Research on Wind Turbine Data Preprocessing Method Combined with Multiple Algorithms Based on Box-Cox Transformation
Due to the influences of wind curtailment,power curtailment,environmental interference and other factors,there will be abnormal data in the original data collected by the supervisory control and data acquisition(SCADA)system,the accurate and ef-fective data preprocessing of the original data is the basis for the subsequent fault early warning work.Based on the data collected by the SCADA system,the preprocessing method for wind turbine operation data is improved and studied,and a method that combines the Box-Cox transformation with the outlier cleaning algorithm premised on normal distribution is proposed to preprocess the original data.The Box-Cox transformation is used to combine the Bin algorithm,Chauvenet criterion,Dixon's criterion and Grubbs'criterion to preprocess the data.The example evaluates that the algorithm of the Chauvenet criterion is simple and short detection time,and poor cleaning effect on abnormal data;the algorithms of the Dixon's criterion and Grubbs'criterion have the features of good cleaning effect on abnormal data,and long processing time.The applicability of two methods is poor for large wind massive data.Compared with other methods,the Bin algorithm has obvious advantages.
Box-Coxwind power generationdata preprocessingSCADAfault early warning
万方数据
维普
