Exploring the PV Power Forecasting at Building Façades Using Gradient Boosting Methods

Solar power forecasting is of high interest in managing any system based on solar energy. In the case photovoltaic (PV) systems, and building integrated PV (BIPV) particular, it may help to better operate grid manage load storage. Power directly time series has some advantages over irradiance first modeling afterwards. this paper, for BIPV systems a vertical façade studied using machine learning algorithms decision trees. The scheme employs skforecast library from Python environment, which facilitates implementation different schemes both deterministic probabilistic applications. Firstly, hourly was performed with XGBoost Random Forest cases, showing an improvement accuracy when exogenous variables were used. Secondly, combined Bootstrap method. results paper show capabilities gradient boosting algorithms, such as XGBoost, work regressors power. Mean absolute error forecast, most influencing variables, around 40% close below 30% south east array, respectively.