Control Method Based On Neural Network Prediction

Along with advancement of industrial control demand, development of control theory and computer technology, a predictive control algorism is produced with effective control and strong robustness, which is applicable to complex industrial processes and the control system, and is successful applied in petroleum, chemical industry, metallurgy and mechanism, and have a good prospect in application. In this paper, it expounds the basal theory and structure of predictive control, discusses detailedly dynamic matrix predictive control including its predictive model, its methods of revising feedback and receding horizon optimization, its stability and robustness, it analyzes how the parameters affect controlling effect. On the basis of pointing out the problems, we propose the idea that the dynamic matrix control combines with neural networks. Actually it uses the neural networks as the predictive models to produce the predictive signals, the control law is solved by optimized algorithm, accordingly control nonlinear system. We choose BP network and RBF network as predictive models because they can approach the nonlinear function very well. As the recursive error is accumulated largely when the single BP network is regarded as predictive model, we propose the paratactic predictive control algorithm based on multi-BP networks. Many single BP networks make up of predictive model paratactically, each predictive model can not be affected each other, so the error is not accumulate. At the same time, we make the simulation. As the structure of RBF network is confirmed easily, fast convergence and the capability approaching the nonlinear function very well, we propose a nonlinear dynamic matrix control algorithm based on RBF neural network, make the simulation, the result shows its application. After discussing the basic theory of generalized predictive control, we make neural network, generalized predictive control and pole placement technique combined very well, propose a generalized predictive pole placement control algorithm based on second-order Adaline neural network, analyze its rationality in theory. This algorithm affords the new measure and approach for solving the control problem of complex nonlinear control system.