Distinguishing Household Groupings within a Precinct Based on Energy Usage Patterns Using Machine Learning Analysis
Open Access
Abstract
The home can be a complex environment to understand, as well as to model and predict, due to inherent variability between people’s routines and practices. A one-size-fits-all approach does not consider people’s contextual and institutional influences that contribute to their daily routines. These contextual and institutional factors relate to the household structure and relationship between occupants, as well as the working lifestyle of the occupants. One household can consume resources and live quite differently compared to a similar size household with the same number of occupants due to these factors. Predictive analysis of consumption data can identify this difference to create household-specific modelling to predict occupant routines and practices. Using post-occupancy data from the Fairwater Living Laboratory in Sydney that monitored 39 homes built in a green-star community, this research has utilised machine learning approaches and a K-Means clustering method complemented by t-distributed Stochastic Neighbour Embedding (t-SNE) to show how households follow different daily routines and activities resulting in resource consumption. This analysis has identified energy usage patterns and household groupings with each group following similar daily routines and consumption. The comparison between modelling the precinct as a whole and modelling households individually shows how detail can be lost when aggregating household data at a precinct/community level. This detail can explain why policies or technologies are not as effective as their design due to ignoring the delicate aspects of household routines and practices. These household groupings can provide insight for policymakers to help them understand the different profiles that may be present in the community. These findings are useful for net-zero developments and decarbonization of the built environment through modelling occupant behaviour accurately and developing policies and technologies to suit.Keywords
Funding Information
- Australian Renewable Energy Agency (ARENA) (2018/ARP020)
- Curtin University (2018/ARP020)
This publication has 69 references indexed in Scilit:
- A high spatial resolution residential energy model based on American Time Use Survey data and the bootstrap sampling methodEnergy and Buildings, 2011
- A systematic procedure to study the influence of occupant behavior on building energy consumptionEnergy and Buildings, 2011
- The Short List: The Most Effective Actions U.S. Households Can Take to Curb Climate ChangeEnvironment: Science and Policy for Sustainable Development, 2008
- Energy productivity improvements and the rebound effect: An overview of the state of knowledgeEnergy Policy, 2007
- A long term study of residential home heating consumption and the effect of occupant behavior on homes in the Pacific Northwest constructed according to improved thermal standardsEnergy, 2006
- Consumption and Theories of PracticeJournal of Consumer Culture, 2005
- Household preferences for energy-saving measures: A conjoint analysisJournal of Economic Psychology, 2003
- The theory of planned behaviorOrganizational Behavior and Human Decision Processes, 1991
- Linking Life-Styles and Energy Use: A Matter of Time?Annual Review of Energy, 1989
- Self-Perception Based Strategies for Stimulating Energy ConservationJournal of Consumer Research, 1982