Abstract: To adapt to the large-scale wind power penetration, this paper proposes a real-time power dispatch where the units and storages jointly participate in the automatic generation control. Besides, a point estimate-based affinely adjustable robust optimization is put forward for the wind uncertainty. Different from the traditional affinely adjustable robust optimization which optimizes the base operation cost, the proposed point estimate-based affinely adjustable robust optimization optimizes the expected operation cost to improve the economical efficiency. In this paper, a deterministic point estimation method is introduced to quickly and accurately evaluate the expected operating cost. The proposed model is a mixed integer bilinear constraint problem, which uses a "estimation-correction" convexified method to solve the problem. In the estimation stage, the state variables of storages are relaxed, while in the correction stage, the state variables are directly corrected. Finally, a difference-of-convex optimization is introduced to convexify the bilinear constraint problems in these two stages in order to improve the solution quality of the real-time dispatch with storages. The simulations in the IEEE 39, 118 and 300 node systems verify the effectiveness of the proposed model and method.