The invention discloses a power distribution network multi-time scale optimization operation method considering uncertainty. The method comprises the steps that 1, a min-max-min two-stage three-layerrobust optimization dispatching model is established, a column constraint generation algorithm (CCG) is utilized for solving the robust optimization dispatching model, and the operation state of a slow motion device in the worst scenario is determined; 2, based on short-term prediction data of gray prediction, an objective function is set, in combination with the system operation constraint conditions and the determined operation state of a slow action device, an optimization operation model of the active distribution network under the short time scale is established; 3, based on ultra-short term prediction data of gray prediction, and by means of comprehensive consideration of limit of operation times and operating time of adjustable controllable equipment, an objective function is set, in combination with the system operation constraint conditions and the determined operation state of the slow action device, an optimization operation model of the active distribution network under theultra-short time scale is established. The safety of the system can be well guaranteed in distributed power supply high-penetration areas.