BuildingsBench

NREL's BuildingsBench platform enables researchers studying time series foundation models to benchmark their methods on a challenging load forecasting problem.

BuildingsBench datasets consist of:

• A large-scale dataset of 900,000 buildings—statistically representative of the entire U.S. building stock--for pretraining models on the task of short-term load forecasting
• Seven real residential and commercial building datasets for benchmarking two downstream tasks evaluating generalization: zero-shot short-term load forecasting and transfer learning for short-term load forecasting
• A collection of various open-source residential and commercial real building energy consumption datasets, providing an evaluation benchmark.

Contact