Building simulation based on physical modeling is commonly adopted for performance prediction,but the high time costs hinder its application in the early design stage of buildings.Data-driven surrogate models have bee...Building simulation based on physical modeling is commonly adopted for performance prediction,but the high time costs hinder its application in the early design stage of buildings.Data-driven surrogate models have been proposed as a means to replicate computationally expensive simulation models.However,existing surrogate models for sustainable residential block design are limited in scope,focusing either on individual buildings or on specific cases within multi-block projects.This study leverages graph neural networks to develop optimal surrogate models incorporating inter-building effects to predict multiple indicators of sustainable performance for residential blocks at a region level.A graph schema is proposed to represent the general geometric features and relations among buildings in residential block design.A regional dataset is generated for model training and testing,using real residential zones in Hong Kong.The surrogate models are developed and evaluated,using two kinds of architectures(individual architectures for specific indicators and an integrative architecture)and three different neural networks(graph attention network(GAT),graph convolutional network,and artificial neural network).The results showed that the surrogate models using the individual architectures and GAT outperform the models using other architectures and neural networks.These surrogate models achieve a high prediction accuracy with CV(RMSE)s of 11.79%,7.63%,and 8.00%in terms of energy consumption,indoor thermal comfort,and daylighting,respectively,on the regional test set.Moreover,they enable a significant acceleration of the performance evaluation,reducing the calculation time from 6.346 min to 1.565 ms(243,297 times)per case compared to physics-based simulations.展开更多
Natural ventilation is particularly important for residential high-rise buildings as it maintains indoor human comfort without incurring the energy demands that air-conditioning does.To improve a building’s natural v...Natural ventilation is particularly important for residential high-rise buildings as it maintains indoor human comfort without incurring the energy demands that air-conditioning does.To improve a building’s natural ventilation,it is essential to develop models to understand the relationship between wind flow characteristics and the building's design.Significantly more effort is still needed for developing such reliable,accurate,and computationally economical models instead of currently the most popular physics-based models such as computational fluid dynamics(CFD)simulation.This paper,therefore,presents a novel model developed based on physics-based modelling and a data-driven approach to evaluate natural ventilation in residential high-rise buildings.The model first uses CFD to simulate wind pressures on the exterior surfaces of a high-rise building.Once the surface pressures have been obtained,multizone modelling is used to predict the air change per hour(ACH)for different flats in various configurations.Data-driven prediction models are then developed using data from the simulation and deep neural networks that are based on mean absolute error,mean absolute percentage error,and a fusion algorithm respectively.These data-driven models are used to predict the ACH of 25 flats.The results from multizone modelling and data-driven modelling are compared.The results imply a high accuracy of the data-driven prediction in comparison with physics-based models.The fusion algorithm-based neural network performs best,achieving 96%accuracy,which is the highest of all models tested.This study contributes a more efficient and robust method for predicting wind-induced natural ventilation.The findings describe the relationship between building design(e.g.,plan layout),distribution of surface pressure,and the resulting ACH,which serve to improve the practical design of sustainable buildings.展开更多
文摘Building simulation based on physical modeling is commonly adopted for performance prediction,but the high time costs hinder its application in the early design stage of buildings.Data-driven surrogate models have been proposed as a means to replicate computationally expensive simulation models.However,existing surrogate models for sustainable residential block design are limited in scope,focusing either on individual buildings or on specific cases within multi-block projects.This study leverages graph neural networks to develop optimal surrogate models incorporating inter-building effects to predict multiple indicators of sustainable performance for residential blocks at a region level.A graph schema is proposed to represent the general geometric features and relations among buildings in residential block design.A regional dataset is generated for model training and testing,using real residential zones in Hong Kong.The surrogate models are developed and evaluated,using two kinds of architectures(individual architectures for specific indicators and an integrative architecture)and three different neural networks(graph attention network(GAT),graph convolutional network,and artificial neural network).The results showed that the surrogate models using the individual architectures and GAT outperform the models using other architectures and neural networks.These surrogate models achieve a high prediction accuracy with CV(RMSE)s of 11.79%,7.63%,and 8.00%in terms of energy consumption,indoor thermal comfort,and daylighting,respectively,on the regional test set.Moreover,they enable a significant acceleration of the performance evaluation,reducing the calculation time from 6.346 min to 1.565 ms(243,297 times)per case compared to physics-based simulations.
基金supported by the Hong Kong University of Science and Technology Research Grant(project no.IGN17EG04).
文摘Natural ventilation is particularly important for residential high-rise buildings as it maintains indoor human comfort without incurring the energy demands that air-conditioning does.To improve a building’s natural ventilation,it is essential to develop models to understand the relationship between wind flow characteristics and the building's design.Significantly more effort is still needed for developing such reliable,accurate,and computationally economical models instead of currently the most popular physics-based models such as computational fluid dynamics(CFD)simulation.This paper,therefore,presents a novel model developed based on physics-based modelling and a data-driven approach to evaluate natural ventilation in residential high-rise buildings.The model first uses CFD to simulate wind pressures on the exterior surfaces of a high-rise building.Once the surface pressures have been obtained,multizone modelling is used to predict the air change per hour(ACH)for different flats in various configurations.Data-driven prediction models are then developed using data from the simulation and deep neural networks that are based on mean absolute error,mean absolute percentage error,and a fusion algorithm respectively.These data-driven models are used to predict the ACH of 25 flats.The results from multizone modelling and data-driven modelling are compared.The results imply a high accuracy of the data-driven prediction in comparison with physics-based models.The fusion algorithm-based neural network performs best,achieving 96%accuracy,which is the highest of all models tested.This study contributes a more efficient and robust method for predicting wind-induced natural ventilation.The findings describe the relationship between building design(e.g.,plan layout),distribution of surface pressure,and the resulting ACH,which serve to improve the practical design of sustainable buildings.
