Fault Early Warning Of Wind Turbine Based On Big Data Analysis

The traditional method of fan fault warning is commonly realized by setting the constant threshold of a single variable,however it is probably to lead to the failure of false alarm or not alarm with little troubleshooting time under the actual complex conditions.The running performance of fans are different under virious ambient conditions,the power generation、grid connected power and temperature value are not same,so it fails to meet the requirements of fault early warning under changing conditions just relying on the constant warning value.Aiming at this problem,the paper put forward a new method that identify the abnormal fans and predict the operating condition based on large data analysis.Through analysing the operating conditions of all fans of a certain large-scale wind farm in chigu area of hebei,the cluster analysis method is adopted to divide the community of fans which the conditions is similar in the wind field,combining with the analysis of relevant historical data in the field SCADA system.Then draw the boxtype distribution drawing of fan temperature parameters in each fan community,identifing the abnormal fan of the community according to the distribution characteristic of the outliers in the box diagram.On this basis,j udge the abnormal fan whether the differernce is significant based on the significant difference analysis method so as to identify the abnormal characteristics of the outlier fans.In order to eliminate the interference caused by accidental factors,use the method of statistical analysis of abnormal rate of sliding window to eliminate the interference of singular points of wind turbines,realizing the identification of abnormal fan among community.In the hadoop big data analysis platform,analyze the whole wind field based on the distributed storage and parallel computing method,realizing the identification of abnormal fans in all communities.To predict the variation characteristics of abnormal fans,the method of linear regression analysis is used to model history data of abnormal fans and analyze the prediction residual.Combining with field experience,set reasonable prediction residual warning threshold to realize the fault warning of abnormal fans.