To enhance the accuracy of wind power prediction, this paper presents a wind power prediction method that incorporates feature selection and error correction techniques. Initially, the impact of wind speed, temperature, humidity, wind direction and other features on wind power output is comprehensively analyzed. A feature selection approach based on orthogonalization maximal information coefficient (OMIC) combined with a prediction model is proposed to optimize the feature dimension of the model. In addition, to address the inherent errors that may arise during the prediction model training, a dynamic mode decomposition (DMD) approach is employed to track the spatiotemporal mode of error data. DMD is advantageous due to its data-driven nature, which eliminates dependence on any parameter settings and prior assumptions, thereby enabling faster and easier error prediction. By optimizing the prediction model through feature selection and error correction, the proposed method yields more accurate prediction results. To evaluate the effectiveness of the proposed method, actual data from a single wind turbine in a wind farm located in the north is employed. The proposed method is combined with a deep learning model, and relevant prediction indicators are compared. Simulation results demonstrate that the proposed method effectively improves prediction accuracy.