FUZZY-NEURON INTELLIGENT COORDINATION CONTROL FOR A UNIT POWER PLANT

Similar documents
A Boiler-Turbine System Control Using A Fuzzy Auto-Regressive Moving Average (FARMA) Model

Improve Performance of Multivariable Robust Control in Boiler System

Linear Analysis and Control of a Boiler-Turbine Unit

Humanoid Based Intelligence Control Strategy of Plastic Cement Die Press Work-Piece Forming Process for Polymer Plastics

Steam-Hydraulic Turbines Load Frequency Controller Based on Fuzzy Logic Control

Application of Dynamic Matrix Control To a Boiler-Turbine System

Fuzzy PID Control System In Industrial Environment

Simulation Model of Brushless Excitation System

Type-2 Fuzzy Logic Control of Continuous Stirred Tank Reactor

Improved Identification and Control of 2-by-2 MIMO System using Relay Feedback

A Study on Performance of Fuzzy And Fuzyy Model Reference Learning Pss In Presence of Interaction Between Lfc and avr Loops

NEW CONTROL STRATEGY FOR LOAD FREQUENCY PROBLEM OF A SINGLE AREA POWER SYSTEM USING FUZZY LOGIC CONTROL

EXCITATION CONTROL OF SYNCHRONOUS GENERATOR USING A FUZZY LOGIC BASED BACKSTEPPING APPROACH

A New Internal Model Control Method for MIMO Over-Actuated Systems

NonlinearControlofpHSystemforChangeOverTitrationCurve

A New Improvement of Conventional PI/PD Controllers for Load Frequency Control With Scaled Fuzzy Controller

Journal of Chemical and Pharmaceutical Research, 2014, 6(3): Research Article

Design of Decentralised PI Controller using Model Reference Adaptive Control for Quadruple Tank Process

Introduction to System Identification and Adaptive Control

Lyapunov Stability of Linear Predictor Feedback for Distributed Input Delays

Lyapunov Function Based Design of Heuristic Fuzzy Logic Controllers

AN INTELLIGENT HYBRID FUZZY PID CONTROLLER

Modeling and Control Overview

DESIGN OF AN ON-LINE TITRATOR FOR NONLINEAR ph CONTROL

GAIN SCHEDULING CONTROL WITH MULTI-LOOP PID FOR 2- DOF ARM ROBOT TRAJECTORY CONTROL

FUZZY CONTROL OF NONLINEAR SYSTEMS WITH INPUT SATURATION USING MULTIPLE MODEL STRUCTURE. Min Zhang and Shousong Hu

3.1 Overview 3.2 Process and control-loop interactions

FUZZY LOGIC CONTROL Vs. CONVENTIONAL PID CONTROL OF AN INVERTED PENDULUM ROBOT

Active Disturbance Rejection Control of Waste Heat Recovery Systems With Organic Rankine Cycles

Comparison of Necessary Conditions for Typical Takagi Sugeno and Mamdani Fuzzy Systems as Universal Approximators

Analysis of Nonlinear Characteristics of Turbine Governor and Its Impact on Power System Oscillation

The Rationale for Second Level Adaptation

RBF Neural Network Adaptive Control for Space Robots without Speed Feedback Signal

Robust PID and Fractional PI Controllers Tuning for General Plant Model

Controller Tuning for Disturbance Rejection Associated with a Delayed Double Integrating Process, Part I: PD-PI Controller

CHAPTER 5 ROBUSTNESS ANALYSIS OF THE CONTROLLER

HYDRAULIC LINEAR ACTUATOR VELOCITY CONTROL USING A FEEDFORWARD-PLUS-PID CONTROL

LPV Decoupling and Input Shaping for Control of Diesel Engines

PID controllers, part I

Norm invariant discretization for sampled-data fault detection

ABOILER TURBINE system provides high-pressure

DECENTRALIZED PI CONTROLLER DESIGN FOR NON LINEAR MULTIVARIABLE SYSTEMS BASED ON IDEAL DECOUPLER

ROBUSTNESS COMPARISON OF CONTROL SYSTEMS FOR A NUCLEAR POWER PLANT

Design of Multivariable Neural Controllers Using a Classical Approach

Research Article Smooth Sliding Mode Control and Its Application in Ship Boiler Drum Water Level

