Abstract: In order to improve the utilization rate of clean energy such as wind power and reduce carbon emissions in integrated energy system, this paper proposes an integrated energy system optimization scheduling model based on the two-stage output prediction of new energy, complementary ensemble empirical mode decomposition, sparrow search algorithm and long short-term memory networks. Firstly, the empirical mode decomposition of complementary sets is performed on historical data, and along-term short-term memory network prediction model based on sparrow search algorithm optimization is constructed. Secondly, this prediction model is used to predict the power of wind power and photovoltaic in two stages of day-ahead and intra-day, respectively. Finally, taking the minimum daily operating cost including carbon penalty cost and curtailment penalty cost as the optimization goal, the IES two-stage scheduling model based on the forecast results of wind power and photovoltaic output is constructed, and the scheduling plan is formulated by CPLEX. Moreover, a park is taken as an example, and the results show that this two-stage scheduling model reduces the energy purchase cost of the system by 9.40%, the cost of carbon emission penalty by 14.05%, and the total daily operating cost by about 12.53%, which effectively improves the economic and environmental performance of the integrated energy system.