This paper describes a data driven urban canopy model that can be coupled with detailed building energy models. The data driven model is used to assess the outdoor air temperature and humidity in a street canyon considering as inputs weather conditions at the atmospheric layer, the surface temperature of surrounding building facades and the street, and the heat released from the use of air-conditioning. Predictions made by the model were tested using measurements of the outdoor air temperature and humidity collected between April and August 2019 in Singapore. Results show that the model estimates the outdoor air temperature with a similar accuracy than others that were validated using the same input and test data, while providing estimates with a higher temporal resolution and considering urban morphology with a higher fidelity. They also demonstrate that the model can predict the impact of waste heat releases and cool pavement on the outdoor air temperature and building energy consumption. In the future, vegetation could be considered as an input of the model if the land surface temperature is measured using an infrared thermal camera. Another improvement would be to define weather conditions at the atmospheric layer from rooftop measurements or a climate model.

Urban building energy modelling aims at studying different approaches to estimate the energy consumed by buildings. However, most urban building energy models introduced in the literature ignore the impact that buildings have on their outdoor conditions, and inversely. Therefore, the paper determines whether interactions between buildings and outdoor conditions affects simulations performed using an urban building energy model, in particular during their calibration. Results suggest that interactions between buildings and their outdoor conditions affect the sensitivity of outcomes provided by an urban building energy model towards uncertain parameters and the probability to calibrate the model in accordance to some standards.

This paper suggests a method to simulate interactions between buildings and their outdoor conditions at the city-scale using a coupled scheme whose physical parameters are entirely assessed from data of the indoor and outdoor built environment. The coupled scheme consists of a reduced order building energy model and a single layer urban canopy model. In a previous study, it was proven that physical parameters of a single layer urban canopy model can be assessed using measurements of the outdoor temperature and humidity in a street canyon. For the coupled scheme to be fully data driven, the next step is to demonstrate that the reduced order building energy model can estimate the cooling consumption and exterior wall surface temperature in good agreement with measurements or simulated data after being trained using machine learning. Indeed, results show that a multi objective genetic algorithm can find values for physical parameters of the reduced order building energy model. Estimates of the cooling consumption and exterior wall surface temperature provided by the trained model achieve a CV-RMSE below 10% and a RMSE lower than 2.5 Kelvin, respectively, with respect to data generated from EnergyPlus. The last step towards a full data driven coupled scheme for city-scale simulations would be to iteratively train the reduce order building energy model with the single layer urban canopy model and show the convergence and accuracy of their respective outputs.