Adaptive Multiagent Control of Distributed Electrolyzers in Renewable-Powered Microgrids

Abstract

This study presents the development and validation of an adaptive multi-agent control system for distributed electrolyzers operating within microgrids powered by variable renewable energy sources. Using a network of PEM and alkaline electrolysers equipped with real-time telemetry and programmable power inputs, the researchers simulated dynamic operating conditions, including fluctuating solar generation, temperature variations, and power supply interruptions. A hierarchical agent-based architecture was implemented, comprising local controllers, regional energy brokers, and a central coordination unit, enabling autonomous adjustment to external disturbances. Experimental results demonstrate significant improvements in hydrogen production efficiency, system responsiveness, and load distribution stability. Key performance indicators show a 4.3-percentage-point gain in average energy efficiency (from 67.9% to 72.2%), an 8.7% reduction in specific energy consumption (from 58.3 to 53.2 kWh/kg H₂), a shorter disturbance response (6.2 → 3.9 s), and an 11.4% increase in hourly hydrogen yield. Robust operation was preserved under up to 500 ms communication delays and injected telemetry noise (tested at 5% and 10% levels), while current ripple amplitude decreased from 18.2% to 4.6% relative to the PI baseline. The multi-agent system reduced the current fluctuation amplitude from 18.2% to 4.6% and increased hydrogen generation by 11.4% compared to centralized control schemes. Furthermore, the system maintained stable operation despite up to 10% noise in telemetry data and 500 ms communication delays. The findings confirm that multi-agent control enhances the resilience, scalability, and energy efficiency of decentralized hydrogen production systems, supporting their integration into future smart energy infrastructures.