Research And Application Of Predictive Control Based On Elman Neural Network

Nonlinear and time-delay are phenomenon during many industrial processes, especially in petroleum,chemical industry and metallurgy. As the core equipments in chemical process,continuous stirred tank reactors(CSTR) are widly used in polymerization industry. In dye,medicine,reagent,foodstuff used widespread. CSTR have many variables and contains characteristics such as noninear, time-varying, time-delay. And it is very difficult to control the process using the classical control theory.To solve the problems mentioned above, in this paper at first,it discusses on the basic structure and theories of internal model control(IMC) and propose the idea that IMC based on neural network. After analyze CSTR system, there will be much amount heat during the process, so decrease the amount heat to guarantee the process safety is the main purpose int this paper, it contructures IMC controller based upon neural network. Both the CSTR internal model and the controller were constructed via Elman neural network. LMBP algorithm was used to train the weights of the neural network.The simulation of CSTR system based upon MPCE-1000 demonstrates the algorithm's good performance and better than PID algorithm. The phenomenon of nonlinear and time-delay have been fixed and CSTR system have better dynamic and static performance.