TNT Loss: A Technical and Nontechnical Generative Cooperative Energy Loss Detection System

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Data Science

Background:

• Electricity theft and technical inefficiencies lead to significant energy losses.
• Existing methods struggle with accurate identification and localization of diverse loss types.
• Smart metering data offers potential for advanced loss detection but requires sophisticated analysis.

Purpose of the Study:

• To develop a robust method for technical/nontechnical electricity loss detection, identification, classification, and localization.
• To propose an AI architecture integrating cooperative and non-cooperative modules for enhanced data analysis.
• To establish a clear, rapid mapping between loss signatures and specific loss types.

Main Methods:

• An architecture of three generative cooperative AI modules and two non-cooperative AI modules for knowledge sharing.
• Embedding expert consumption-based knowledge and feature collaboration into an AI classification algorithm.
• Implementing an anomaly pooling mechanism for one-to-one mapping of signatures to loss types.
• Utilizing robotic process automation for loss type localization.

Main Results:

• Achieved high-accuracy technical/nontechnical loss detection, differentiating from other grid anomalies.
• Demonstrated a simple and rapid method for exact loss type to signature mapping.
• Showcased the effectiveness of reactive energy load profiles in enhancing loss signatures.
• Validated the system's performance through experimental testing matching theory and practice.

Conclusions:

• The proposed AI system effectively detects, classifies, and locates electricity losses with high accuracy.
• The generative cooperative modules and anomaly pooling mechanism provide a novel approach to loss detection.
• The system leverages comprehensive smart metering data, outperforming standalone algorithms in field conditions.