Fixing the ACOPF rectangular formulation and updating datasets

[hhijazi]

Description

This PR fixes #123 and corrects the rectangular ACOPF formulation as follows:

• Formulate the problem without auxiliary variables to avoid feasibility violations.
• Make sure to add voltage magnitude and phase angle bound constraints in the original space (e,f).
• Make sure to remove buses that are disconnected (status==4).
• Make sure to use the correct initial point (setting evar to 1).
• Compute the voltage angles and magnitudes correctly when using the rectangular formulation.

Warning

The pstart initialization does not work with Gurobi 11, only with Gurobi 12. Therefore the performance on ACOPF will be bad until Gurobi 12. For instance, Gurobi 11.0.2 cannot find a feasible solution on network IEEE69.mat (from #123) after 600 seconds.

[simonbowly]

I added some comments. Happy to help with the changes, just let me know. The main issue is that the tests do not pass. Have you run them locally?

[hhijazi]

@simonbowly I finished cleaning up the tests and I added the AC-LOCAL option.
Using the python file from #123 and replacing the command below:

result = opf.solve_opf(case, opftype="AC-LOCAL")

we get:

Objective value = 82.31523524020653.
Solution quality statistics for model 'AC_Formulation_Model' :
  Maximum violation:
    Bound       : 5.50004486e-13 (s_1_1_2)
    Constraint  : 2.27396420e-09 (Qdef_1_1_2)
    Integrality : 0.00000000e+00
sum(abs(sinfeas)) = 4.98547332045843e-09.

Not sure why the doc-tests are failing though.

[simonbowly]

@hhijazi I added a lot of comments.

My main concern is that so many of the tests are deleted or overwritten. It makes it impossible to tell whether any of the old cases would still work. The new version should be able to work with data that the old version handled.

[simonbowly]

@hhijazi when going through these comments (and the old ones) can you please mark them as resolved (if they are resolved).

Also, I don't think the description or title of this PR is accurate any more. This isn't just a fix. Can you update these to better describe the new functionality?

[simonbowly]

@hhijazi this looks good, thanks. I added some tests for the "aclocal" mode. It's expected that these do not find solutions with Gurobi 10 or 11, right?

What should we do there? Error out if a user chooses aclocal with gurobi<12? Or take our chances but print a better warning message if/when it fails?

[simonbowly]

What should we do there? Error out if a user chooses aclocal with gurobi<12? Or take our chances but print a better warning message if/when it fails?

I'll merge this and figure it out afterwards.