Diffuse increases on transposition

[S.Faulkner]

Revised question

I understand that for transposition the diffuse irradiance is split into circumsolar and isotropic (and horizon) components and the circumsolar is treated like the direct irradiance in the transposition.

I had previously assumed that the circumsolar component was actually subtracted from the diffuse irradiance and added to the direct irradiance, and therefore treated like direct irradiance for all the subsequent calculations.

I gather from the data* that this isn't done, and that PVsyst therefore treats the circumsolar component as diffuse for all subsequent calculations (reflection and shading being affected in particular).

Is that correct, and is there any particular reason for not treating circumsolar the same as direct in subsequent calculations?

* In PVsyst the diffuse irradiance generally increases from horizontal to plane of array. The isotropic component decreases as a result of some being behind the array, and the circumsolar component is (generally) increased in the same way as the direct irradiance. The circumsolar component must therefore be still included in the diffuse irradiance.

Diffuse on POA vs Diffuse on horizontal, Perez, colour-coded by Kn (red is high, Blue is low)

[André Mermoud]

You are right. For all diffuse calculations (namely integrals for shading or IAM factor), PVsyst assumes the strong hypothesis that the diffuse is isotropic.

And the circumsolar, included in the transposed diffuse value, is treated as isotropic in the same way.

However these are approximations. A more realistic calculation associating the circumsolar component to the beam component for these treatments would seriously complexify the simulation.

Moreover, this component is rather well defined with the Hay model (but not necessarily realistic), but not evident to extract when using the Perez model. Its exact determination is extremeny dependent on the Linke coefficient (especially the aerosols).

This approximation may be justified by the fact that the circumsolar contribution is relatively low: in your diagram, it would correspond to the difference between your red and blue points, and for high Kt only.

The evaluation of induced errors could be a nice subject for an academic research project.