Developing a surrogate model for naturally ventilated cellular offices in Brazil

Demand for artificial cooling in buildings is increasing worldwide, and it is expected to continue to grow in the upcoming decades. To mitigate energy use, the adoption of passive cooling strategies such as natural ventilation is a solution. In this study, a metamodel is developed to estimate thermal performance in naturally ventilated cellular offices in Brazil. Metamodels help to overcome the complexity of building simulation, although some caveats should be considered when developing such models. Simulation parameters were based on a database of cellular office buildings from São Paulo. The output of the simulations was the fraction of hours with operative temperatures above ASHRAE’s Standard 55 adaptive model. Two different modeling approaches were analyzed, a single-zone and a multi-zone. Sensitivity analyses were conducted to identify what parameters are the most influential on simulation results, and what parameters compromise the accuracy of the single-zone approach. Window opening factor, walls’ transmittance and the condition of exposure of walls and windows were the most influential parameters on the simulations. Although the single-zone approach is simpler, it does not consider heat transfer between different offices adequately. From the two modeling approaches, two metamodels were developed. The single-zone metamodel had a more accurate performance when validated on single-zone simulations. However, the errors related to this modeling approach compromises the performance when compared to multi-zone simulations. It is concluded that it is important to understand the limitations and accuracy of a surrogate model before applying it, and that such tool could be helpful for building designers.