Research On Wind Energy Characteristics Based On Wind Farm Data Fusion

In order to solve the problems of fossil energy shortage and environmental pollution,all countries in the world are vigorously developing new energy power generation industry.Because of the advantages of clean and reliable wind power,its development and utilization have received extensive attention.How to effectively utilize wind energy and ensure the safe and effective operation of wind turbines are the core research issues of wind power generation.To solve this problem,it is necessary to study the wind energy characteristics of wind farms in depth.To study wind energy characteristics,it is necessary to sort out and analyze the huge operation data of wind farms.Traditional statistical analysis is difficult to achieve in-depth study of wind energy characteristics.As a typical representative of intelligent methods,machine learning has powerful data analysis and calculation capabilities,which provides a new way of thinking for the study of wind energy characteristics.The appropriate wind speed probability model can describe the wind speed distribution and fluctuation more accurately,which is of great significance for wind energy characteristic evaluation and wind turbine grid-connected stable operation.In this paper,the Weibull distribution probability model is used to fit the wind speed probability of the wind farm studied.In this paper,the particle swarm optimization(PSO)algorithm in intelligent algorithm is used to estimate the scale parameters of Weibull model,and the goodness of fit is compared with that of numerical method.The results show that the goodness of fit of PSO parameter estimation model is better than that of numerical method in judging coefficient,correlation coefficient and chi-square test.At present,experts and scholars mainly focus on wind speed correction and wind speed prediction,and few people study wind speed measurement devices.Wind speed of wind farm depends only on the anemometer at the tail of the engine room.The anemometer is very important for the safe and stable operation of wind turbine.The working state of the anemometer can not be accurately obtained through the operation data.Based on the feature engineering,this paper proposes a wind speed identification model based on genetic algorithm optimization supportvector machine,which uses the large data of wind turbine operation to carry out the performance of the model.The validation results show that the model has more advantages than the neural network and traditional support vector machine.The recognition rate of wind speed is 92.15%.The model can realize real-time detection and correction of the working state of the anemometer.In view of the abnormal power generation of some units in wind farm,this paper analyses the relationship between yaw error angle and power of abnormal units,and concludes that the mechanical anemometer can not accurately adjust the yaw of units due to the existence of yaw error,which leads to abnormal output power.In this paper,a yaw correction model based on random forest is constructed.The yaw error interval is divided based on the yaw principle.The wind turbine operation data equipped with lidar wind measuring device are used as validation data.The validation results show that the accuracy of the model is as high as 96.2%,which can realize the yaw correction of wind turbines.