Best practices for modeling cell temperature with bifacial modules

[Will Hobbs]

I have a few questions about best* practices for modeling cell temperature with bifacial systems. Assuming I wanted to use the Faiman (or possibly PVsyst, faimain_rad) temperature model:

1. Should modeled (or measured) rear POA irradiance be included in the effective irradiance that goes into the temperature model?
2. If the answer to #1 is "yes", should it be after bifaciality (e.g., 0.8) is applied to rear POA?
3. If the answer to #1 is "yes", how might default values of coefficients (U1, U0, etc.) be changed?

On #1, rear irradiance obviously matter for cell temperature, but it also matters (some) for monofacial modules, and I don't think it's common to calculate rear irradiance for monofacial and then apply some sort of backsheet absorptivity factor (e.g., 0.1 or something small but non-zero) to it. More importantly, I don't think Faiman or PVsyst models originally considered this.

On #2, bifaciality typically refers to electrical performance, not absorptivity, so I'm not sure it's the right value to apply. As an extreme example, First Solar Series 6 Plus Bificial has a bifacility of 0.15, but I don't think that means that it is 85% reflective. 

On #3, I think this is the most important question. I think the answer to #1 is "yes" and the answer to 2 is "yes, sort of, it's better than nothing, maybe just use 0.8 or 0.9 no matter what the spec sheet says". Then the key thing is what to do about coefficients if you don't have time or data to fit your own. Re-using the same defaults could bias any monofacial-vs-bifacial modeling exercises. 

This study [1] seems relevant. They fit coefficients, and it looks to me like scaling U0 and U1 up by about 5-10% might be a good approach. But how much the coefficients increase goes up with albedo, so it gets a bit circular: the more rear POA you have added to effective irradiance, the more you reduce the impact of irradiance in the temperature model. 

FWIW, this pvlib gallery example 1) includes rear POA in effective, 2) scales it by a bifaciality factor, and 3) uses default coefficients: https://pvlib-python.readthedocs.io/en/latest/gallery/bifacial/plot_bifi_model_pvwatts.html.

Last, the reason I ask is that I'm trying to figure out what to do in my pvlib-based modeling function. Related GitHub issue is here: https://github.com/williamhobbs/pv-system-model/issues/31.

Thanks!

Will

*"best" could mean many things here: "common", "worth-trying", "probably-not-terrible", etc...

[1] https://doi.org/10.1002/pip.3892 (preprint https://doi.org/10.22541/au.172115390.06323163/v1)

[cwh...@sandia.gov]

My first reactions:

1. Yes.

2. No. Bifaciality is usually measured as a ratio of short circuit currents. That's not equivalent to a ratio of absorbed heat. But, see comment below.

3. My first thought was, I don't see why the coefficients would change. U0 represents radiative and zero-convection, conduction heat transfer; that should already account for heat expelled from both surfaces. But then looking at the reference, it occurs to me that the change in coefficient value is an artifact of fitting the equation to measurements **using different predictors**.

They fit e.g. the Faiman equation to the same module temperature Tf, ambient temperature Ta and two different irradiances: G (front only), and G+ (front + bifaciality * rear).

The difference Tf - Ta is what is predicted by irradiance / (U0 + Uv * WS). If irradiance increases, then U0 has to also increase, because the temperature difference doesn't change.

So, my second reactions:

1. Yes.

2. Your preference, but

3. You must use coefficients fitted to data that reflect #2.

Cliff

[Will Hobbs]

Thanks, Cliff. Do you (or anyone else) have thoughts on obtaining reasonable default coefficients that do include rear POA? Either with or without a bifaciality adjustment?

And is it fair to assume that monofacial module temperature models would benefit from including rear POA? (in both calculation and the fitting of new coefficients...). This absolutely makes sense to me, but I'm not sure I've seen it proposed before. But now that we commonly calculate rear irradiance, maybe we should always calculate and use it.

--
You received this message because you are subscribed to the Google Groups "pvlib-python" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pvlib-python...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/pvlib-python/ee647e39-3ff2-409e-a8bd-a41c1200cd93n%40googlegroups.com.

[Mark Campanelli]

Hi Will,

  I have floated around the idea of using a reference module with a calibrated single-diode model (that is sufficiently accurate for the technology) in order to infer the effective cell-junction temperature (and irradiance) based on I-V curve measurements (or even just two I-V point such as Isc and Voc, or perhaps Pmp and Voc). I say "effective" here, because the typical SDM makes homogeneity assumptions about the cells in the module.

  Why bother doing this? Because it could be an effective way to calibrate/validate the upstream models that ultimately predict cell-junction temperature (not back of module temperature, which is perhaps even more challenging with bifacials).

  I think I may have already shared with you the draft paper that I wrote on this technique. My biggest blocker has been finding a data-provider partner that can handle and deploy the modules in order to validate the technique's merits and work out a good measurement protocol. There is also the cost of the IEC 61873-1 module calibrations, which should probably be done pre- and post-experiment due to degradation effects that could confound the results.

As always, happy to discuss further,

Mark

[Will Hobbs]

Mark,

Have you talked with Atonometrics (now Kipp & Zonen) about this? It seems like something they might be interested in. They could probably address part of this.

Will

To view this discussion visit https://groups.google.com/d/msgid/pvlib-python/2281dfe2-5234-4614-aff8-12b0cb5e9415n%40googlegroups.com.

[cwh...@sandia.gov]

I don't think its a matter of including or excluding rear POA. The rear POA is the same on both a monofacial and bifacial module. The primary difference affecting cell temperature is the thermal mass (glass-backsheet vs. glass-glass). A secondary difference is the removal of energy from rear POA as electricity in a bifacial module, which would have been retained as heat in a monofacial module. But this difference is small: (front + bifaciality x rear POA) x efficiency, vs front x efficiency.

I'm of the opinion that U0/U1 coefficients should be similar for modules with similar thermal mass, and in similar mounting, without regard to bifaciality. If coefficients are found, then the input irradiance needs to be what was used to fit the model (most likely, front only.)

Cliff

On Friday, July 31, 2026 at 10:15:33 AM UTC-6 will....@gmail.com wrote: