Recent advances in detection and prediction of customers energy consumption patterns through the use of machine learning techniques

Due to the increasing volume of energy losses from power companies, identifying fraudulent consumers who use illegal practices to gain advantages over energy consumption has become an important process. By analizing data on energy consumption collected from customers, it is possible to trace usage patterns and further employ them to build smart algorithms capable of detecting illegal usage of power, thus aiding companies to fix irregularities and recover the damages caused by fraudulent consumers. In this paper, we present and discuss recent researches that have been published in the literature regarding energy losses, energy consumption forecasts and prediction of illegal consumption. Most of the studies in electricity consumption field aim for trying to detect fraudulent customers based on their kwh consumption, comparing the oldest pattern of the customer analyzed or neighborhood pattern, identifying possible load deviation due to fraudulent usage.