LBNL FDD PFPU Data Sets

The DOE and Berkeley Lab have partnered across the national laboratory complex and with the research community to curate, validate, and publish the world’s largest set of labeled time-series data representing commercial HVAC systems operating in faulted and fault-free states. The data sets currently cover rooftop units, single-duct air handler units, dual-duct air handler units, variable air volume boxes, fan coil units, chiller plant, and boiler plant. Each data set includes from 20 to more than 100 data points that are commonly monitored in today’s buildings. This operational data is paired with ground truth information indicating which faults are present during which time periods.

美国能源部（Department of Energy, DOE）与劳伦斯伯克利国家实验室（Berkeley Lab）依托国家实验室集群，并联合科研共同体，共同甄选、校验并发布了全球规模最大的带标注时间序列数据集，该数据集涵盖处于故障与无故障运行状态的商用暖通空调（Heating, Ventilation, and Air Conditioning, HVAC）系统。当前数据集覆盖屋顶式空调机组、单风道空气处理机组、双风道空气处理机组、变风量末端装置（Variable Air Volume, VAV boxes）、风机盘管机组、冷水机组系统以及锅炉机组系统。每套数据集包含20至100余个当前楼宇中常规监测的数据测点。该运行数据配套有真实标签（ground truth）信息，可指明各时段内对应的故障类型。

搜集汇总

背景与挑战

背景概述

该数据集是由美国能源部和伯克利实验室合作整理的全球最大标记时间序列数据集，专注于商业暖通空调（HVAC）系统在故障和无故障状态下的运行数据。它覆盖多种HVAC设备类型，每个数据集包含20至100多个常用监测数据点，并配有故障时间的地面真实信息，适用于故障检测与诊断研究。数据集发布于2023年，采用CC BY 4.0许可证，属于建筑工程领域。