Abstract: The interference of wake in a large-scale wind farm decreases its overall power output. It is therefore necessary to establish an equivalent model of the wind farm and optimize its wake distribution to increase the overall power. The authors presented a novel yaw optimization framework for wind farms, including a radial basis function neural network algorithm improved by parameter supervised learning and regularization techniques, and an intelligent equivalent model for the power conversion of the wind farm targeting multi-degree-of-freedom yaw control. An optimization problem maximizing the power output of the wind farm with the yaw angles as decision variables was defined based on the present model. An enhanced elitist multiple population genetic algorithm was proposed and used to solve the optimization problem, and the optimum yaw angles of each wind turbines minimizing the wake interference were obtained. Numerical simulations were carried out to validate the equivalent models using an actual wind farm layout and realistic wind conditions. The results show good accordance between the present intelligent equivalent model and the actual characteristics of the wind farm. Under specific wind conditions, yaw optimization control leads to a power increase of 718.79 kW for the wind farm. For continuous wind conditions, the proposed optimization control method shows significant improvement under different wind conditions, resulting in an average power increase of 1 208 kW for the wind farm. The superiority of the proposed novel yaw optimization framework in enhancing the overall power output of wind farms has been validated.