To investigate methods of conducting flame exposure tests (bonfire tests) on high-pressure hydrogen gas cylinders that are safe and have high accuracy across repeated tests, we used numerical simulation and experiments to analyze the feasibility of using substitutive gases for filling as well as the effects of the burners used as the fire source. Through a series of virtual experiments using substitutive gases, flame scales, and filling pressure as parameters, we examined the maximum internal pressure, the rate of pressure rise, and the starting time of Pressure Relief Device (PRD) activation. Because substitutive gas properties differ from those of hydrogen gas, we concluded that using substitutive gases would be inappropriate. In addition, we observed that when the flame scale was small, the cylinder's internal pressure before the thermal-activated PRD activation, the rate of pressure rise, and the starting time of PRD activation all increased rapidly. Therefore, it is necessary to either maintain a constant value for the fire source's fuel flow rate, or increase the flame scale, in order to reduce the variance between repeated tests.