Deep Learning Research On Wind Direction Forecasting And Yaw Optimization Of Wind Farm

As a non-polluting,recyclable energy source,wind energy is rich in reserves and widely distributed.However,in practical application,it was found that wind energy has objective defects such as insufficient continuity,strong uncertainty and strong randomness.Yaw control mainly adjusts the yaw direction of the wind turbine according to the real-time change of wind direction,which is the basic guarantee for the efficient and safe operation of the wind turbine.Therefore,accurate wind direction prediction and yaw control are the basis for high-precision control of wind turbines.In response to the above problems,the research is mainly carried out from the following aspects:(1)In view of the characteristics of wind direction sequence such as volatility and chaos,the study analyzes the chaotic characteristics of the corresponding wind direction data;on this basis,a comprehensive data analysis method combined with multiple data processing algorithms is used to analyze the input wind direction data Information feature extraction;(2)In order to achieve high-precision wind direction prediction,the research proposes a deep confidence network based on an improved optimization algorithm to perform deep prediction on wind direction data;that is,the network parameters of a three-layer deep confidence network based on an improved FPA algorithm Carry out optimization and use the optimized network to realize wind direction data prediction.Finally,the study uses an error analysis method based on LSSVM to further improve the model accuracy;(3)On the basis of the above-mentioned wind direction model,the research explores the implementation of yaw optimization,that is,using the multi-objective optimization algorithm-NSGA2 algorithm to maximize wind power generation efficiency and minimize yaw adjustment load as the optimization goal,establish wind power The yaw system optimization model optimizes the yaw adjustment in continuous time periods.This topic uses a model built by the actual monitoring data of a wind farm in the north.The training results show that the multi-processing combination model proposed in this topic has high prediction accuracy,good model stability,and can achieve comprehensive prediction of wind direction,which is then yaw Optimization provides protection;it has certain significance for the economic and safe production of wind power enterprises.