192 lines
9.2 KiB
Python
192 lines
9.2 KiB
Python
import gym
|
|
import numpy as np
|
|
import pandas as pd
|
|
from gym import spaces
|
|
from module import *
|
|
from parameters import *
|
|
from data_manager import *
|
|
|
|
|
|
class ESSEnv(gym.Env):
|
|
def __init__(self, **kwargs):
|
|
super(ESSEnv, self).__init__()
|
|
self.excess = None
|
|
self.shedding = None
|
|
self.unbalance = None
|
|
self.real_unbalance = None
|
|
self.operation_cost = None
|
|
self.current_output = None
|
|
self.final_step_outputs = None
|
|
self.data_manager = DataManager()
|
|
self._load_year_data()
|
|
self.month = None
|
|
self.day = None
|
|
self.TRAIN = True
|
|
self.current_time = None
|
|
self.episode_length = kwargs.get('episode_length', 24)
|
|
self.battery_parameters = kwargs.get('battery_parameters', battery_parameters)
|
|
self.dg_parameters = kwargs.get('dg_parameters', dg_parameters)
|
|
self.solar_parameters = kwargs.get('solar_parameters', solar_parameters)
|
|
self.wind_parameters = kwargs.get('wind_parameters', wind_parameters)
|
|
self.penalty_coefficient = 50 # control soft penalty constrain
|
|
self.sell_coefficient = 0.5 # control sell benefits
|
|
|
|
self.grid = Grid()
|
|
self.battery = Battery(self.battery_parameters)
|
|
self.dg1 = DG(self.dg_parameters['gen_1'])
|
|
self.dg2 = DG(self.dg_parameters['gen_2'])
|
|
self.dg3 = DG(self.dg_parameters['gen_3'])
|
|
self.solar = Solar(self.solar_parameters)
|
|
self.wind = Wind(self.wind_parameters)
|
|
|
|
self.action_space = spaces.Box(low=-1, high=1, shape=(5,), dtype=np.float32) # 已增加调节电压动作
|
|
self.state_space = spaces.Box(low=-np.inf, high=np.inf, shape=(10,), dtype=np.float32)
|
|
|
|
def reset(self, *args):
|
|
self.month = np.random.randint(1, 13) # choose 12 month
|
|
if self.TRAIN:
|
|
self.day = np.random.randint(1, 20)
|
|
else:
|
|
self.day = np.random.randint(20, Constant.MONTHS_LEN[self.month] - 1)
|
|
self.current_time = 0
|
|
self.battery.reset()
|
|
self.dg1.reset()
|
|
self.dg2.reset()
|
|
self.dg3.reset()
|
|
self.solar.reset()
|
|
self.wind.reset()
|
|
return self._build_state()
|
|
|
|
def _build_state(self):
|
|
soc = self.battery.SOC()
|
|
dg1_output = self.dg1.current_output
|
|
dg2_output = self.dg2.current_output
|
|
dg3_output = self.dg3.current_output
|
|
time_step = self.current_time
|
|
|
|
price = self.data_manager.get_price_data(self.month, self.day, self.current_time)
|
|
house_load = self.data_manager.get_load_cons_data(self.month, self.day, self.current_time)
|
|
temperature = self.data_manager.get_temperature_data(self.month, self.day, self.current_time)
|
|
irradiance = self.data_manager.get_irradiance_data(self.month, self.day, self.current_time)
|
|
wind_speed = self.data_manager.get_wind_data(self.month, self.day, self.current_time)
|
|
# print('house_load:', house_load)
|
|
pv_generation = self.solar.step(temperature, irradiance)
|
|
wd_generation = self.wind.step(wind_speed)
|
|
generation = pv_generation + wd_generation
|
|
net_load = house_load - generation
|
|
|
|
obs = np.concatenate((np.float32(time_step), np.float32(price), np.float32(soc), np.float32(net_load),
|
|
np.float32(dg1_output), np.float32(dg2_output), np.float32(dg3_output),
|
|
np.float32(temperature), np.float32(irradiance), np.float32(wind_speed)), axis=None)
|
|
return obs
|
|
|
|
def step(self, action): # state transition: current_obs->take_action->get_reward->get_finish->next_obs
|
|
# put action into each component
|
|
current_obs = self._build_state()
|
|
temperature = current_obs[7]
|
|
irradiance = current_obs[8]
|
|
wind_speed = current_obs[9]
|
|
self.battery.step(action[0]) # execute the state-transition part, battery.current_capacity also changed
|
|
self.dg1.step(action[1])
|
|
self.dg2.step(action[2])
|
|
self.dg3.step(action[3])
|
|
self.solar.step(action[4], temperature, irradiance)
|
|
self.wind.step(wind_speed)
|
|
self.current_output = np.array((self.dg1.current_output, self.dg2.current_output, self.dg3.current_output,
|
|
-self.battery.energy_change, self.solar.current_power, self.wind.current_power))
|
|
actual_production = sum(self.current_output)
|
|
price = current_obs[1]
|
|
netload = current_obs[3]
|
|
# print('actual_production:', actual_production, 'netload:', netload)
|
|
unbalance = actual_production - netload
|
|
|
|
reward = 0
|
|
excess_penalty = 0 # 过多
