With the increasing number of wind turbine generators（WTGs） connected to the electricity network via power electronic devices

the issue of system frequency stability needs attention. This paper presents a model predictive control method for wind farm to participate in primary frequency regulation while minimizing rotor speed distortion. The proposed control theme is made fully data-driven by fitting a dynamic model recursively from real-time measurements of WTGs and optimizing active power dispatch under model predictive control framework. Through coordinating the operation of WTGs as a group

the wind farm is capable of providing primary frequency regulation with flexible frequency droop characteristics

and the rotor state inside each WTG is dynamically optimized and stabilized to maintain the operation lifetime. Moreover

this coordinated control strategy can follow the dispatch signal from the upper control center by online adjusting the droop ratio of wind farm. The effectiveness of the method is verified by the case study of IEEE9 bus.