Reproduce “A Meta-Analysis of Solar Forecasting Based on Skill Score” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Thi Ngoc Nguyen, Felix Müsgens
This meta-analysis by Brandenburgische Technische Universität Cottbus-Senftenberg researchers examines deterministic solar forecasting accuracy using the Skill Score (SS) metric, quantifying the impacts of ten influencing factors and identifying horizon-specific optimal strategies. The study reveals a significant and accelerating increase in forecast accuracy over time, with up to a 10 percentage point improvement in recent years for certain forecast horizons.
Key insights
- The forecast horizon is the most critical variable, exhibiting a non-linear impact on SS and functionally segmenting forecasts into distinct intra-hour, intra-day, and day-ahead domains.
- Optimal training data length exists (around 2000 days), beyond which additional data may lead to decreased accuracy due to overfitting, while longer test sets (at least 365 days) reduce result cherry-picking.
- The effectiveness of input variables (historical, meteorological, NWP, spatial-temporal) and model types (e.g., ensemble-hybrid, image-based) is highly dependent on the specific forecast horizon.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.