def __init__(self, env):
# =============== Init Observation Space =================
self.env = env
self.gym_obs_space = GymObservationSpace(self.env)
self.all_obs = []
obs = self.env.reset()
gym_obs = self.gym_obs_space.to_gym(obs)
for key in gym_obs.items():
self.all_obs.append(key)
self.obs_list = [
"prod_p", "prod_v", "load_p", "load_q", "rho", "topo_vect"
]
init_obs = self.convert_obs(obs)
#print("obs_shape = ", init_obs.shape) # (39,)
# =============== Init Action Space =================
self.converter = grid2op.Converter.IdToAct(self.env.action_space)
# action number 0 = DoNothing
self.converter.init_converter(
set_line_status=False, # 40
change_line_status=True, # 8
change_bus_vect=False, # 59
set_topo_vect=True, # 58
)
self.gym_action_space = GymActionSpace(action_space=self.converter)
ACT_SIZE = len(self.converter.all_actions)
#print("action space size= ", ACT_SIZE) # 68
#print("gym_action_space = ", self.gym_action_space) # Dict(action:Discrete(68))
gym_action = self.gym_action_space.sample()
#print("sample_action = ", gym_action) # OrderedDict([('action', 34)])
encoded_action = self.gym_action_space.from_gym(gym_action) # 34
self.num_actions = ACT_SIZE
def test_creation(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make("l2rpn_wcci_2020", test=True) as env:
# test i can create
converter = self.init_converter(env)
act_space = GymActionSpace(converter)
act_space.sample()
def test_json(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make("l2rpn_wcci_2020", test=True) as env:
# test i can create
obs_space = GymObservationSpace(env)
act_space = GymActionSpace(env.action_space)
obs_space.seed(0)
act_space.seed(0)
self._aux_test_json(obs_space)
self._aux_test_json(act_space)
class GridAgent():
def __init__(self, env):
# =============== Init Observation Space =================
self.env = env
self.gym_obs_space = GymObservationSpace(self.env)
self.all_obs = []
obs = self.env.reset()
gym_obs = self.gym_obs_space.to_gym(obs)
for key in gym_obs.items():
self.all_obs.append(key)
self.obs_list = [
"prod_p", "prod_v", "load_p", "load_q", "rho", "topo_vect"
]
init_obs = self.convert_obs(obs)
#print("obs_shape = ", init_obs.shape) # (39,)
# =============== Init Action Space =================
self.converter = grid2op.Converter.IdToAct(self.env.action_space)
# action number 0 = DoNothing
self.converter.init_converter(
set_line_status=False, # 40
change_line_status=True, # 8
change_bus_vect=False, # 59
set_topo_vect=True, # 58
)
self.gym_action_space = GymActionSpace(action_space=self.converter)
ACT_SIZE = len(self.converter.all_actions)
#print("action space size= ", ACT_SIZE) # 68
#print("gym_action_space = ", self.gym_action_space) # Dict(action:Discrete(68))
gym_action = self.gym_action_space.sample()
#print("sample_action = ", gym_action) # OrderedDict([('action', 34)])
encoded_action = self.gym_action_space.from_gym(gym_action) # 34
self.num_actions = ACT_SIZE
# grid observation -> gym observation
def convert_obs(self, obs):
gym_obs = self.gym_obs_space.to_gym(obs)
obs = []
for key, val in gym_obs.items():
if key in self.obs_list:
obs.extend(val)
obs = tf.convert_to_tensor(obs, dtype=tf.float32)
return obs
# grid action id (int) -> grid action
def id_to_act(self, encoded_act):
grid2op_action = self.converter.convert_act(
encoded_act) # action for grid2op (abstract)
return grid2op_action
def __init__(self,
env,
units=[32, 32],
buffer_size=5000,
learning_rate=0.01,
init_epsilon=1.,
epsilon_decay=0.99,
min_epsilon=0.01,
gamma=0.98,
batch_size=16,
target_update_iter=300,
train_nums=5000,
start_learning=64,
max_iter=200):
# =============== Init Observation Space =================
self.env = env
self.gym_obs_space = GymObservationSpace(self.env)
self.all_obs = []
obs = self.env.reset()
gym_obs = self.gym_obs_space.to_gym(obs)
for key in gym_obs.items():
self.all_obs.append(key)
self.obs_list = [
"prod_p", "prod_v", "load_p", "load_q", "rho", "topo_vect"
]
init_obs = self.convert_obs(obs)
#print("obs_shape = ", init_obs.shape) # (39,)
# =============== Init Action Space =================
self.converter = grid2op.Converter.IdToAct(self.env.action_space)
# action number 0 = DoNothing
self.converter.init_converter(
set_line_status=False, # 40
change_line_status=True, # 8
change_bus_vect=False, # 59
set_topo_vect=True, # 58
)
self.gym_action_space = GymActionSpace(action_space=self.converter)
ACT_SIZE = len(self.converter.all_actions)
#print("action space size= ", ACT_SIZE) # 68
#print("gym_action_space = ", self.gym_action_space) # Dict(action:Discrete(68))
gym_action = self.gym_action_space.sample()
#print("sample_action = ", gym_action) # OrderedDict([('action', 34)])
encoded_action = self.gym_action_space.from_gym(gym_action) # 34
self.num_actions = ACT_SIZE
'''
print("=======================alll action ==================")
for i in range(ACT_SIZE):
print("action number = ", i)
print(self.converter.convert_act(i))
print("buffer_size : ", buffer_size)
print("env : ", self.env)
'''
self.units = units
