Promising AI Applications in Power Systems: Explainable AI (XAI), Transformers, LLMs

This paper aims to analyze and identify the most promising opportunities for Artificial Intelligence (AI) applications in the Power Systems (PS) domain. It identifies major challenges faced in PS and explores the corresponding technical tasks: forecasting and optimal control. Then, the paper investigates the key AI techniques commonly employed in PS for these tasks, e.g. reinforcement learning (RL) and time series forecasting. It also highlights promising methods with great potential in advancing PS solutions: attention-based models (Transformers, LLMs) and explainable AI (XAI) approaches. This study’s primary contribution lies in identifying critical research gaps in AI for PS, highlighting areas where research and development may have the biggest impact. Additionally, the paper provides a structured literature overview, serving as a valuable resource for researchers and practitioners in the field.