空间电力负荷预测中异常数据的辨识与处理

Abnormal Data Identification and Treatment in Spatial Electric Load Forecasting

空间电力负荷预测中所使用数据的质量将直接影响预测的效果和准确性。未经处理的馈线历史负荷数据中经常存在各种异常情况,导致预测精度的降低,因此预测前需将原始数据进行预处理。将馈线历史负荷数据中出现的异常情况进行分类并利用负荷的横向相似性和纵向相似性对传统横向比较法进行改进,提出了一种新的馈线历史负荷数据预处理方法 -双向比较法。以吉林市城市电网实际馈线为例进行分析,结果表明该方法是有效的。

The quality of data used by spatial electric load forecasting will directly affect the forecasting results and accuracy. Untreated feeder load data usually exist in various anomalies, which lead the forecasting preci- sion to be low. Therefore, the original data should be handed before forecasting In this paper, the abnormal situ- ations of feeder historical load data are classified and the traditional model is improved according to horizontal similarity and longitudinal similarity of the load. Finally, this paper puts forward a new method used for pre- processing the feeder historical load data （ Bidirectional comparison method）. The validities of the method are demonstrated by analysising the actual power system in Jilin city.