Large horizontal openings and tall spaces in calculated natural ventilation
Posted by Andy Tindale on 21-Jan-10 06:40 PM

There are in principle two possible approaches when modelling tall spaces such as atria using DesignBuilder with calculated natural ventilation:

1. Model the atrium as a single tall space or,

2. Model the atrium as a series of horizontal slices with air flow connections between them.

Option 1, where the atrium is modelled as a single space is the best option because:

a) There is no reliable way to get solar radiation pass through multiple horizontal slices in the general case.

b) EnergyPlus allows you to set up a dynamically varying temperature gradients to improve the accuracy of simulations in tall spaces. This can be accessed in DesignBuilder using the Temperature distribution data on the HVAC tab. Be warned though that, in current versions, the EnergyPlus temperature distribution model is not applied to airflow which occurs in the airflow network simulation so it is not possible for example to include the effect of relatively warmer air being extracted from a vent at the top of an atrium. In the simulation air would be extracted at the mean temperature. For more information on air temperature distribution in DesignBuilder/EnergyPlus see:

So at the moment we have to live with some inaccuracy in EnergyPlus models due to temperature gradients in tall spaces. Otherwise there is no problem. DesignBuilder also includes a CFD module which calculates spatially distributed temperatures within zones as a 'snapshot' but we are some way away from dynamic CFD simulations.

In summary, there are inaccuracies involved with large horizontal openings in airflow network simulations (a and b above), but these may not be 'show stoppers', especially if the openings are small.  It is always worth checking that hourly temperature results in the various zones look believable. Considering all the other inaccuracies involved in building energy simulations you may find that the approximation involved with using algorithms intended for LVO to model LHO may not cause serious problems.