Facing infeasibility issue when upgrading to PyPSA 1.0.5

[nguyenha...@gmail.com]

Dear group,

We are currently testing PyPSA 1.0.5 before upgrading PyPSA-SPICE. When running our dummy data with the new version, we encounter an unexpected issue: adding `ramp_limit_up` and `ramp_limit_down` with a value of 1 for generators (e.g., solar or wind) results in an infeasible problem, with certain constraints not being satisfied (see below).

What puzzles us is that we also specify ramp limits for links (e.g., coal power plants), but those do not appear to cause any issues*. The violated constraints also don't point to any of the generators for which we added ramp limits. The same model solves without problems when using PyPSA versions older than 1.0.

Do you have any insight into what might be causing this or how we could resolve it?  Are there any changes to the ramp limit constraints in the new version that we should be aware of?

I look forward to hearing from you.

Best regards,

Long

*Update: Ramp limits on links are also causing infeasibility issues in some future years when running the rolling myopic with an almost identical log. However, when ramp limits are applied to both generators and links in a minimal PyPSA network, the model solves without issues, which suggests there may be a issue with our model setup.

[nguyenha...@gmail.com]

Sorry, the infeasibility log image was not properly attached in previous email. I share it again here:

[Iegor Riepin]

Dear Long,

PyPSA has a suite of unit tests for unit commitment problems, including various ramp_limit_up/down cases. It could be you've hit an untested edge case with your toy model, if an issue is on PyPSA's side. Could you please create an issue in PyPSA github repo and add a code snippet to reproduce your problem? If it is confirmed to be a bug, it will be fixed asap.

Kindly Iegor

[nguyenha...@gmail.com]

Dear iegor,

Our colleague, Saeed, has identified the problem/bug and wrote an issue + small PR on PyPSA github repo (https://github.com/PyPSA/PyPSA/issues/1493). Can you help review them?

Best regards,

Long