Abstract: In order to enhance the flexibility and voltage stability of wind farms facing external uncertainties, a multi-scenario reactive voltage optimization control strategy based on self-adaptive model predictive control (MPC) was proposed. A reactive voltage prediction model was established using sensitivity matrices, and scenario segmentation was conducted based on evaluation criteria derived from short-term wind speed forecasts. The control objective was to minimize voltage deviations and active power losses at multiple time points within the prediction horizon. Adaptive adjustment of weight ratios in various scenarios facilitated rolling optimization. Crawfish optimization algorithm (COA) was employed to solve the optimization model and derive reactive power optimization schedules. Feedback correction based on optimization results and prediction errors was implemented. Simulation was conducted using a wind farm in the "Three-North" regions of China. Results show that compared with traditional open-loop control and MPC, the proposed strategy reduces daily average voltage deviations by 82.46% and 67.74%, respectively, validating the feasibility and effectiveness of this control strategy.