DQNAgent.__init__(self,
self.env,
obs=init_obs,
num_actions=ACT_SIZE,
buffer_size=buffer_size,
learning_rate=learning_rate,
init_epsilon=init_epsilon,
epsilon_decay=epsilon_decay,
min_epsilon=min_epsilon,
gamma=gamma,
batch_size=batch_size,
target_update_iter=target_update_iter,
train_nums=train_nums,
start_learning=start_learning)
#print(self.gym_action_space)
self.max_iter = max_iter
class GridAgent(DQNAgent):
def __init__(self,
env,
units=[32, 32],
buffer_size=5000,
learning_rate=0.01,
init_epsilon=1.,
epsilon_decay=0.99,
min_epsilon=0.01,
gamma=0.98,
batch_size=16,
target_update_iter=300,
train_nums=5000,
start_learning=64,
max_iter=200):
# =============== Init Observation Space =================
self.env = env
self.gym_obs_space = GymObservationSpace(self.env)
self.all_obs = []
obs = self.env.reset()
gym_obs = self.gym_obs_space.to_gym(obs)
for key in gym_obs.items():
self.all_obs.append(key)
self.obs_list = [
"prod_p", "prod_v", "load_p", "load_q", "rho", "topo_vect"
]
init_obs = self.convert_obs(obs)
#print("obs_shape = ", init_obs.shape) # (39,)
# =============== Init Action Space =================
self.converter = grid2op.Converter.IdToAct(self.env.action_space)
# action number 0 = DoNothing
self.converter.init_converter(
set_line_status=False, # 40
change_line_status=True, # 8
change_bus_vect=False, # 59
set_topo_vect=True, # 58
)
self.gym_action_space = GymActionSpace(action_space=self.converter)
ACT_SIZE = len(self.converter.all_actions)
#print("action space size= ", ACT_SIZE) # 68
#print("gym_action_space = ", self.gym_action_space) # Dict(action:Discrete(68))
gym_action = self.gym_action_space.sample()
#print("sample_action = ", gym_action) # OrderedDict([('action', 34)])
encoded_action = self.gym_action_space.from_gym(gym_action) # 34
self.num_actions = ACT_SIZE
'''
print("=======================alll action ==================")
for i in range(ACT_SIZE):
print("action number = ", i)
print(self.converter.convert_act(i))
print("buffer_size : ", buffer_size)
print("env : ", self.env)
'''
self.units = units
DQNAgent.__init__(self,
self.env,
obs=init_obs,
num_actions=ACT_SIZE,
buffer_size=buffer_size,
learning_rate=learning_rate,
init_epsilon=init_epsilon,
epsilon_decay=epsilon_decay,
min_epsilon=min_epsilon,
gamma=gamma,
batch_size=batch_size,
target_update_iter=target_update_iter,
train_nums=train_nums,
start_learning=start_learning)
#print(self.gym_action_space)
self.max_iter = max_iter
# grid observation -> gym observation
def convert_obs(self, obs):
gym_obs = self.gym_obs_space.to_gym(obs)
obs = []
for key, val in gym_obs.items():
if key in self.obs_list:
obs.extend(val)
return np.array(obs)
# grid action id (int) -> grid action
def id_to_act(self, encoded_act):
grid2op_action = self.converter.convert_act(
encoded_act) # action for grid2op (abstract)
return grid2op_action
def grid_evaluation(self, action_type='network', epi=0):
done = False
obs, done, ep_reward, steps = self.env.reset(), False, 0, 0
msg = 'SmartGrid Episode {}'.format(epi)
for t in notebook.tqdm(range(1, 2017), desc=msg):
converted_obs = self.convert_obs(obs)
#print("convt_obs : ",converted_obs)
if action_type == "network":
action = self.model.action_value(converted_obs[None])
elif action_type == "random":
action = np.random.randint(self.num_actions)
elif action_type == "do_nothing":
action = 0
#print("act : ",action)
converted_act = self.id_to_act(action)
#print("convt_act : ",converted_act)
obs, reward, done, _ = self.env.step(converted_act)
ep_reward += reward
steps += 1
if done:
break
self.env.close()
return steps, ep_reward
def test_to_from_gym_act(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make("l2rpn_wcci_2020", test=True) as env:
converter = self.init_converter(env)
act_space = GymActionSpace(converter)
act_space.seed(0)
converter.seed(0)
gym_act = act_space.sample()
act = act_space.from_gym(gym_act)
self._aux_test_json(act_space, gym_act)
gym_act2 = act_space.to_gym(act)
act2 = act_space.from_gym(gym_act2)
g2op_act = converter.convert_act(act)
g2op_act2 = converter.convert_act(act2)
assert g2op_act == g2op_act2
act_space.seed(0)
for i in range(10):
gym_act = act_space.sample()
act = act_space.from_gym(gym_act)
self._aux_test_json(act_space, gym_act)
gym_act2 = act_space.to_gym(act)
act2 = act_space.from_gym(gym_act2)
g2op_act = converter.convert_act(act)
g2op_act2 = converter.convert_act(act2)
assert g2op_act == g2op_act2
def test_to_from_gym_act(self):
with warnings.catch_warnings():
warnings.filterwarnings("ignore")
with make("l2rpn_wcci_2020", test=True) as env:
act_space = GymActionSpace(env.action_space)
act = env.action_space()
gym_act = act_space.to_gym(act)
self._aux_test_json(act_space, gym_act)
assert act_space.contains(gym_act)
act2 = act_space.from_gym(gym_act)
assert act == act2
act_space.seed(0)
for i in range(10):
gym_act = act_space.sample()
act = act_space.from_gym(gym_act)
self._aux_test_json(act_space, gym_act)
gym_act2 = act_space.to_gym(act)
act2 = act_space.from_gym(gym_act2)
assert act == act2