Definition of NTL Zones Using Georeferenced Data and Dimensionality Reduction with UMAP

Inhalt:
non-technical losses at the distribution level are calculated as the difference between the electricity purchased by distribution system operators (DSOs) and the amount billed to their consumers. These differences are primarily caused by electricity theft and fraudulent connections. In addition to impacting the concessionaire’s revenue, they can also affect the level of technical losses. This paper presents an improvement to a method previously proposed by the authors. The method applies geospatial analysis to electricity theft with the primary goal of identifying areas with a high density of irregularities. It incorporates the null hypothesis test, spatial autocorrelation (Global and Local Moran’s I), Uniform Manifold Approximation and Projection (UMAP), self-organizing maps (SOM), and K-means clustering. The results reveal clustering patterns that are relevant for directing targeted inspection efforts. Socioeconomic data and historical utility records were used as input datasets for the algorithms.