An Integrated Power System Machine Learning Model for Detecting Smart Meter Frauds in Distribution System

Case study of PHED Market

  • Biobele Alexander Wokoma, Dr. Rivers State University, Port Harcourt, Rivers State, Nigeria. https://orcid.org/0000-0002-1116-5789
  • Elijah Olusanmi Olubiyo, Mr. Department of Electrical Engineering, University of Nigeria, Nsukka, Enugu State, Nigeria.
  • Maxwell Nwoku, Dr. Rivers State University, Port Harcourt, Rivers State, Nigeria.
  • Ibitroko Biobele Alexander Wokoma, Engr. Port-Harcourt Electricity Distribution Company, Port Harcourt, Rivers State, Nigeria.
DOI: https://doi.org/10.29138/ijeeit.v7i2.2730
Keywords: Billing, Fraud, Neural Networks, Power Distribution, Smart Meter

Abstract

Smart meter billing systems represent the modern state-of-the-art in pre-paid billing as the move from estimated bills within the Nigerian state now, becomes a priority to circumvent the heavy fraud situation. However, the detection of fraud in real-time smart meter systems presents an opportunity for more dynamic approaches to be employed, due to the unique nature of this billing system. In this paper, an emerging auditory inspired neural technique that supports continual learning and adaptive processing of temporal streaming data states is proposed for the detection of smart meter fraud. The technique is integrated in a simulated power system program and applied to the detection of frauds in two feeder units of a Port-Harcourt Electricity Distribution (PHED) business district. Simulation results showed that the proposed approach is a potential candidate for detecting fraud in smart meter billing systems with minimal mean absolute percentage errors.