A TinyML-based system for Electricity Theft Detection By Using Novel Smart Meter Data

In electricity grids, energy theft has many adverse effects on equipment and infrastructures of grids, reducing power quality and causing many financial losses. Nowadays, electricity theft causes vast financial loss both in developed and developing countries all over the world. In smart grids, by combining data related to the energy consumption of customers and energy flow, various methods introduced to detect electricity theft. Each of these methods has advantages and disadvantages and may not perform well in different situations. With the help of fine-grained datasets gathered from smart meters, which are an inseparable part of smart grids, it is possible to overcome the shortcomings of the existing methods. This work presents a TinyML-based system for electricity theft detection. In this system, the data related to the energy consumption of the customers is checked in the meter itself, and if suspicious behavior or anomalies are detected, warnings and messages are sent to the homeowner and network operators. The performance of the proposed system is evaluated with data gathered from the proposed novel smart meter. For electricity theft detection based on these consumption data, The performance is acceptable and the system achieved an f۱-Score of ۰.۷۴.

Keywords:

Machine Learning , Electricity Theft Detection , Tiny Machine Learning(TinyML) , Smart meters