Uncertainties in airflow network modelling to support natural ventilation early stage design

Despite a lot of Integrated Design Process guidelines and procedures have been developed in the last few years, more specific energy design procedures are needed to push the implementation of passive design techniques.
As natural ventilation design influences strongly the building shape and aspect, it has to be considered since the early design stages and its effect should be correctly predicted and proved by means of suitable tools and methods. In this respect, airflownetwork models seems a promising modelling techniques as they are already integrated in the existing dynamic building simulation tools and they have a quick solver. On the other hand, the simplicity of these models implies uncertainties in a lot of input parameters, first of all the wind conditions. The urban wind environment is a stochastic phenomenon, and as consequence the ventilation performance in a naturally ventilated building changes. An accurate wind analysis should be supported by weather data collected on site and by an external CFD simulation at urban scale. Discharge coefficients and external convection coefficients could be accurately estimated by laboratory tests. This would mean additional design costs (and time) and need of expertise in the field. A key issue of this work is to assess the thermal-airflow model reliability in airflow prediction when accurate estimation of input data is not feasible.
This paper presents an uncertatinty analysis performed on a dynamic simulation model of a new office building naturally ventilated during the night.
Full-factorial parametric analysis have been performed to assess the influence of the input parameters on the model reliability. Possible input ranges have been estimated and organized in a tree-structure to investigate the effect of dependent parameters like wind velocity profile, wind pressure coefficients, discharge coefficients, and external convection coefficients.
Significant variations in air change rates are shown that reflect an uncertainty of +/- 2% on total cooling loads in respect to the base case simulation result. No direct correlations between outdoor environmental conditions and air change rates have been found as the discharge coefficients affect significantly the results.
The design proposal supported by quantitative analysis and results prediction uncertainty assessment can be taken into consideration more seriously by the design team. The obtained results can be generally extended to airflownetworks with similar airflow paths.