Buildings-900K
收藏资源简介:
Buildings-900K是一个大规模的模拟数据集,由国家可再生能源实验室创建,包含900,000个代表美国建筑库存的模拟建筑。该数据集通过精心校准和验证的物理基础建筑模拟创建,旨在为整个美国建筑库存的能源消耗提供统计代表性的数据库。数据集还包括一个评估平台,结合了7个开放数据集,总计超过1,900个真实建筑。Buildings-900K支持研究大规模预训练,以及零样本短期负荷预测和迁移学习任务,旨在促进可泛化的短期负荷预测研究。
Buildings-900K is a large-scale simulation dataset developed by the National Renewable Energy Laboratory, containing 900,000 simulated buildings representing the U.S. building stock. Constructed via carefully calibrated and validated physics-based building simulations, this dataset aims to provide a statistically representative database for the energy consumption of the entire U.S. building stock. The dataset also includes an evaluation platform that integrates seven open datasets, totaling over 1,900 real-world buildings. Buildings-900K supports research on large-scale pre-training, zero-shot short-term load forecasting and transfer learning tasks, with the goal of advancing research on generalizable short-term load forecasting.




