INVESTIGATING HOW BUILDING ARCHETYPES ARE CREATED BASED ON THEIR COOLING ENERGY CONSUMPTION FOR BUILDINGS WITHIN THE NUS CAMPUS

Abstract

With more building owners striving to be energy efficient, Urban Building Energy Models (UBEMs) have emerged as an attractive area due to their ability to simulate and generate insights on the ideal energy conservation measures for buildings. However, developing accurate models of large urban areas require extensive calibration, which makes them time-consuming and costly to develop. While using clustering to generate building archetypes from building stock is widely used to reduce the computational effort to develop such models, there is limited research where clustering was used in the tropics. The main objectives of this paper are to develop a simplified clustering approach that could be used to develop UBEMs and recommend how the archetypes generated could be further used.

The dataset comprised of weather, cooling demands, WiFi count and GFA data of buildings within the NUS Kent Ridge campus. A six-step framework was developed to create the archetypes. In order to reduce the computational effort required during clustering, feature selection was included to select features that provide useful and relevant information. The elbow method and silhouette index were incorporated to determine the optimal number of clusters and ensure that the dataset is well clustered, respectively.

Based on the findings, the WiFi count and cooling demands were features used in the clustering process. To ensure a fair comparison among buildings, these features were normalised before clustering the dataset to derive three main archetypes. The study also analysed trends based on the clustering results. Although the results were insufficient to allocate a specific building type to an archetype, the archetypes provided insights into the rate of change of the features.

The study then concludes by suggesting how the framework and archetypes developed could be used for future studies.