Abstract: With the increase of wind power penetration rate, the problems of predictability and poor schedulability of wind farms have become prominent. Aiming at the problem that the above-mentioned factors lead to the reduction of wind power consumption level, this paper proposes an active-power layered control strategy combined with error feedforward prediction for wind power clusters. Firstly, an error-feedforward prediction model considering wind power variation trend is proposed, which is combined with wavelet packet decomposition and persistence method prediction model to form an ultra-short-term power prediction model, and the error feedforward limit is given according to the training situation of historical data. Secondly, on the basis of this prediction model, a  multiple temporal and spatial scales  active-power layered control strategy is proposed. The strategy divides the control layer  into cluster layer, farm group layer and sub-farm layer under the premise of existing scheduling instructions.  It realizes coordinated control of various wind farms. Finally, on the basis of the actual operation data of a wind power in Northeast China, simulation analysis is carried out by MATLAB and CPLEX. The results show that the proposed method improves the wind power accommodation level and the coordination output of wind farms energy storage.

Key words: wind power accommodation, error feedforward prediction, active-power layered control, ultra-short-term power prediction