Applying artificial intelligence for forecasting behavior in a liquefied hydrogen unit

Abstract

Liquid H₂ production using the Claude cycle requires an appropriate refrigeration cycle for initial cooling of gaseous H₂. Additionally, the modified generator-absorber heat exchanger cycle exhibits higher coefficient of performance among refrigeration cycles and can operate at very low temperatures approaching negative 30 °C. In this work, a modified generator-absorber heat exchanger cycle, geothermal energy, and Claude cycle are combined to produce and store liquid H₂. A comprehensive analysis is conducted, including thermodynamic, economic, sustainability, net present value, environmental perspectives, and multi-objective optimization using an artificial neural network and the multi-objective grey wolf optimizer. Compressor pressure has no effect on the amount of produced liquid H₂, remaining constant at 12.55 kg/s. Key results show an exergetic efficiency increase to 24.27%, a reduction in liquefied H₂ cost to 1.482 $/kg, and a decrease in system cost rate to 45.71 $/h. The liquefied H₂ mass flow rate is optimized at 4.178 kg/s, while compressor power consumption dropped to 214.2 MW, and turbine power output reached 13.54 MW. The payback period is shortened to 3.39 years, with the net present value exceeding 40 M$ over 20 years. Sensitivity analysis revealed that evaporator temperature (51.4%) and compressor pressure (57.1%) are the most influential factors.