Explainable AI for Solar PV Power Forecasting: From Model Selection to Trustworthy, Reproducible Deployment
November 20, 2026
A reproducible workflow combining automated model search with explainable AI reduces solar PV forecasting errors to an 11.6% nRMSE on a 16.32 kWp plant. The methodology replaces standard random splits with a structured deployment pipeline to ensure feature attributions and model selection are comparable against smart persistence baselines.