基于数值模拟与机器学习建模优化住宅用质子交换膜燃料电池微型热电联产系统中的氢气储存

Abstract This study explores the integration and optimization of a hydrogen-based energy system, emphasizing the use of metal hydride (MH) storage coupled with Proton Exchange Membrane Fuel Cell Micro Combined Heat and Power (PEMFC MCHP) system for residential applications. MH storage coupled to a heat pump, operates at charging and discharging pressures of 10 bar. COMSOL model in 6.1 version using heat transfer in solids and fluids in brinkman equations modules is validated by experimental data and uses machine learning (Feedforward Neural Networks) for predictive modeling of MH dynamics. Smaller 500 NL tanks were found to have high mass-specific heat demand but faster hydrogen gas kinetics, reaching (∼77 % capacity in one hour), whereas larger 6500 NL (∼57 %/hour) absorb hydrogen gas more gradually but reduce thermal management intensities. Using 13 × 500 NL tanks reach ∼25 % discharge in 1 h but require ∼2170 Wh heating, whereas one 6500 NL tank only attains ∼48.5 % discharge yet uses ∼1750 Wh, illustrating a trade-off between faster kinetics and lower thermal load. A genetic algorithm identified an optimal configuration of two 6500 NL tanks that covered ∼68 % of total hydrogen gas consumption and 65 % of production at a maximum of 2.4 kW heating and 2.45 kW cooling. Additional comparisons with 170 bar compressed storage revealed lower instantaneous thermal requirements for high-pressure gas tanks. Adding a 170 bar compressed H 2 alongside the 10 bar MH system, hydrogen gas coverage rose from ∼70 % to ∼97 % when storage expanded to 200 Nm 3 , but at the cost of higher compression energy. The proposed MH-based approach, especially at moderate pressures with carefully planned tank geometries, achieves enhanced operational flexibility for a residential 120 m 2 building’s space heating and hot water, while machine learning optimizations further refine charge–discharge performance.