CHARACTERIZATION OF URBAN MORPHOLOGY AND ITS EFFECT ON WEATHER UNCERTAINTY IN BUILDING ENERGY SIMULATION

Abstract

The application of building simulation serves to assess the performance of a building throughout its lifetime. But the proper use of these applications relies heavily on the boundary conditions under which the behavior of a model is simulated. One of the most important inputs for simulation models is the stimulus by the weather conditions (actual or typical) in which it is supposed to operate. Traditionally, weather data for building simulation is a composed of 8760 hourly values of weather variables (temperature, humidity, solar insolation etc.) derived through statistical means from historical weather data acquired conventionally from remote (usually airport) weather stations. The derived data is taken to represent a typical weather year for a city. However, due to rapid increase in urbanization, weather in city centers with high urban density is significantly different from rural areas, a large part of which is due to localized effects, e.g. urban heat islands, increased albedo of man-made surfaces and anthropogenic emissions. This thesis investigates the relative importance of spatial weather variability in predicted building performance simulation outcomes. Ranking the importance cannot be looked at in isolation but needs to be determined relative to all other sources of uncertainty, predominantly in the parameters of the energy model which in this thesis is EnergyPlus. The latter stem from lack of information or ignorance about many physical and scenario of use parameters. Together they are the ensemble of sources of uncertainties that need to be recognized in any simulation. A sensitivity analysis is conducted to reveal their relative ranking. An inspection of the resulting rank of the effect of spatial weather variability reveals whether the knowledge of local weather, in contrast to the assumption of uniform weather throughout the city, significantly reduces the overall uncertainty in the outcomes of the simulation. It should be recognized that there is only limited availability of localized weather data that reflect variability of urban contexts throughout a city. This recognition leads to the first contribution of this thesis: the development of a high fidelity statistical urban weather model fitted on local urban morphology and recorded weather. This is accomplished with a Multiple Tensor on Tensor (MTOT) regression model. The model can be applied universally and enables building modelers to create synthetic meso scale weather data for their site, essentially putting the individual building in the urban fabric of the city. The resulting model is a new cornerstone in the uncertainty analysis of the building simulation with inclusion of spatial weather variability. It is consequently used to inspect the role of spatially diverse weather in two critical applications. First, at the single building scale it is verified in three applications whether spatially diverse weather plays an important role when the assessment is conducted for a non-specific location in the city. Secondly, the role of spatial variability is tested in a three urban decision making cases where the question is answered whether decisions should be diversified per location. The thesis offers answers to both questions that elevate our understanding of the role of meso scale weather information in building simulation practice.