Abstract: In an actual industrial process, the control system is often affected by the time-variant disturbances, so the minimum variance control design based on a single disturbance model as the benchmark is no longer useful to assess the time-variant disturbance control loop performance. If several disturbances appear at the same time, a conventional method of multi-model switching could result in intermittent switching and a larger transient error, failing to evaluate the current system performance accurately. Therefore this paper proposes a multi-model mixing minimum variance control for performance assessment. Firstly, one minimum variance controller is designed for each disturbance characteristic, such that a multiple model minimum variance controller can be formed. Then at each time point, multiple minimum variance controllers are mixed, and the output variance is the final performance evaluation benchmark. This method fully considers the feature of each disturbance and avoids the intermittent switching and larger transient error caused by the conventional multi-model switching method, realize accurate of and reliable assess of the time-variant disturbance performance. Finally, the effectiveness of the multi-model mixing minimum variance control as the benchmark to assess time-variant disturbance control systems is verified through simulations.