Add Batzelis 2017 simple nonlinear PV model

[kandersolar]

• Closes Datasheet-only nonlinear model for Pmp/Imp/Vmp/Isc/Voc #2562
• 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.

[kandersolar]

This PR currently adds three new functions. As they are related to each other I thought it would be helpful to review them together, but I can split the PR into smaller pieces to make it more digestible if needed.

To summarize the three functions:

• pvlib.ivtools.sdm.fit_desoto_batzelis: estimate De Soto SDM parameters from datasheet values
• pvlib.singlediode.batzelis_keypoints: estimate Pmp/Imp/Vmp/Isc/Voc from SDE parameters
• pvlib.pvarray.batzelis: estimate Pmp/Imp/Vmp/Isc/Voc from datasheet values

The third, effectively, combines the first two with calcparams_desoto, but with some algebraic manipulations to make the computation more compact. I think there is also a minor difference in the calcparams_desoto calculations. All three come from the paper linked in #2562.

Questions for reviewers:

• Currently alpha_sc has units [1/K] in some places and [A/K] in others. Better to diverge from the reference (which uses [1/K]) and use [A/K] everywhere for consistency with the rest of pvlib?
• Parameter names likely also should be changed for consistency. i_sc instead of isc0, etc?
• Are the current modules the best places for these new functions?

[cwhanse]

Some reactions.

[cwhanse]

A question about the method: does batzelis produce the same results as fit_desoto_batzelis + batzelis_keypoints? It is not clear to me that should be the case.

[kandersolar]

A question about the method: does batzelis produce the same results as fit_desoto_batzelis + batzelis_keypoints? It is not clear to me that should be the case.

Close but not exactly the same, in my testing. The latter has to be fit_desoto_batzelis + calcparams_desoto + batzelis_keypoints, and I think there is at least a difference in the calcparams_desoto step (regarding dEgdT). I think there are differences elsewhere too, but I can't point to any specifics.

[adriesse]

* Currently `alpha_sc` has units [1/K] in some places and [A/K] in others.  Better to diverge from the reference (which uses [1/K]) and use [A/K] everywhere for consistency with the rest of pvlib?

I think this is the right decision for pvlib because of precedent, although personally I much prefer the normalized values.

* Parameter names likely also should be changed for consistency.  `i_sc` instead of `isc0`, etc?

+1

* Are the current modules the best places for these new functions?

+1

[markcampanelli]

@kandersolar Was adding this SDE solver to the method options in pvsystem.singlediode discussed?

[cwhanse]

I don't think we discussed it. I'm not in favor of adding it as a method option to pvsystem.singlediode because singlediode.batzelis solves an approximation of the single diode equation, rather than the SDE itself.