Merge pull request Grid2op#61 from BDonnot/bd-dev

Fix an issue when environment got opponent (when generating new data)

chronix2grid/grid2op_utils/utils.py

@@ -19,6 +19,8 @@
 import grid2op
 from grid2op.Parameters import Parameters
 from grid2op.Chronics import ChangeNothing, FromNPY
+from grid2op.Action import DontAct
+from grid2op.Opponent import NeverAttackBudget, BaseOpponent
 from lightsim2grid import LightSimBackend
 import numpy as np
 from numpy.random import default_rng
@@ -598,13 +600,19 @@ def _adjust_gens(all_loss_orig,
         try:
             prob.solve()
         except cp.error.SolverError as exc_:
-            error_ = RuntimeError(f"Pypsa failed to find a solution at iteration {iter_num}, error {exc_}")
+            error_ = RuntimeError(f"cvxpy failed to find a solution at iteration {iter_num}, error {exc_}")
             res_gen_p = None
             quality_ = None
             break
 
         # assign the generators
         gen_p_after_optim = real_p.value
+        if gen_p_after_optim is None:
+            error_ = RuntimeError(f"cvxpy failed to find a solution at iteration {iter_num}, and returned None.")
+            res_gen_p = None
+            quality_ = None
+            break
+
         id_redisp = 0
         for gen_id, gen_nm in enumerate(env_for_loss.name_gen):
             if env_for_loss.gen_redispatchable[gen_id]:
@@ -625,7 +633,12 @@ def _adjust_gens(all_loss_orig,
                 data_feeding_kwargs={"load_p": load_p,
                                      "load_q": load_q,
                                      "prod_p": 1.0 * res_gen_p,
-                                     "prod_v": gen_v}
+                                     "prod_v": gen_v},
+                opponent_budget_per_ts=0.,
+                opponent_init_budget=0.,
+                opponent_class=BaseOpponent,
+                opponent_budget_class=NeverAttackBudget,
+                opponent_action_class=DontAct,
                 )
         diff_ = np.full((env_fixed.max_episode_duration(), env_fixed.n_gen), fill_value=np.NaN)
         all_loss[:] = np.NaN
@@ -921,7 +934,16 @@ def handle_losses(path_env,
 
     return res_gen_p_df, error_, quality_
 
-def save_generated_data(this_scen_path, load_p, load_p_forecasted, load_q, load_q_forecasted, prod_p, prod_p_forecasted,
+def save_generated_data(this_scen_path,
+                        load_p,
+                        load_p_forecasted,
+                        load_q,
+                        load_q_forecasted,
+                        prod_p_generated,
+                        prod_p_after_dispatch,
+                        prod_p,  # after everything, the one the agent sees
+                        prod_p_forecasted,
+                        debug,
                         sep=';',
                         float_prec=FLOATING_POINT_PRECISION_FORMAT):
     """This function saves the data that have been generated by this script.
@@ -938,22 +960,30 @@ def save_generated_data(this_scen_path, load_p, load_p_forecasted, load_q, load_
         _description_
     load_q_forecasted : _type_
         _description_
+    prod_p_generated: _type_
+        generators just after renewable generation (non renewables gens are Nans here !)
     prod_p : _type_
-        _description_
+        generators after dispatch and losses !
     prod_p_forecasted : _type_
         _description_
     sep : str, optional
         _description_, by default ';'
     float_prec : _type_, optional
         _description_, by default FLOATING_POINT_PRECISION_FORMAT
     """
-    for df, nm in zip([load_p, load_p_forecasted, load_q, load_q_forecasted, prod_p, prod_p_forecasted],
-                      ["load_p", "load_p_forecasted", "load_q", "load_q_forecasted", "prod_p", "prod_p_forecasted"]):
+    li_dfs = [load_p, load_p_forecasted, load_q, load_q_forecasted, prod_p, prod_p_forecasted]
+    li_nms = ["load_p", "load_p_forecasted", "load_q", "load_q_forecasted", "prod_p", "prod_p_forecasted"]
+    if debug:
+        li_dfs.append(prod_p_generated)
+        li_nms.append("prod_p_generated")
+        li_dfs.append(prod_p_after_dispatch)
+        li_nms.append("prod_p_after_dispatch")
+    for df, nm in zip(li_dfs, li_nms):
         df.to_csv(os.path.join(this_scen_path, f'{nm}.csv.bz2'),
                   sep=sep,
                   float_format=float_prec,
                   header=True,
-                  index=False)
+                  index=False)       
 
 
 def save_meta_data(this_scen_path,
@@ -1062,7 +1092,9 @@ def generate_a_scenario(path_env,
                         PmaxErrorCorrRatio=0.9,
                         RampErrorCorrRatio=0.95,
                         threshold_stop=0.05,
-                        max_iter=100):
+                        max_iter=100,
+                        debug=True  # TODO more feature !
+                        ):
     """This function generates and save the data for a scenario.
     
     Generation includes:
@@ -1190,7 +1222,16 @@ def generate_a_scenario(path_env,
         this_scen_path = os.path.join(output_dir, scenario_id)
         if not os.path.exists(this_scen_path):
             os.mkdir(this_scen_path)
-        save_generated_data(this_scen_path, load_p, load_p_forecasted, load_q, load_q_forecasted, res_gen_p_df, res_gen_p_forecasted_df)
+        save_generated_data(this_scen_path,
+                            load_p,
+                            load_p_forecasted,
+                            load_q,
+                            load_q_forecasted,
+                            final_gen_p,  # generated, before economic dispatch
+                            gen_p_after_dispatch,  # generated, after economic dispatch (and possibly curtailment)
+                            res_gen_p_df,
+                            res_gen_p_forecasted_df,
+                            debug=debug)
         total_load = float(load_p.sum().sum())
         total_gen = float(res_gen_p_df.sum().sum())
         gen_p_per_step = res_gen_p_df.sum(axis=1)