Abstract: The daily optimal scheduling of the traditional integrated energy system (IES) exists low energy utilization rate and poor economic and social benefits due to the over-simplification of equipment modeling, the low accuracy of the prediction model, and the delay of information capture and transmission. To solve the above problems, this paper proposes a real-time optimization scheduling strategy of the integrated energy system based on the digital twins and the dynamic energy efficiency model. Firstly, aiming at the problem of insufficient accuracy of the traditional predictions, a prediction model combining the long and short-term memory neural network and the similar day weather searching algorithm is constructed. Secondly, considering that the equipment energy efficiency coefficient is apt to be disturbed by the load rate and the environmental factors, a dynamic energy efficiency model of the equipment is established. Then, the physical model and the digital twins of the system are established respectively, and the output and load predictions and the IES real-time optimization of the new energy unit are carried out through the twin data of the environment and load. Finally, a test model of the integrated energy system in the park is established based on the Cloudpss platform to verify the effectiveness of the proposed strategy. The results show that the proposed strategy realizes the real-time information capture and transmission and improves the calculation accuracy of the model. It increases the energy efficiency while taking into account the operation costs, which has a guiding significance for the practical engineering.