Non-technical loss detection using data mining algorithms

Abstract

The non-technical losses are an important problem for the electric networks in the Region. However, its detection is possible using data mining. This work presents the implementation of clustering algorithms to detect non-technical losses using demand daily curves obtained from Advanced Metering Instruments (AMI). Three different clustering algorithms are compared, and their ability to identify outliers profiles is discussed. The study used synthetic data created with the Gaussian Hidden Markov Model adjusted with a common residential demand pattern from Guayaquil residential users. Results evidence the detection of 68% of the non-technical losses.