Non-technical losses identification using Optimum-Path Forest and state estimation

This work proposes a new method to identify non-technical losses in typical Electrical Power Distribution Systems (EPDS). Aiming to improve the efficiency of on-site inspection for fraudulent clients, the consumers are classified in suspects and non-suspects using the Optimum-Path Forest (OPF) classifier. The authors propose an improvement to this method using results of a Distribution State Estimator (DSE) and geographic information. The method is tested in the IEEE 123 Bus Test Feeder considering a group of consumers for each load. The consumers are based on typical Brazilian residential consumers. The results indicate that the proposed method improves the performance of previously developed techniques and is suitable for both typical and smart EPDS. The best results are obtained when the DSE and OPF are associated using normalized data as the input for this hybrid approach, with expected success rate of up to 72.43% for on-site inspections in the EPDS test system.