The energy output of wind power generation systems is greatly affected by external factors. Usually, battery packs are used to cut peaks and fill valleys to stabilize power. This paper proposes a two-level control of the battery. First, the extended Kalman particle filter (EKPF) algorithm is used to estimate the state of charge of the battery pack, which reduces the system"s requirements for battery capacity and improves working accuracy. After that, the adaptive inversion sliding mode control strategy is applied to track the reference power based on accurate battery charge state data, which reduces the output power fluctuation of the battery and enables stable tracking. Two-stage control of the battery pack improves the sensitivity and stability of the battery pack. The effectiveness and superiority of the proposed control are verified by simulation and experiments.