Schneider’s AI cut hydrogen costs
Schneider Electric says its AI control software ran more than 6,000 hours on a 20 kW solid‑oxide electrolyser and cut the levelised cost of hydrogen by about 10%. (h2-view.com) The company frames this as part of an “open software‑defined automation” push with Microsoft that treats optimisation as continuously updatable software rather than locked hardware, and it argues the same operating model could reshape other industrial control stacks. (fuelcellsworks.com)
Hydrogen is usually made by splitting water with electricity, and the machine that does it is called an electrolyser. The expensive part is not the water; it is the electricity bill and the wear on the machine when temperatures and loads keep moving around. (energy.gov) A solid oxide electrolyser is a high-temperature version of that machine. It uses steam and heat so it can need less electrical energy than lower-temperature designs, but the trade-off is that the equipment has to survive very hot operation and repeated thermal stress. (energy.gov) That is why control software matters here. Running a hot industrial system is a bit like driving a car engine up a mountain: if you floor it, brake hard, and floor it again, you burn extra fuel and wear out parts faster. (energy.gov) Schneider Electric says it used artificial intelligence software to keep one of these systems steadier in real time. In trials on a 20 kilowatt solid oxide electrolyser from Indian company h2e Power, the software ran for more than 6,000 hours across part-load and full-load conditions. (h2-view.com) The software adjusted thermal balance, hydrogen flow, energy inputs, safety, and equipment health while the system was operating. Schneider says that smoother operation cut stack wear and improved energy efficiency by about 10%. (globalhydrogenreview.com) The number the company is pushing is the levelised cost of hydrogen. That is the all-in cost of making hydrogen over the life of the equipment, rolled together from power use, capital cost, maintenance, and operating life, and Schneider says its trial cut that figure by around 10%. (h2-view.com) That kind of cut gets attention because green hydrogen is still expensive. A recent United States Department of Energy program record put clean hydrogen from today’s proton exchange membrane electrolysers at roughly $5 to $7 per kilogram without subsidies, and electricity price and utilisation are two of the biggest levers in that math. (hydrogen.energy.gov) Schneider is also using this project to sell a bigger idea about factory control systems. Its pitch is “open software-defined automation,” which means separating the control software from the hardware underneath so plants can update logic more like software on a server than like a fixed box on a wall. (se.com) Microsoft is part of that pitch because Schneider says the engineering and control layer used Microsoft Azure AI Foundry. Microsoft describes Azure AI Foundry as a platform for building and managing artificial intelligence applications and agents, and Schneider is framing that cloud software layer as something that can keep improving after the equipment is installed. (azure.microsoft.com) The hydrogen pilot is small at 20 kilowatts, and Schneider has not published a full independent cost breakdown from the trial. But if software can really lower energy use and slow stack degradation on hot electrolysers, the same playbook could spread to other industrial systems where heat, power, and equipment life are all fighting each other every hour. (h2-view.com)
