Nonlinear State Forecasting And Stability Control Of Hydro Turbine Regulation System

In environmental economics,environmental resources including environmental quality are categorized as amenity resources.Due to its importance to human welfare,the amenity resources theoretical study and valuation is an ongoing issue at the academic frontier in the environmental economics area.Water resources are important renewable and clean resources,and priority is given to the development of hydropower.With the rapid development of the hydropower industry,more than half of the world’s hydroelectric generating units with a capacity of 700,000 kilowatts or more are in China.Such a large generator set makes the hydropower system face higher requirements in terms of stability and safety in operation.Generator,governor,generator port side power load,water diversion system,etc.together constitute a turbine regulation system,with a variety of typical operating conditions,belonging to a complex nonlinear system that integrates electricity,machine,water,etc.Whether it is operating normally or not will have an impact on the reliability and safety of the overall operation of the hydro-generator unit or the grid system.In the production practice,large and medium-sized hydroelectric generating sets at home and abroad have different degrees of stability problems.Therefore,the stability analysis and effective control of the turbine governing system are of great significance for the safe operation of hydropower stations.The state prediction of the turbine governing system is the premise and basis of stability control.This paper focuses on the key issues in the nonlinear prediction and control of state parameters of hydraulic turbine regulation system.The main work is as follows:(1)The RBF prediction of the turbine governing system was studied.In view of the superiority of RBF neural network in nonlinear time series prediction and the effective information extraction function of phase space reconstruction,the RBF neural network prediction model is combined with phase space reconstruction to study the time series prediction model of the Francis turbine regulation system.Firstly,using the Runge-Kutta method to solve the nonlinear equations of the hydraulic turbine,the chaotic time series is obtained,and the chaotic characteristics are verified by the Poincare section method.Secondly,using the two-parameter independent evaluation method and the two-parameter simultaneous expansion method to obtain the two important parameters of the time-phase reconstruction of the turbine operating system’s operating state time series,namely the embedding dimension and the delay time,The RBF neural network single-step and multi-step prediction models are used to substitute the embedding dimension and delay time obtained by the two methods into the RBF neural network single-step and multi-step prediction models,and analyze the difference between the valuation method and the extended method.Predicting the operating state of the turbine governing system with embedded dimensions and delay times.The numerical simulation results show that the RBF neural network model has a good predictive effect on the Francis turbine adjustment system.(2)The genetic wavelet neural network prediction of the turbine governing system is studied.In view of the good multi-input parallel processing and nonlinear mapping ability of BP neural network,the training speed is slow and it is easy to fall into the local minimum.The wavelet theory has good time-frequency local characteristics and zooming ability,and the genetic optimization algorithm has excellent global optimization characteristics.Based on the BP neural network prediction model of the turbine governing system,the wavelet neural network is constructed by using wavelet elements instead of neurons,and the weights and thresholds of each layer of the wavelet neural network are optimized by genetic algorithm,and the existing global optimization search is obtained.The genetic wavelet neural network prediction model with good local optimization solution performance shows that the genetic wavelet neural network has high nonlinear approximation ability,and has more prominent prediction accuracy and convergence speed.The effect is better in optimizing BP neural network performance.It can effectively solve the problem that the state of the turbine governing system is complex,the common neural network prediction model is difficult to predict accurately,and the training speed is slow and it is easy to fall into the local minimum.(3)The nonlinear prediction of hydraulic turbine governing system based on clustering empirical mode and genetic support vector machine is studied.Due to the application of clustering empirical mode decomposition,the nonlinear and non-stationary signals can be decomposed into stationary narrow-band signals with different feature scales.The support of vector machines can also make the modeling of systems affected by many factors complicated.The problem was solved properly.And with the following characteristics,such as the ability to highlight generalization,global optimization,etc.,a new prediction method based on the combination of empirical mode decomposition and genetic support vector machine is proposed.Combined with spline interpolation,the discrete adjustment data is a uniform sampling of multiple data,and the fitted smooth curve is obtained.Then,using the spatiotemporal filtering characteristics of clustering empirical mode decomposition,the intrinsic mode and residual component reflecting the data trend are obtained.In addition,based on noise-assisted signal processing,combined with the addition of a small-amplitude white noise equalization signal,the human factor influence and modal aliasing problem at the time of decomposition are solved.The intrinsic modal components processed by empirical mode decomposition are input into the GA-SVM model as samples,that is,the kernel function and the optimal parameters are selected,and the components are predicted.By superimposing all component prediction results,the overall prediction result is formed.And analyze this.Based on numerical simulation,it can be recognized that the GA-SVM regression model uses the weighting method of different error coefficients to make the generalization performance and accuracy of the model prediction significantly improved.(4)The finite-time terminal sliding mode control method for the turbine governing system is studied.Considering that in actual operation,the turbine regulation system is often affected by the load change of the power system,and the random disturbance is introduced on the basis of the established nonlinear turbine regulation system model.In order to overcome the singular problems in the traditional sliding surface design,based on Lyapunov stability theorem and finite time lemma,a novel robust terminal sliding mode controller is designed for the turbine governing system by using the terminal sliding mode control strategy.It shows that the system can achieve stability quickly in a limited time,and verifies the effectiveness of the proposed control strategy.In view of the fact that fractional-order systems are more capable of describing actual physical processes than integer-order systems,a fractional-order mathematical model of the turbine governing system is given.Unlike integer-order systems,the development of fractional-order system stability theory is not yet mature.In proving the stability of the system,a method of transforming the fractional-order model into an equivalent integer-order model is proposed by applying the frequency distribution model.Finally,based on the sliding mode control strategy,this paper designed a fractional-order robust terminal sliding mode controller for the turbine governing system.The effectiveness of the fractional-stage turbine regulation system is verified by numerical simulation.(5)Exploring the control method of nonlinear fuzzy prediction function of the system.Because the system has many characteristics,such as non-minimum phase and nonlinearity,the T-S fuzzy model is used to describe the parameter-indetermined integer-order hydraulic turbine governing system with random disturbance.The corresponding control law is proposed by combining parallel distributed compensation technology.The Lyapunov stability theorem is theoretically proved by the effect of the R&D state feedback controller,and the gain matrix is solved by the application of the linear matrix inequality.The fuzzy neural network decoupling method is used to uncouple the systems,and the linearized turbine regulation system is transformed from a multivariable system to a single variable system.For the non-minimum phase system,in order to ensure its control performance,a zero-point configuration method is proposed,and the characteristics of the zero-point configuration system and the rationality of the zero-point configuration are analyzed.Combined with the proposed T-S fuzzy control and predictive function control method,a suitable nonlinear fuzzy predictive function control method for hydraulic turbine governing system is proposed,which provides a reference for the stable control of hydropower station systems.