Research On BGNN Model Identification, Control And Controller Parameter Optimization Of Hydraulic Turbine

The hydraulic turbine is a complex system which is involved with hydrodynamics, mechanical and electrical dynamics. Owing to the strong nonlinear characteristics, the modeling and control problems of hydraulic turbine always attract many scholars’ attention. In this dissertation, the Bayesian-Gaussian neural network (BGNN) and Bacterial Foraging Optimization Algorithm (BFOA) were investigated and explored intensively and furthermore the new identification, control strategies of hydraulic turbine system and optimization method of controller parameters were proposed. The contents of the dissertation are listed as follows:(1) The background and significance of this dissertation are given first before introducing of the development about the controller and control strategies of hydraulic turbine tuning system. The research on system identification of hydraulic turbine is described next. The Bayesian modeling technology and Bacterial Foraging Optimization Algorithm (BFOA) are summarized in the dissertation and the research goal and arrangement of every chapter are also introduced.(2) For the limitation of the hydraulic turbine actual model experiments method, the computer simulation provides an effective procedure for the research of the hydraulic turbine. The conventional mathematical models of components of hydraulic turbine system are described in details. The linear and nonlinear dynamic model of hydraulic turbine are established based on the modularization simulink soft in the MATLAB circumstance which provides a research platform for the contents in the dissertation.(3) The Bayesian-Gaussian identification theory is analyzed. The off-line threshold matrix training algorithm based on swarm intelligence is proposed and the online identification algorithm of the Bayesian-Gaussian neural network is also presented which is driven by the sliding data windows method. The identification experiments on some classical nonlinear system validate the effectiveness of the given Bayesian-Gaussian neural network.(4) The Bayesian-Gaussian neural network is applied to the model indemnification of hydraulic turbine. Based on the description of the identification problem of hydraulic turbine, the Bayesian-Gaussian model structures of hydraulic turbine are designed and the identification methods based on BP and RBF neural network are also introduced for the comparison research with the Bayesian-Gaussian neural network. The simulation experiments on the models identification of hydraulic turbine show that the Bayesian-Gaussian neural network owns the characteristics of easy determination structure, few tuning parameters and satisfactory identification results.(5) The nonlinear predictive control strategy of hydraulic turbine tuning system is designed. The predictive model used in the control strategy adopts the Bayesian-Gaussian predictive model. The nonlinear predictive simulation experiments are conducted on the control of hydraulic turbine and the results validate the proposed nonlinear predictive control strategy.(6) The theory analysis is conducted on the Bacterial Foraging Optimization Algorithm (BFOA) and an improved E.Coli foraging optimization algorithm (IEFOA) is proposed which introduces an optimal value tracing operator to enhance the convergence of the basic BFOA. Five benchmark functions are used to test the effectiveness of the IEFOA. Then the IEFOA is employed for the optimization design of the controller parameters of the hydraulic turbine tuning system. The simulation results denote the feasibility of IEFOA in the optimization design of the hydraulic turbine tuning controller.