ECE Introduction to Artificial Neural Network and Fuzzy Systems

Simulation based Modeling and Implementation of Adaptive Control Technique for Non Linear Process Tank

Pass Balancing Switching Control of a Four-passes Furnace System

CHAPTER 6 CLOSED LOOP STUDIES

NONLINEAR BLACK BOX MODELING OF A LEAD ACID BATTERY USING HAMMERSTEIN-WIENER MODEL

Multi-Input Multi-output (MIMO) Processes CBE495 LECTURE III CONTROL OF MULTI INPUT MULTI OUTPUT PROCESSES. Professor Dae Ryook Yang

Control of MIMO processes. 1. Introduction. Control of MIMO processes. Control of Multiple-Input, Multiple Output (MIMO) Processes

CONTINUOUS processes in power plant and power station

A Survey for the Selection of Control Structure for Distillation Columns Based on Steady State Controllability Indexes

inputs. The velocity form is used in the digital implementation to avoid wind-up [7]. The unified LQR scheme has been developed due to several reasons

A Systematic Study of Fuzzy PID Controllers Function-Based Evaluation Approach

Repetitive control mechanism of disturbance rejection using basis function feedback with fuzzy regression approach

Chapter 2 Review of Linear and Nonlinear Controller Designs

Adaptive Robust Control for Servo Mechanisms With Partially Unknown States via Dynamic Surface Control Approach

Estimation of the Pre-Consolidation Pressure in Soils Using ANN method

Improving the Control System for Pumped Storage Hydro Plant

Multi-Objective Optimization and Online Adaptation Methods for Robust Tuning of PSS Parameters

Design of Decentralized Fuzzy Controllers for Quadruple tank Process

Self-tuning Control Based on Discrete Sliding Mode

Iterative Controller Tuning Using Bode s Integrals

Process Identification for an SOPDT Model Using Rectangular Pulse Input

Multipredictive Adaptive Control of Arc Welding Trailing Centerline Temperature

Fuzzy Compensation for Nonlinear Friction in a Hard Drive

Parameter Estimation of Single and Decentralized Control Systems Using Pulse Response Data

Subject: Introduction to Process Control. Week 01, Lectures 01 02, Spring Content

Design and Stability Analysis of Single-Input Fuzzy Logic Controller

On Convergence of Nonlinear Active Disturbance Rejection for SISO Systems

A New Robust Decentralized Control Method for Interconnected Nonlinear Systems Based on State Extension and Adaptive Tracking

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION Vol. XVII - Analysis and Stability of Fuzzy Systems - Ralf Mikut and Georg Bretthauer

Robust Actuator Fault Detection and Isolation in a Multi-Area Interconnected Power System

Results on stability of linear systems with time varying delay

Analyzing Control Problems and Improving Control Loop Performance

Reduced Size Rule Set Based Fuzzy Logic Dual Input Power System Stabilizer

9. Two-Degrees-of-Freedom Design

Parameterized Linear Matrix Inequality Techniques in Fuzzy Control System Design

PERIODIC signals are commonly experienced in industrial

Review on Aircraft Gain Scheduling

MULTIVARIABLE ROBUST CONTROL OF AN INTEGRATED NUCLEAR POWER REACTOR

A FEEDBACK STRUCTURE WITH HIGHER ORDER DERIVATIVES IN REGULATOR. Ryszard Gessing

Control Of Heat Exchanger Using Internal Model Controller

Temperature Prediction Using Fuzzy Time Series

Intelligent Systems and Control Prof. Laxmidhar Behera Indian Institute of Technology, Kanpur

Dynamic Modeling, Simulation and Control of MIMO Systems

Overview of the Seminar Topic

Control Systems I. Lecture 2: Modeling. Suggested Readings: Åström & Murray Ch. 2-3, Guzzella Ch Emilio Frazzoli

