Medium and Long-Term Wind Power Forecasting Based on XGBoost Algorithm

Accurate mid- to long-term wind power forecasting is crucial for grid dispatch and energy planning. As the proportion of installed wind power capacity continues to increase, its inherent volatility and intermittency bring significant challenges to the power system’s peak and frequency regulation, backup capacity allocation, and energy infrastructure planning. Traditional physical models rely on high-precision meteorological numerical forecast data, but medium and long-term forecast errors accumulate over time. To solve the above problems, this paper proposes a medium- and long-term wind power prediction method based on XGBoost algorithm. Through the multi-source data fusion strategy, the meteorological elements such as wind speed, temperature, and air pressure generated by numerical weather forecasting (NWP) are deeply coupled with historical power series, and a multi-dimensional feature system including temporal features (such as seasonal cycles, holiday effects), statistical features (moving average, volatility) and physical derivative features (wind power density and wind speed cubes) is constructed. In model construction, XGBoost’s gradient lifting mechanism is used to automatically capture the nonlinear relationship of data, and hyperparameters such as tree depth, learning rate, and regularization parameters are optimized through grid search and cross-validation to enhance the model’s adaptability to non-stationary data.