Intelligent On-Demand Green Hydrogen Production for Synthetic Fuels via PSO- and GA-Optimized Inverse Neural Controllers

Green hydrogen is a key energy carrier in Power-to-Liquid (PtL) pathways for the production of sustainable synthetic fuels, contributing to the decarbonization of the industrial and transport sectors. However, the intermittent nature of renewable energy sources and the variable hydrogen requirements needed to maintain the appropriate stoichiometric ratio for synthesis processes necessitate regulating hydrogen production according to process demand, rather than maximizing its generation. This article proposes an intelligent control strategy for alkaline water electrolysis, in which the hydrogen production target is determined from the stoichiometric requirements of synthetic methanol production, based on available carbon dioxide. ANN models were developed using the experimental data, incorporating both classical and conformable activation functions in the hidden layer. Based on the selected models, the ANNi was formulated, and PSO and GA were used to determine the required feed current according to hydrogen demand. The proposed methodology was evaluated under a dynamic hydrogen-demand profile derived from the stoichiometric requirements of methanol synthesis. The results show that the proposed controllers closely track changes in hydrogen demand. After each change in the setpoint, the H2/CO2 ratio returned to a ±2% band around the stoichiometric setpoint in approximately 0.98 s for ICANNi-PSO and 0.96 s for ICANNi-GA. Furthermore, some conformable activation functions achieved performance comparable to that of classical activation functions while using fewer neurons in the hidden layer. Both optimization algorithms provided comparable tracking performance under the evaluated conditions.