Building a comprehensive dataset for the validation of daylight simulation software, using complex "real architecture"
This research focused on building a comprehensive dataset for use in validation studies of daylight simulation software. The aim of the set is to add to existing validation data to better cover a wide range of complexities and weather conditions. This will allow for not only the validation of simulation software, but the comparison of multiple simulators in their general strengths and weaknesses as well as feasibility for early ‘sketch’ design stages and complete building simulations. The set can also aid in the creation of valid simulation parameter starting points for designers. The research examined the current ‘gold standard’ validation dataset from the BRE-IDMP, and found that while it provides excellent validation opportunities for simulators that can support its detailed patch-based sky model; an equally high quality dataset is needed for simulators that support more simplified skies. This is essential as most of the weather data for annual daylighting simulations available to designers, such as the US-DOE’s collection of TMY data, can only be used in mathematical sky models such as the Perez all-weather model. It is also essential that real world, complex light-path scenarios commonly found in buildings be addressed by validation in addition to the simple single room, single opening tests which are prevalent in the daylight simulation field. A dataset suite is proposed, similar to the BESTEST suite for energy simulation, which covers basic analytical test cases for lighting simulators, simple office scenarios and a complex shaded classroom in a tropical climate. The dataset is valuable for the testing of daylight simulators which make use of the common CIE general or Perez all-weather skies. These datasets were used in a trial validation of Autodesk’s 3ds Max Design and Radiance, which included significant sensitivity testing of the two empirical datasets included in the suite. This demonstrated the usefulness of each dataset, and any issues with their data. It also highlighted the key inputs of any simulation model where designers must take significant care.