AI for Science in Next-Gen Power Systems: A Perspective

In this perspective, we present a concise, visionary abstract that highlights the challenge of ever‑increasing complexity in future power grids and our proposed artificial intelligence for science (AI4S) framework. Next‑generation grids, with high levels of renewables and distributed generation, create unprecedented uncertainty that strains traditional control and forecasting methods. Our AI4S framework integrates cutting‑edge artificial intelligence (AI), including large language models (LLMs), physics-informed neural networks (PINNs), deep reinforcement learning (DRL), and autonomous multi-agent systems, within a domain‑aware architecture that embeds grid physics, operational constraints, and human oversight into learning. By fusing data‑driven models with power‑system laws, AI4S yields trustworthy, explainable AI that enables fast, real‑time decision making and richer situational awareness via digital twins and advanced forecasting. We distinguish which AI4S elements are already demonstrated versus conceptual, and we envision benefits like accelerated simulation, improved resilience, and greater operator trust.