Multi-Rate Fast Frequency Regulation With High-Fidelity Neural Network Model of the Electrolyzer Process

Frequency regulation is crucial for grid stability, especially with the increasing penetration of inverter-based resources (IBRs). Hydrogen electrolyzers, as flexible loads, offer a promising solution by dynamically adjusting power consumption to counteract frequency deviations. While existing studies primarily focus on electrolyzer stacks and power converters, the role of auxiliary process components such as pumps and coolers remains largely unexplored. This paper proposes a multi-rate frequency response (MRFR) model for hydrogen electrolyzer process plants, distinguishing frequency regulation contributions from both the electrolyzer stack and auxiliary components. A high-fidelity neural network (HiFiNN) model is developed using a detailed alkaline water electrolysis (AWE) process model inAspen HYSYS, accurately capturing the interactions between electrical and thermodynamic parameters. The HiFiNN model is integrated with a model predictive controller (MPC) to optimize frequency regulation while maintaining process stability. The proposed framework is validated through simulations and real-time controller-hardware-in-the-loop (CHIL) experiment to validate its feasibility for practical deployment. Additionally, its effectiveness is demonstrated under varying IBR penetration levels, highlighting its capability to maintain frequency stability in low-inertia power systems.