Abnormal electricity consumption detection based on ensemble learning
Zhou Fang, Qing Cheng, Mou Li, Hongyun Qin, Houpan Zhou, Jiuwen Cao · 2019
Abnormal electricity consumption (AEC) seriously affects the management of the power grid marketing department, causing huge economic losses to the power supply enterprises. An accurate abnormal electricity consumption detection system is thus vital to reducing the non-technical losses and the operating cost of state Grid. Comparing to the normal case, the amount of abnormal electricity consumptions is relatively small, leading to a typical imbalance problem. In this paper, an ensemble learning algorithm based on an artificial neural network (ANN) is developed for the abnormal electricity consumption detection. We studied 6 typical AECs, and an monthly amount of electricity consumptions is firstly analyzed for feature extraction and classifier learning. Then, the ensemble learning scheme built on the extreme learning machine (ELM) algorithm and the majority voting method is proposed for the classifier training and AEC classification. Experiments on the electricity consumption data of State Grid Zhejiang Electric Power Corporation are conducted to demonstrate the effectiveness of the proposed algorithm.