An intelligent hybrid model for reducing non-technical losses in electrical industry

Abstract

Non-technical losses, specially electricity theft is a major concern for electricity distribution companies all over the world due to its huge financial impact on their revenue. For developing countries such as Iran, the case is even more important because of the fact that the use of advanced metering infrastructure have not been implemented completely. Since the data mining and its techniques have been widely used nowadays and proven to be useful in so many cases like fraud detection problems, it has been decided to use this science in order to tackle non-technical losses problem for an Iranian company by focusing on meter tampering or it may also be referred as fraud in electricity which plays a huge role in creating none-technical losses. The proposed model proved to be applicable for the company by having a much better performance using multi agent ensemble model of SVM, RF, and NN, in comparison to the company’s traditional statistical solution which could merely predict less than ten percent of fraud cases correctly.