Modelling of Primary Frequency Control and Effect Analyses of Governing System Parameters on the Grid Frequency. Zhixin Sun

CONVERGENCE BEHAVIOUR OF SOLUTIONS TO DELAY CELLULAR NEURAL NETWORKS WITH NON-PERIODIC COEFFICIENTS

Incorporating Feedforward Action into Self-optimizing Control Policies

A TSK-Type Quantum Neural Fuzzy Network for Temperature Control

New Feedback Control Model in the Lattice Hydrodynamic Model Considering the Historic Optimal Velocity Difference Effect

An Adaptive LQG Combined With the MRAS Based LFFC for Motion Control Systems

Performance Of Power System Stabilizerusing Fuzzy Logic Controller

Gain Scheduling Control with Multi-loop PID for 2-DOF Arm Robot Trajectory Control

IN THIS PAPER, we consider a class of continuous-time recurrent

THERE exist two major different types of fuzzy control:

Transcription:

57 Asian Journal of Control, Vol. 3, No. 1, pp. 57-63, March 2001 FUZZY-NEURON INTELLIGENT COORDINATION CONTROL FOR A UNIT POWER PLANT Jianming Zhang, Ning Wang and Shuqing Wang ABSTRACT A novel fuzzy-neuron intelligent coordination control method for a unit power plant is proposed in this paper. Based on the complementarity between a fuzzy controller and a neuron model-free controller, a fuzzy-neuron compound control method for Single-In-Single-Out (SISO) systems is presented to enhance the robustness and precision of the control system. In this new intelligent control system, the fuzzy logic controller is used to speed up the transient response, and the adaptive neuron controller is used to eliminate the steady state error of the system. For the multivariable control system, the multivariable controlled plant is decoupled statically, and then the fuzzyneuron intelligent controller is used in each input-output path of the decoupled plant. To the complex unit power plant, the structure of this new intelligent coordination controller is very simple and the simulation test results show that good performances such as strong robustness and adaptability, etc. are obtained. One of the outstanding advantages is that the proposed method can separate the controller design procedure and control signals from the plant model. It can be used in practice very conveniently. KeyWords: A unit power plant, intelligent coordinated control, fuzzy-neuron compound control, model-free control. Brief Paper I. INTRODUCTION A unit power plant is a complicated multivariable controlled plant, which consists of a boiler, a steam turbine, a generator, power network, and loads. The boiler and the steam turbine have their own regulating systems respectively. Considering the features of their running together, we must keep them run in a coordinated fashion. When the unit runs under the coordination control strategy, we should not only adapt its out power to the load changes of power network as quickly as possible, but also keep the vapour pressure prior to the steam turbine within the allowed limits at the same time. In recent years, some traditional control methods, such as optimal control, and Manuscript received May 25, 2000; revised October 17, 2000; accepted December 28, 2000. The authors are with National Key Laboratory of Industrial Control Technology and Research Institute of Advanced Process Control, Zhejiang University, Hangzhou, Zhejiang Province, 310027, P. R. China. This work was supported by the National Natural Science Foundation of P. R. China (No.69774023) and by the Zhejiang Provincial Natural Science Foundation (No.698034). modern frequency domain methods have been applied in the coordination control for a unit power plant [1-7]. Several approaches with self-tuning PID parameters are also studied [8-11]. And, a new control method which integrates fuzzy reasoning with adaptive control and decoupling techniques is proposed in [12]. Because of the uncertainties and the time-variability existing in the system structure and parameters, it is hardly to establish exact model of the unit power plant. Most conventional controllers are designed based on the mathematical model of the plant. This complicates the design of controllers and degrades practical operation and robustness of the designed controllers. It is not surprising that the designed control system will not perform well, even become unstable. However, use of model-free control method based on neuron may cope with these difficulties. An effective method to handle the modeling uncertainties and time-variability of parameters is to use the model-free control strategy. Being as the two main items of intelligent control, both the neuron control and the fuzzy control are model-free control methods. They have been applied in the industrial process control successfully in recent years [13-17]. However, the precision of fuzzy control system is lower than conventional control methods.

