A review of data-driven building performance analysis and design on big on-site building performance data. (September 2021)

• Record Type: Journal Article
• Authors: Tian, Zhichao; Zhang, Xinkai; Wei, Shen; Du, Sihong; Shi, Xing
• Is Part Of: Journal of building engineering. Volume 41(2021)
• Publication Date: 2021-09
• Subjects: Building performance design -- Data-driven -- Building energy -- Building performance data; Building -- Periodicals; 690.05
• DOI: 10.1016/j.jobe.2021.102706
• Languages: English
• ISSNs: 2352-7102

Abstract

Abstract: Building performance design (BPD) is a crucial pathway to achieve high-performance buildings. Previous simulation-based BPD is being questioned due to the performance gaps between simulated and measured values. In recent years, accumulated on-site building performance data (OBPD) make it possible to analyze and design buildings with data-driven methods. This article makes a review of previous studies that conducted data-driven building performance analysis and design on a large amount of OBPD. The covered studies are summarized by the applied techniques, i.e., statistics, regression, classification, and clustering. The data used by these studies are compared and discussed emphasizing the data size and public availability. A comprehensive discussion is given about the achievements of existing studies, and challenges for boosting data-driven BPD from three aspects, i.e., developing data-driven models, the availability of building performance data, and stimulation of industrial practices. The review results indicate that data-driven methods were commonly applied to estimate energy consumptions, and explore energy trends, determinant features, and reference buildings. Identifying determinant features is one of the most successful applications. This study highlights the future research gaps for boosting data-driven building performance design. Highlights: This study focused on data-driven studies that fulfilled exclusively on on-site building data. Data-driven studies were classified by statistics, regression, classification, clustering. Major on-site building performance databases were summarized. Challenges were discussed on data-driven models, availability of data, industrial practices. Exploring determinant features and prediction of energy consumption are usual applications.