Heterogeneous multi-agent proximal policy optimization based control of solid oxide electrolysis and wind turbine generator for fast frequency response

As the transition toward net-zero emissions accelerates, the demand for intelligent, energy-aware grid management strategies is growing rapidly. The increasing share of renewable energies, particularly wind power, introduces grid stability challenges due to the low inertia of power-electronic-interfaced resources. Providing reliable fast frequency response (FFR) is vital to maintaining system resilience under such dynamic conditions. This paper proposes a novel heterogeneous multi-agent proximal policy optimization (HAPPO)-based control scheme for coordinating a hybrid energy system comprising wind turbine generators (WTs) and solid oxide electrolytic cells (SOECs). The proposed HAPPO framework leverages deep reinforcement learning to enable real-time decision-making and adaptability under dynamic grid conditions. The WTs provide FFR by dynamically adjusting their power output using stored kinetic energy, while the SOECs regulate hydrogen production to support frequency stability. Also, a fuzzy-based control scheme is developed as a benchmark to evaluate the effectiveness of AI-based control strategies. The proposed schemes are tested and validated on a two-area power system model under varying wind and load conditions. Simulation results demonstrate that the HAPPO scheme significantly improves frequency nadir and recovery over 33% compared to classical PID and fuzzy-based controllers, particularly under large power imbalance events. The proposed scheme was further validated using hardware-in-the-loop (HIL) real-time simulation, confirming real-time feasibility. The findings highlight the potential of AI-driven hybrid systems for enhancing grid stability and advancing renewable energy integration in smart, IoT-enabled energy infrastructures.