Post by Alinane Brown

📢Happy to share our latest publication in IEEE Transactions on Power Systems: Physics-Shielded Deep Reinforcement Learning for Adaptive Emergency Frequency Control. Conventional under-frequency load shedding (UFLS) schemes are great, but modern power systems demand more adaptive and scalable solutions. In this work, we propose a physics-shielded reinforcement learning (RL) framework where governing system equations actively guide RL learning rather than simply enforcing safety/physical constraints. Unlike conventional shielding that relies on static system characteristics as filters, our method uses dynamic, scenario-aware system behaviour to support learning and accelerate convergence. Combined with deep learning and real-time system coherence detection, the framework reduces unnecessary exploration (minimises excessive sampling), improves training efficiency, and scales effectively to large-scale power systems. Special thanks to my co-authors, Waqquas Bukhsh and Panagiotis Papadopoulos, for the excellent collaboration. 📄Paper link: https://lnkd.in/eQki8jAh #ReinforcementLearning #PowerSystems #AI #Dynamics #Stability #UFLS #SmartGrid #EnergySystems #MachineLearning #DeepLearning