Abstract: As energy internet is constantly developing, it has become a study focus for a virtual power plant to give full play of the green value and the flexibility of distributed generation, distributed energy storage, micro-grid and other resources by using the users' demand response (DR) However, it is difficult to quantify the DR intentions of diverse users in various scenarios, which makes it a tough job for virtual power plants to use the users' DR. To solve this problem, firstly this paper constructs an accurate demand response model for virtual power plants. Secondly, in order to correctly reflect the demand response characteristics of different users, two LSTM networks are built to calculate the benefit coefficients of the demand response. On this basis, by combining and solving the accurate demand response model and the virtual power plant optimal dispatching model the optimal dispatching results are obtained. The simulation results show that the demand response coefficients established in this paper can correctly reflect the user's demand response intentions. The accurate demand response model can make the full use of the user side demand response and improve the overall economy of the virtual power plant.