A New Automated Energy Meter Fraud Detection System Based on Artificial Intelligence
ResumoEnergy meter frauds are increasingly innovative and complex, making them harder to detect and demanding different solutions while bringing undesirable costs for energy companies and consumers. Most of the current solutions use energy measurements differences to find tampering in meters, and many of them cannot be applied in field inspections, causing the costs to escalate even more until detected. To deal with this problem, this paper proposes a combination of solutions based on Artificial Intelligence that together tries to automatize and simplify the inspection process and increase the assertiveness in field fraud tracking. The first one is based on a combination of the Time Domain Reflectometry (TDR) technique with an Artificial Neural Network (ANN) to classify the meter’s internal impedance and determine if modifications were made in its circuit. To optimize the processing time, the implementation is done with a Hardware/Software co-design. The second solution tracks visual irregularities within the meter area using a Convolutional Neural Network (CNN). Both solutions are integrated into a mobile application, allowing them to be applied in field inspections. Results indicate that the TDR ANN technique was able to identify every tampered meter tested, and the CNN solution, even with a final F1-score of 0.67, which may not be high due to the small and unbalanced datasets, it is still good enough for field inspections.
Palavras-chave: Meters, Visualization, Costs, Computer architecture, Inspection, Software, Convolutional neural networks, Time Domain Reflectometry, Hardware-Software Codesign, Fraud Detection, Artificial Intelligence, Energy Meter
KLOCK, João Pedro; CORRÊA, Jhonatan; BESSA, Miguel; ARIAS-GARCIA, Janier; BARBOZA, Felipe; MEINERTZ, Carmo. A New Automated Energy Meter Fraud Detection System Based on Artificial Intelligence. In: SIMPÓSIO BRASILEIRO DE ENGENHARIA DE SISTEMAS COMPUTACIONAIS (SBESC), 11. , 2021, Evento Online. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2021 . p. 135-142. ISSN 2237-5430.