Zhang, Ruijun
(2022)
Modelling of neighbourhood effect in cities by coupling computational fluid dynamics and building energy simulation techniques.
PhD thesis, University of Nottingham.
Abstract
Building energy simulation (BES) is widely applied to assess indoor comfort and building energy demand, which is reported to encompass a one-third share of the world’s energy demand in 2021. There is a range of worldwide accepted building energy simulation packages in hand, such as EnergyPlus ©, Revit©, DOE-2©. These tools comprise series of subroutines to predict the behaviour of systems within the buildings. The calculations in these programs are based on the combination of well-defined laws (i.e., energy and mass balance) and empirical algorithms (e.g., convective heat transfer coefficient). In particular, despite numerous updates over the empirical algorithms of convective heat transfer coefficient (CHTC), the current packages are still considered not proper in representing the local CHTC values in many urban occasions as they simplify the surrounding airflows with homogeneous patterns. Furthermore, it has been reported that the inadequate understanding of outdoor airflow can lead to up to 20 – 40 % error in building energy predictions. This weakness, thus, initiated a subject of the research in the past decades to couple BES with computational fluid dynamics (CFD) tools, which are known for their strengths in airflow modelling, especially in representing the neighbourhood effects in an urban area.
Dynamic coupling of BES to computational fluid dynamics (CFD) techniques are a common strategy to improve the simulation performance. The precedent research of CFD-BES coupling mainly focuses on the indoor environment, but only a few consider the outdoor microclimate conditions. This research aims to investigate the neighbourhood effect on the convection of buildings' exterior surfaces and enhance their presence in the local convective heat transfer coefficient format in building energy modelling.
Among the dynamic coupling strategies, the fully dynamic coupling is understood as computationally intensive and impractical in medium-to-long-term modelling or even short-term (hourly, daily or weekly) modelling of naturally ventilated scenarios on a neighbourhood scale. Therefore, though it provides a more accurate assessment than the quasi-dynamic approach, it is less popular than the latter one. In this study, frameworks of fully dynamic coupling and virtual dynamic coupling are proposed for short-term and medium-to-long-term (monthly, seasonally or annually) modelling, respectively.
Three case studies are performed for
1) short-term modelling of scenarios with all buildings sealed from outdoors without natural ventilation (sealed scenarios);
2) short-term modelling of scenarios with all buildings under the natural ventilation during the night-time (ventilated scenarios);
3) medium-to-long-term modelling of sealed scenarios.
where ‘sealed’ here means the rooms are sealed from outdoors with the windows closed all the time.
The first case study proves the feasibility of the developed benchmark coupling framework. After that, the second case study expands the framework for application in scenarios of natural ventilation with fast calculations. The last case study provides the virtual dynamic coupling of the CFD and BES with artificial neural network for medium-to-long-term prediction of CHTCs. All case studies are performed in typical hot weather for a simple city community in Los Angeles, U.S.
The results highlight the importance of the neighbourhood effect. For short-term modelling of sealed scenarios, the difference between the prediction of the hourly averaged external convection using the coupling method and that of the standalone conventional BES models is up to 64 %. Furthermore, for short-term modelling of ventilated scenarios, the proposed model substantially reduces the computational cost of the dynamic coupling procedure, taking almost 1/200 of time as the conventional method. Concerning the medium-to-long-term modelling using a virtual dynamic coupling, the predictions of the local CHTCs on the external surfaces are found satisfactory with an accuracy of 0.88. Moreover, ten is the effective number of days to train the neural network tools for a one-month simulation—the proposed approach saves approximately 2/3 of the required computational time using an ordinary approach.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
