Model-Less Non-Technical Loss Detection Using Smart Meter Data

Abstract

This paper proposes a novel method for estimating non-technical losses (NTL), typically due to electricity theft caused by bypassed meters in low-voltage distribution networks. First, the voltage sensitivity coefficients are estimated by a weighted least squares based approach using the residuals obtained from an ordinary least squares-based approach. Then, the voltage measurements of smart meters and the estimated sensitivity coefficients are used to estimate the actual consumption of customers. The differences between the measured and the estimated consumption values are compared against a threshold and customers whose differences exceed the threshold are flagged as fraudulent customers. The performance of the algorithm is evaluated on the IEEE European Low Voltage (LV) network in various case studies, and on large randomly generated distribution networks. The results exhibit the capability of the proposed algorithm in correctly identifying fraudulent meters, together with their actual consumption values.