The BRE-IDMP dataset: a new benchmark for the validation of illuminance prediction techniques

Scale models are generally believed to be a reliable tool for illumination modelling and are often used to predict daylight factors. Recent work however has revealed that scale models generally over-predict illuminance by a significant margin for overcast skies. For nonovercast skies the divergence between model and real building performance is greater still. Advances in lighting simulation and physical modelling allow for the possibility of a daylighting evaluation that is based on a range of nonovercast sky luminance patterns including sun. Using either technique it is now a practical possibility to predict hourly internal illuminance levels for a full year under realistic sky and sun conditions. Illuminance predictions for these conditions need to be validated using the best possible data. The uncertainties associated with scale modelling suggest that the technique is insufficiently reliable for validation of predictions under non-overcast skies, and that a new benchmark is needed. The BRE-IDMP validation dataset contains simultaneous measurements of sky luminance patterns, solar illuminance and internal illuminance in two full-size mock offices. With this dataset it is possible to specify to an unprecedented degree of precision the conditions at the time of measurement. The dataset, believed to be the only one of its type in existence, is complex and does contain occurrences of potentially unreliable entries. These had to be identified so that a true assessment of the accuracy of the illuminance predictions can be made. This paper describes how this was achieved. Predictions from lighting simulation are validated using the dataset but the findings are also applicable to the validation of physical modelling approaches using the new generation of sky simulators.