Error using redispatching action

[CayphasFrancois]

Hello,

I am trying to implement a simple redispatching action on an environnement such as "l2rpn_neurips_2020_track1_small" or "l2rpn_2019" but at the moment I don't see any redispatching. Actual_dispatch stays at 0 even if "is_illegal" and "is_ambiguous" are false. Could you tell why is the environnement sort of avoiding my action and how to correct it ?

Here is my code and output :

Code :
`import grid2op
import os
import numpy as np
from grid2op.PlotGrid import PlotMatplot
import matplotlib.pyplot as plt
from tqdm import tqdm
display_tqdm = True

from grid2op.Agent import DoNothingAgent
from grid2op.Agent import BaseAgent
from grid2op.Agent import PowerLineSwitch

import warnings
import pandas as pd
pd.options.mode.chained_assignment = None

from grid2op.Runner import Runner
from grid2op.Action import BaseAction, PlayableAction
from grid2op.Chronics import GridStateFromFileWithForecasts #classe qui gère les données temporelles (les chroniques) dans Grid2Op, mais avec une fonctionnalité spécifique : les prévisions.
from grid2op.Reward import L2RPNReward #def d'un reward basé sur la charge des lignes
from grid2op.Chronics.readPypowNetData import ReadPypowNetData
from grid2op.Action import PlayableAction
from grid2op.Parameters import Parameters

env_name = "l2rpn_case14_sandbox"
p = Parameters()
p.ENV_DOES_REDISPATCHING = False
p.ALLOW_DISPATCH_GEN_SWITCH_OFF = True
env = grid2op.make("l2rpn_neurips_2020_track1_small",action_class = BaseAction,param=p)
path = os.path.join(grid2op.get_current_local_dir(), env_name)

env.set_id(1)
env.seed(1)
obs = env.reset()

print(obs.gen_pmax)
action_redisp = env.action_space({"redispatch": [
(0, +1.4), (2, +1.4),
(4, +1.4), (10, +1.4),
(19, -2.8), (21, -2.8)
]})
print(obs.gen_p)

next_obs, reward, done, extra_information = env.step(action_redisp)
print(action_redisp)
print(extra_information['is_illegal'])
print(extra_information['is_ambiguous'])
print(next_obs.gen_p)
print(next_obs.actual_dispatch)`

Output :

[BDonnot]

Hello

You set the flag "ENv_DOES_REDISPATCHING=False".

So yes the redispatching routine will be completely turned off. This is probably why it's not working.

Have you tried without this parameters?

[CayphasFrancois]

Without the parameter set as "False", the dispatching is realised by the environnement not by the action that I am applying to my environnement. As the documentation explained it, this parameter must be set to False if you want to implement redispatching action.

[BDonnot]

Then the doc is totally misleading. How would you rephrase it ?

I see this on the doc

"""
Setting this parameter to False is not advised if your agent really depends on redispatching, curtailment or actions on storage units OR if your backend does not have a distributed slack (at least). (note that we do not recommend to use it even if your backend has a distributed slack).
"""

Would rephrasing like "keep it the default (True) if you want your agent to be able to do redispatching" ? It's much more direct that's for sure 😊

Basically what this feature does is to allow the redispatching. If you disable it then the redispatching is also disabled.

[CayphasFrancois]

Indeed, I mainly focused on this sentence : "Whether to let the environment do the redispatching instead of relying on the backend internal implementation (usually done with the “slack buses”).". Then, I understood that the environnement would be doing the redispatching instead of my agent.

Thanks for the help

[BDonnot]

No problem. I'll add a GitHub issue to make sure this is updated for next release.