Development of modular hardware and software system for fraud detection using machine learning and data mining techniques with applications in power systems energy thefts

Date
2022-05
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Electricity is a critical commodity that powers the technology that nations and their inhabitants heavily rely on for economic and social activities. However, the issue of electricity theft threatens the looms large for these countries as these misdemeanours cost electricity utilities close to $20 billion worldwide [4]. Ghana is taking the needed steps to combat electricity theft such as employing smart meters and altering supply frequency. There are no methods that completely eradicate the issue of electricity theft, but the solutions deployed by Ghana's electricity utility are also expensive. This project presents a novel method of detecting electricity theft using data mining and machine learning utilising non-intrusive load monitoring techniques (NILM) on a low-cost smart metering device. NILM enables one to know the power consumption levels and durations of individual appliances from the total consumption measured. As the artificial intelligence algorithm is enhanced by the NILM techniques to understand the behaviour of the consumer, there is a greater possibility of making an accurate power fraud prediction and making a fruitful inspection. The final accuracy of the machine learning model that detects theft was tagged at 91.8% and the non-intrusive load monitoring correctly identified the ON/OFF states of appliances 91% of the time but performed not as well in predicting their energy consumption. Furthermore, the smart meter provides electricity utility customers with a device-level breakdown of their consumption for energy monitoring purposes.
Description
Capstone Project submitted to the Department of Engineering, Ashesi University in partial fulfillment of the requirements for the award of Bachelor of Science degree in Computer Engineering, May 2022
Keywords
non-intrusive load monitoring, power fraud detection, electricity theft
Citation
DOI