Considering the demand for wind power to participate in the power market, a mid-to-long term daily wind power generation prediction method based on GRU-Dlinear model is proposed. First, the trend feature encoder and seasonal feature encoder are built respectively based on GRU to extract the fluctuation characteristics and seasonal characteristics, and then the power prediction decoder is constructed based on Dlinear to predict the daily wind power generation in the next 30 days by integrating the fluctuation characteristics and seasonal characteristics. Using power generation data of two wind farms as case studies, the proposed prediction model shows the best performance among 5 prediction models in both two wind farms. The proposed GRU-Dlinear model can effectively predict daily wind power generation in the next 30 days with high accuracy. The prediction result can be a reference for power market transactions and improve dispatching efficiency of the power system.