Asian Journal of Control, Vol. 3, No. 1, March 2001 58 In order to achieve the high degree control precision, the fuzzy rule base should be set up as perfect as possible. Many neural network control methods have also their weaknesses such as the complex learning algorithm, slow convergence and local minimum, etc. Hence, Wang et al. proposed the neuron model-free control method [14] that is very simple and can give good performances. Being used in hydraulic turbine generators and some industrial process plants [18-19], the neuron model-free control has reached its success. By combining fuzzy controller and the neuron model-free controller, the fuzzy-neuron compound control method is proposed for Single-In- Single-Out system in this paper. Taking the concept of decentralized control, the above method is developed into a fuzzy-neuron intelligent coordination controller for a unit power plant. The effectiveness of this new model-free control is demonstrated with several simulation tests. This paper is organized as follows. Section II presents the dynamic characteristics of a unit power plant, and gives its linearized mathematical model and the demands for designing control system. Section III illustrates the characteristics of neuron control and fuzzy control in detail, followed by the presented fuzzy-neuron compound control schemes. Only relying on static decoupling, the fuzzy-neuron intelligent control method for the power unit is presented in section IV. In section V, several simulation tests for a unit power plant are made and the corresponding results are given. At the end, we conclude the paper with some remarks. II. DYNAMIC CHARACTERISTICS OF A UNIT POWER PLANT To a steam drum boiler, we assume that the fuel systems work well. Then, the unit power plant can be simplified as a controlled plant with two inputs and two outputs, that is N P = G Nu T G Nu B G Pu T G Pu B u T u B. (1) G Nu T = 68.81s (1 + 12s)(1 + 82s) G Pu T = 2.194(0.064 + 0.936 1 +124s ). (2) G Nu B = 1 (1 + 83s) 2 G Pu B = 2.194 (1 + 80s) 2 III. FUZZY-NEURON COMPOUND CONTROL 3.1 Neuron model-free control [14] In [14], an adaptive neuron model-free control system is proposed as in Fig. 1, where E is the surroundings of the neuron. The Transfer turns the signals stemmed from E to the neuron inputs x i (t). The neuron output u(t) can be written as u(t)=k n w i (t) x i (t). (3) where K > 0 is the neuron proportional coefficient, x i (t)(i = 1, 2,, n) denote the neuron inputs, and w i (t) are the connection weights of x i (t) and are determined by some learning rules. It is widely believed that a neuron self-organizes by modifying its synaptic weights. According to the wellknown hypothesis proposed by D. O. Hebb, the learning rule of a neuron can be formulated as w i (t + 1) = w i (t) + dp i (t), (4) where d > 0 is the learning rate and p i (t) denotes the learning strategy. In [14], the associative learning strategy is suggested for control purposes as follows p i (t) = z(t)u(t)x i (t). (5) where u T and u B are the manipulated variables. In particular, u T is the opening position of the steam turbine regulating valve, u B is the boiler input, P denotes the vapour pressure prior to the regulating valve of the steam turbine, N denotes the output power of the steam turbine generator. Both P and N are the controlled variables. Note that G NuT, G PuT, G NuB, G PuB are the transfer functions. A test was made on the 125 MW intermediate reheating drum boiler-turbogenerator unit in a power plant. The linearized mathematical model is obtained as follows [20], Fig. 1. The general neuron control system.

