Ultra-Short-Term Wind Power Forecasting Method Based on an Improved XGBoost Algorithm Combined with the Smoothed Stochastic Indicator

As renewable energy rapidly develops, wind power is becoming increasingly vital in energy systems. This study proposes an ultra-short-term wind power forecasting method using machine learning and time series correlations to enhance accuracy and real-time performance. By employing an improved XGBoost algorithm and introducing the Smoothed Stochastic Indicator, the method reduces reliance on meteorological data, achieving superior accuracy and faster convergence on the Biliuhe Wind Farm dataset.