Update 2015: This research is now available as a paper in Solar Energy journal!
Citation: Engerer, Nicholas A and Franklin P. Mills. Validating Nine Clear Sky Radiation Models in Australia. Solar Energy. 120, October 2015, pp. 9-24.
Have you ever wondered which clear-sky solar radiation model you should use in your research project or solar energy simulation? When I was designing my KPV method for estimating PV system power output I needed to figure out which clear-sky model would be the best one to use. But there was a problem - I couldn't find a single validation study for clear-sky radiation modelling for Australia!
So in the paper, I had to do a quick model validation using one year of radiation data from Wagga Wagga, from which I decided on the Esra model. But that simple validation left me wondering which model REALLY was the best?
Thus I embarked on a scientific journey to discover which model was the best for use in Australia using the solar radiation data from 14 sites in Australia:
First, I set a few ground rules. I wasn't going to use any radiation model that was overly complicated, nor was I going to use atmospheric variables that were difficult to obtain.
This meant using climatological values for input values such as the Linke Turbidity coefficient or ozone content - rather than using direct measurements from a photometer (because who honestly has spectral data?). I think this is important, because a validation study should focus on models that are widely applicable so that it is widely useful.
In the end, I chose nine models from the options from both beam (direct) and global radiation:
In the many Australian presentations and publications I've read and attended over the past few years, the most common clear-sky radiation model used is the Ineichen-Perez model.
However, my research shows that the Ineichen-Perez model is not even in the top-three best choices. So we really shouldn't be using it in our research as it is introducing unnecessary errors into our collective research knowledge.
What models are the best to use? For the beam models, the top three choices are the Iqbal-C, Esra and REST2 models. And for the global models, they are the Solis, Esra and REST2 models.
If we were to chose the best overall model, that would be the Esra model. Which edges out the REST2 model, due to is large errors at high zenith angles.
Until then, the quick answer is...