Abstract:
According to statistics, the developing countries all over the world have suffered significant Non-Technical Losses (NTLs) in natural gas and electricity distribution companies. Developed World has moved on to the Smart Grid System which has significantly reduced Non-Technical Losses (NTLs) and with the development in Artificial Intelligence the developed countries has taken its next step in controlling NTLs through Machine Learning (ML) and Data Analytics (DA). Solutions to mitigate NTLs used by developed countries are not directly applicable to the developing countries because of the poor state of data collection and technological advancement. Therefore, a tailored solution based on machine learning to mitigate NTLs in developing countries is presented. Our framework uses Multivariate Gaussian Distribution (MGD) to identify fraudulent consumers. A new feature—to cater social class stratification in developing countries and weather profile—is introduced which has significantly improved the hit-rate. Data of consumers is taken from Power Distribution Companies of Pakistan and output result of Fraudulent Consumers Identification Framework (FCIF) is physically verified by onsite inspection staff of Energy Companies and hit-rate stands high at 75% which is better than the state of the art frameworks.
