In this study, a solar-driven reduction process of nonstoichiometric cerium oxide in a fixed bed is optimized for efficient water splitting via metal-oxide-based redox cycling. Nitrogen is used as sweeping gas to scavenge oxygen from the beds during the reduction process. A transient lumped heat transfer model is developed for the simulation of the process. Parametric analysis and genetic algorithm are used to find the optimal N2 flow rate and establish a novel N2 feeding strategy with variable flow to maximize the thermal efficiency for water splitting. An efficiency close to 13% is estimated without solid-phase heat recovery, which is more than twice that of the best present experimental systems (∼5%). The results are regarded preliminary as a thermodynamic analysis.