In this study, the optimal design conditions of a repository system were investigated, as concerning the effect of initial conditions on a buffer bentonite. A coupled thermal-hydraulic-mechanical numerical model was established, and five influential factors, dry density and water content of each of buffer bentonite and backfill material, and permeability of the surrounding rock were set to determine relationships with buffer temperature, saturation time, and swelling time. After constructing metamodels based on a small number of simulation results, a multi-objective optimization was performed, using the genetic algorithm. As a result of the optimization process, a Pareto optimal set has been derived, and the process and the results provide a guideline for designing a repository system.