A novel initial method of fuzzy wavelet network in once-through steam generator control

Since the outlet pressure of generator becomes a significant variable in the process of controlling to influence the efficiency of secondary loop directly, the PID control system of fuzzy wavelet network (FWN) is proposed to maintain the stabilization of pressure in this paper. By establishing FWN with MIMO structure, the network outputs provide parameters for enhance PID controller online. As the need of learning rate and control precision, a method of parameters initialization based on state estimation in wavelet neural network is presented connecting with system information such as the empirical initial values. In the process of associating with empirical values, state estimation is introduced into weights initialization, and finite element method is applied to obtain a priori probability density function of the state to calculate the initial parameters by Bayesian estimation. The FWN initial method proposed for enhance PID controller could effectively improve the accuracy of the controller parameters by more appropriate FWN initial weights, and thus the dynamic complex could be controlled more adaptively by the FWN enhance PID controller with parameters online.