59 J. Zhang et al.: Fuzzy-Neuron Intelligent Coordination Control for a Unit Power Plant It expresses that an adaptive neuron, whose learning is through integrating Hebbian learning strategy (p i (t) = u(t)x i (t)) and Supervised learning strategy (p i (t) = z(t)x i (t)), makes actions and reflections to the unknown outside with the associative search. It means that the neuron self-organizes the surrounding information under supervising from the teacher s signal z(t) and gives the control signal. According to the neuron model and its learning strategy, the neuron model-free control algorithm is presented as follows [14] n u(t)=[k w i (t) x i (t)] [ w i (t) ] n w i (t +1)=w i (t)+d[r(t) y(t)]u(t)x i (t) (6) where u(t) is the neuron output taken as control signal, r(t) is the set-point of the controlled plant, and y(t) is the plant response. The inputs of the neuron x i (t) can be selected by the demands of the control system design and realized by the Transfer. 3.2 Fuzzy control The key to enhance the performances of a fuzzy control system is to regulate the fuzzy rule base according to the external changes. One method for regulating fuzzy control rule base is proposed by S. Long et al. in [21], in which the fuzzy inference process can be expressed by a simple formula U = <αe + (1 α)ec>, (7) where E and EC are supposed to be the fuzzy input variables of the system error e(t) and its change e(t), respectively, α is the factor of regulating fuzzy rule base, and U is the fuzzy variable of the system output. By changing the factor α, the control rule base can be regulated and the performances of the fuzzy control system can be improved conveniently. It has been proved that formula (7) has the same functions as a conventional fuzzy rule base. Thus, the fuzzy control system can be written as Fuzzifier: E = <k e e(t)>, EC = <k ec e(t)>, (8) fuzzy controller output can be easily calculated according to e(t) and e(t) when the parameters of the fuzzy controller are selected. Although this method is very simple and convenient, it cannot eliminate the steady state error. 3.3 The fuzzy neuron compound controller As mentioned in introduction, a conventional fuzzy controller can not give a satisfactory control performance both in static and dynamic characteristics. However, the neuron model-free controller is able to learn on-line during control process by itself. When the system error is small, the neuron controller has good performance and high control precision. On the contrary, a large system error may deteriorate the learning process of the neuron. Based on the complementarity between fuzzy system and neuron controller, a novel compound model-free controller is set up as in Fig. 2, where u FC (t) is the fuzzy controller output, u NC (t) is the neuron controller output, u(t) is the compound control signal of the controllers. The compound strategy we suggested in this paper is u(t) = u FC (t) + f(e(t))u NC (t), (11) where f(e(t)) is a nonlinear function of the system error e(t) that approximates to 1 or a constant when the error e(t) is small and approximates to zero for a big error. For example, a feasible selection is f(e(t) =a exp( be(t) ), (12) where a, b are constants selected properly for the controlled plant. Therefore, the compound control strategy presented in this paper has the following characteristics: at the beginning of system response, the bigger system error makes f(e(t)) smaller, and the fuzzy controller plays the main role in control action on the plant, which makes the transient response fast. On the contrary, when the error is getting smaller, f(e(t)) rises, and the fuzzy controller is replaced by the neuron controller. From the above analysis, we can see that, in this new compound control system, the fuzzy controller speeds up the transient response, and the neuron controller eliminates the steady state error. Fuzzy inference: U = <αe + (1 α)ec>, (9) Defuzzifier: u FC (t) = k u U, (10) where k e and k ec are supposed to be the fuzzification factors corresponding to the inputs e(t) and e(t), respectively, k u is the defuzzification factor, and u FC (t) is the practical manipulated variable. From Eqs. (8)-(10), it is obvious that a practical Fig. 2. The structure of fuzzy-neuron compound controller.

