Machine learning for energy transition

Abstract

Topology knowledge plays a pivotal role for Distribution System Operators (DSO), offering invaluable insights for network maintenance, fault diagnosis, and capacity planning to meet customer demand efficiently. However, attaining accuratetopology knowledge poses significant challenges. The dynamic nature of the network, coupled with the susceptibility to humanerror in recording changes, complicates maintaining an updated topology. Furthermore, the historical transition of network documentation from paper to digital formats introduces reliability issues. Access limitations and inadequate toolsalso hinder comprehensive data collection, leaving gaps in the records.To overcome these hurdles, a comprehensive approach is required. By amalgamating static data derived from historical records and field observations withdynamic data from smart meters, a more precise topology reconstruction becomesfeasible. This study specifically addresses the augmentation of an already reconstructed network, based solely on static data, with dynamic data from smartmeters. This integrated approach aims to capitalize on the strengths of bothstatic and dynamic data sources to enhance topology accuracy.In the presented case study, conducted in collaboration with a real DSO, thesparse penetration of smart meters in households poses a unique challenge. Additionally, the unavailability of measurements from the Medium-to-Low Voltage Transformer (MLVT)s further complicates the task. Consequently, innovativetechniques are explored to address these dual challenges, ensuring robust topology reconstruction despite data limitations. The evaluation of the results bythe partnering DSO provides valuable feedback, validating the applicability ofthe approach to real-world scenarios and underscoring its potential to addresspressing challenges faced by DSOs. This research bridges the gap between theoretical advancements and practicalapplications, addressing real-world problems faced by DSO. By working closelywith industry partners, the study ensures that the proposed solutions are not confined to laboratory settings but are rigorously evaluated in operational contexts,enhancing their relevance and impact.