Search Results

Abstract: Electricity theft is a major challenge for power utilities because fraudulent consumption can increase non-technical losses, reduce revenue, and affect the accuracy of electrical energy accounting in smart-grid environments. The availability of smart meters and Advanced Metering Infrastructure (AMI) provides detailed time-series electricity consumption data that can be utilized for automated identification of abnormal consumption patterns. In this research, a Convolutional Neural Network (CNN)-based deep learning model is developed for smart meter electricity theft detection. The proposed approach utilizes historical smart-meter electricity consumption data and applies data preprocessing, normalization, and relevant feature preparation before model training. The CNN model is designed to extract important local patterns and variations from electricity consumption sequences and classify consumers into normal and potentially fraudulent categories. The developed model is evaluated using standard classification performance measures, including accuracy, precision, recall, F1-score, and confusion matrix analysis. The framework aims to improve the identification of complex electricity theft patterns while reducing dependence on manual inspection and predefined detection rules. The proposed CNN-based approach can provide an efficient and scalable solution for automated electricity theft detection and support power utilities in reducing non-technical losses, improving energy monitoring, and strengthening the reliability and security of modern smart-grid systems.