Remove unecessary plants

[staadecker]

This PR adds two post-process steps. One of them will remove all emitting plants from the model if there's a zero-emissions constraint. The second will remove all plants that get retired before the first period. In a recent zero-emissions scenario, these post-processing steps reduced the size of my input folder by 21% (307M to 243M) and reduced solve time by 29% (94 min to 67 min). In another run that failed to find a solution within 29+ hours, this post-process step allowed a solution to be found in 7-8 hours.

This pull request solves part of the issues in #120.

• Compare outputs of runs. See comment below.

[staadecker]

Update: Seems like the results were slightly different, I am currently debugging as to why this would be

[staadecker]

Update: I have confirmed on a California-only model that both get_inputs post-process steps don't affect the objective function result. Although dispatch is slightly different, I believe this is because there are multiple optimal solutions since both the total costs are exactly the same to the cent. Further, I've found this post-process step to greatly speed up solve time for more complex models. In a 365 day, full-WECC model with only PV aggregation, I found the model to be infeasible without this step (29 hours and still running) but take only 7 hours with this step.

[pesap]

Hey @staadecker,

I am actually curious why the model for WECC is infeasible. From the runs I have made so far, we found feasible solutions with and w/o aggregated generators. Are you changing something else in the scenario?

[staadecker]

Were your scenarios 365 days with 4-hour timesteps? I remember the model being infeasible during the summer which was the reason we decided to do some aggregation. I don't think I've changed anything else, I can attach the input config.yaml if you want to see for yourself.

[staadecker]

@PatyHidalgo could you review this, I would like to use it in my 1h sample tests since performance is a big concern