R$^2$Energy: A Large-Scale Benchmark for Robust Renewable Energy Forecasting under Diverse and Extreme Conditions
10.7M hourly records from 902 stations expose critical failures during extreme weather events.
Deep Dive
Researchers from China introduced R²Energy, a large-scale benchmark for renewable energy forecasting. It contains 10.7 million hourly records from 902 wind and solar stations and uses a standardized, leakage-free paradigm for fair model comparison. The benchmark reveals a critical 'robustness gap' where models fail under extreme weather, showing reliability depends more on meteorological integration than architectural complexity. This provides a foundation for developing safer, more reliable grid AI.
Why It Matters
As climate change intensifies, robust AI forecasting is critical for maintaining stable power grids and preventing blackouts.