Research And Application Of Generalized Predictive Control Based On RBF Neural Network

Nonlinear and time-delay are phenomenon during many industrial processes, especially in petroleum, chemical industry and metallurgy. There will be much amount heat during the process of continuous stirred tank reactor (CSTR). So decrease the amount heat to guarantee the process safety. The main purpose of the paper is to this paper, we will use generalized predictive control based upon neural network to solve the problems mentioned above.In this paper at first, it discusses on the basic structure and theories of generalized predictive control (GPC), then it analysis the predictive model, methods of feedback adjustment and rolling optimization. On the basis of pointing out the problem of the present difficulty and actuality, we propose the idea that GPC combined by neural networks.Radial basis function (RBF) neural network is chosen to construct multi-steps predictive model, because it has the merits of smaller calculation, fast convergence and without local infini-tesimal values. Considering the model error and uncertain factors in the process, compensated strategy is designed to improve the precision of prediction.In this paper, the simulation of CSTR process results show that GPC based on RBF neural networks provides good adaptation, robust and ability of anti-interference for the system.