Detecting intrusions in Internet of Energy (IoE) systems using an automated machine learning framework

The rapid growth of ICT and the increasing demand for power systems led to the development of the Internet-of-Energy (IoE), which integrates IoT devices for digital control and ICTs for data transmission, enabling real-time energy management. Distributed energy resources (DERs) and battery energy storage systems (BESS) play a crucial role in enhancing grid stability and efficiency. However, the integration of these technologies makes IoE vulnerable to cyber threats like DDoS, SQL injection, and false data injection, which can have serious consequences. Machine learning (ML) has proven effective in detecting such attacks, but traditional ML methods often require extensive manual management. To address this, automated ML (AutoML) frameworks automate the entire ML pipeline, improving efficiency. This work proposes an AutoML-based framework for intrusion detection in IoE systems, specifically targeting BESSs. The framework uses the TVAE algorithm for data balancing, tree-based models for feature selection, Bayesian Optimization for tuning, and stacking-based ensembles for optimal performance. Experimental results show that the proposed AutoML framework outperforms existing models in accuracy, precision, recall, and F1-score. It also reduces training and prediction times, demonstrating its advantages in detection accuracy and computational/inference efficiency.