Firstly, the probability distribution characteristics of demand response resources are expounded. The IES（integrated energy system）operating cost is taken as the optimization goal and the power system, natural gas system operation constraints and energy coupling constraints are considered. The IES optimal energy flow（OFF）model is established, which is used as the input of the power system stability estimator to calculate the generator and coupling link power. Secondly, by setting up the transient simulation model under different load levels, the system stability under fault conditions is obtained. The training of power system stability estimator based on stack denoising auto-encoders（SDAE）is proposed. Finally, the IES comprised of the IEEE?39 node power system and the modified Belgian 20?node natural gas system is used to analyze the intelligent evaluation of power system stability. The simulation results show that the SDAE-based power system stability estimator has high recognition accuracy and excellent computational efficiency.