The integrated cryogenic separation system (ICSS), which includes a chilling train and a demethanization section, is a crucial production system in an ethylene plant. It accounts for up to 50% of the total start-up time and flare emissions of the whole ethylene plant. Traditional start-up scenarios are developed on experience and improved by trial and error, which is inefficient and dangerous. This paper employs rigorous dynamic simulations to examine the potential infeasibilities and operational risks of different start-up scenarios. The best start-up scenario with minimal start-up time and flare emissions is determined. To obtain the initial start-up state with ambient nitrogen filling in the whole system, an initialization algorithm based on parameter modification for the dynamic ICSS model is presented. This novel initialization algorithm greatly saves time and effort of the initialization of the start-up model compared with previous studies. In addition, if time-related variables are important for scenario study or temperature of the dynamic model changes greatly, heat capacity must be added into dynamic start-up model, which was not mentioned in published literatures. They are critical to determine accurate start-up time, control the cooling rate, and verify the feasibility of scenarios during the dynamic start-up process under arbitrary temperature changes. The dynamic simulation provides insight into process dynamic behavior, which is crucial for the plant to evaluate and improve start-up scenarios. A real case study has demonstrated the efficacy of the dynamic simulation with heat capacity added.