Nonlinear Inverse System Self-learning Control Based on Variable Step Size BP Neural Network

The authors present two aspects of the neural network identification and control of nonlinear systems. First, a method of neural network identification learner (NNIL) of the inverse nonlinear system is considered. This enables the investigation of adaptive nonlinear systems based on neural network identification. Second, neural network error controller (NNEC) is constructed simultaneously. The structure of the NNIL and NNEC is the same. The weight matrix of the NNEC is dynamically transferred and updated by NNIL. By this way, the adaptive and self-learning non-model control based on neural network is implemented.This controller is applied to a complex nonlinear system, which includes formidable but realistic nonlinear process. The new method is compared with fuzzy control. The comparison shows the new method has higher effective than that one.