HYGAS-AI turns complex hydrogen plant data into continuously optimised, explainable decisions, grounded in physics and keeping operators in control.

HYGAS-AI is a physics-informed, agent-driven digital twin for hydrogen production plants. It was initially developed and validated using operational data from a real waste-gasification-to-hydrogen pilot plant with a production capacity of one tonne of hydrogen per day and a modular pathway for scaling up to 25 tonnes per day. The platform integrates physics-based reaction kinetics and mass and energy balance models with a hybrid control architecture combining Model Predictive Control and Reinforcement Learning. This enables continuous optimisation of critical operating parameters; including feed rate, steam-to-feed ratio, water-gas shift conversion, and combined heat and power dispatch, rather than relying on fixed operating setpoints. A genetic algorithm further optimises output allocation among four competing CHP technologies operating under shared fuel constraints. Building on this physics-based foundation, HYGAS-AI incorporates specialised AI agents that support plant operation and decision-making. These include a natural-language operator copilot, root-cause diagnosis, automated support for preparing relevant EU regulatory and sustainability documentation, and intelligent vendor sourcing. Although initially developed for waste-gasification-based hydrogen production, HYGAS-AI is designed as a scalable and adaptable platform. Its modelling, optimisation, diagnostic, and decision-support capabilities can be configured for other hydrogen production pathways requiring real-time optimisation and intelligent operational oversight.