|
|
deficient_penalty = 0 # 过少
|
|
sell_benefit = 0
|
|
buy_cost = 0
|
|
self.excess = 0
|
|
self.shedding = 0
|
|
if unbalance >= 0: # now in excess condition
|
|
if unbalance <= self.grid.exchange_ability:
|
|
# sell money to grid is little [0.029,0.1]
|
|
sell_benefit = self.grid.get_cost(price, unbalance) * self.sell_coefficient
|
|
else:
|
|
sell_benefit = self.grid.get_cost(price, self.grid.exchange_ability) * self.sell_coefficient
|
|
# real unbalance that grid could not meet
|
|
self.excess = unbalance - self.grid.exchange_ability
|
|
excess_penalty = self.excess * self.penalty_coefficient
|
|
else: # unbalance <0, its load shedding model, deficient penalty is used
|
|
if abs(unbalance) <= self.grid.exchange_ability:
|
|
buy_cost = self.grid.get_cost(price, abs(unbalance))
|
|
else:
|
|
buy_cost = self.grid.get_cost(price, self.grid.exchange_ability)
|
|
self.shedding = abs(unbalance) - self.grid.exchange_ability
|
|
deficient_penalty = self.shedding * self.penalty_coefficient
|
|
battery_cost = self.battery.get_cost(self.battery.energy_change)
|
|
dg1_cost = self.dg1.get_cost(self.dg1.current_output)
|
|
dg2_cost = self.dg2.get_cost(self.dg2.current_output)
|
|
dg3_cost = self.dg3.get_cost(self.dg3.current_output)
|
|
solar_cost = self.solar.get_cost(self.solar.current_power)
|
|
wind_cost = self.wind.gen_cost(self.wind.current_power)
|
|
|
|
reward -= (battery_cost + dg1_cost + dg2_cost + dg3_cost + solar_cost + wind_cost + excess_penalty +
|
|
deficient_penalty - sell_benefit + buy_cost) / 1e3
|
|
self.operation_cost = (battery_cost + dg1_cost + dg2_cost + dg3_cost + solar_cost + wind_cost + excess_penalty +
|
|
deficient_penalty - sell_benefit + buy_cost)
|
|
self.unbalance = unbalance
|
|
self.real_unbalance = self.shedding + self.excess
|
|
final_step_outputs = [self.dg1.current_output, self.dg2.current_output, self.dg3.current_output,
|
|
self.battery.current_capacity, self.solar.current_power, self.wind.current_power]
|
|
self.current_time += 1
|
|
finish = (self.current_time == self.episode_length)
|
|
if finish:
|
|
self.final_step_outputs = final_step_outputs
|
|
self.current_time = 0
|
|
next_obs = self.reset()
|
|
else:
|
|
next_obs = self._build_state()
|
|
return current_obs, next_obs, float(reward), finish
|
|
|
|
# def render(self, current_obs, next_obs, reward, finish):
|
|
# print('day={},hour={:2d}, state={}, next_state={}, reward={:.4f}, terminal={}\n'.
|
|
# format(self.day, self.current_time, current_obs, next_obs, reward, finish))
|
|
|
|
def _load_year_data(self):
|
|
price_df = pd.read_csv('data/prices.csv', sep=',')
|
|
load_df = pd.read_csv('data/houseload.csv', sep=',')
|
|
irradiance_df = pd.read_csv('data/irradiance.csv', sep=',')
|
|
temperature_df = pd.read_csv('data/temper.csv', sep=',')
|
|
wind_df = pd.read_csv('data/wind.csv', sep=',')
|
|
|
|
price = price_df['Price'].to_numpy(dtype=float)
|
|
load = load_df['houseload'].to_numpy(dtype=float)
|
|
irradiance = irradiance_df['irradiance'].to_numpy(dtype=float)
|
|
temperature = temperature_df['t2m'].to_numpy(dtype=float)
|
|
wind = wind_df['wind_speed'].to_numpy(dtype=float)
|
|
|
|
'''redesign the magnitude for price and amount of generation as well as demand'''
|
|
def process_elements(elements, transform_function, add_function):
|
|
for element in elements:
|
|
transformed_element = transform_function(element)
|
|
add_function(transformed_element)
|
|
|
|
process_elements(price, lambda x: max(x / 10, 0.5), self.data_manager.add_price_element)
|
|
process_elements(load, lambda x: x * 3, self.data_manager.add_load_element)
|
|
process_elements(irradiance, lambda x: x, self.data_manager.add_irradiance_element)
|
|
process_elements(temperature, lambda x: x - 273.15, self.data_manager.add_temperature_element)
|
|
process_elements(wind, lambda x: x, self.data_manager.add_wind_element)
|
|
|
|
# if __name__ == '__main__':
|
|
# env = ESSEnv()
|
|
# env.TRAIN = False
|
|
# rewards = []
|
|
# env.reset()
|
|
# tem_action = [0.1, 0.1, 0.1, 0.1, 0.1]
|
|
# for _ in range(144):
|
|
# print(f'current month is {env.month}, current day is {env.day}, current time is {env.current_time}')
|
|
# current_obs, next_obs, reward, finish = env.step(tem_action)
|
|
# env.render(current_obs, next_obs, reward, finish)
|
|
# current_obs = next_obs
|
|
# rewards.append(reward)
|