The Research Of Neural Network Generalized Predictive Control For Thermal Power Unit

Thermal power unit is a system which is complex, nonlinear, time-variant with long-time delay and difficult to establish accurate model. So it is hard to make system gain optimum running effect with conventional control strategy. Here, Generating Unit coordinated control system and its model are analyzed after reading abundant relevant literature, then the strategy of generalized predictive control combined neural network is proposed, which used the neural network as the identified model of control object to produce predictive signal, the control law was solved by generalized predictive control arithmetic.The math model is established with BP algorithm which combines changing step and inducting momentum to improve convergent rate, when the precision reached a certain extent, the recursive predictive model is achieved through on-line identification, finally the optimized control law is acquired by minimizing the function of performance guideline. This algorithm not only improved the precision of model identification but also avoided on-line computation of Diophantine equations. Furthermore it simplified the calculation of matrix inversion, so the controller?ˉs burden was decreased drastically and the real time character of arithmetic was improved. The result of MATLAB simulation shows that the arithmetic ensures the outputs could follow initialization speedily and stably, and has better robustness than PID control.