Abstract: Deflected yaw control is beneficial to reduce the wake effect of wind turbines, and the power generation of wind farm can be maximized by reducing wake loss through yaw optimization at field-level. The FLORIS wake surrogate model was used to optimize the yaw angles of all turbines to reach the maximum total power of wind farm. The sensitivity of yaw optimization to wake loss and power increase was compared and analyzed in terms of different spacing distance of turbines, number of turbines, turbulence intensity, incoming wind speed and wind direction. Results show that yaw control of wind farm has the best effect on wake optimization when the turbine distance is less than 5D, there are more than 3 turbines at line, and only the first 5 rows need to be optimized, the angle between turbine-connect-line direction and main-frequency wind direction is less than 15°, the wind turbulence is less than 0.1 and wind speed is within the range of "cut-in+2 m/s" and "rated+2 m/s". If only single-turbine yaw control system is used, the yaw error of 3°~5° is beneficial to the overall wind farm power generation.