Application Of Improved Generalized Predictive Control And Performance Evaluation Method In Wind Power System

As a pillar of new energy generation technology,the operational control effect and economic cost of wind power generation are the important focus points in the process of reducing carbon emissions and achieving the goal of carbon neutrality.Because of the development of modern industry and the advances in technologies,more advanced control technologies and intelligent algorithms are gradually replacing traditional control methods to create higher production value and lower cost in actual industrial production,in which there are many problems to overcome and areas for improvement from theory to practical application.In this thesis,we improve the generalized predictive control(GPC)to solve the problem of poor optimization-seeking effect under the input-output limit and apply it to the design of maximum power point tracking(MPPT)strategy for the turbine,and design a SCADA system-based evaluation model for the operation state of wind turbines.The main work of the thesis is carried out in the following aspects:(1)Firstly,the GPC algorithm is improved by the improved PSO algorithm.The aim is to address the problems that PSO is slow in finding the optimal speed and weak in high-dimensional search,as well as easy to fall into local optimum,it is improved by circumventing the speed term update and combining the weight change and particle position update mechanism of the system entropy principle.The improved PSO is combined into the GPC for rolling optimization,which improves the control effect and regulation time of the algorithm.(2)A variable-speed wind turbine maximum power point tracking strategy is designed based on the improved generalized predictive control algorithm.The electromagnetic torque history data is used as a feedforward coefficient to correct the reference trajectory to alleviate the frequent fluctuation of the wind power system output active power,and the NSGA2 genetic algorithm is used to perform multi-objective optimization of the control parameters to make the variable-speed wind turbine output smoother.(3)An online performance evaluation model of wind turbine based on SCADA data is established.The wind speed-power curve is established as a benchmark using BP neural network.The three indicators of power dispersion,wind energy utilization coefficient volatility and dominance ratio after relative degradation degree and normalization are used to find the comprehensive performance indicators by entropy weight method and coefficient of variation method.The kernel density estimation method is also used to find the threshold values,and the condition monitoring of wind turbines is performed in a time-based sliding window.