Two-layer scheduling and model predictive control of a multiphysics alkaline water electrolyzer with integrated energy storage under variable renewables

Abstract

Alkaline water electrolysis (AWE) is a promising technology for large-scale renewable energy storage. However, with the increasing penetration of renewable energy sources (RESs), maintaining economic and safe operation under fluctuating power inputs remains challenging due to the large inertia of AWE systems. To this end, a two-layer scheduling and control strategy is proposed based on a hydrogen production system model that integrates multiple physical domains, including electrochemical processes, fluid dynamics, and heat transfer. At the scheduling level, Li-ion battery and hydrogen storage units are coordinated with the electrolyzer to buffer RES variability, thereby smoothing power fluctuations imposed on the AWE stack and reducing dependence on grid electricity purchases. The energy storage system reduced the PV curtailment by an average of 11.79% and saved 160 kWh of electricity per day. At the control level, controllers are designed for electrolyzer power tracking and lye temperature regulation, subject to hydrogen impurity constraints. Comparative dynamic simulations show that model predictive control (MPC) outperforms tuned PID-based strategies: MPC limits peak overtemperature to 1.72 °C, and achieves a 0.65-kW mean power deviation. The results of the inertial analysis show the power response lagging behind the scheduling command by 43.15 min, and hydrogen production response further lagging behind the power change by around 12.02 min.

The authors' abstract, as published at the source. Applied Energy, 2026 · DOI ↗