Study On Control Of Water Level In Marine Boiler Based On Fuzzy Neural Network

Boiler is a necessary power plant in the process of industry. In order to ensure the production and security, the auto-control of it is very important. The drum water level is an important variable to be controlled, it is hard to get the mathematic model of the water level with adjustive process, it is characteristic of nonlinearity, instability and time lag. The traditional control mode in the drum water level mostly use PID, the effect of it is dissatisfactory.In this paper, Based on fundamentally theoretical studies about the basic principles and theories of fuzzy neural networks(FNN) and genetic algorithm, in order to overcome the drawbacks of conventional control method, the new method of control based on compensatory fuzzy neural networks is presented .Through the introduction of compensatory fuzzy inference and quick arithmetic, the compensatory FNN is designed; Around this structure of FNN, an optimization strategy with genetic algorithm, in which same genetic operators in traditional genetic algorithm is changed, is presented, in order to optimize the structure and parameter of FNN.The FNN was mainly applied in control of water level in marine boiler in this thesis .the compensatory FNN controller was designed to instead of PID controller used widely in the area of water level control. At last simulation was also performed to the boiler control method of using compensatory fuzzy neural net work by Matlab software. The results indicate that the control arithmetic presented in this paper is effective, the control effect and the performance for tracking parameter perturbation and disturbances is better than that of the PID controller.The main research of the paper is as follows:Firstly, relative theories of FNN are summarized. A kind of FNN using compensat- ory fuzzy operators is studied in this paper, including structure of networks,learning algorithm, and so on.Second, genetic algorithm is studied. On the basis of traditional genetic algorithm, the paper presented a hybrid genetic algorithm through the analysis on the shortage of traditional genetic algorithm. The hybrid algorithm adopted the self-adaptive adjustment of crossover probability and mutation probability, combines the gradient descend method to improve the efficiency of the searching progress.Third, an optimization strategy based on genetic algorithm with two steps is presented. The clustering method is used to automatically separate the space of input-output data, rough estimates of the parameters describing the fuzzy sets. Then optimize the structure and parameter of controller with improved genetic algorithm.Fourth, the model of controller of water level in marine boiler based on Compensa- tory FNN had been built. The paper signed the compensatory FNN Controller based on optimization strategy presented above.Fifth, Simulation. Simulation was also performed to the boiler control method of using compensatory fuzzy neural net work by Matlab software and the result proved its validity.