Building Data Genome Project

本文描述了来自1,636座非住宅建筑的3,053个电能表的开放数据集，范围为整整两年（2016年和2017年），每小时频率（每米17,544次测量，约5360万次测量）。这些仪表是从北美和欧洲的 19 个地点收集的，每栋建筑都有一个或多个仪表测量整个建筑的电力、加热和冷却水、蒸汽和太阳能以及水和灌溉仪表。2019 年 10 月至 12 月，美国供暖、制冷和空调工程师协会 （ASHRAE） 主办的 Great Energy Predictor III （GEPIII） 竞赛中使用了部分数据。GEPIII是一项用于长期预测的机器学习竞赛，适用于测量和验证。本文描述了时间序列仪表数据、建筑物元数据和补充天气数据的数据收集、清理和收敛过程。该数据集可用于进一步的预测基准测试和原型设计，以及异常检测、能量分析和建筑类型分类。

This paper presents an open dataset of 3,053 electricity meters from 1,636 non-residential buildings, covering a full two-year period (2016 and 2017) at an hourly sampling frequency, with 17,544 measurements per meter and approximately 53.6 million total measurements. These meters were collected from 19 sites across North America and Europe. Each building is equipped with one or more meters measuring whole-building electricity, heating and cooling water, steam, solar energy systems, as well as water and irrigation usage. Portions of this dataset were used in the Great Energy Predictor III (GEPIII) competition hosted by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) from October to December 2019. GEPIII is a machine learning competition focused on long-term forecasting for measurement and verification purposes. This paper details the data collection, cleaning, and convergence processes for the time-series meter data, building metadata, and supplementary weather datasets. This dataset can be utilized for further forecasting benchmarking and prototyping, as well as anomaly detection, energy analysis, and building type classification.

搜集汇总

背景与挑战

背景概述

Building Data Genome Project是一个包含1,636座非住宅建筑的3,053个电能表的开放数据集，涵盖2016年和2017年两年的每小时频率测量数据（约5360万次测量）。这些数据来自北美和欧洲的19个地点，曾用于ASHRAE的Great Energy Predictor III竞赛，适用于机器学习预测、异常检测和建筑类型分类等应用。