarchical ordering of fuzzy rules is used to reduce the size of the inference engine. 2. Real-time implementation, or on-line simulation of fuzzy controllers can help reduce the burden of large-sized rule sets by fusing

FUZZY CONTROL

1625

sensory data before imputing the systems output to the inference engine. A concerted effort has been made to formally reduce the size of the fuzzy rule base to make fuzzy control attractive to interconnected systems. Two of the difficulties with the design of any fuzzy control system are:
. .

The shape of the membership functions. The choice of fuzzy rules.

The properties that a fuzzy membership function is used to characterize are usually fuzzy. Therefore, we may use different membership functions to characterize the same description. Conceptully, there are two approaches to determine a membership function. The first approach is to use the knowledge of human experts. Usually this approach can only give a rough formula of the membership function; fine-tuning is required. In the second approach, data are collected from various sensors to determine the membership functions. Specifically, the structures of the membership functions are specified first. Then fine-tuning of the membership function parameters should be implemented based on the collected data [8]. In this paper, we contribute to the further development of intelligent control techniques of interconnected systems. It provides a new approach to fuzzy control design for interconnected systems. The approach consists of two stages: In the first stage, a group of local state estimators is constructed to generate the data base of inputoutput pairs. In the second stage, an array of feedback fuzzy controllers is designed and implemented to ensure the asymptotic stability of the interconnected system. Simulation studies on a large-scale system with unstable eigenvalues are carried to illustrate the features and capability of the new approach.

2. FUZZY SYSTEMS BACKGROUND Fuzzy control is by far the most successful application of fuzzy sets and systems theory to practical problems. Numerous applications of fuzzy logic controllers to a variety of consumer products and industrial systems have been recorded [4,9].

1626

M.S. MAHMOUD et al.

Fuzzy systems are linguistic knowledge based system. The heart of a fuzzy system is what is so-called fuzzy IF-THEN rules. These rules are statements in which some words are described by a continuous membership function (Fig. 1). For example,

IF vessel temperature is high THEN small opening of fuel value is required: IF vessel temperature is low THEN wide opening of fuel value is required: 1

In general, the starting point of constructing a fuzzy system is to obtain a collection of fuzzy IF-THEN rules from human experts, experiments or based on domain knowledge. The next step is to combine these rules into a single system. There are three types of fuzzy systems that are commonly used: 1. Pure fuzzy systems, 2. TakagiSugenoKang (TSK) fuzzy systems, and 3. Fuzzy systems with fuzzifier and defuzzifier. The three systems are described briefly hereinafter. The configuration of a pure fuzzy system is illustrated in Fig. 2. The fuzzy rule base represents the collection of fuzzy IF-THEN rules. The fuzzy inference engine combines these fuzzy IF-THEN rules into a mapping from fuzzy set in the input space U & Rn to fuzzy sets in the output space V & R based on fuzzy logic principles. If the dashed

FIGURE 1 a. Temperature membership functions; b. Valve opening membership functions.

FUZZY CONTROL

1627

FIGURE 2 Basic configuration of pure fuzzy systems.

feed back line in Fig. 2 is exists, the system becomes a fuzzy dynamic system (FDS). The main disadvantage in the pure fuzzy system is that its input and output is fuzzy set, whereas in design and engineering the input and output are real-valued variables. Takagi, Sugeno and Kang [11,12] introduced another fuzzy system whose input and outputs are real-valued variables. This system uses rules in the following: IF the input x is high then the output y cx: 2

Where the word high has the same meaning as in (1), and c is a constant. Comparing (1) and (2) we can see that the THEN part of the rule changes from logistic to into a simple mathematical formula which leads to combine the rule easier. In fact, the TSK fuzzy system is a weighted average of the value in the THEN parts of the rules. Figure 3 shows the basic configuration of TSK fuzzy system. The main problem with TSK fuzzy system is its THEN part, which may not reflect a good framework to represent human knowledge. To solve this problem, the third type of fuzzy systems is used. Figure 4 illustrates the main structure of the fuzzy system with fuzzifier and defuzzifier. Comparing this system with a pure fuzzy system, we can see that the only difference between the two systems is that are the fuzzifier that transfer the real-valued variable into a fuzzy set, and the defuzzifier that transfer the fuzzy set into a real-valued variable.

