Study On Nonlinear Identification And Control Optimization Of Pumped Turbine Governing System

Energy is the foundation that supports the survival and development of humankind.With the continuous development of China’s national economy and the continuous improvement of people’s living standards,the demand for electricity grows rapidly and the load peak and off-peak difference increases continuously,thus the lack of peak regulating ability has become the prominent problem that constraints the development of power system.The unique function of pumped storage power station for peak clipping and valley filling and its operational characteristics of rapid response to dispatching requirements makes it important in China’s energy structure adjustment and development layout.However,because of the complex layout of the diversion system,the frequent transformation of the working conditions,the reverse operation of the unit,and the inherent "S" characteristic of the pump turbine,the problems faced by the unit’s safe and stable operation have become increasingly prominent,the dynamic coupling effect of pumped storage units in waterelectricity-energy conversion process has become more and more complex and the problems faced by the safe and stable operation of the generating units has become increasingly prominent.Under this background,aimed at solving the scientific problems and engineering technical difficulties faced by the safe,stable and efficient operation of pumped-storage power generation systems,this paper takes the accurate identification modeling and optimization control of the pumped-storage generating units as an breakthrough point.Based on the deep analysis of the complex nonlinear characteristics for the segments of governing system and supported by the nonlinear system identification theory,the advanced control law and the intelligent optimization algorithms,the nonlinear modeling,the parameter identification,the uncertainty analysis,the robust system identification and the multi-objective optimization control of the pumped turbine governing system is researched and analyzed.An integrated system of the identification-control optimization of the pumped storage unit system is the proposed.The main research results and innovations of the paper are as follows:(1)The modeling theory and method of the different parts of the pumped turbine governing system are synthetically studied.The nonlinear characteristics of the servomechanism,the pump turbine etc.are analyzed emphatically.In view of the difficulty in modeling the "S" characteristic region of the full-characteristic curve of pump-turbine,based on the introduction of improved Suter transform for curve preprocessing,an extreme learning machine model based on automatic encoder-partial least-squares regression is proposed to realize the high precision modeling of the full characteristic of the pump turbine.Further,the research work sets up a nonlinear simulation model for the regulating system of pumped storage unit,which lays a foundation for the research of identification and optimization control of the nonlinear system.(2)In order to solve the problem of accurate parameter identification caused by the uncertainties and strong nonlinear characteristics of the pumping storage unit,the nonlinear system parameter identification theory based on intelligent optimization algorithms is studied,an improved backtracking search algorithm combined with the orthogonal initialization technique,the chaotic local search algorithm,the elastic boundary processing and the adaptive mutation scale coefficient strategy is introduced,which realizes the high precision identification of the parameters of the nonlinear pumped turbine governing system.(3)In order to quantify the uncertainty of the identification results caused by random observation noise and the lack of prior knowledge,a stochastic inversion framework based on the Bayesian theory is studied and constructed.The parameters set to be identified is combined with the system observation data,the theoretical model and the empirical knowledge information.The differential evolution adaptive Metropolis method is introduced to sample and traverse the posterior probability density function of the identification parameter,and a posterior probability distribution containing uncertainties of parameter solution sets is obtained,which describes and evaluates the uncertainty of the identification results intuitively and provides a new idea for the improvement and development of the existing parameter identification strategies.(4)In view of the complex and changeable operating environment of pumped storage power station and the noise and outlier interference in the modeling data,this study proposes a sparse robust least squares support vector machine identification model for the pumped turbine governing system.By introducing the maximal linear independent group,the sparsity of the support vector is realized,and the complexity of the model is reduced.The robustness of the model to the noise and outliers is enhanced based on the weighted function of the improved normal distribution.In order to further improve the accuracy and generalization performance of the model identification of the storage unit adjustment system,the overall optimization of the model input variables,the kernel function parameters and the regularization parameters is realized by introducing a binary-real coded hybrid backtracking search optimization algorithm.Experiments on multiple nonlinear systems show that the proposed HBSA-S-R-LSSVM model not only has high identification accuracy,but also good robustness and generalization performance.(5)Considering that the traditional PID control law of the pumped turbine governing system is not adaptable to changes in environment and working conditions,and the low and middle head no-load operation is susceptible to the frequency oscillation caused by the “S” characteristic region,the fractional PID control strategy is introduced and a multi-objective optimization framework that takes into account the control stationarity,quickness and the robustness under multiple operating conditions is constructed.A multi-objective nondominated sorting genetic algorithm(LCNSGA-III)based on the Latin hypercube experimental design and the chaos optimization theory is proposed to achieve the efficient solution of the optimization problem.The test results of multi working conditions show that the proposed multi-operating condition and multi-objective fractional order PID optimization control strategy based on LCNSGA-III can effectively improve the dynamic response performance of the pumped storage unit and improve the control quality and adaptive ability of the governing system.