Linear shade loss model - continuation of pr 1725

[echedey-ls]

• Closes Add linear shade loss model for thin film #1690
• Substitutes Linear shade gh1690 #1725 (originally from @mikofski )
• I am familiar with the contributing guidelines
• Tests added
• Updates entries in docs/sphinx/source/reference for API changes.
• Adds description and name entries in the appropriate "what's new" file in docs/sphinx/source/whatsnew for all changes. Includes link to the GitHub Issue with :issue:`num` or this Pull Request with :pull:`num`. Includes contributor name and/or GitHub username (link with :ghuser:`user`).
• New code is fully documented. Includes numpydoc compliant docstrings, examples, and comments where necessary.
• Pull request is nearly complete and ready for detailed review.
• Maintainer: Appropriate GitHub Labels (including remote-data) and Milestone are assigned to the Pull Request and linked Issue.

Addresses the linear thin film module losses.

Comment from @kandersolar at #1962:

you are saying that we need a substantive reference that says linear power loss models are valid for thin film modules. Is that right?

What I'd really like to reference is a paper that deals with specifically the equation implemented in this PR: power_loss_fraction = shaded_fraction1d * (1 - diffuse_fraction).

@mikofski do you have a reference for that or should I change to irradiance losses and a SAM reference?

Wait for #1962 to be merged before this one so the docstring makes sense.

[mikofski]

I found this by googling:
https://energy.sandia.gov/wp-content/gallery/uploads/26-Littmann-Tracking-and-Diffuse-Shading.pdf0_.pdf

It’s for series 4 but explains why linear shade for orientation perpendicular to cells

[echedey-ls]

I found this by googling:
https://energy.sandia.gov/wp-content/gallery/uploads/26-Littmann-Tracking-and-Diffuse-Shading.pdf0_.pdf
It’s for series 4 but explains why linear shade for orientation perpendicular to cells

I don't really get to see where it explains the shade is perpendicular to the cells. I mean, I see the statement but not the reasoning.

I would prefer some other reference, although I'll stick to that one. I will update the PR by the end of the week.

[mikofski]

Maybe if you combine this reference with the FirstSolar series 4 user guide. The point is the cells are long and thin and strung in series from side to side. The user guide explains that they should only be oriented so that the cells extend from bottom to top, not left to right or damage will occur. There is also a series 6/7 user guide.

[echedey-ls]

Thanks for that insightful manual. I like how the docstring is turning out!

These references don't have the model equation nor address what the model modifies (power yield or irradiance). I think it doesn't matter as long as we find some other source that deals with the model eventually. Maybe some code snippet on the public domain under First Solar authorship?

[echedey-ls]

I have the feeling that the reference still is an issue. If you maintainers think so, do you think the same reference as #2063 and #2070 , equation (3), is a sustainable reference?

https://www.sciencedirect.com/science/article/abs/pii/S092702481000454X

Pro: it compares the results in Table 2 and Fig. 4.
Cons: it is compared for non-monolithic modules

EDIT: content

[mikofski]

Hi @echedey-ls that reference isn’t really valid for firstsolar modules because firstsolar don’t use bypass diodes and the power loss isn’t exactly the same as the fraction of direct beam shaded, because there is still diffuse sky and ground. I think the best references are Kevin’s & Adam’s paper and the poster from PVPMC 2023 (https://pvpmc.sandia.gov/download/5381/?tmstv=1719027091). Saying the loss is linear is just industry jargon four saying there is no electrical mismatch due to shade, so the only loss is the linear decrease in direct beam.

[kandersolar]

I have the feeling that the reference still is an issue.

For me, yes. I don't think we have a reference that clearly describes the specific model implemented here. I spent some time previously trying to locate a better reference, but was unsuccessful.

Without a good reference, I suggest we close this PR. It's easy enough to pick it back up again if/when a good reference becomes available in the future.

[mikofski]

I disagree. The first solar series 4, 6 & 7 manuals suffice. I will ask first Solar if they have something more definitive.

Is the question:

1. What is the orientation of cells in a first solar module and how are they strung?
2. Is it, do first solar modules have bypass diodes?
3. Or is it, how the current of the cell is affected by light as it is progressively shaded from zero to 100%?

There are references available for all of these questions already. Please do not close this PR. My preference would be to figure out how to complete it or merge as is. I don’t understand the blocker, but I do know that I am seeing more first solar modules on terrain and it will likely continue due to concerns over domestic manufacturing and tariffs

[mikofski]

what the model modifies (power yield or irradiance)

The equations affect direct beam irradiance only, so add whatever diffuse light before evaluating effect on power.

The reference for the equation is the poster from PVPMC 2023

[echedey-ls]

Hey @mikofski , I plan on finishing development on all my PRs to pvlib for now unless straight interest is shown in certain proposals and I find the time to work on them. I will be closing this for these reasons, but feel free to copy this branch and continue development if you wish.

Now that I better understand pvlib and the PV context, my two cents to consider that poster not enough to include it are the following:

• Whether the losses factor is applied to power or irradiance:

  The equations affect direct beam irradiance only

  but the poster states

  (r2r) shade affects all cells uniformly and reduces
  output linearly proportional to the fraction of shade

  I understand output is power output, since irradiance is not an output.

  So if the poster was to be used as a reference, I think that a matching statement between the implementation and it is essential.

• For either possibility, a reference equation would be nice, at least to double check the implementation and document all the inputs.

• Probably pvlib.shading should conform to the same API in terms of shading losses factor: irradiance or power losses, and current code (edit: at main) matches the latter. So the equation is not as direct as the linear loss presented here deduced from the text (edit: in terms of a irradiance loss).

• I would love to see a comparison of different models, à la https://www.sciencedirect.com/science/article/abs/pii/S092702481000454X , fig 4:

  
  Thou I understand the underlying effect and it seems intuitive, I wonder if there are any non-linearities that may affect the power yield (more or less question n. 3), and even if there aren't, just a validation of the model.
  I expect references to require this validation stage, just to know how much difference I can assign to each modelling step.
  [as a non-expert on what your expertise is and how DNV works, would it be very difficult to obtain real data and model the losses from the shaded fraction?, maybe publish at arxiv or similar]

Again, feel free to pick up this work and continue from there - I can support you with reviews; and thanks for showing interest in improving these capabilities of pvlib, I learnt a lot through this journey.