GA based fuzzy neural network generalized predictive control method

摘要

A recurrent fuzzy neural network (RFNN) modeling based on genetic algorithm is designed to control the pH neutralization nonlinear process by generalized predictive controller (GPC). GA is used to optimize the number of fuzzy rules, the centers and widths of Gaussion membership function. The recurrent least square method is utilized to obtain the weights of sequent part in the RFNN. Thus, the dynamic RFNN model is obtained with high precision to pH neutralization process. The linearized model at every control period is derived and the nonlinear optimization problem in generalized predictive control is simplified. Simulation results of pH neutralization process show that the proposed method overcome the complexity of neural network based generalized predictive control, and ensure the control precision and robustness.