The issue of forecast accuracy of green energy sources (wind, solar and waves) is becoming more and more important as energy production from green sources continues to increase year after year. Having accurate forecasts for the energy market clashes with intrinsic difficulties of forecasts due to, e.g., the coarse resolution of Numerical Weather Prediction models and the intermittent nature of all green sources (wind in primis). We study new AI-based calibration strategies able to overcome the abovementioned problems. The basic idea is to learn the NWP model error from a long track of past model forecasts where observed data are known (from in-situ measurements and/or from distributed observations coming, e.g., from satellites). Once the model error is learnt via suitable Machine Learning algorithms (deep or shallow) the latter is applied downstream a NWP model output to produce accurate forecasts.
Andrea Mazzino, Francesco Ferrari, Mattia Cavaiola, Peter Enos Tuju, Daniele Carnevale
G. Casciaro, F. Ferrari, M. Cavaiola, A. Mazzino, Novel
strategies of Ensemble Model Output Statistics (EMOS) for calibrating wind speed/power forecasts, Energy Conversion and Management 271, 116297 (2022)
G. Casciaro, F. Ferrari, D. Lagomarsino-Oneto, A. Lira-Loarca, A. Mazzino, Increasing the skill of short-term wind speed ensemble forecasts combining forecasts and observations via a new dynamic calibration, Energy 251, 123894 (2022)
D. Lagomarsino-Oneto, G. Meanti, N. Pagliana, A. Verri, A. Mazzino, L. Rosasco, A. Seminara, Physics informed machine learning for wind speed prediction, Energy 268, 126628 (2023)