Asian Journal of Control, Vol. 3, No. 1, March 2001 60 IV. FUZZY-NEURON INTELLIGENT COORDINATION CONTROL SYSTEM In the coordination control of a unit power plant, the output power N(t) is a main controlled variable, and should follow the load changes of the power network as quickly as possible. The vapour pressure prior to the steam turbine P(t) is allowed to change in a limit range, but its final value should be the set-point. It is of beneficial to increase the adaptability of the unit to load changes, when the vapour pressure P(t) changes in a limit range for varying loads, because the changes of P(t) can vary the thermal deposition. Thus a roughly static decoupling method used in the coordination control for a unit power plant can give satisfying steady performances. The intelligent coordination control system using fuzzy-neuron compound controller is presented as follows. 4.1 Static decoupling for multivariable systems Usually, plants can be modeled by using static gains or dynamic gains. In most cases, the static gains of a controlled plant are more important than the dynamic gains of the plant. And the static gains can be easily obtained by means of many simple methods. Therefore, we can obtain a diagonal superiority plant by using a static decoupling matrix. Rewrite Eq. (1) as Y = GU, (13) where, Y=[N P] T, U=[u T u B ] T, G= G Nu T G Nu B G Pu T G Pu B. The static outputs of the multivariable system (13) can be written as Y(0) = G(0)U(0), (14) where G(0) is the static gain matrix of the plant. Suppose G(0) 0, let 4.2 Fuzzy-neuron intelligent coordination control system The static decoupling method described as above is specially available for the fuzzy-neuron intelligent modelfree control method that we suggested for the unit power plant coordination control. When the static decoupling method is used, the intelligent control system can be designed as two single-variable systems. This method only relying on the static gains of a multivariable plant to get the static decoupled system is very simple because it is merely to know the static gains roughly. However, when conventional control methods are used to design the controller for the decoupled system, it is still very complicated. Without knowing the exactly linearized model of the decoupled system, it is impossible to design the control system with traditional control methods. The situation changes when the fuzzy-neuron intelligent controller is used in each input-output path for the decoupled system. The controller design becomes very simple due to modelfree nature. The fuzzy-neuron intelligent coordination control system for the unit power plant is set up as in Fig. 3, where the fuzzy-neuron controller I and II adopt the control structure described in section III. There is a decoupling matrix between the controller and the plant for decoupling the multivariable system. When necessary, the decoupling matrix can be obtained by calculating or estimating the plant static gains easily. 4.3 The fuzzy-neuron intelligent coordination control algorithm According to the structure of the fuzzy-neuron intelligent control system for multivariable plant shown in Fig. 3, the corresponding intelligent coordination control algorithm can be described as follows. (i) When necessary, find out the decoupling matrix D by calculating or estimating the static gains of the multivariable system. (ii) Calculate the control signal using the following formulae U = DV, (15) where D is a constant matrix, and let D = G 1 (0), (16) From (13) and (15), the static decoupled system is Y = GDV. (17) Fig. 3. The fuzzy-neuron intelligent coordination control system.

61 J. Zhang et al.: Fuzzy-Neuron Intelligent Coordination Control for a Unit Power Plant v N i (t)=[k i 3 K i j =1 w i j (t) x i j (t)] 3 [ w i j (t) ] w j i (t +1)=w j i (t)+d i (r i (t) y i (t))u i (t)x j i (t) E i =<k i i e e i (t)>, EC i =<k ec e i (t)> U F i =<α i E i +(1+α i )EC i > v i F (t)=k i i u U F v i (t)=v F i (t)+f i (e i (t))v N i (t). (18) 3 sec. When the plant model changes from the normal case (Eq. (2)) to the following case (Eq. (20)) as well as the sample period time changes from 3 sec. to 8 sec., the robustness of the fuzzy-neuron intelligent controller is verified. G= 50.6s (1 + 16s)(1 +60s) 1.5 (1 + 35s) 2 3.12(0.064 + 0.936 1 +160s ) 3.12 (1 + 60s) 2. (20) where, i = 1, 2, j = 1, 2, 3, α i [0, 1], v N i (t) are the i-th neuron controller output, v F i (t) are the i-th fuzzy controller output, v i (t) are the fuzzy-neuron compound controller outputs, x j i (t) is the j-th input of the i-th neuron, w j i (t) denotes the connection weight of x j i (t), d i > 0 and K i > 0 are the learning rates and the neuron proportional coefficients, respectively, u(t) = [u 1 (t), u 2 (t)] T are two control signals determined by U(t) = DV(t), where V(t) = [v 1 (t), v 2 (t)] T are the outputs of the fuzzyneuron compound controller in each channel and D is the static decoupling matrix of the controlled plant. To the unit power plant, the set-points and the system outputs are [r 1 (t) r 2 (t)] T =[N 0 (t) P 0 (t)] T, [y 1 (t) y 2 (t)] T =[N(t) P(t)] T. where N 0 (t) and P 0 (t) are the set-points of N(t) and P(t), respectively. For the unit power plant, the neurons transfer are chosen as From Eq. (2), the decoupling matrix D is D =G 1 (0) = 0 1 2.194 2.194 1 = 1 0.45581 1 0. (21) Using the two control methods, the simulation tests are implemented and are shown in Figs. 4-6, where, Fig. 4 shows the coordination control results for the normal case (Eq. (2)), and Fig. 5 shows the results of the robust test (I) when the plant model is changed suddenly from normal case to the changing case (Eq. (20)) at time t = 6 min. When the sample period time is 8 sec., the robust test (II) results of normal case are shown in Fig. 6. It can be seen that the PID controller has grievous oscillation; however, the fuzzy-neuron intelligent coordination controller seems to have nothing changed when being compared with the a forementioned sample period. From the comparison of the simulation results, we can see that the proposed control method has the advantages in the following aspects. x i 1 (t)=r i (t) x i 2 (t)=r i (t) y i (t). (19) x i 3 (t)=x i 2 (t) x i 2 (t); (,2) V. SIMULATION TESTS AND RESULTS To verify the proposed control method, the simulation tests and a comparison between the PID controller and the fuzzy-neuron controller are made. The fuzzy-neuron controller (Eqs. (18)-(19)) parameters are chosen as K 1 = 1.61, K 2 = 6.0, d 1 = d 2 = 100, k 1 e = k 1 ec = 5, k 1 u = 0.1, k 2 2 e = k ec = 5, k 2 u = 0.2, α 1 = 0.8, α 2 = 0.7. And, the PID controller parameters are selected as k P1 = 1.8, k I1 = 0.022, k D1 = 0.6, k P2 = 3.9, k I2 = 0.007, k D2 = 0.9. The sample period is T = Fig. 4. The results under the normal case. Solid: fuzzy-neuron; Dashed: PID.

