This paper proposes a frequency support control scheme for large-scale wind farm (WF) by considering distributed layout of wind turbines (WT) and wake effects inside the WF. The frequency support of WF is formulated as optimization problems. A novel clustering-based optimization method is proposed to coordinate all the WTs so that the whole WF can work at the optimal operating states. In large-scale WF, number of WTs is numerous. With the proposed method, the WTs are first clustered into groups according to their wind profiles. By dispatching the same control commands to the WTs belonging to the same group, each group can be further seen as one single WT. As a result, the number of the WTs and control variables reduces significantly and hence the optimization problems are highly simplified. A case study is carried out to verify the effectiveness of the proposed method.