Merge pull request #15 from BDonnot/nan_tovect

fixing bug for Nan in observation after some powerline change bus

CHANGELOG.rst

@@ -1,6 +1,11 @@
 Change Log
 =============
 
+[0.4.1] - 2019-12-17
+--------------------
+- [FIXED] Bug#14 : Nan in the observation space after switching one powerline [PandaPowerBackend]
+- [UPDATED] plot now improved for buses in substations
+
 [0.4.0] - 2019-12-04
 --------------------
 - [ADDED] Basic tools for plotting with the `PlotPlotly` module

docs/conf.py

@@ -22,7 +22,7 @@
 author = 'Benjamin Donnot'
 
 # The full version, including alpha/beta/rc tags
-release = '0.4.0'
+release = '0.4.1'
 version = '0.4'
 
 

grid2op/BackendPandaPower.py

@@ -407,12 +407,13 @@ def apply_action(self, action: Action):
                         nb_bus_now = len(np.unique(actual_topo))
                         if nb_bus_before > nb_bus_now:
                             # i must deactivate the unused bus
-                            self._grid.bus["in_service"].iloc[sub_id + self.n_substations] = False
+                            self._grid.bus["in_service"][sub_id + self.n_substations] = False
                             # print("I deactivated bus {} [before: {}; after {}]".format(sub_id,nb_bus_before,nb_bus_now))
                         elif nb_bus_before < nb_bus_now:
                             # i must activate the new bus
-                            self._grid.bus["in_service"].iloc[sub_id + self.n_substations] = True
-                            # print("I reactivated bus {} [before: {}; after {}]".format(sub_id,nb_bus_before,nb_bus_now))
+                            self._grid.bus["in_service"][sub_id + self.n_substations] = True
+                            # print("I reactivated second bus of substation {} [before: {}; after {}]".format(sub_id,nb_bus_before,nb_bus_now))
+
                         # now assign the proper bus to each element
                         for i, (table, col_name, row_id) in enumerate(self._what_object_where[sub_id]):
                             self._grid[table][col_name].iloc[row_id] = sub_id if actual_topo[i] == 1 else sub_id + self.n_substations

grid2op/PlotPlotly.py

@@ -516,6 +516,11 @@ def _draw_topos(self, observation, fig):
                     else:
                         buses_z[this_el_bus] += dict_el["z"]
             buses_z = [el / nb for el, nb in zip(buses_z, nb_co)]
+            theta_z = [cmath.phase((el - z_sub)) for el in buses_z]
+            m_ = np.mean(theta_z) - cmath.pi / 2
+            theta_z = [el-m_ for el in theta_z]
+            buses_z = [z_sub + (self.radius_sub-self.bus_radius)*0.75*cmath.exp(1j * theta) for theta in theta_z]
+
             # TODO don't just do the average, but afterwards split it more evenly, and at a fixed distance from the
             # center of the substation
 

grid2op/__init__.py

@@ -6,7 +6,7 @@
 import os
 import pkg_resources
 
-__version__ = '0.4.0'
+__version__ = '0.4.1'
 
 __all__ = ['Action', "BackendPandaPower", "Agent", "Backend", "ChronicsHandler", "Environment", "Exceptions",
            "Observation", "Parameters", "GameRules", "Reward", "Runner", "main", "Utils", "PlotPlotly"]

grid2op/tests/test_Action.py

@@ -12,11 +12,14 @@
 from Action import HelperAction, Action
 from GameRules import GameRules
 
+from grid2op.MakeEnv import make
+
 # TODO test that "twice change" is reset to normal. when i update an action twice, nothing is done.
 # TODO test for all class of Action
 
 # TODO clean the test to have it for all class of actions without recoding everything each time
 
+
 class TestLoadingBackendFunc(unittest.TestCase):
     def setUp(self):
         """

grid2op/tests/test_Agent.py

@@ -90,7 +90,7 @@ def setUp(self):
                                names_chronics_to_backend=self.names_chronics_to_backend,
                                rewardClass=self.rewardClass)
 
