基于Box-Cox变换结合多种算法的风电机组数据预处理方法研究

Research on Wind Turbine Data Preprocessing Method Combined with Multiple Algorithms Based on Box-Cox Transformation

由于弃风限电、环境干扰等因素的影响,SCADA系统采集的原始数据中会存在异常数据,对原始数据进行精确有效的数据预处理,是后续故障预警工作的基础;基于SCADA系统采集的数据,对风电机组运行数据的预处理方法进行改进和研究,提出了一种将Box-Cox变换与以正态分布为前提的异常值清洗算法相结合的方法,对原始数据进行预处理;运用Box-Cox变换分别与Bin算法、肖维勒准则、狄克逊准则和格拉布斯准则相结合的方法进行数据预处理,经过实例验证:肖维勒准则的算法简单且检测时间短,但是对于异常数据的清洗效果较差;狄克逊准则和格拉布斯准则对于异常数据的清洗效果较好,但是处理时间较长,对大型风电场海量数据,这种方法的实用性较差;相比于其他算法,Bin算法的优势较为明显。

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 effective 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.