Virtual power plants (VPPs) integrate distributed energy effectively and overcome geographical restrictions. This could improve the broad benefits of distributed energy generation. Owing to the intermittency of distributed wind and photovoltaic power at different time scales, when multiple VPPs exist, the interactive energy management strategy should be addressed in terms of two dimensions, namely multi-operator and multi-time-scale (day-ahead (24 hours), intraday (4 hours), and real-time (1 hour)). In this study, an interactive energy management framework for multiple VPPs was designed. Furthermore, a three-level energy-coordinated management model was employed for multi-VPP optimal operation including day-ahead cooperative scheduling, intraday noncooperative bidding, and real-time cooperative reserve. To solve this model, a chaotic search algorithm was applied to improve the ant-colony optimization algorithm. The resulting improved chaotic ant-colony optimization algorithm is proposed to increase the optimization velocity and solution efficiency. Finally, a demonstration project, namely the Sanliduo's VPP in Guangxi Province in China, was taken as an example. The following points can be stressed according to the obtained results: (a) the proposed model implements dispatching-balance-reserve interactive optimization such that the output of wind and photovoltaic power increased by 3.42% and 1.85%, respectively, and the revenue increased by 2.53%; (b) the proposed solution algorithm develops the optimal global strategy: compared with the ant-colony algorithm, the average convergence time increased by 44 seconds, and the supply cost decreased by 1.2%; (c) the robust coefficient provides effective risk decision-making tools for different types of decision-makers considering the operational characteristics of VPPs; (d) the openness of power markets is positively correlated to the balance revenue, and higher levels are beneficial for VPPs. Novelty statement Designed a two-layer multi-time scale and multi-agent energy management framework. Analyzed the interactive game behavior of different operation agents in multi-time scale. Proposed a three-level dispatching-bidding-reserve energy interactive management model. Constructed a greedy mutation strategy-based chaotic ant colony group intelligent algorithm. Discussed the influence of various sensitive variables on the optimal energy interactive strategy. Highlights Designed an interactive energy management framework for the optimal dispatching of virtual power plants (VPPs). Characterize the energy interactive behavior of multiple virtual power plants at multi-time scale. Proposed a three-level coordinated optimization model for multi-VPPs' dispatching-bidding-reserve interactive management. Constructed a greedy mutation strategy-based chaotic ant colony group intelligent algorithm. Calculate the threshold value of key decision variables for getting the optimal energy management strategy.