Comparative Analysis of Energy Theft Detection in a Power System using Support Vector Machine and Quantitative Technique

This study compares energy theft detection methods using qualitative analysis techniques and a Support Vector Machine (SVM) model. With a precision of 97.86%, a recall of 99.93%, an F1-score of 98.88%, and an accuracy of 97.94%, the findings show that SVM performed well across all key evaluation metrics. However, qualitative analysis revealed an average consistency of 80% across these indicators, indicating a higher risk of misclassification and lower reliability. The results demonstrate that, compared to traditional qualitative methods, SVM provides better detection accuracy, reduces false alarms, and ensures comprehensive identification of theft cases. When compared to other related works on an overall basis, the results were superior. These findings highlight the potential of machine learning models, particularly SVM, as a scalable and dependable approach to preventing electricity theft in modern power grids.