Post-processing techniques and principal component analysis for regional wind power and solar irradiance forecasting

• Wind power and solar irradiance forecasting techniques are tested on two wide areas.
• Principal Component Analysis (PCA) allows reducing the dimension of the datasets.
• PCA combined with postprocessing reduces computational costs and forecast errors.

This work explores a Principal Component Analysis (PCA) in combination with two post-processing techniques for the prediction of wind power produced over Sicily, and of solar irradiance measured by Oklahoma Mesonet measurements’ network. For wind power, the study is conducted over a 2-year long period, with hourly data of the aggregated wind power output of the Sicily island. The 0–72 h wind predictions are generated with the limited-area Regional Atmospheric Modeling System (RAMS), with boundary conditions provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) deterministic forecast. For solar irradiance, we consider daily data of the aggregated solar radiation energy output (based on the Kaggle competition dataset) over an 8-year long period. Numerical Weather Prediction data for the contest come from the National Oceanic & Atmospheric Administration – Earth System Research Laboratory (NOAA/ESRL) Global Ensemble Forecast System (GEFS) Reforecast Version 2. The PCA is applied to reduce the datasets dimension. A Neural Network (NN) and an Analog Ensemble (AnEn) post-processing are then applied on the PCA output to obtain the final forecasts. The study shows that combining PCA with these post-processing techniques leads to better results when compared to the implementation without the PCA reduction.