Use scipy.constants

[cwhanse]

• Closes Add a public constants module and include scipy as base requirement for pvlib #483
• I am familiar with the contributing guidelines
• 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.

In review of #1573 @markcampanelli pointed out that the precisely calculated solutions to the single diode equation should use the latest values of physical constants (Boltzmann and elementary charge). In 2019, the various committees agreed that the published values of these two constants were exact, without uncertainty. (I link to the U.S. agency NIST, other national agencies also published the same values).

With this agreement I see no reason for pvlib to use approximate values anywhere. Scipy 1.4.0 published the agreed-upon exact values in scipy.constants. Advancing to scipy 1.4.0 does not require pvlib to also advance the numpy version.

[kandersolar]

The new minimum scipy must also be specified in setup.py's INSTALL_REQUIRES.

I didn't do an exhaustive search, but at least pvlib.temperature.fuentes has a hardcoded boltz variable that could be updated here as well. One reason to keep the current value in that case is to match the original model implementation from which ours was derived, but I wouldn't object to using scipy's value.

[cwhanse]

I searched but am not confident, because I didn't find that instance in temperature.py. I'm inclined to leave that one as is, with a comment.

[kandersolar]

One more place to update the minimum scipy version: /benchmarks/asv.conf.json.

[wholmgren]

Should we note the minimum version change in the release notes? I can go either way.

[kandersolar]

Should we note the minimum version change in the release notes? I can go either way.

I vote yes and am a fan of the format we've used previously: Minimum pandas version increased to v0.25.0, released July 18, 2019. (#1448)

[markcampanelli]

Thanks everyone for seeing this through, including chasing down all the related “homework” items.