Allow for fixing investments

[jokochems]

Closes #1005

• Introduces an attribute fix_investments to class Model
• Iterates over investment flows resp. variables and fixes them.
• Raises an error in case model has not yet been solved.

[pep8speaks]

Hello @jokochems! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2023-12-08 16:40:39 UTC

[jokochems]

This is a draft with yet untested functionality, just to get a first impression. I need to add tests as well as to touch the GenericInvestmentStorageBlock as well as the SinkDSM investment blocks.

[p-snft]

I do see your personal need for this adjustment. However, I fear that this is introducing code dept for a very limited use case. There are several use cases that require a very similar functionality but would not be possible with this one:

• Optimise investment as well as operation including demand rate for electricity. Then fixing the investments but leaving the peak demand as a variable. This would not work as optimising the peak load is not distinguished from investments.
• Optimise day ahead energy trading based on a forecast. Then switch to slightly different energy demands and optimise intra-day trading.
• Run a multi-period investment model, then fix the sizes just of the technologies that have a long planning phase (but not e.g. small scale PV installations) to emulate multiple medium-term scenarios that an still work out once an initial decision is taken.

Unfortunately, I am also lacking the time to implement this in a more generic way. Thus, I propose to leave this open as a reference or until someone find the time to do it.

[jokochems]

I do see your personal need for this adjustment. However, I fear that this is introducing code dept for a very limited use case. There are several use cases that require a very similar functionality but would not be possible with this one:

• Optimise investment as well as operation including demand rate for electricity. Then fixing the investments but leaving the peak demand as a variable. This would not work as optimising the peak load is not distinguished from investments.
• Optimise day ahead energy trading based on a forecast. Then switch to slightly different energy demands and optimise intra-day trading.
• Run a multi-period investment model, then fix the sizes just of the technologies that have a long planning phase (but not e.g. small scale PV installations) to emulate multiple medium-term scenarios that an still work out once an initial decision is taken.

Unfortunately, I am also lacking the time to implement this in a more generic way. Thus, I propose to leave this open as a reference or until someone find the time to do it.

Hi, thanks for the response. I totally understand it and find it a good suggestion to simply leave it open. (Btw: I'm still facing some numerical issues which lead the model to be solved again to be infeasible. But maybe I just have to choose an even higher rounding precision for my particular use case.)

You did not request any particular code changes, did you? This was more on the general need for a revision and it to be more generic.

What would be a cool thing would be to provide a list of variables you want to fix for another run / units you want to exclude from fixing. That might also be dispatch-related ones and allow you to exclude dedicated investment variables as for the example you provided above.

[p-snft]

Yes, I did not come up with specific requests.

What would be a cool thing would be to provide a list of variables you want to fix for another run / units you want to exclude from fixing. That might also be dispatch-related ones and allow you to exclude dedicated investment variables as for the example you provided above.

If this was implemented, the approach should be general enough to be merged, I guess.

[jokochems]

I currently do not have the capacities to tackle this Pull Request. Also, I faced numerical issues and infeasibilities when I wanted to test it in a large scale model. Thus, I'll close this pull request and the dedicated issue.

@oemof-developer Feel free to reopen in case someone has a use case or capacities to retackle it.