Smart Energy Guardian: A Hybrid Deep Learning Model for Detecting Fraudulent PV Generation

2024-10-29

2024 IEEE International Smart Cities Conference (ISC2)

AI Summary

Conventional electricity theft detection (ETD) methods struggle to model complex temporal dependencies and integrate heterogeneous multi-source data, limiting their effectiveness against sophisticated, adaptive, and stealthy electricity theft in smart-city residential photovoltaic (PV) systems. Method: This paper proposes an end-to-end fraud detection framework featuring a novel CNN-LSTM-Transformer hybrid architecture for multi-scale temporal fusion, coupled with a joint embedding mechanism that unifies time-series power measurements and discrete temperature variables. Contribution/Results: Evaluated on a real-world residential PV dataset, the framework achieves significantly improved detection accuracy and markedly reduced false alarm rates. Its robust modeling of nonlinear, time-varying theft patterns enhances reliability in operational settings. The approach effectively supports demand-supply balancing and energy equity in smart grids by enabling timely, precise identification of anomalous consumption behavior.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

With the proliferation of smart grids, smart cities face growing challenges due to cyber-attacks and sophisticated electricity theft behaviors, particularly in residential photovoltaic (PV) generation systems. Traditional Electricity Theft Detection (ETD) methods often struggle to capture complex temporal dependencies and integrating multi-source data, limiting their effectiveness. In this work, we propose an efficient ETD method that accurately identifies fraudulent behaviors in residential PV generation, thus ensuring the supply-demand balance in smart cities. Our hybrid deep learning model, combining multi-scale Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Transformer, excels in capturing both shortterm and long-term temporal dependencies. Additionally, we introduce a data embedding technique that seamlessly integrates time-series data with discrete temperature variables, enhancing detection robustness. Extensive simulation experiments using real-world data validate the effectiveness of our approach, demonstrating significant improvements in the accuracy of detecting sophisticated energy theft activities, thereby contributing to the stability and fairness of energy systems in smart cities.

Problem

• Detects fraudulent residential PV generation in smart grids
• Overcomes limitations of traditional electricity theft detection methods
• Ensures supply-demand balance and energy system stability

Innovation

• Hybrid deep learning model combining CNN, LSTM, Transformer
• Data embedding integrates time-series and temperature variables
• Accurate detection of residential PV generation fraud