Closed marota closed 2 years ago
return_action_path https://github.com/marota/Oracle4Grid/blob/87eebf1980fffee65ef12663b6996d4cf816812e/oracle4grid/core/graph/compute_trajectory.py#L42 should actually return not the whole action needed to reach the configuration, but the "delta" action between the configuration at timestep t and timestep t+1 => take the atomic action difference between the 2 oracle_actions
As it is now, actions could be declared illegal between timestep, as you are not supposed to play again on the same substation for few timesteps
Quand on "annihile" une action précédente en revenant vers la configuration initiale, ce n'est peut-être pas suffisant de faire la diff entre les actions atomiques. IL faut sans doute aussi la configuration initiale
Cooldown was adjusted in previous commits. See https://github.com/marota/Oracle4Grid/blob/79ca824e97cd8f954670bfc305008421e8d905cc/oracle4grid/core/replay/agent_replay.py#L21
Currently NB_TIMESTEP_COOLDOWN_SUB=0 and NB_TIMESTEP_COOLDOWN_LINE=0 when running replay. But there should be considered as in challenge mode
Difficulty = "challenge" (default) NO_OVERFLOW_DISCONNECTION: False NB_TIMESTEP_OVERFLOW_ALLOWED: 3 NB_TIMESTEP_COOLDOWN_SUB: 3 NB_TIMESTEP_COOLDOWN_LINE: 3 HARD_OVERFLOW_THRESHOLD: 2.0 NB_TIMESTEP_RECONNECTION: 12 IGNORE_MIN_UP_DOWN_TIME: true ALLOW_DISPATCH_GEN_SWITCH_OFF: True ENV_DC: False FORECAST_DC: False MAX_SUB_CHANGED: 1 MAX_LINE_STATUS_CHANGED: 1