1628

M.S. MAHMOUD et al.

FIGURE 3 Basic configuration of TakagiSergenoKang (TSK) fuzzy systems.

FIGURE 4 Basic configuration of fuzzy systems with fuzzifier and defuzzifier.

3. STATE ESTIMATION OF INTERCONNECTED SYSTEMS In the sequel, the terms large-scale and interconnected are used interchangeably. The term large-scale system (LSS) does not have a unique established meaning, but it covers systems that possess several particular features, such as multiple subsystem, multiple control agents, multiple objectives, decentralized and/or hierarchical information structures [1,4]. Any LSS includes many variables but their control is faced by a well-known fact [3] that the states are not always available for measurement and state must be estimated. Many authors have considered the state estimation of LSSs in input decentralized fashion. Here we summarize one convenient algorithm [2]. Let the state model of the ith subsystem be described by XN
i6j

xi t Ai xi t Bi ui t

Gij xj ;

3 4

yi t Ci xi t, i, j 1, 2, . . . , N :

Where all vectors and matrices are appropriately defined and gi(.) is the interaction function between the ith subsystem and the rest of

FUZZY CONTROL

1629

the system. It is considered that (Ci, Ai) is completely observable for i 1, 2, . . . , N . The following algorithm finds the optimal states of a LSS based on decentralized estimation and control [4]:

Algorithm 1 Step 1: Read the matrices Ai, Bi and select Qi ! 0 and Ri > 0 as weighted matrices. Step 2: Solve the following 2N algebraic Raccati equations for Hi, Ki Hi AT i Ii Ai Ii Hi Hi Di Hi Qi 0; Ki AT i Ii Ai Ii Ki Ki Si Ki Qi 0: 5 6

1 T Where Di CiT Ci , Si Bi R j B :

Step 3: Integrate the following set of N simultaneous equations for ei(t), i 1, 2, . . . , N, using the initial condition ei(0) xi(0) 2 3 2 32 3 2 3 A1 S 1 K 1 . . . GIN B 1 v1 e1 e1 .. . 6 . 7 6 76 . 7 6 . 7 . 4 . 54 . . 54 . 5 4 . . 5 . . eN eN B N vN GNI AN H N DN

Step 4: Integrate the following set of n simultaneous equations for x1(t), i 1, 2, . . . , N 2 3 2 A1 S 1 K 1 x1 . 6 . 7 6 . 4 . . 54 . xN 2 G1N S 1 K 1 . . . AN S N I N 0 3 .. . 0 3 7 5 SN K N 8

GN 1 3 2 x1 B1 v1 6 . 7 6 . 7 4 . . 54 . . 5 xN BN vN

^ i ci x ^ i g. Step 5: Generate the inputoutput pairs fvi , y

1630

M.S. MAHMOUD et al.

4. INTERCONNECTED SYSTEM Assume the following interconnected system of order 10: 1:5 6 0:1 6 6 0 6 6 0:6 6 6 0:4 A6 6 r211 6 6 r221 6 6 r231 6 4 r241 r251 2 0:3 0:25 0:1 0 0 0:2 0:2 1 0 0:1 0:25 2 0:2 1 0:5 r212 r213 r214 r222 r223 r224 r232 r233 r234 r242 r243 r244 r252 r253 r254 2 0:5 0 0:4 0 0:1 r215 r225 r235 r245 r255 0 0 0 1 1 0 0 0 0 0 0 1 0 0 r111 r121 r131 r141 r151 1:5 0:1 0 0:6 0:4 0 0 0 0 0 0 1 0 0 0 0 0 1 0 r112 r113 r122 r123 r132 r133 r142 r143 r152 r153 0:3 0:25 0 0 0:2 1 0:1 0:25 0:2 1 r114 r124 r134 r144 r154 0:1 0:2 0 2 0:5 3 r115 r125 7 7 r135 7 7 r145 7 7 r155 7 7 0: 5 7 7 0 7 7 0: 4 7 7 0 5 0: 1

