The traditional model-based stochastic control method needs a precise system model, but, it's difficult to build up precise mathematical model of plants with highly nonlinearity and uncertainty. There are many troubles and difficult to control this kind of nonlinear stochastic systems. so we studied some control strategy based on the neural networks and fuzzy systems. At first, some control methods based on the plant's neural network forward model and inverse model, such as inverse model control and nonlinear predictive control are implemented and used in the simulation of nonlinear stochastic systems control. Then, a linear model is extracted from a nonlinear neural network model, we can use the minimum variance control method and pole placement method on this linear model to realize the self-tuning controller for nonlinear system. This extend the linear design method to nonlinear system. A kind of minimum variance control method based on nonlinear Takagi-Sugeno fuzzy model is proposed in this paper, this method extend the minimum variance control form linear system to nonlinear system. Fuzzy controller have strong disturbance rejection ability, but its rules and member functions are based on the expert's experience, ANFIS not only have the learning ability of neural networks but also have the knowledge representation ability of fuzzy inference systems, so we suggest use ANFIS to learn the expert's experience, ANFIS can automatically generate the fuzzy rules and member functions and realize the slef-learning controller. This paper implemented these control methods based on Matlab, NNSYSID, fuzzy logic and other toolbox, the simulation results show the efficiency of these methods. |