Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

The impact of occupancy on baseline building energy modelling performance

Chowdhury, Md Rezaul Karim (2022) The impact of occupancy on baseline building energy modelling performance. Other thesis, Murdoch University.

[img]
Preview
PDF - Whole Thesis
Download (6MB) | Preview

Abstract

For promoting energy efficiency practices in the building sector, energy conservation measures are of utmost importance. The energy conservation measures implemented through Energy performance contracting (EPC) are predominantly linked with the performance of baseline building energy models. While occupants are recognized as one of the most important driving factors of energy use in buildings, current research has failed to identify if building occupancy rate can be an influential independent variable to predict baseline energy use in buildings. This research aims to identify the influence of considering occupancy rate as an explanatory variable on the modelling performance of baseline building energy.

Six multivariate machine learning approaches (e.g., linear regression, regression trees, ensembles of trees, Gaussian Process, support vector machine, nonlinear Autoregressive Exogenous model (NARX)) and one univariate (e.g., reformed ten-parameter change point model) inverse modelling approach were employed in the baseline model development process of building heating and cooling energy use and electricity. The specified multivariate baseline modelling approaches were investigated to better isolate the impact of occupancy on building energy performance. NARX outperformed other baseline modelling approaches in terms of model predictive accuracy and data fitting capabilities. On the contrary, the proposed adapted change point model demonstrates the capability of providing operational insight into the case study building.

The hourly fifteen-month worth of energy use data used in baseline models was extracted from the building management system (BMS) server of a real case study building. The prediction period was defined as the most recent six months of the available data representing the COVID lockdown period. The models were trained using the nine-month worth of data that immediately preceded the prediction period. The arrangement of different input parameters selected by a forward sequential feature selection approach was considered an important step to identify the influence of individual parameters on baseline energy use. The influence of occupancy on the accuracy of baseline models was quantitively evaluated from this analysis. The results show that baseline model performance slightly improves when occupancy data are considered as an explanatory variable. However, occupancy data can significantly influence the performance of a baseline energy use model in an occupant-centric building. The assessment of hourly energy data and associated occupancy data for the case study building indicates the necessity of implementing occupant-centric control strategies to improve its energy performance.

Item Type: Thesis (Other)
Murdoch Affiliation(s): Engineering and Energy
United Nations SDGs: Goal 7: Affordable and Clean Energy
Notes: This is a Research Masters with Training (RMT) thesis
Supervisor(s): Urmee, Tania, Whale, Jonathan, Shahnia, Farhad and Gunay, Burak
URI: http://researchrepository.murdoch.edu.au/id/eprint/64692
Item Control Page Item Control Page

Downloads

Downloads per month over past year