PREDICTIVE ANALYTICS FOR ELECTRICITY THEFT DETECTION USING MACHINE LEARNING

In recent years, efficient energy utilization has become critically important, particularly in the context of electricity consumption. One of the major contributors to non-technical losses (NTLs) in power distribution networks is electricity theft, which poses significant challenges to power grids by degrading supply quality and reducing operational revenue. With the deployment of Advanced Metering Infrastructure (AMI), smart meters (SMs) are installed at the consumer end to transmit fine-grained energy consumption data at regular intervals for purposes such as load monitoring, energy management, and billing.However, malicious consumers may launch cyber-attacks by manipulating or falsifying meter readings to reduce their electricity bills unlawfully. Such activities not only result in financial losses but also adversely affect grid performance, as these readings are used for critical grid management decisions. To detect such fraudulent behavior, existing approaches often employ machine learning models based on detailed consumption data. Nevertheless, these methods raise significant privacy concerns, as they may inadvertently reveal sensitive information about consumers’ lifestyles.In this paper, we propose an efficient and privacy-preserving scheme for electricity theft detection, billing, and load monitoring. The proposed approach utilizes symmetric encryption to secure consumer data, ensuring privacy, while employing the K-means clustering algorithm to identify anomalous consumption patterns indicative of electricity theft.