def test_dqn(EV_list_DQN_REF, CS_list_DQN_REF, graph, env, agent ):
agent.epsilon = 0
for e, pev in enumerate(EV_list_DQN_REF):
done = False
score = 0
path=[]
state, source, destination = env.test_reset(pev, graph, CS_list_DQN_REF)
print("\nEpi:", e, agent.epsilon)
print(source,'->', destination)
print('sim time:', env.sim_time)
path.append(source)
while not done:
action = agent.get_action(state)
next_state, next_node, reward, done = env.step(action)
score += reward
state = next_state
path.append(next_node)
if done:
print('sim time:', env.sim_time)
print('Distance:', pev.totaldrivingdistance)
print('Driving time:', pev.totaldrivingtime)
print(pev.charged, pev.curr_location, pev.init_SOC, pev.curr_SOC)
print(path)
while pev.curr_location != pev.destination:
came_from, cost_so_far = rt.a_star_search(graph, pev.curr_location, pev.destination)
path = rt.reconstruct_path(came_from, pev.curr_location, pev.destination)
path_distance = graph.get_path_distance(path)
# print("evcango: {} path dist: {}".format(evcango, path_distance))
pev.next_location = path[1]
pev.path.append(pev.next_location)
env.sim_time, time = rt.update_ev(pev, graph, pev.curr_location, pev.next_location, env.sim_time)
pev.curr_location = pev.next_location
def step(self, action, done):
cs, pev_SOC, front_path, total_d_time, waiting_time, charging_time = self.path_info[
action]
# print(self.sim_time, self.pev.curr_location, cs.id, pev_SOC)
self.target = cs
if pev_SOC <= 0.0:
print('error soc', pev_SOC)
done = 1
reward = -5
return np.zeros((1, self.state_size)), -1, reward, done
if len(front_path) > 1:
next_node = front_path[1]
self.sim_time, time = rt.update_ev(self.pev, self.graph,
self.pev.curr_location,
next_node, self.sim_time)
if self.sim_time == 0 and time == 0:
print('time idx error')
done = 1
reward = -5
return np.zeros((1, self.state_size)), -1, reward, done
done = 0
reward = -1 * (time)
self.pev.curr_location = next_node
self.path_info = rt.sim_main_first_time_check(
self.pev, self.CS_list, self.graph, self.action_size)
next_state = [self.pev.curr_location, self.pev.curr_SOC]
for path in self.path_info:
cs, pev_SOC, front_path, total_d_time, waiting_time, charging_time = path
next_state += [total_d_time, waiting_time, charging_time]
next_state = np.reshape(next_state, [1, self.state_size])
return next_state, next_node, reward, done
elif self.pev.curr_location == cs.id:
self.pev.before_charging_SOC = self.pev.curr_SOC
self.pev.cscharingenergy = self.pev.maxBCAPA * self.pev.req_SOC - self.pev.curr_SOC * self.pev.maxBCAPA
self.pev.cschargingcost = self.pev.cscharingenergy * cs.price[int(
self.sim_time / 5)]
self.pev.curr_SOC = self.pev.req_SOC
self.pev.cschargingtime = charging_time
self.pev.cschargingwaitingtime = waiting_time
self.pev.charged = 1
self.pev.cs = cs
self.sim_time += charging_time * 60
self.sim_time += waiting_time * 60
# print(waiting_time, waiting_time * 60)
self.pev.csdrivingtime = self.pev.totaldrivingtime
self.pev.csdistance = self.pev.totaldrivingdistance
self.pev.cschargingwaitingtime = self.pev.cschargingwaitingtime
self.pev.cschargingtime = self.pev.cschargingtime
self.pev.cssoc = self.pev.curr_SOC
done = 1
reward = -1 * (waiting_time + charging_time)
print(done, '충전소야')
return np.zeros((1, self.state_size)), -1, reward, done
else:
print("???")
