IAM change in PVsyst versions - unexpected result

[Grae]

Hi,

We have a simulation in Australia, where we found a change in IAM values between PVsyst version 6.42 and version 6.52, which we investigated. We ensured that the .pan file and IAM was the same between the two variants along with the meteo file to ensure a like for like test.

We found whilst analysing the results that although the main GII on the PVsyst report was the same, the GII value in the 'Annual Yield', 'Detailed study of IAM (Incidence Angle Modifier)', this was an unexpected outcome.

We see in the release notes that some work has been done on the IAM coding in 6.43, ‘Fixed errors’ point 2. Please can you confirm if this is the reason why we are seeing the changes.

Many thanks,

Grae

[André Mermoud]

From version 6.40 to 6.42, the IAM profile was indeed computed with a cubic spline between point#2 and point#3, which may lead to an artefact curve depending on how the points are aligned (if Point#1 and Point#2 are 1). This could give a discrepancy up to 0.3% to 0.4% in some cases. Now all the points interpolations up to the first <1 are computed with a linear interpolation.

[Jen H]

Hi André,

Thanks for this response, that's useful to know. Could we please clarify what is meant by "all the points interpolations up to the first <1" just so we are clear between which points the linear interpolation applies?

Thank you!

[Soldnerkugel]

From version 6.40 to 6.42, the IAM profile was indeed computed with a cubic spline between point#2 and point#3, which may lead to an artefact curve depending on how the points are aligned (if Point#1 and Point#2 are 1). This could give a discrepancy up to 0.3% to 0.4% in some cases. Now all the points interpolations up to the first <1 are computed with a linear interpolation.

Do I get you right: this discrepancy only appears when using a "customised" IAM parameterisation (either manually entered in PVsyst or as information in a .pan-file). As a standard the ASHRAE parameterisation is used, but it is not calculated by using a cubic spline.

In older versions of PVsyst a cubic spline was calculated using the nodal points either manually entered or from the pan-file. So if two (or more) consecutive points were of value "1", the calculation of the cubic spline returned false values (because cubic splines are defined only in sections using the delimiting nodal points and between two points of the same value you cannot define a curve but only a straight). But - as I already mentioned - this only appears using a "customised" IAM parameterisation.