The integration of nuclear energy systems (NESs) with hydrogen production via high-temperature steam electrolysis (HTSE) offers a promising pathway for efficient hydrogen generation and enhanced operational flexibility. However, tightly coupled nuclear–hydrogen systems exhibit complex multi-timescale dynamics, stringent constraints, and strong grid interactions that challenge conventional control approaches.

This paper proposes a grid-aware, online, and real-time deep reinforcement learning-based optimization and control framework for coordinated operation of nuclear–hydrogen integrated systems. A continuous-control twin-delayed deep deterministic policy gradient algorithm is employed to learn optimal control policies that regulate HTSE thermal and electrical demand. This enhances NES flexibility, maintains voltage stability, and provides grid support under time-varying load and intermittent generation. A physics-informed environment captures HTSE electro-thermal dynamics, nuclear operational limits, and uncertain power generation from intermittent energy resources and grid power flow constraints, enabling realistic closed-loop training and real-time deployment.

The proposed framework enables dynamic power dispatch coordination between nuclear generation and hydrogen production, improves operational efficiency, and supports grid services such as voltage regulation and demand response. The Results demonstrate that this approach provides a scalable and adaptive control solution for integrated energy systems with high penetrations of variable energy resources and uncertain loads.