Combating electricity theft in smart grids: accurate location identification using machine learning and the fast Walsh-Hadamard transform

CİVELEK Ö., GÖRMÜŞ S., OKUMUŞ H. İ., Yilmaz H.

ELECTRICAL ENGINEERING, 2024 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Basım Tarihi: 2024
  • Doi Numarası: 10.1007/s00202-024-02824-8
  • Dergi Adı: ELECTRICAL ENGINEERING
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Compendex, INSPEC, DIALNET
  • Karadeniz Teknik Üniversitesi Adresli: Evet

Özet

The decarbonization of electricity generation is a crucial goal today. To achieve this, energy grids must be managed efficiently and intelligently. Smart Grid systems optimize energy flow, minimize losses, and integrate renewable energy sources through technologies such as advanced metering infrastructure. However, non-technical losses (NTL) make efficient energy use challenging. Reducing NTL is a key goal of Smart Grid systems. Advanced algorithms and sensors in these systems detect and prevent electricity theft, thereby enhancing the grid efficiency and sustainability. This study examines a pilot distribution line in Malatya, T & uuml;rkiye, and using field data. A dataset recording three-phase current data from supply lines under various scenarios is compiled. The current patterns are extracted using the fast Walsh-Hadamard Transform and used to train and test machine learning models to detect illegal load locations. Among the algorithms tested, the three layer neural network method showed the best performance, with a validation R2 value of 0.92, an RMSE of 70.06, and an MAE of 38.33. The test data results yielded the following performance ratios: R2 value of 0.98, RMSE value of 27.41, and MAE value of 19.51.This research highlights the potential of machine learning algorithms to combat electricity theft in smart grids and enhance grid security and efficiency.