input()
def step(self, action, done):
cs, pev_SOC, front_path, rear_path, front_path_distance, rear_path_distance, front_d_time, rear_d_time, rear_consump_energy, waiting_time, charging_time = self.path_info[
action]
# print(self.sim_time, self.pev.curr_location, cs.id, pev_SOC)
self.pev.path.append(self.pev.curr_location)
if len(front_path) > 1:
next_node = front_path[1]
self.sim_time, time = rt.update_ev(self.pev, self.graph,
self.pev.curr_location,
next_node, self.sim_time)
if self.sim_time == 0 and time == 0:
done = 1
reward = -10
return np.zeros((1, self.state_size)), -1, reward, done
if pev_SOC <= 0.0:
done = 1
reward = -10
return np.zeros((1, self.state_size)), -1, reward, done
done = 0
reward = -1
# reward = -time
self.pev.curr_location = next_node
self.path_info = rt.get_feature_state(self.sim_time, self.pev,
self.CS_list, self.graph,
self.action_size)
next_state = [
self.pev.source, self.pev.destination, self.pev.curr_SOC
]
for path in self.path_info:
cs, pev_SOC, front_path, rear_path, front_path_distance, rear_path_distance, front_d_time, rear_d_time, rear_consump_energy, waiting_time, charging_time = path
# next_state += [front_d_time, rear_d_time, waiting_time, charging_time]
next_state += [front_path_distance, rear_path_distance]
next_state = np.reshape(next_state, [1, self.state_size])
return next_state, next_node, reward, done
elif self.pev.curr_location == cs.id:
self.pev.before_charging_SOC = self.pev.curr_SOC
self.pev.cscharingenergy = self.pev.maxBCAPA * self.pev.req_SOC - self.pev.curr_SOC * self.pev.maxBCAPA
self.pev.cschargingcost = self.pev.cscharingenergy * cs.price[int(
self.sim_time / 5)]
self.pev.curr_SOC = self.pev.req_SOC
self.pev.cschargingtime = (
self.pev.cscharingenergy /
(cs.chargingpower * self.pev.charging_effi))
self.pev.cschargingwaitingtime = cs.waittime[int(self.sim_time /
5)]
self.pev.charged = 1
self.pev.cs = cs
self.pev.csid = cs.id
self.sim_time += self.pev.cschargingtime * 60
self.sim_time += self.pev.cschargingwaitingtime * 60
# print(waiting_time, waiting_time * 60)
self.pev.fdist = self.pev.totaldrivingdistance
self.pev.csdrivingtime = self.pev.totaldrivingtime
self.pev.csdistance = self.pev.totaldrivingdistance
self.pev.cschargingwaitingtime = self.pev.cschargingwaitingtime
self.pev.cschargingtime = self.pev.cschargingtime
self.pev.cssoc = self.pev.curr_SOC
done = 1
# reward = -1 * (self.pev.waitingtime + self.pev.chargingtime + rear_d_time + (rear_consump_energy/(cs.chargingpower*self.pev.charging_effi)))
reward = -1 * len(rear_path)
# print(done, '충전소야')
return np.zeros((1, self.state_size)), -1, reward, done
else:
print("???")
input()
scores.append(score)
episodes.append(e)
steps.append(step)
pev = env.pev
while pev.curr_location != pev.destination:
came_from, cost_so_far = rt.a_star_search(graph, pev.curr_location,
pev.destination)
path = rt.reconstruct_path(came_from, pev.curr_location,
pev.destination)
path_distance = graph.get_path_distance(path)
# print("evcango: {} path dist: {}".format(evcango, path_distance))
pev.next_location = path[1]
pev.path.append(pev.next_location)
env.sim_time, time = rt.update_ev(pev, graph, pev.curr_location,
pev.next_location, env.sim_time)
pev.curr_location = pev.next_location
EV_list_TA = copy.deepcopy(EV_list)
CS_list_TA = copy.deepcopy(CS_list)
# ta.one_time_check_timeweight(EV_list_TA, CS_list_TA, graph, 313)
ta.one_time_check_distweight(EV_list_TA, CS_list_TA, graph, 313)
EV_list_TEA = copy.deepcopy(EV_list)
CS_list_TEA = copy.deepcopy(CS_list)
# ta.every_time_check_timeweight(EV_list_TEA, CS_list_TEA, graph, 313)
ta.every_time_check_distweight(EV_list_TEA, CS_list_TEA, graph, 313)
ta.sim_result_text(resultdir,
EV_list_DQN=EV_list_DQN,
EV_list_TA=EV_list_TA,