Multi-Task Learning Empowered Anomaly Detection for Internet of Power Systems

Abstract

The integration of the Internet of Things (IoT) with power systems, referred to as the Internet of Power Systems (IoPS), has significantly enhanced the efficiency and reliability of energy distribution and management. However, this integration introduces complexity and vulnerability to anomalies that can disrupt system functionality and security. Traditional anomaly detection methods, while effective to a degree, often struggle with the scale and diversity of data generated by IoPS. Motivated by this, we propose a novel anomaly detection framework based on multi-task learning (MTL) to address these challenges in this paper. MTL leverages shared representations across multiple related tasks, improving detection performance and robustness compared to single-task systems. We present a comprehensive methodology for implementing this framework, including model architecture, data handling, and evaluation metrics. Our experimental results demonstrate that our MTL approach significantly outperforms traditional methods in accuracy and efficiency. This research aims to advance IoPS security and, meanwhile, sets a foundational approach for future explorations into smart grid analytics. The paper concludes by discussing the implications of our findings for the development of more resilient IoPS and suggesting directions for further research.