Enhanced encoding technique for identifying abnormal energy usage pattern

Transmission and Distribution (T&D) of electricity from a power generation station involve substantial losses. T&D losses include technical as well as nontechnical losses (NTL). Most portion of the NTL constitutes of electricity theft. This paper explains the significance of the evaluation of customer energy consumption profiles for identification of illegal consumers. To reduce the complexity of the instantaneous energy consumption data for evaluation, this paper proposes and implements a data encoding technique. This encoding technique maps instantaneous energy consumption data into irregularities in consumption. In addition, exclusivity in each customer's energy consumption has been preserved. After the encoding process, the data has been inputted to a support vector machine (SVM) classification model that classifies customers into three categories: genuine customers, illegal consumers or suspicious customers. Classification accuracy of the SVM model with the encoded data is 92%. The obtained results demonstrate that this encoding procedure is significantly quick and robust in identifying (classifying) illegal consumers.