Asian Journal of Control, Vol. 3, No. 1, March 2001 62 tests are made. The comparison of the simulation results shows that the proposed control method has better performances, especially when the plant model changes considerably and suddenly or when some big disturbance occurs. The performances of the fuzzy-neuron intelligent controller can not be reached easily by the methods of both conventional control and traditional intelligent control. The outstanding advantage of this novel controller is modelfree. Only the static gains of the plant need to be estimated or calculated roughly for static decoupling. This method is very simple and very useful in practice. REFERENCES Fig. 5. The robust test (I). Solid: fuzzy-neuron; Dashed: PID. Fig. 6. The robust test (II). Solid: fuzzy-neuron; Dashed: PID. (1) In the control system design, the selection of the PID controller parameters is closely related to the mathematic model of the plant. But the fuzzy-neuron controller parameters do not have to be selected strictly. (2) The robustness of the proposed control method is much better than that of the PID control method. VI. CONCLUSIONS The fuzzy-neuron intelligent coordination control method for a unit power plant is proposed in this paper. In this control system, the fuzzy controller is taken as a main controller during transient response and alternated by the neuron model-free controller during steady state process. With the example of a unit power plant, a lot of simulation 1. Cori, R. and C. Maffezzoni, Practical Optimal Control of a Drum Boiler Power Plant, Automatica, Vol. 20, No. 2, pp. 163-173 (1984). 2. Hu, K., Z. Wang and Z. Qian, A Frequency Method for the Design of Coordinated Control System of a Unit Power Plant, J. Southeast Univ., Vol. 19, No. 1, pp. 69-77 (1989) ( In Chinese). 3. Johansson, L. and H. N. Koivo, Inverse Nyquist Array Technique in the Design of a Multivariable Controller for a Soiled-fuel Boiler, Int. J. Contr., Vol. 40, No. 6, pp. 1077-1088 (1984). 4. Mortensen, J.H., T. Molbak, P. Andersen and T.S. Pedersen, Optimization of Boiler Control to Improve the Load-Following Capability of Power Plant Units, VGB Kraftwerkstechnik (German Edition), Vol. 80, No. 5, pp. 87-93 (2000). 5. Garduno-Ramirez, R. and K.Y. Lee, Multiobjective Optimal Power Plant Coordinated Control Through Pressure Set-point Scheduling, Proc. Power Eng. Soc. Summer Meet., Seattle, WA, USA, Vol. 1, pp. 205 (2000). 6. Sindelar, R. Feed-forward Control of the Electrical Power Output of a Power Plant Unit, Based on a Controlled Process Model, Automatizace, Vol. 43, No. 5, pp. 320-325 (2000). 7. Zhao, H., W. Li, C. Taft and J. Bentsman, Robust Controller Design for Simultaneous Control of Throttle Pressure and Megawatt Output in a Power Plant Unit, Proc. IEEE Int. Conf. Contr. Appl., Kohala Coast, HI, USA, Vol. 1, pp. 802-807 (1999). 8. Cao, J., Multivariable Self-tuning Feedforward PID Control and Its Application in Coordinated Control, Proc. Chinese Soc. Electr. Eng., Vol. 9, No. 1, pp. 48-54 (1989) (in Chinese). 9. Shen, J. and L. Chen, Automatic Tuning of Coordinated Control System Based on Intelligent Decoupling, Proc. Chinese Soc. Electr. Eng., Vol. 13, No. 4, pp. 13-19 (1993) ( in Chinese). 10. Shen, J. and L. Chen, A Frequency Domain Method for the Design and Auto-Tuning of the Coordinated Control System, J. Southeast Univ., Vol. 23, No. 2, pp. 125-129 (1993) ( in Chinese).

