INTELLIGENT TECHNIQUE FOR ELECTRICITY THEFT IDENTIFICATION USING AUTOREGRESSIVE MODEL
Various studies have investigated electricity theft, an illegally act, perpetrated to the detriment of the electricity power providers, however, less attention has been given to identification of the types of electricity theft. Data were acquired from the Consumer Load Prototype developed at two different levels using Sensor-A connected to the Pole Terminal Unit and Sensor-B connected to the Consumer Terminal Unit. The output of the sensors were connected to BNC-2110 device and linked to the PCI 6420E channel, which log the data in the computer for further analysis. LABVIEW (2012) software was programmed to acquires data at a sampling frequency of 500Hz and decimated at 10s interval before logging into the computer hard disk. The feature extraction of the data acquired was achieved using autoregressive technique and model order selectionwas based on minimum description length. The model coefficient AR (20), data acquired and predicted data were used for theft identification. Meter-bypassing theft was identified when the energy consumption from sensor A and sensor B are different, however sensor B reads zero value and there are disparities in the model coefficients. Illegal connection before the meter theft was identified whenever there is difference in energy consumption as evaluated form sensor A and sensor B and there is no zero value recorded from sensor B, while Meter tampering was detected when the energy consumption as evaluated form sensor A and sensor B are different and there are no disparities in the model coefficients.