Detection of abnormal trends in electrical data

Abnormal detection of electrical data has been widely used in the electric power industry. However, traditional abnormal detection algorithms mainly focus on the abnormal value in data of power consumption. Electrical data, which describes electricity consumption of different regions in different time, implies the tendency of the electricity consumption in different areas. By focusing on the change of trend in electricity data, this paper presents an algorithm to detect the abnormal change of electricity trend. By using backtracking dynamic window model, the proposed algorithm can find the abnormal situations of electricity trend that occur under windows with different lengths. Experiments on the real electrical data sets verify the effectiveness of the algorithm.