63 J. Zhang et al.: Fuzzy-Neuron Intelligent Coordination Control for a Unit Power Plant 11. Shen, J. and L. Chen, On-line Auto-tuning of Thermal Power Unit Load Control System, Proc. Chinese Soc. Electr. Eng., Vol. 14, No. 2, pp. 13-19 (1994) ( In Chinese). 12. Chai, T., H. Liu, J. Zhang, et al, Novel Design Method for The Coordinated Control System Based on Fuzzy Reasoning and Adaptive Control and Its Application, Proc. Chinese Soc. Electr. Eng., Vol. 20, No. 4, pp. 14-18 (2000) ( in Chinese). 13. Special issue on Fuzzy Logic and Neural Networks, IEEE Trans. Neural Networks, Vol. 3, No. 5 (1992). 14. Wang, N., J. Tu and J. Chen, Intelligent Control Using a Single Adaptive Neuron, Proc. 3rd Congr. Chinese Assoc. Autom., Beijing, pp. 173-177 (1991) (in Chinese). 15. Garduno-Ramirez, R. and K.Y. Lee, Fuzzy Scheduling Control of a Power Plant, Proc. IEEE Power Eng. Soc. Winter Meet., Singapore, Vol. 1, pp. 441-445 (2000). 16. Garduno-Ramirez, R. and K.Y. Lee, Fuzzy Coordinated Control of a Power Unit, Proc. IEEE Power Eng. Soc. Winter Meet., New York, NY, USA, Vol. 1, pp. 134-139 (1999). 17. Zhang Huaguang and Zeungnam Bien. Fuzzy System Identification and Predictive Control of Load System in Power Plant, Proc. IEEE Int. Conf. Fuzzy Syst. & IEEE World Congr. Comput. Intell., Anchorage, AK, USA, Vol. 1, pp. 342-347 (1998). 18. Wang, N., J. Chen and J. Wang, Neuron Intelligent Control for Hydraulic Turbine Generators, Proc. IEEE Int. Conf. Ind. Technol., Guangzhou, China, pp. 288-292 (1994). 19. Wang, N., J. Tu and J. Chen, Neuron Intelligent Control for Electroslag Remelting Processes, ACTA AUTOMATICA SINICA, Vol. 19, No. 5, pp. 634-636 (1993) (In Chinese). 20. Zhang, W. and L. Chen, Analysis and Tuning of Generating Unit Load Feedback Control System, as Turbine Is Equipped with Power-Frequency Control System, Proc. Chinese Soc. Electr. Eng., Vol. 5, No. 3, pp. 10-20 (1985) (in Chinese). 21. Long, S. and P. Wang, Studies on Self-regulation of Fuzzy Control Rule, Fuzzy Math., Vol. 2, No. 3, pp. 105-111 (1982) (in Chinese).