Study On Wavelet Networks And Its Application For Hydro-Turbine Governing System

This dissertation mainly discusses issues on the applications of wavelet networks in the hydro-turbine governing system, which does give a new approach of intelligent identification and control strategies for the hydro-turbine governing system based on wavelet networks. The major contents are summarized as following:In Ch.1, a detail summary is given on the hydro-turbine governor and governing system control strategy as well as the development and research state of wavelet analysis and wavelet networks in control fields, the aims studied and the significance of this dissertation are also presented.In Ch.2, a new method that uses genetic algorithm and orthogonal experimental technique to optimize structure of wavelet network is presented and proved to be the effectiveness. In equal conditions the wavelet networks have advantages than the neural networks is also proved.In Ch.3, the wavelet network identifiably of stable SISO nonlinear discrete dynamic system is proved in theory. The chapter presents new algorithms, namely greed+proportion differential , changeable forgetting factor recursive predictive error method and its localized method of single right, and proves convergence of algorithms, and the simulation results are given for hydraulic turbo-generators online identification. Based on the time-frequency domain property of wavelets, a method for affirmation first value of wavelet network parameters is given.In Ch.4, the real nonlinear dynamic simulation model of the hydro-turbine governing system is established in simulink soft. The model can react nonlinear internal character of the hydro-turbine governing system and becomes a research flat of realizing senior intelligent control for the hydro-turbine governing system.In Ch.5, for nonlinear systems, the sufficient conditions of the plant inverse model existence are given by introducing the wavelet networks. A aim function of broad sense optimized is defined and a adaptive inverse control method of the hydro-turbine governing system using wavelet network as well as a hybrid control strategy for PID control and adaptive inverse control combine are proposed. The simulation model experiment results show that the effectiveness for the control strategy.In Ch.6, the main research achievements in this dissertation are summarized, and some trends are surveyed and proposed for further.