fix some bugs with forecasted env + add tests for the previous known connected state

Signed-off-by: DONNOT Benjamin <[email protected]>

Author: BDonnot

CHANGELOG.rst

@@ -140,6 +140,8 @@ Native multi agents support:
 - [FIXED] a bug in "obs.simulate" and "obs.get_forecast_env" : when a line was disconnected
   and the user tried to reconnect it (without specifying on which bus) it could do something
   different than "env.step" (with the same action)
+- [FIXED] a powerflow is run when the environment is first created even before the initial "env.step"
+  function is called. This is to ensure proper behaviour if env is used without being reset.
 - [ADDED] possibility to set the "thermal limits" when calling `env.reset(..., options={"thermal limit": xxx})`
 - [ADDED] possibility to retrieve some structural information about elements with
   with `gridobj.get_line_info(...)`, `gridobj.get_load_info(...)`, `gridobj.get_gen_info(...)` 

grid2op/Environment/_env_prev_state.py

@@ -6,6 +6,7 @@
 # SPDX-License-Identifier: MPL-2.0
 # This file is part of Grid2Op, Grid2Op a testbed platform to model sequential decision making in power systems.
 
+import copy
 from typing import Optional, Type, Any
 import numpy as np
 from grid2op.Space import GridObjects
@@ -103,6 +104,8 @@ def update_from_backend(self,
 
     def update_from_other(self, 
                           other : "_EnvPreviousState"):
+        if not self._can_modif:
+            raise Grid2OpException(f"Impossible to modifiy this _EnvPreviousState")
         for attr_nm in ["_load_p",
                         "_load_q",
                         "_gen_p",
@@ -115,7 +118,7 @@ def update_from_other(self,
             tmp = getattr(self, attr_nm)
             if tmp.size > 1:
                 # works only for array of size 2 or more
-                tmp[:] = getattr(other, attr_nm)
+                tmp[:] = copy.deepcopy(getattr(other, attr_nm))
             else:
                 setattr(self, attr_nm, getattr(other, attr_nm))
 

grid2op/Environment/baseEnv.py

@@ -1380,7 +1380,6 @@ def _has_been_initialized(self):
         if np.min([self.n_line, self.n_gen, self.n_load, self.n_sub]) <= 0:
             raise EnvironmentError("Environment has not been initialized properly")
         self._backend_action_class = _BackendAction.init_grid(bk_type, _local_dir_cls=self._local_dir_cls)
-        self._backend_action = self._backend_action_class()
 
         # initialize maintenance / hazards
         self._time_next_maintenance = np.full(bk_type.n_line, -1, dtype=dt_int)
@@ -1480,20 +1479,23 @@ def _has_been_initialized(self):
         if self._init_obs is None:
             # regular environment, initialized from scratch
             try:
+                self.backend.runpf(is_dc=self._parameters.ENV_DC)
                 self._previous_conn_state.update_from_backend(self.backend)
             except Exception as exc_:
                 # nothing to do in this case
                 self.logger.warning(f"Impossible to retrieve the initial state of the grid before running the initial powerflow: {exc_}")
                 self._previous_conn_state._topo_vect[:] = 1  # I force assign everything to busbar 1 by default...
             self._cst_prev_state_at_init = copy.deepcopy(self._previous_conn_state)
-            self._cst_prev_state_at_init.prevent_modification()
+            self._backend_action = self._backend_action_class()
         else:
             # environment initialized from an observation, eg forecast_env
+            # update the backend
+            self._backend_action = self.backend.update_from_obs(self._init_obs)
+            self._backend_action.last_topo_registered.values[:] = self._init_obs._prev_conn._topo_vect
             self._cst_prev_state_at_init = copy.deepcopy(self._init_obs._prev_conn)
-            self._cst_prev_state_at_init.prevent_modification()
-            self._previous_conn_state.update_from_other(self._cst_prev_state_at_init)
-            # TODO this has to be done also in "_reset_to_orig_state" but I don't know why...
+            self._previous_conn_state.update_from_other(self._init_obs._prev_conn)
 
+        self._cst_prev_state_at_init.prevent_modification()
         # update backend_action with the "last known" state
         self._backend_action.last_topo_registered.values[:] = self._previous_conn_state._topo_vect
 

grid2op/Environment/environment.py

@@ -534,14 +534,6 @@ def _init_backend(
         # reset everything to be consistent
         self._reset_vectors_and_timings()
 
-        # 1.11 and detachement: forecasted env
-        if self._init_obs is not None:
-            # update the backend
-            self._backend_action = self.backend.update_from_obs(self._init_obs)
-            self._backend_action.last_topo_registered.values[:] = self._init_obs._prev_conn._topo_vect
-            self._cst_prev_state_at_init.force_update(self._init_obs._prev_conn)
-            self._previous_conn_state.update_from_other(self._init_obs._prev_conn)
-        
     def max_episode_duration(self):
         """
         Return the maximum duration (in number of steps) of the current episode.