Using Machine Learning Analytics to Detect Abnormalities and Electricity Theft

Abstract

Abnormalities and Electricity Theft are a major concern for power and economic chaos of one's country. The reason here is, the fraudulent usage of electricity power by customers and broken electric meters or billing errors. Currently, electrical transmission and distribution losses remain a hurdle to the development and sustainability of the sector despite several techniques of energy conservation and electricity distribution analysis that have been employed. While technical failures are regular and predictable, non-technical losses, which are responsible for 80% energy losses, are random and hard to identify and evaluate. Hence it requires more advanced technology. For these reasons, the problem has attracted research interests in many fields, including artificial intelligence, including machine learning and expert knowledge approaches. Here, we have used a linear regression method for anomaly detection. The project therefore showed that the method has improved detection accuracy, sensitivity and reduced magnitude of data required.