母线负荷预测中样本数据预处理的新方法

A New Data Preprocessing Method for Bus Load Forecasting

选择SCADA量测数据作为原始数据源,针对目前母线负荷数据中3类典型的异常数据,提出了一种样本数据预处理方法。采用改进的数据横向比较法识别并修正数据丢失点和由突发事件等原因引起的异常突变点,随后采用db4小波阀值去噪法处理由信道噪声等原因引起的数据波动,使负荷曲线平滑化。该方法能够有效识别连续数据丢失点和异常突变点,在保持原有负荷曲线变化趋势的基础上剔除异常波动数据,实现平滑处理,为下一步直接进行母线负荷预测提供高质量的样本数据,在一定程度上提高最终的预测精度。

Selecting the measured data from SCADA as original data source, a data pre-processing method for sample data is proposed to deal with three kinds of typical abnormal data in bus load data. The improved horizontal comparison is used to recognize and modify the points where data is lost and the abnormal catastrophe points of data caused by sudden event; and then the data fluctuation due to channel noise and other reasons are dealt with by db4 wavelet threshold denoising method to smooth the load curve. The proposed method can effectively recognize continuous points where data is lost and abnormal catastrophe points of data, and on the basis of keeping variation trend of original load curve the abnormal fluctuate data is rejected, thus the smoothing processing is implemented and the pre-processed high quality sample data can be provided for subsequent bus load forecasting, and the accuracy of final bus load forecasting can be improved to a centain extent.