Identification of theft in a power distribution network: A tree-based approach

Electricity is an essential commodity, yet it is not freely available to everyone in the country due to widespread electricity theft. This theft results in significant financial losses for Distribution Companies (DISCOMS), hindering their ability to provide reliable services to consumers. Although there are methods to curb electricity theft, DISCOMS have struggled to control it effectively. The most successful current method is the manual investigation of nodes; however, it has its drawbacks, such as a low probability of theft detection when nodes are randomly sampled and excessive time consumption when all nodes are investigated. To address these issues, a systematic approach utilizing the capture, storage, and analysis of data is essential. This paper presents a coloring-based tree approach to identify power theft without altering the existing network infrastructure. The proposed approach aims to reduce the dataset for manual inspection, thereby increasing the probability of theft detection while substantially reducing investigation time. By discouraging blind checking of nodes and providing a basis for targeted node inspections, this method benefits both DISCOMS and general consumers. In this paper, the existing methods of theft detection have been compared with the proposed approach of tree-based theft detection. The application of the proposed approach removed certain nodes from the set of nodes to be investigated for theft detection.