0 61 6 60 6 60 6 60 B6 60 6 60 6 60 6 40 0 1 60 C6 40 0 2 1 0 0 0 0 1 0 0 0 0 0 0

3 0 07 7 07 7 07 7 07 7 07 7 07 7 07 7 15 1 0 0 1 0 0 0 0 1 0 0 0 0 3 0 07 7 05 1

which is considered to be composed of two-coupled subsystems; each of order 5. The coupling parameters are r1jk and r2jk where j and k take values of 1, 2, 3, 4 and 5. In the sequel, we refer to the structure of the interconnected system model as: 2 6 x4 A11 . . . G21 r2 3 2 3 . . . G12 r1 B1 7 .. 5x 4 . . . 5v . B2 A22

10

FUZZY CONTROL

1631

Where G12(r1) and G21(r2) are the coupling matrices with r1 {r111, . . . , r155} and r2 {r211, . . . , r255}. For typical values of r115 0.1, r124 0.1, r142 0.2, r222 0.1, r242 0.15, r251 0.11 and all other values of coupling parameters being zeros, we examined the stability of the system by computing the eigenvalues of matrix A. They are { 1.0915, 1.0641, 0.477 j 0.0206, 0.477 j 0.00206, 0.022 j 0.0544, 0.022 j 0.0544, 1.8709 j 0.1713, 1.8709 j 0.1713, 1.9306 j 0.1413, 1.9306 j 0.1413}, and it is quite clear that there are four eigenvalues lying in the open right half of the complex plane, and thus the interconnected system is unstable. Further, it is easy to check that the interconnected system is both controllable and observable.

4.1. Estimation of the System State Variables and Outputs A Matlab program is written to implement the computational algorithm (1) of section 3 on the interconnected system. Different positive and negative step inputs are applied to estimate the outputs. The results of two cases are illustrated in Figs. 5 and 6. It is observed that the outputs tend to track conveniently the input signals.

FIGURE 5 Simulation results (case 1).

1632

M.S. MAHMOUD et al.

FIGURE 6 Simulation results (case.2).

4.2. Design of an Array of Fuzzy Controller We are going to treat the interconnected system at hand as being composed of two identical and coupled subsystems. The control system to be designed is such that each subsystem has its own fuzzy negative feedback controller array with its input being the output of the respective subsystem (Fig. 7). The array of fuzzy controllers is built from a collection of individual controllers based on single-input single-output (SISO) design. A schematic of the subsystem fuzzy controller is shown in Fig. 7. In order to build each fuzzy controller, the following steps are implemented. Step 1: The range of the inputs to each fuzzy controller [i, i] are driven from the estimated value of the respective subsystem outputs, where i 1, 2, 3, 4. Step 2: 2N 1 fuzzy set MiL in [i, i] that are normal, consistent and complete with triangular membership functions, as shown in Fig. 6 are defined for each controller, where L 1, 2, . . . , 2Ni 1. That is we use Ni fuzzy set Mi1 , . . . , MiNi to cover the negative internal (i, 0), the other Ni fuzzy sets MiNi2 , . . . , Mi2Ni1 to cover the positive internal (0, ], and the center of fuzzy set MiNi1 at zero.

FUZZY CONTROL

1633

FIGURE 7 Block diagram of the fuzzy feedback controller array.

Step 3: The following 2Ni 1 rules are considered: IF yai is MiL or ybi is MiL THEN u is KiL : 11

Where L 1, 2, . . . , 2Ni 1, and ai, bi are the input to the fuzzy conL L troller i, and the center yL ai and yb of the fuzzy set Ki are chosen such that 8 < 0 for 1 1, . . . , Ni L L y and y 0 for 1 Ni 1 12 ai bi : !0 for 1 Ni 2, . . . , 2Ni 1: Step 4: Product inference engine, singleton fuzzyfier, and center average defuzzifier are selected to design the fuzzy controller. 4.3. Simulation Results The behavior of the interconnected system outputs after implementing the fuzzy controllers with unity step function input are shown in Figs. 8 and 9. It is clearly evident that the system becomes asymptotically stable by using the negative fuzzy feedback controller array. 4.4. Performance of the Fuzzy Feedback Controller Array Now, we examine the effect of coupling matrices on the performance of fuzzy controlled interconnected system. Five additional cases with

1634

M.S. MAHMOUD et al.

FIGURE 8 Output y1 vs y2.

