Identification And Control Of Variable Pitch Wind Turbine System Based On Neural Network

With the continuous update and development of wind power generation technology,variable-pitch wind turbines have been widely used in the field of fans due to its stable system,small structure and convenient shutdown.However,it is worth noting that the variable-pitch wind turbine system involves knowledge of aerodynamics,mechanics,electronics,and energy,moreover,it is influenced by external disturbances and changes in its own parameters,making it difficult to establish accurate mathematical model during operation.Many advanced control technologies are based on the controlled object model,which makes the identification of the nonlinear model of variable pitch wind turbine system become the focus and difficulty of wind power technology research.This thesis focus on the output power control and system identification of variable-pitch wind turbine systems based on the MATLAB and BLADE platforms.The main research contents include:1.The research about identification of Variable Pitch Wind Turbine system(1)Based on the analysis of the nonlinear characteristics of the variable pitch wind turbine system,combined with the neural network has a strong ability of self-learning,and can approach the characteristics of a nonlinear system with arbitrary accuracy,therefore,the method of neural network identification was applied to the identification of a complex variable-pitch wind turbine system.Aiming at the problem of BP neural network is easy to fall into the local minimum when using log-sigmoid function,and the training iteration time is long in pitch-wind turbine non-system fitting training,a local approximation RBF neural network algorithm is proposed to solve the above problems.(2)Aiming at the problem that improper learning rate easily caused the deviation for network during the process of weight training.A combination of gradient descent method and error dynamic feedback optimization algorithm is proposed to optimize the learning rate.Experimental results show that the identification accuracy of pitch-variable wind turbine system can be improved through the proposed optimization algorithm.(3)Aiming at the problem that the network structure is difficult to determine and the real-time performance is not strong,the sensitivity analysis method is adopted to optimize and adjust its structure in this thesis,which delete the nodes that have less impact on the output,and update the nodes and parameters that have a greater impact on the output.Simulation experiments show that the proposed algorithm can improve the identification accuracy of variable pitch system of wind turbines.2.The research about Output Power Control of Pitch Wind Turbine System.(1)High-precision control is the premise of stable output of fan power.Therefore,a suitable controller is designed to control the output power of the fan based on the identified model.However,the traditional PID control can't adjust its parameters adaptively,resulting in unstable controller output.The fuzzy system can be used in the control of nonlinear systems with fuzzy linguistic rules,robustness and strong antiinterference ability,so it is combined with the radial basis neural network with strong learning ability.Due to the advantage of the highly symmetrical and the smooth zero point,the Gaussian function is selected as its membership function to dynamically modify the PID parameters.A PID controller based on fuzzy radial basis neural network is designed to control the variable pitch wind turbine system.(2)Aiming at the problem that the initial value of the parameters of the fuzzy neural network PID controller is difficult to be selected,the genetic algorithm is used to optimize the initial value of the PID parameter,and the square of the control variable is added to the optimization index to prevent overshoot.Eventually,experimental results show that,the fuzzy RBF neural network PID controller has better performance with less training iterations and higher efficiency compared with traditional PID controller.