Wind power generation is an important part of renewable energy, and its output power has high volatility and uncertainty to address the nonstationary, multiscale, and intertwined trend-periodic characteristics of wind power forecasting time series, This article proposes the information-WM-GRU(multi-scale decomposition, wavelet decomposition, GRU and Informer).The proposed model uses wavelet transform analysis to decompose the original series into trend and seasonal components, which are modeled by GRU and Informer, respectively. GRU is used for long-term dependence modeling to accurately depict the slow change trend of wind power. The seasonal component has global dependence, and the informant based on probabilistic sparse attention mechanism can effectively extract the features of the seasonal component. The experimental results show that the proposed method is superior to the existing mainstream models on the real wind power data set, showing stronger generalization ability and prediction accuracy.