Due to the enhanced permeability of renewable energy, the power current distribution of the distribution network changes, and the variability of renewable energy generation results in power fluctuations on the distribution network, which significantly impacts the operation of distribution network. Therefore, this paper proposes multi-timescale coordinated scheduling approach considering wind generation correlation analysis and improved model predictive control (MPC). Firstly, based on copula correlation analysis, correlation model of predictive error between multiple wind farms and different time periods is formed, aiming at capturing the stochastic character of wind generation output more accurately. According to the accuracy of prediction data at different time scales, this paper selects different optimization methods. Furtherly, a novel multi-timescale dispatching method is established. In the day-ahead and intraday scheduling stages, stochastic optimization method is adopted to minimize operation cost based on the scenario method. In the real-time stage, deterministic optimization method is adopted to minimize the power fluctuation between the distribution network and the transmission network. Therefore, it can ensure that the distribution network has the minimal power fluctuation between the transmission network under the premise of the lowest operating cost. Then, due to the characteristic of MPC scheme and system operation, an improved MPC method is proposed to further limit the fluctuation of exchange power between the transmission and distribution networks. Finally, the simulation verifies the results.