According to the adjustment agility of the load aggregator, the interrupt load is divided into two categories, and a virtual power plant model based on demand response is constructed.On this basis, the operating cost of the new energy power system with minimum wind power is the optimization goal, and a twostage optimal scheduling model involving demand response virtual power plant is established. Based on the premise of optimizing the model to meet the system backup, the scheduling of different types of interruptible loads in the day before and during the day is studied in the range of 24 time periods. In particular, the effects of different wind power prediction error admission thresholds on interruptible load calls are considered. The results of the example show that the proposed model can reduce the operating cost of the new energy power system and determine the number of annual interruptions of the interruptible load.