A Study Of Fault Diagnosis Algorithm For Wind Turbines

In recent years,the problem of energy crisis has received increasing attention from the community.As a kind of clean and renewable energy,wind energy attracts more and more attention.However,the operation and maintenance cost of the wind turbines accounts for about 15%of the total cost of wind power generation,which immensely increases the cost of wind energy employment.Repairing and replacing the broken parts after wind turbine failure accounts for the majority of the cost.Therefore,to reduce the expenses of wind turbine maintenance,researching on fault diagnosis method of wind turbines has become a hot topic in wind power generation industry.At the beginning of the paper,the status of domestic and foreign wind turbine fault diagnosis research is studied.Then common fault characteristics of wind turbines and features of supervisory control and data acquisition system are introduced.Sequentially,frequently used machine learning methods are understood.The paper then proposes a wind turbine fault diagnosis algorithm based on support vector regression with learning samples of scada data.Compared to traditional machine learning algorithms,the support vector regression algorithm has the advantages of strong convergence,good generalization performance and fast training speed.After introducing the learning strategy of the model,the support vector regression model is established by using the wind turbine scada data provided by the State Power New Energy Ltd,and the performance of the model is tested.Test results show that the support vector regression algorithm based on scada data in this paper achieves a fault recognition rate of 80%,which is satisfactory.