In order to increase the flexibility and economy of electricity retailers participating in the electricity market and improve the utilization efficiency of decentralized resources, a multi time scale rolling optimization model of electricity retailers is proposed. This is based on the aggregation management mode of virtual power plants. In the day-ahead stage, the optimization goal is to maximize the revenue of electricity retailers. Based on the multi scenario analysis method, the strategy of purchasing and selling electricity and the daily dispatching plan of each virtual power plant are developed. In the intra-day phase, more accurate distributed photovoltaic and load forecasting are considered, and an optimal dispatching model is constructed with the objective of minimizing the dispatching cost of virtual power plant. In the real-time stage, the capacity reserve of the virtual power plant is considered in the form of an opportunity constraint, and the output of some units of the virtual power plant is corrected, and the output of the uncertain unit is fed back to the intra-day stage. Simulation results show that the proposed multi time scale optimization model can improve the operation strategy of electricity retailers, allocate resources reasonably and maximize their own economic benefits. This work is supported by the National Natural Science Foundation of China (No. 51407104) and Open-end Fund of Fujian Provincial University Engineering Research Center (No. KF-X19016).