AN EFFICIENT ELECTRICITY THEFT DETECTION USING XG BOOST

Abstract

Electricity theft is a significant occurrence that is happening about all nations from one side of the planet to the other. This adversely affects the nation's economy. Theft identification in the power sector is a difficult task for power distribution organizations overall, since this power theft prompts to monetary losses just as loss of electric energy. There are numerous ways by which power is stolen like controlling energy meters or tapping cables at the consumer's end, and so on. Since this robbery is going on in enormous amount, manual examination of such burglary is a hectic task. So, automatic detection of power theft is a vital need of the hour. This paper presents an electricity theft detection model using smart grid meter data dependent on Extreme Gradient Boosting (XGBoost) and OCR. Feature Selection is made used of in the model in order to pick the most relevant features from the dataset to that of our electricity theft detection model. XGBoost is remarkable as it utilizes a more regularized model formalization to have command over-fitting which results in better execution significantly quicker. OCR is employed in order to know what object led to electricity theft detection.