ArchEGraph/ArchEGraph
收藏资源简介:
PACK是一个为基于图和天气条件学习而组织的建筑能源数据集。数据集包含49,326个案例,涉及5,481个独特的建筑和64个独特的天气ID。每个案例的时间步数范围从968到8,760,空间数范围从1到231。数据集包含manifest.csv文件(所有案例的索引)、building/文件夹(5,481个文件)、geometry/文件夹(5,482个文件)、weather/文件夹(64个文件)、energy/文件夹(49,326个文件,嵌套在子文件夹如00/下)和split/文件夹(预定义的分割CSV文件)。每个案例在manifest.csv中有一行记录,包含样本ID、来源标签、天气ID、建筑ID、能源文件路径、时间步数和空间数等信息。数据集还提供了多个分割文件,用于不同的训练和测试场景。
PACK is a building-energy dataset organized for graph-based and weather-conditioned learning. The dataset contains 49,326 cases, involving 5,481 unique buildings and 64 unique weather IDs. The range of time steps (n_steps) is from 968 to 8,760, and the range of spaces (n_spaces) is from 1 to 231. The dataset includes manifest.csv (index of all cases), building/ (5,481 files), geometry/ (5,482 files), weather/ (64 files), energy/ (49,326 files; nested under subfolders like 00/), and split/ (predefined split CSV files). Each case in manifest.csv contains sample_id, source_job_tag, weather_id, building_id, energy_file, n_steps, and n_spaces. The dataset also provides several split files for different training and testing scenarios.




