The invention discloses a day-ahead robust scheduling method considering load fluctuation of a power distribution system. The day-ahead robust scheduling method comprises the following steps: step 1:establishing an advanced adiabatic compressed air energy storage AA-CAES power station operation and wind power generation system power generation model and determining the constraints of the model; step 2: establishing the objective function and using a robust dynamic programming method to solve the model; and step 3: performing simulation and verification of the method in the IEEE-RBTS test system according to the step 1 and the step 2 and performing simulation of only considering two scenes, wherein one is considering the uncertainty of wind turbine only, the other is considering the uncertainty of wind turbine and load. The CAES robust scheduling method based on fan and load prediction errors is provided so that the load fluctuation of the whole system is smaller. Different weights ofunderpower and uncertainty set are compared and the influence of the prediction error of the system on the reliability is analyzed.