The potential solution to hydrogen storage and transportation challenges is provided by liquid fuel-based on-line hydrogen production. An online hydrogen production device has been constructed and the stable output of hydrogen at 6.8 bar under dynamic conditions of 6–13 Nm3/h has been achieved. To achieve optimal economic and thermodynamic performances, the mathematical models for mass, energy, and techno-economics are constructed based on field test data. The Non-Dominated Sorting Genetic Algorithm is employed to update temperature splitting points between multi-stage heat exchangers, targeting multi-objective optimization of heat exchange area, system energy efficiency, and annual average cost. Pareto optimal solutions indicate that such optimization can improve system energy efficiency by 0.03%–3.76% and decrease the hydrogen production cost from 0.004 $/kg to 0.599 $/kg. This work demonstrates the feasibility and effectiveness of multi-objective optimization for online hydrogen production system design and offers important guidance for decision-making in heat exchange network schemes.