Neural network-based PID predictive control for nonlinear time-delay systems

A novel recurrent neural network-based PID predictive control structure is proposed for a class of nonlinear systems with large time-delay. In order to overcome the drawbacks of autocorrelation of the prediction errors in direct prediction approach, a new direct cutting-error multi-step prediction method is developed. Based on nonlinear Smith predictor, a PID-type long-range prediction cost function is introduced. In the control process, dynamic recurrent neural networks are used and the weights of the network are trained by the DBP. Then, a simulation example is provided to show the effectiveness of the proposed control strategy.