House 18 in ENERTALK dataset

Zip files enertalk-dataset-{house_number} contain a directory for each houses. Each directory holds a set of subdirectories that contain Parquet files for the daily aggregate and appliance-level data. The naming convention for these subdirectories is “” (e.g. “20161124” for November 24, 2016). The Parquet files are named “\_.parquet.gzip” (e.g. “01\_fridge.parquet.gzip”). In these names, the two-digit integer is uniquely associated with a distinct measuring device in a house. Each Parquet file consists of three columns: “timestamp,” “active\_power,” and “reactive\_power.” The “timestamp” column contains Unix timestamps in milliseconds, such that 1000 corresponds to one second. The “active\_power” column represents active power in watts and the “reactive\_power” column represents reactive power in VAR (volt-ampere reactive) units.

名为"enertalk-dataset-{house_number}"的压缩包中，为每户住宅分别设有一个专属目录。每个住宅目录下包含若干子目录，用于存储日聚合用电数据与设备级用电数据对应的Parquet文件。此类子目录的命名规则为YYYYMMDD格式（例如"20161124"代表2016年11月24日）。Parquet文件的命名格式为"{两位整数}_{设备标识}.parquet.gzip"（例如"01_fridge.parquet.gzip"）。文件名中的两位整数，唯一对应住宅内的一台独立测量设备。每份Parquet文件均包含三列数据："timestamp"（时间戳）、"active_power"（有功功率）与"reactive_power"（无功功率）。其中"timestamp"列存储以毫秒为单位的Unix时间戳，即数值1000对应1秒；"active_power"列以瓦特（W）为单位表征有功功率，"reactive_power"列则以VAR（无功伏安，volt-ampere reactive）为单位表征无功功率。