XGBoost–SFS and Double Nested Stacking Ensemble Model for Photovoltaic Power Forecasting under Variable Weather Conditions

Sustainability can achieve a balance among economic prosperity, social equity, and environmental protection to ensure the sustainable development happiness of current future generations; photovoltaic (PV) power, as clean, renewable energy, is closely related sustainability providing reliable energy supply for development. To solve problem with difficulty PV power forecasting due its strong intermittency volatility, which influenced by complex ever-changing natural factors, this paper proposes method based on eXtreme gradient boosting (XGBoost)–sequential forward selection (SFS) double nested stacking (DNS) ensemble model improve stability accuracy forecasts. First, analyzes variety relevant features affecting correlation between these then constructs two features: global horizontal irradiance (GHI) similar day power. Next, total 16 types feature data, such temperature, azimuth, ground pressure, are preprocessed optimal combination selected establishing an XGBoost–SFS build multidimensional climate dataset. Then, DNS model. Based decision tree (GBDT), XGBoost, support vector regression (SVR), base set, new constructed again metamodel already in order make more robust reliable. Finally, station data from 2019 used example validation, results show that proposed effectively integrate multiple factors better nonlinear relationships features. This applicable case variable climates have higher requirements.