-    def _aux_test_agent(self, agent):
+    def _aux_test_agent(self, agent, i_max=30):
         done = False
         i = 0
         beg_ = time.time()
@@ -105,7 +105,7 @@ def _aux_test_agent(self, agent):
             time_act += end__ - beg__
             cum_reward += reward
             i += 1
-            if i > 30:
+            if i > i_max:
                 break
 
         end_ = time.time()
@@ -134,9 +134,9 @@ def test_1_powerlineswitch(self):
 
     def test_2_busswitch(self):
         agent = TopologyGreedy(self.env.helper_action_player)
-        i, cum_reward = self._aux_test_agent(agent)
-        assert i == 31, "The powerflow diverged before step 30 for greedy agent"
-        assert np.abs(cum_reward - 619.999011) <= self.tol_one, "The reward has not been properly computed"
+        i, cum_reward = self._aux_test_agent(agent, i_max=10)
+        assert i == 11, "The powerflow diverged before step 30 for greedy agent"
+        assert np.abs(cum_reward - 219.99795) <= self.tol_one, "The reward has not been properly computed"
 
 
 if __name__ == "__main__":

grid2op/tests/test_PandaPowerBackend.py

@@ -16,10 +16,12 @@
 from Environment import Environment
 from Exceptions import *
 from GameRules import GameRules
+from MakeEnv import make
 
 PATH_DATA_TEST = PATH_DATA_TEST_PP
 import pandapower as pppp
 
+
 class TestLoadingADN(unittest.TestCase):
     def setUp(self):
         self.tolvect = 1e-2
@@ -635,11 +637,11 @@ def test_topo_change_2sub(self):
         assert np.all(topo_vect[self.backend.gen_pos_topo_vect[gen_ids]] == arr2[self.backend.gen_to_sub_pos[gen_ids]])
 
         after_amps_flow = self.backend.get_line_flow()
-        after_amps_flow_th = np.array([  564.86831329,   313.93253616, 14222.67349148, 29227.00333159,
-                                       18393.56319792, 13649.04803367, 13649.04803367,   311.98084347,
-                                         252.73768739,    80.124112  ,    90.55351819,   220.8797656 ,
-                                       28364.56104918, 30242.87395452,     0.        ,   118.014306  ,
-                                          63.71939605,   147.6468084 ,   702.15602054,  1067.37181783])
+        after_amps_flow_th = np.array([  596.97014348,   342.10559579, 16615.11815357, 31328.50690716,
+                                       11832.77202397, 11043.10650167, 11043.10650167,   322.79533908,
+                                         273.86501458,    82.34066647,    80.89289074,   208.42396413,
+                                       22178.16766548, 27690.51322075, 38684.31540646,   129.44842477,
+                                          70.02629553,   185.67687123,   706.77680037,  1155.45754617])
         assert self.compare_vect(after_amps_flow, after_amps_flow_th)
 
         try:
@@ -865,6 +867,35 @@ def test_nb_timestep_overflow_disc2(self):
                 assert (not grid_tmp.get_line_status()[self.id_2nd_line_disco])
 
 
+class TestChangeBusAffectRightBus(unittest.TestCase):
+    def test_set_bus(self):
+        env = make()
+        env.reset()
+        # action = env.helper_action_player({"change_bus": {"lines_or_id": [17]}})
+        action = env.helper_action_player({"set_bus": {"lines_or_id": [(17, 2)]}})
+        obs, reward, done, info = env.step(action)
+
+        assert np.all(np.isfinite(obs.v_or))
+        assert np.sum(env.backend._grid["bus"]["in_service"]) == 15
+
+    def test_change_bus(self):
+        env = make()
+        env.reset()
+        action = env.helper_action_player({"change_bus": {"lines_or_id": [17]}})
+        obs, reward, done, info = env.step(action)
+        assert np.all(np.isfinite(obs.v_or))
+        assert np.sum(env.backend._grid["bus"]["in_service"]) == 15
+
+    def test_change_bustwice(self):
+        env = make()
+        env.reset()
+        action = env.helper_action_player({"change_bus": {"lines_or_id": [17]}})
+        obs, reward, done, info = env.step(action)
+        obs, reward, done, info = env.step(action)
+        assert np.all(np.isfinite(obs.v_or))
+        assert np.sum(env.backend._grid["bus"]["in_service"]) == 14
+
+
 # TODO test also the methods added for observation:
 """
         self._line_status = copy.copy(backend.get_line_status())

setup.py

@@ -7,7 +7,7 @@
 }
 
 setup(name='Grid2Op',
-      version='0.4.0',
+      version='0.4.1',
       description='An environment that allows to perform powergrid optimization.',
       long_description='Built with modularity in mind, this package allows to perform the same operations independantly of the software used to compute powerflow or method to generate grid states or forecasts.',
       classifiers=[