Design and Implementation of an Electrical Tamper Detection System

Electrical power theft is one of the main causes of Non-Technical Losses (NTLs) in an electrical power system. NTLs in electrical power grids refers to the transmission and distribution loses originating from electrical theft and various other illegal uses of electricity. According to Eskom, South Africa loses at least R20 billion a year due to electricity theft, three-quarters of which is reported to be losses suffered by municipalities. Thus, to address this issue, this paper presents the design and implementation of a device capable of detecting electricity theft by identifying the small changes between the real-time measurements and that of a trained model. The energy metering system consists of a voltage and current sensing circuits while the theft detection algorithm consists of a Support Vector Machine (SVM) model trained and used to classify users as either a clean or fraudulent user. The data that was used to train the SVM model was obtained from a test bench. The test bench consisted of 100 appliances that were emulated using purely resistive loads. Each appliance has its own probability function which consists of a PRNG that generates pseudorandom numbers according to a chosen distribution type. The probability function also considers the probability of a specific appliance being used for that day and for how long the appliances are active. The detection device was implemented using a Raspberry Pi 3 and several evaluations were carried out. Evaluation result showed that the device can accurately detect electricity theft.