Electricity Loss and Fraud Prediction with Deep Learning

Abstract

In developing countries, energy usage has increased with remaining population, industry, widespread technology and the increasing trend of economy. The main energy source of this increase is electricity. From this perspective, the forecasting of electricity fraud has an important role in the control of this trend to support, run, plan of distribution network's investment. High percentage of fraud in this region damages both the region economy growth and also electricity distribution network. The main source of Fraud usage comes from industry so fraud detection is very hard. So with the correct analysis of daily usage, the usage before theft and last usage of electricity which retrieved from Automatic Meter Reading System (AMRS), we can forecast future theft with Deep Learning. If we use more than one method so we can decide to use one that gives us the best proven.