Green hydrogen is expensive (at best, $ 4 per kg); The two key promoters of costs (electrolyisms and electricity) are the focus of research, but the economy cannot discuss with physics. There is only much that can do to mark costs.
Or that was thought.
A new route opens to tame the recalcitrant costs: Artificial Intelligence (AI) and its automatic subset learning (ML). Ai/mL can help in each step of production analysis in the real analysis of operating parameters in real time, such as temperature, pressure and input electricity; Predict failures; and help discover catalysts that are alternative to rare expensive materials such as platinum and iridium (one gram costs $ 5,000).
The green hydrogen review of international energy agencies 2023 ‘estimated that more than 50 green hydrogen projects used AI, which implied an investment of $ 500 million. In many pilot projects, the use of AI has achieved efficiency or 5-15 percent gains; This could be higher with more operational data, notice a June 2024 document by Ashish Saxena, Master of Technology with Amazon, Seattle. He points out that Siemens Energy incorporated “reinforcement learning” (RL), a type of ML, to allow online adjustment or 70,000 parameters for its electrolysis process to achieve an improvement of 8-10 percent. “The models used equipment control variables and performance measures to create the best control strategies,” says Saxena, noting that the fall in the cost of software, sensors and IoT connectivity is making AI affordable.
Experts such as Saxena underline the value of AI in predictive maintenance and process optimization. When they are continuously executed, the AI models feed on real time performance data to anticipate failures.
Catalyst search engine
A more important role that AI could play is (RE) in electrolisers to identify affordable catalysts for hydrogen evolution reactions. Catalysts, which accelerate a chemical reaction without consumed in the process, help in the evolution of hydrogen and oxygen in electrolysters.
The performance of a catalyst is influenced by the interaction of the structure, composition, electronic configuration, geometry of the adsorption site, etc. of alloy materials. Analyzing the parameters of hundreds of materials to determine the correct combination is a discouraging task, but IA could do Asier.
But this bee the question of why catalysts for hydrogen evolution reactions (she) thrown by AI are not common. In fact, AI has suggested some interesting catalysts, as we will see later, but the problem is the cost and complexity of the generation of experimental data necessary to feed the AI models.
Professor Ke Vipin and Professor Prahallad Padhan de Iit-Madras are developing new ML methodologies to predict and design profitable and high performance intermetallic catalysts for her and the oxygen evolution reaction (REA).
They used a data set extracted from the catalysis cate database, which represents a significant compilation of alloy catalysis data for it and primal. It involved 16,226 different data points, with 8,856 entries focused on its catalysis and 7,370 entries dedicated to the Catalysis REA, they say in an article, which has not yet been reviewed by pairs. Vipin and Padhan created a “systematic approach driven by data for the catalyst design that can exceed the limitation of traditional test methods and Figor”. By the way, the model is good to design any catalyst, not only for the evolution of hydrogen.
Using regression algorithms such as the random forest, the XGBOOST and the regression of the support vector, the model can “precisely predict the free energy of hydrogen adsorption on bimetallic alloy surfaces, a key description for the catalytic activity ‘vipin’ vipin Quantum. Gibbs free energy is a thermodynamic amount that tells us if a chemical process can occur spontaneously, without extra energy.
“While we carry” experimentally validated the new catalysts yet, we are using the model trained to detect and predict promising new bimetallic candidates, “said Vipin. These predictions are evaluated through simulations of functional density theory to confirm its theoretically potential before going to experimental tests.
The IA -driven catalyst design has already led several emerging candidates, including a ‘quantum carbon points incorporated with nickel’ (based on a Bayesian genetic algorithm); a mixed rust catalyst in Ruthenium – Manganese – Calcio – Paseodimio; and a copper-aluminum catalyst.
First key step
It should be noted that designing a catalyst is only the first step (but indispensable): it must cross the ‘Death Valley’ between the research and industry laboratories. It must work well for long years and not collapse after a performance explosion. You must integrate the sides with the existing systems.
AI is helping to win this much of the battle. Once you have systems that help increase the efficiency of electrolysters (in fact, the entire green hydrogen value chain, including renewable energy assets) and an affordable catalyst, then green hydrogen can reach the reach of trade.
“The way to a sustainable hydrogen economy is complex and multifaceted, which requires innovative solutions that transcend disciplinary limits,” write Vipin and Padhan.
“Through advanced automatic learning approaches, we stand on the cusp of a possible advance that could revolutionize the production of green hydrogen, which brings us a lot to a decarbonized and sustainable global energy system,” Id. ADD.
More like this
Posted on April 20, 2025