Electricity Theft Detection Using Machine Learning Techniques to Secure Smart Grid

Abstract

Non Technical Losses (NTL) is major problem in power system and cause big revenue losses to the electric utility. The Electricity Theft Detection (ETD) is important topic of research over the years and achieves great success in efficiently detecting the electricity thieves. Further research is needed to improve the existing work and to overcome the problems of data imbalance and detection accuracy of electricity theft. In this paper, we propose a solution to address the above two challenges. The propose solution is consists of Long Short Term Memory (LSTM) and Random Under Sampling Boosting (RUSBoost) technique. Firstly, the data is pre-processed using data normalization and data interpolation. The pre-processed data is further given to LSTM module for feature extraction. Finally, refined features are passed to RUSBoost module for classification. This technique is efficient in solving the data imbalance problem without causing the loss of information and overfitting problems. For evaluation, the proposed model is compared with the state-of-the-art techniques. The experimental results show that our proposed model has achieved high performance in terms of F1-score, precision, recall and Recieving Operating Characteristics curve. The proposed technique is efficient and performs better for recovery of revenue losses in electric utilities.