As an essential renewable energy source, the fast-growing penetration of wind power over the past decades has played an essential role in both industrial and domestic spheres. Nevertheless, the stochastic nature of wind results in pronounced volatility and intermittent behavior in wind power generation, posing a significant challenge to its widespread deployment. Consequently, precise short and medium-term wind power generation predictions can significantly enhance wind power generation systems' management, allocation, and overall stability of the power system. To achieve the above goal, this paper presents a hybrid model combining WDlinear and lightGBM methodologies, then applying Kalman filtering to obtain the final prediction results. The experimental outcomes demonstrate the superior predictive performance of the proposed hybrid model compared to the baseline models when forecasting 7-day wind power generation. Furthermore, the hybrid model exhibits noteworthy traits of precision and stability across various data scenarios, affirming its efficacy and superiority.