改进Bayesian后验比的异常风速值检测方法  

Anomaly Wind Speed Detection Method with Improved Bayesian Posterior Ratio

作　　者：陈伟[1] 吴布托[1] 裴喜平[1] 王懿喆[1] 

机构地区：[1]兰州理工大学电气工程与信息工程学院,甘肃兰州730050

出　　处：《电网与清洁能源》2017年第2期104-111,116,共9页Power System and Clean Energy

基　　金：国家重点研发计划(2016YFB0601600);国家自然科学基金项目(51267012);甘肃省科技支撑工业计划项目(1504GKCA033)~~

摘　　要：风电场运行数据中含有异常风速值,为了优化风电数据的质量,提出了组合预测与Bayesian后验比的异常值检测方法。为了降低预测误差,先对风速序列建立Adaboost-BP网络和EMD-LV-SVM的组合预测模型,利用预测值与测量值的偏差得到含有粗大误差的残差序列;为了提高检测方法的可靠性,采用Bayesian后验比的检验方法识别残差序列中粗大误差,从而确定异常风速值的位置,并利用ARIMA方法修正异常风速值。RBF预测结果表明,所提方法能准确识别异常值,从而提高了风电场短期风速预测精度。As wind speed data contains abnormal values from wind farms, in order to optimize the quality of wind power data, this paper proposes an outlier detection method with the improved Bayesian posterior ratio. To reduce the prediction error, the paper establishes a combination forecasting model based on BP network and least square support vector machine. The residual error sequence is obtained by calculating the deviation between the predicted value and the measured value. The gross errors in the residual series are identified by the test method of Bayesian posterior ratio,and this approach can improve the reliability and determine the location of the abnormal value. Finally, we use the ARIMA method to correct the abnormal wind speed. RBF prediction results show that the proposed method can accurately identify outliers, thus improving the forecasting accuracy of wind speed.

关 键 词：异常风速值检测 组合预测模型 残差分 Bayesian后验比 

分 类 号：TM614[电气工程—电力系统及自动化]