Issues reproducing PVsyst separation model results

[spelland74]

Hello,

I have just sent an email to support@pvsyst.com with multiple files. This concerns the separation model used by PVsyst to calculate diffuse (and direct) irradiance when GHI only is imported using the ascii meteorological data import tool.

Here is the situation:

-We are importing GHI only (because this was adapted to ground measurements) and importing using ascii import. PVsyst calculates diffuse and direct irradiance.

-I am checking whether the diffuse and direct calculated by PVsyst make sense. My understanding based on the PVsyst help was that the Liu and Jordan algorithm is being used.

-However, I calculated diffuse irradiance using that algorithm, and the values are very different from what is in PVsyst (I used the clearness index from PVsyst). In terms of annual total, the PVsyst DIF value is 27% higher than what I find using Liu and Jordan.

-The equations I used are fits to Figure 7 in the original (1960) Liu and Jordan paper taken from a 2015 article by Christian Gueymard (see reference below). I checked that the equations indeed match Figure 7 extremely well. Here are the equations (Kt is clearness index):

If Kt<=0.75

(DIF/GHI)=1+0.006381*Kt-3.2315*Kt^2+2.2448*Kt^3+0.081882*Kt^4

else

(DIF/GHI)=0.16

I was wondering if you could please help us in understanding what’s going on. Ideally, if you can give more details about the equations you were using, that would be much appreciated.

Best regards,

Sophie

Gueymard 2015: https://www.researchgate.net/publication/283570186_Extensive_worldwide_validation_and_climate_sensitivity_analysis_of_direct_irradiance_predictions_from_1-min_global_irradiance

[spelland74]

Simply writing here to indicate I received a reply by email, and PVsyst uses the Erbs model to estimate diffuse from global horizontal (not the Liu and Jordan model, which is what was mentioned in the online help that I consulted). When I use the Erbs model, I can reproduce the PVsyst diffuse values properly.