add mlfm

[steve-ransome]

• Closes Adding the Loss Factors Model and Mechanistic Performance Models to PVLIB #925
• I am familiar with the contributing guidelines
• Tests added
• Updates entries to docs/sphinx/source/api.rst 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 and Milestone are assigned to the Pull Request and linked Issue.

[markcampanelli]

Because there are several mflm-related tutorial notebooks and supporting files, we may want to add a general docs/tutorials/mflm/ directory (with a data sub-directory) to hold everything.

[steve-ransome]

We already do that. , there are directories for reference data, meas_gtw measurements+weather, figs and outputs

[markcampanelli]

We already do that. , there are directories for reference data, meas_gtw measurements+weather, figs and outputs

I may have not been clear. I was thinking in terms of how this structure might better grow and coexist with other multi-file tutorials.

[markcampanelli]

@steve-ransome In the example notebooks, don't use Windows-specific path separators, which don't work on Mac/Linux. For example, instead of
ref_file_name = (root_dir + '\\mlfm_data\\ref\\' + 'mlfm_reference_modules.csv'),
use
ref_file_name = os.path.join(root_dir, 'mlfm_data', 'ref', 'mlfm_reference_modules.csv')

[steve-ransome]

Thanks, have updated mlfm to use that

[markcampanelli]

@steve-ransome Would it be helpful at this point for me to review all the comments that I initiated and resolve/follow up with them?

[steve-ransome]

Thanks @markcampanelli that would be very useful. I think I've got as far as I can for the moment.

[markcampanelli]

@steve-ransome I walked through the comment history, and there seems to be some unresolved ones from all the reviewers (including me). Not sure if the maintainers are open to arranging a conference call to walk through all of these?

Personally I have few outstanding small issues, whereas my main issue is that the coefficient-solver information is not propagated back to the user, including a potential convergence failure.

In particular, I think the whatsnew info is still missing. This depends on the intended release version, and I wonder if the MLFM addition warrants a "new feature" version bump to v0.10.0, which would likely queue it behind the pending v0.9.4 release.

[adriesse]

Personally I have few outstanding small issues, whereas my main issue is that the coefficient-solver information is not propagated back to the user, including a potential convergence failure.

This could be a more general question about how we would like fitting functions in pvlib to behave. curve_fit will raise a warning or exception if things go terribly wrong. Is that suitable and/or sufficient? Admittedly the user will have to have a higher level of expertise to figure out what might be the problem. If this is not good enough, then maybe we shouldn't be using curve_fit but rather some lower level function like least_squares or minimize.

[kandersolar]

This PR is in a sorry state. Looks like some bad merges got in here somehow. If this ever goes forward, I think it would be easiest to just start a fresh PR, so I'm closing this one.

FWIW I also think this feature is complicated enough that it would need a more experienced pvlib contributor to champion it.

[cwhanse]

Been a while since I tried to help this one. My memory is that a primary (perhaps fatal) deficiency is a lack of clear references.