Publication Type

Journal Article

Version

publishedVersion

Publication Date

4-2022

Abstract

Building energy use benchmarking is the process of measuring the energy performance of buildings relative to their peer group for creating awareness and identifying energy-saving opportunities. In this paper, we present the design and implementation of BEEM, a data-driven energy use benchmarking system for buildings in Singapore. The peer groups for comparison are established using a public energy disclosure data set. We use an ensemble tree algorithm for accurately modeling building energy use and for identifying the most influential factors. Our models reduce the prediction error from 24.39% to 6.04%, on average, when compared to the baseline linear regression models, which were used in the previous energy efficiency labeling program in Singapore, and outperforms ten other recent models. Using the prototype implementation of BEEM, we benchmarked three building types, office (290), hotel (203), and retail (125), and compared their rating. The code repository and the accompanying data set are released as an open-source project for community use.

Keywords

Building energy benchmarking, Building energy labeling, Regression analysis, Gradient boosting trees, Feature interactionInterpretable machine learning

Discipline

Asian Studies | Energy Policy | Engineering

Research Areas

Integrative Research Areas

Publication

Energy and Buildings

Volume

260

First Page

1

Last Page

10

ISSN

0378-7788

Identifier

10.1016/J.ENBUILD.2022.111869

Publisher

Elsevier

Additional URL

https://doi.org/10.1016/j.enbuild.2022.111869

Download

Share

COinS