Abstract
Rapid urbanisation has led to significant population growth and increased demand for residential building capacity, resulting in greater energy consumption and elevated greenhouse gas emissions. In the context of the increasing interest in urban sustainability, it is essential to accurately understand and assess the energy demand of buildings. Among the various factors, occupant behaviour has been identified as one of the key factors influencing building energy consumption. Accordingly, in recent years, occupant behaviour in buildings has gradually attracted the attention of researchers and policy makers aiming to reduce energy consumption, promote green buildings, and achieve low-carbon urban goals.Currently, occupant behaviour research in the field of building energy consumption and carbon emission is concentrated on two directions: the investigation of low carbon behaviour strategies based on the occupant perspective and the development of occupant behaviour models for building energy simulation, respectively. The former is intended to reveal the mechanism of occupants' individual behaviours in relation to the macro-level carbon reduction targets. It emphasises the need for a deeper understanding of the specific sources of CO2 emissions, particularly at the individual level, in response to global climate change. Although the current literature extensively discusses the driving factors of individual or household carbon emissions, how to achieve carbon reduction through adjustments in individual behaviour within the current market mechanism remains a research gap that should be addressed. The latter, on the other hand, is devoted to the assessment of building energy consumption through occupant behaviour modelling. However, traditional models often rely on static standard profiles that fail to capture the dynamic and stochastic nature of occupant behaviour. Meanwhile, privacy concerns further limit the applicability of stochastic models to public buildings, with limited implementation in residential buildings. Furthermore, in energy consumption simulations of urban residential buildings, the widely used bottom-up approach is primarily based on prototype buildings, gradually constructing citywide energy models. These prototype buildings typically include basic building geometry but lack detailed functional spaces. Moreover, the stochastic of occupant behaviour and the combined effects of microclimate factors (such as surrounding air temperature) at the urban scale remain a significant challenge for urban-scale simulations. Overall, accurately modelling these complexities is essential for improving the reliability and applicability of urban building energy simulations.
To address these challenges, this study establishes a research framework focused on occupants' individual behaviours and residential building energy consumption reduction and emission reduction using Time Use Survey (TUS) data. On the level of individual occupant behaviour, based on the TUS dataset, path analysis is used to analyse the relationship between occupants' basic characteristics and their energy consumption/carbon emissions in various behaviours. T-tests are used to quantify the variations in carbon emissions caused by individual differences, and then carbon emission reduction recommendations are provided accordingly. At the level of energy consumption simulation in residential buildings, this study develops a stochastic occupancy simulation method for different functional spaces in residential buildings based on the TUS dataset using Markov chains and probabilistic sampling algorithms. The developed method is further applied at the urban level. Within the bottom-up urban building energy simulation framework, this study employs the k-means algorithm to identify prototype buildings of the study area. Subsequently, an urban-scale building energy simulation is established. It integrates stochastic occupant behaviour modelling with consideration of microclimatic conditions (with a focus on incorporating the air temperature). The key results and contributions of this study are as follows:
1） Through a detailed analysis of energy consumption related to personal activities of residents across multiple countries, this study found that, beyond basic demographic factors such as gender and age, various driving factors significantly affect the individual energy consumption in different countries. Therefore, effective carbon reduction strategies must account for country-specific characteristics. Based on an in-depth analysis using the UK TUS dataset, this study reveals that age is a predominant driving factor influencing carbon emissions related to employment, studying, and mass media consumption. Gender significantly impacts household and family care, social life and entertainment as well as personal care activities, while household income levels influence emissions from hobbies and computing. These insights highlight the importance of developing tailored carbon reduction strategies based on an individual behavioural patterns. The analysis provides quartile results for different behaviours, aiding policymakers in devising carbon reduction strategies customised to occupants' emissions profiles. The scenario analysis suggests that such suggestions have a carbon reduction potential ranging from 12.93% to 67.90%.
2） A novel model construction approach is proposed to simulate occupant movements and room occupancy in residential buildings. The model's accuracy is ascertained through ten-fold cross-validation, achieving an average R2 value of 0.91 across key functional rooms (bedroom, bathroom, kitchen, living room).
3） Prototype buildings with detailed functional zoning are developed for Xiamen Island. Meanwhile, high-resolution microclimate temperature data is generated using an interpolation method. The validation results show that the temperature obtained using this method has a minimum correlation of 0.99 with the real values, significantly outperforming the temperature data provided by the EnergyPlus Weather File (with a maximum correlation of only 0.83). By integrating the prototype building models, the 1km×1km grid temperature dataset and the stochastic occupant behaviour generation method, the study generated energy consumption data for eleven different residential communities in six zones in Xiamen. It is worth noting that in four of the six study regions the distribution of modelled and measured data presented no statistically significant differences (p>0.05), with negligible |δ| values. Considering the entire urban area collectively, the overall p=0.674 and |δ| value is -0.002 (negligible difference). All the quantitative evidence collectively indicate that the model constructed in this study has a reliable predictive accuracy at the urban spatial scale. Also, the simulation results show that the simulation can, on average, cover 96.26% of the actual energy consumption data, demonstrating a high level of accuracy. Meanwhile, the simulation results for the annual energy use intensity of each community are consistent with the range of results from previous studies. In summary, the model not only reflects regional differences in urban building energy consumption, but also provides more accurate energy simulation results.
Overall, this research explores occupant behaviour at multiple scales. Firstly, it investigates on the behaviour patterns of individual occupants to facilitate targeted behavioural interventions and policy formulation aimed at carbon reduction. Then, it develops a scalable and precise occupant behaviour simulation methodology that accurately reflects occupant stochasticity, serving as an effective tool for energy simulation and sustainable residential building design. Finally, it proposes an urban building energy simulation method that considers detailed building models, occupant behaviour and the surrounding microclimatic conditions. It offers a dynamic, realistic simulation that is closer to the real-life scenario, and provides technical support for city-level energy assessment and urban sustainability planning.
| Date of Award | 13 Jul 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Tongyu Zhou (Supervisor), Hong Ye (Supervisor) & Jo Darkwa (Supervisor) |