Research On Anomaly Identification Method Of Wind Turbine Power Curve Based On Data Driven

The severe problems of energy consumption and environmental protection faced by today's society are seriously related to sustainable development and the rational survival of human beings,which makes the development of wind energy become an important development direction of global renewable energy,and correspondingly,so ensuring the healthy,rapid and stable development of China's wind power industry has become a hot research topic.As an important technical index to evaluate the operation characteristics of wind turbines,the power curve can be regarded as the research object of abnormal running state of wind turbines.Wind turbines will produce a large number of related data in the course of operation,how to manage and make good use of these data for power curve modeling analysis has become a wind power development must face the issue.Data mining technology can be based on historical data analysis,early detection of potential abnormal characterization and effectively reduce the failure rate.Therefore,it is not only of profound theoretical value,but also of wide practical application value to complete the research on the anomaly recognition method of wind turbines' power curve based on data-driven.In order to ensure the validity of the data,on the basis of the in-depth study of the data acquisition characteristics and numerical characteristics of wind speed and output power of wind farms,the improved K-means method is used to clean the data for the absence and abnormality of historical data.Aiming at improving the accuracy and practicability of the power curve model,based on the deep and systematic study of the power curve of the wind turbine,a mathematical model of parameter optimization based on historical data is established for the complex and changeable characteristics of the power curve.The model uses the parametric logistic equation for heuristic expression,and then introduces the particle swarm algorithm to optimize and determine the parameters.In the further study of the output power of wind turbines,it is found that the output power is affected by various factors,which is uncertain.In order to describe the inherent kinetic mechanism more accurately,the power curve model of BP neural network based on genetic algorithm optimization is established,and the model is trained with the measured data of wind turbine,so that it has the function of prediction.Based on the analysis of the running state of wind turbines,the state abnormal detection and conformity calculation method of wind turbines are put forward on the basis of the power curve model.The wind farm in Northeast China is taken as an example to verify the above methods,and the results show that the methods are effective and feasible,and have high reference value.