WIP: Flexibile Loads

[DEUCE1957]

Attempt to extend Redispatching by enabling Environment's to have flexible loads. See "getting_started" / "12_Flexibility.ipynb" for example use. Related to issue #623 (will move onto load / generator shedding once this works).

Currently working:

1. Able to define "flexibility" actions (env.action_space({"flexibility":[("load_0_0", 0.1]}) with correct print string for BaseAction
2. Able to load flexible load attributes from "flex_loads_charac.csv"
3. "InvalidFlexibility" error if flexibility exceeds size of load
4. Observable change in load after executing flexibility action in environment
5. Integration with redispatch routine (flexibility and redispatch are solved together in one optim. problem)
6. Observable change in actual/target flexibility after executing flexibility action in environment
7. Automated tests

Conventions:
Positive flexibility: Reduce consumption (less demand in the grid)
Negative flexibility: Increase consumption (more demand in the grid)
target_flex: Sum of all flexibility at a load
actual_flex: Sum of all physically feasible flexibility at a load (determined by Environment)

[DEUCE1957]

Most tests passing now, remaining error (in ci/circleci:test) appears to be from CI itself:

ERROR: test_alert_gym_compat (unittest.loader._FailedTest)
----------------------------------------------------------------------
AttributeError: type object '_FailedTest' has no attribute 'test_alert_gym_compat'

----------------------------------------------------------------------

Appreciate any review / feedback to solve this @BDonnot

[BDonnot]

Hello,

Sorry for the delay.

This is so weird. I'll check the CI ASAP

[BDonnot]

Hello,

I think the CI is working on the dev_1.10.4 now, can you synch and see if it solves your issue ?

[DEUCE1957]

Afraid I am still getting the same error as above for the CI. @BDonnot

[BDonnot]

I'll try to update the doc but actually point 4 above is two different things:

• either the backend diverge (cannot compute the flows, no feasible solution found)
• or the "grid2op routine" to compute the "redispatching" fails

It is two distinct things.

[BDonnot]

I would avoid doing that if possible

[BDonnot]

It was not necessary before, and tests used to pass on the CI, maybe this is what caused the issue. Can you remove all these changes in the test suite and run it again ?

[DEUCE1957]

Trying that now, was a hotfix to get the tests running locally but seems I forgot to remove it later

[BDonnot]

Hi,

I see you added

import sys
from pathlib import Path
sys.path.append(str(Path(__file__).parent))

In a lot of files. That should not be necessary (did not used to be there when tests passed). Have you tried to remove it and see if tests pass ? Maybe you are missing the gym / gymnasium import when you do that (it's in general not a good idea to "mess with" python sys.path, it often leads to weird behaviour). Maybe it's not the cause of the problem, but maybe it is :-)

[BDonnot]

Also (to keep track of it, I noticed it and did not want to "discover this again later")

I think you did not update the https://grid2op.readthedocs.io/en/latest/modeled_elements.html page of the doc nor the https://grid2op.readthedocs.io/en/latest/mdp.html, for example https://grid2op.readthedocs.io/en/latest/mdp.html#action-space or https://grid2op.readthedocs.io/en/latest/mdp.html#state-space

[BDonnot]

Hello,

I fixed the issues caused by the install of gymnasium 1.0 in dev_1.10.4, I think once you'll have synched this will no longer be an issue :-)

Benjamin

[sonarqubecloud]

Quality Gate failed

Failed conditions
E Reliability Rating on New Code (required ≥ A)

See analysis details on SonarCloud

Catch issues before they fail your Quality Gate with our IDE extension SonarLint

[DEUCE1957]

Closing due to migration to dev_1.11.0.
See new Pull Request: #654