Abstract：Historical load data is the basis of load forecasting of power system, the abnormal historical data affect the accuracy and effectiveness of load forecasting, hence, it is necessary to identify the abnormal load data. This paper takes a node load data for research object, and presents a method for abnormal load data identification based on two times clustering algorithm, using fuzzy clustering algorithm combining with validity index to cluster the load curve for the first time; using the clustering results combined with neural network to cluster the load curve for the second time and extract daily load characteristic curve; according to the similarity and smoothness of load curve to identify the abnormal load data. The results of bad data identification in examples indicate that the proposed method is feasible and effective. The example analysis result shows that this method is effective.
王珂, 王天秀. 基于二次聚类的电力负荷异常数据辨识[J]. 电气技术, 2014, 15(11): 1-3.
Wang Ke, Wang Tianxiu. Identification of Abnormal Power Load Data Based on Two Times Clustering. Electrical Engineering, 2014, 15(11): 1-3.
2014, Vol. 15 
