View Proposal


Proposer
Medhdhar Salem Ali Muthanna Al- Gaashani
Title
Data-Driven Electricity Theft Detection Using Deep Learning on Smart Meter Data
Goal
Build a machine-learning system that detects electricity theft (non-technical losses) from smart-meter consumption data, classifying users as normal or fraudulent to help utilities reduce financial losses and improve grid safety
Description
Electricity theft causes significant financial losses, while manual inspection is costly and inefficient. This project develops a data-driven detector using historical smart-meter consumption data. Students preprocess and normalize the data, extract useful consumption features, handle class imbalance, and compare a traditional model such as Random Forest or SVM with a deep-learning model such as a 1D CNN or LSTM. Performance is evaluated using precision, recall, F1-score, AUC, and confusion matrices. Optionally, SHAP can be used to explain suspicious consumption patterns, and a simple dashboard can display flagged.
Resources
Chen, J., Nanehkaran, Y. A., Chen, W., Liu, Y., & Zhang, D. (2023). Data-driven intelligent method for detection of electricity theft. International Journal of Electrical Power & Energy Systems, 148, 108948 — https://doi.org/10.1016/j.ijepes.2023.108948 ; Paper code and data: https://github.com/xtu502/electricity-theft-detection ; SGCC real-world benchmark dataset
Background
Url
Difficulty Level
Moderate
Ethical Approval
None
Number Of Students
1
Supervisor
Medhdhar Salem Ali Muthanna Al- Gaashani
Keywords
deep learning, machine learning, smart grid, time-series, anomaly detection
Degrees
Bachelor of Science in Computational Sciences and Software Engineering