Non-Technical Loss Detection in Power Grids with Statistical Profile Images Based on Semi-Supervised Learning

Area of Science:

• Electrical Engineering
• Data Science
• Cybersecurity

Background:

• Smart grids, the largest sensor systems globally, utilize millions of smart meters for operational monitoring.
• Non-technical losses (NTLs) pose significant security risks and revenue loss concerns for power grids.
• Detecting NTLs is challenging due to the vast scale of data and diverse anomaly characteristics.

Purpose of the Study:

• To propose a novel methodology for accurately detecting abnormal electricity consumption patterns indicative of NTLs.
• To address the challenges of NTL detection in large-scale smart grid data.

Main Methods:

• Time-series electricity consumption data is transformed into image representations to capture long-term user behavior.
• A deep learning model, inspired by computer vision object detection, is designed to process these images.
• The model employs a semi-supervised learning approach to handle limited labeled data, especially for abnormal consumption.

Main Results:

• The proposed method demonstrated significant improvements in NTL detection compared to existing state-of-the-art techniques.
• The image-based representation effectively captures user consumption behaviors relevant to anomaly detection.
• The semi-supervised deep learning model achieved high accuracy on field-verified samples.

Conclusions:

• The developed methodology offers a promising solution for enhancing NTL detection in smart grids.
• Transforming time-series data into images combined with deep learning is effective for anomaly detection.
• This approach contributes to improved power grid security and operational efficiency by better identifying revenue losses.