With distributed generation (DG) connected to the distribution network, in order to reduce voltage instability caused by high permeability DG, while considering the output fluctuation of DG and flexible load scheduling flexibility, a two-layer optimal dispatching model was proposed. A compensation mechanism based on the user's joint landscape benefit distribution model was constructed in the upper layer to change the user's consumption behavior, in order to improve the voltage quality as the goal. A dynamic active-reactive power optimization strategy of distribution network taking demand-side response into account was proposed. Firstly, the voltage fluctuation of distribution network was optimized by demand-side reactive response and reactive device compensation output, then the reactive power coupled active power was scheduled on the demand side to suppress DG fluctuation and reduce the network loss, the integrated operation cost of the distribution network was minimized. The satisfaction model was constructed in the lower-level, and the active power was obtained by the optimization of the upper-level demand response, which was used as the decision variable of the lower level user satisfaction, and the user satisfaction index was quantified. For the model features, an improved multi-objective gray wolf optimizer algorithm was proposed to generate Pareto optimal scheduling strategy. Finally, the IEEE-33 node system was taken as an example to verify the effectiveness of the proposed strategy.