FIGURE 9

Output y3 vs y4.

deferent coupling ranks are implemented. Fine tuning of membership functions was required to adjust their ranges. The following table summarizes the test cases:
Case No. A11, A22 Norm 1 2 3 4 5 6 (Figs. (Figs. (Figs. (Figs. (Figs. (Figs. 8,9) 10,11) 12,13) 14,15) 16,17) 18,19) 2.2529 2.2529 2.2529 2.2529 2.2529 2.2529 G12 Sparsity 3/25 12/25 3/25 1 3/25 1 Norm 0.2 0.4712 .2 3.0361 .2 3.0361 G21 Sparsity 3/25 3/25 12/25 3/25 1 1 Norm 1.8028 0.1803 0.5341 0.1803 3.0364 3.0417 System Stability without with controller controller Unstable Unstable Unstable Unstable Unstable Unstable Stable Stable Stable Stable Stable Stable

*Sparsity equals the number of non zero elements divided by the total number of elements.

The following figures illustrate the above test cases:

FIGURE 10 Case 2 Output y1 vs y2.

FIGURE 11 Case 2 Output y3 vs y4.

FUZZY CONTROL

1635

FIGURE 12 Case 3 Output y1 vs y2.

FIGURE 13 Case 3 Output y3 vs y4.

FIGURE 14 Case 4 Output y1 vs y2.

FIGURE 15 Case 4 Output y3 vs y4.

FIGURE 16 Case 5 Output y1 vs y2.

FIGURE 17 Case 5 Output y3 vs y4.

1636

M.S. MAHMOUD et al.

FIGURE 18 Case 6 Output y1 vs y2.

FIGURE 19 Case 6 Output y3 vs y4.

5. CONCLUSIONS This paper developed a new fuzzy control design approach to an interconnected system. It is shown that the approach consists of two stages: In stage 1, a group of local state estimators was constructed to generate the inputoutput database. Then an array of feedback controllers was designed and implemented to guarantee the overall asymptoticaly system stability. Extensive simulation studies were performed to support the developed design approach.

References
[1] M.S. Mahmoud, M.F. Hassan and M.G. Darwish (1985). Large Scale Control Systems: Theories and Techniques, Marcel Dekker Inc., New York. [2] D.D. Silijak (1991). Decentralized Control of Complex Systems, Academic Press, Boston. [3] M.S. Mahmoud (1991). Computer-Operated Systems Control, Marcel Dekker Inc., New York. [4] M. Jamshidi (1997). Large Scale Systems: Modeling, Control and Fuzzy Logic, Prentice-Hall, New Jersey. [5] M.S. Mahmoud, S.Z. Eid and A.A. Abou-Elseoud (September/October 1989). A Real Time Expert System for Dynamical Processes, IEEE Transactions Systems, Man and Cybernetics, Vol. SMC-19, No. 5, pp. 11011105. [6] M.S. Mahmoud, S. Kotob and A.A. Abou-Elseoud (January 1992). A Learning Rule-Based Control System. Journal of Information and Decision Technologies, Vol. 18, No. 1, pp. 5566. [7] M.S. Mahmoud, A.A. Abou-Elseoud and S. Kotob (April 1992). Development of Expert Control Systems: A Pattern Classification and Recognition Approach. Journal of Intelligent and Robotic Systems, Vol. 5, No. 2, pp. 129146. [8] L.X. Wang (September 1994). A Supervisory Controller for Fuzzy Control Systems that Guarantees. IEEE Trans. Automatic Control, Vol. 39, No. 9, pp. 18451847.

FUZZY CONTROL

1637

[9] K.M. Passino and S. Yurkovick (1998). Fuzzy Control, Addison Wesley, California. [10] C.C. Lee (March/April 1990). Fuzzy Logic in Control Systems: Fuzzy Logic Controller, Parts I and II. IEEE Systems, Man and Cybernetics, Vol. 20, No. 2, pp. 404435. [11] M. Sugeno and G.T. Kang (1998). Structure Identification of Fuzzy Model, Fuzzy Sets and Systems, 28, 1533. [12] Takagi and M. Sugeno (1985). Fuzzy Identification of Systems and its Applications to Modelling and Control, IEEE Trans. On Systems, Man and Cybernetics, Vol. 15, pp. 116132.

Copyright of Systems Analysis Modelling Simulation is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.