EEBD_AI
/
wgz_decision
11
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

refactor models

This commit is contained in:
chenxiaodong 2025-02-17 10:18:29 +08:00
parent 425477fcab
commit d0ea09d663
4 changed files with 37 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
models/env.py
View File

@ -1,9 +1,10 @@
import gym import gym
import numpy as np
import pandas as pd import pandas as pd
from data_manager import * from models.data_manager import *
from module import * from models.module import *
from parameters import * from models.parameters import *
class WgzGym(gym.Env): class WgzGym(gym.Env):
@ -29,18 +30,18 @@ class WgzGym(gym.Env):
self.heat_a = 0.6 self.heat_a = 0.6
self.power_a = 0.4 self.power_a = 0.4
self.EC_parameters = kwargs.get('EC_parameters', EC_parameters) # 电解水制氢器 self.EC_parameters = kwargs.get('EC_parameters', EC_parameters) # 电解水制氢器
self.HST_parameters = kwargs.get('dg_parameters', dg_parameters) # 储氢罐 self.HST_parameters = kwargs.get('HST_parameters', HST_parameters) # 储氢罐
self.grid = Grid() self.grid = Grid()
self.EC = EC(self.EC_parameters) self.EC = EC(self.EC_parameters)
self.HST = HST(self.HST_parameters) self.HST = HST(self.HST_parameters)
self.action_space = gym.spaces.Box(low=-1, high=1, shape=(3,), dtype=np.float32) self.action_space = gym.spaces.Box(low=-1, high=1, shape=(2,), dtype=np.float32)
''' '''
时间 光伏 温度（湿度暂未考虑） 电需 热需（转化为对应热水所需瓦数） 人数 电价 7 时间 光伏 温度（湿度暂未考虑） 电需 热需（转化为对应热水所需瓦数） 人数 电价 7
电解水制氢功率 储氢罐容量占比 市电功率（注意标准化） 3 电解水制氢功率 储氢罐容量占比 2 市电功率（注意标准化）->舍(由供需控制)
''' '''
self.state_space = gym.spaces.Box(low=0, high=1, shape=(10,), dtype=np.float32) self.state_space = gym.spaces.Box(low=0, high=1, shape=(9,), dtype=np.float32)
def reset(self, *args): def reset(self, *args):
self.month = np.random.randint(1, 13) # choose 12 month self.month = np.random.randint(1, 13) # choose 12 month
@ -56,12 +57,12 @@ class WgzGym(gym.Env):
def _build_state(self): def _build_state(self):
hst_soc = self.HST.current_soc hst_soc = self.HST.current_soc
ec_out = self.EC.get_hydrogen() ec_out = self.EC.get_hydrogen()
grid_ex = self.grid grid_ex = self.grid.trade_energy
time_step = self.current_time time_step = self.current_time
if self.TRAIN: if self.TRAIN:
price = self.data_manager.get_price_data(self.month, self.day, self.current_time) price = self.data_manager.get_price_data(self.month, self.day, self.current_time)
temper = self.data_manager.get_temperature_data(self.month, self.day, self.current_time) temper = self.data_manager.get_temper_data(self.month, self.day, self.current_time)
solar = self.data_manager.get_solar_data(self.month, self.day, self.current_time) solar = self.data_manager.get_solar_data(self.month, self.day, self.current_time)
load = self.data_manager.get_load_data(self.month, self.day, self.current_time) load = self.data_manager.get_load_data(self.month, self.day, self.current_time)
heat = self.data_manager.get_heat_data(self.month, self.day, self.current_time) heat = self.data_manager.get_heat_data(self.month, self.day, self.current_time)
@ -76,8 +77,7 @@ class WgzGym(gym.Env):
obs = np.concatenate((np.float32(time_step), np.float32(price), np.float32(temper), obs = np.concatenate((np.float32(time_step), np.float32(price), np.float32(temper),
np.float32(solar), np.float32(load), np.float32(heat), np.float32(solar), np.float32(load), np.float32(heat),
np.float32(people), np.float32(ec_out), np.float32(hst_soc), np.float32(grid_ex)), np.float32(people), np.float32(ec_out), np.float32(hst_soc)), axis=None)
axis=None)
return obs return obs
def step(self, action): def step(self, action):
@ -85,7 +85,7 @@ class WgzGym(gym.Env):
current_obs = self._build_state() current_obs = self._build_state()
self.EC.step(action[0]) self.EC.step(action[0])
self.HST.step(action[1]) self.HST.step(action[1])
self.grid.step(action[2]) # self.grid.step(action[2], self.EC.power_max)
price = current_obs[1] price = current_obs[1]
temper = current_obs[2] # 用途待补充 temper = current_obs[2] # 用途待补充
solar = current_obs[3] solar = current_obs[3]
@ -120,11 +120,11 @@ class WgzGym(gym.Env):
economic_cost = hst_cost + ec_cost + solar_cost - sell_benefit + buy_cost economic_cost = hst_cost + ec_cost + solar_cost - sell_benefit + buy_cost
demand_cost = self.heat_a * heat_penalty + self.power_a * power_penalty demand_cost = self.heat_a * heat_penalty + self.power_a * power_penalty
eco_benifit = self.EC.less_carbon() - self.grid.get_carbon() eco_benifit = self.EC.less_carbon() - self.grid.get_carbon(power_gap)
reward = - self.a * demand_cost - self.b * economic_cost + self.c * eco_benifit reward = - self.a * demand_cost - self.b * economic_cost + self.c * eco_benifit
self.unbalance = power_gap + heat_gap self.unbalance = power_gap + heat_gap
final_step_outputs = [self.HST.current_soc, self.HST.get_power(), self.EC.current_power, self.grid.current_power] final_step_outputs = [self.HST.current_soc, self.HST.get_power(), self.EC.current_power]
self.current_time += 1 self.current_time += 1
finish = (self.current_time == self.episode_length) finish = (self.current_time == self.episode_length)
if finish: if finish:
@ -140,7 +140,7 @@ class WgzGym(gym.Env):
solar = data_df['solar_power'].to_numpy(dtype=float) solar = data_df['solar_power'].to_numpy(dtype=float)
temper = data_df['temper'].to_numpy(dtype=float) temper = data_df['temper'].to_numpy(dtype=float)
energy = data_df['energy_demand'].to_numpy(dtype=float) energy = data_df['energy_demand'].to_numpy(dtype=float)
water = data_df['water_demand'].to_numpy(dtype=float) heat = data_df['water_demand'].to_numpy(dtype=float)
people = data_df['people_count'].to_numpy(dtype=float) people = data_df['people_count'].to_numpy(dtype=float)
price = data_df['price'].to_numpy(dtype=float) price = data_df['price'].to_numpy(dtype=float)
@ -151,9 +151,9 @@ class WgzGym(gym.Env):
transformed_e = transform_function(e) transformed_e = transform_function(e)
add_function(transformed_e) add_function(transformed_e)
process_elements(solar, lambda x: x, self.data_manager.add_load_element) process_elements(solar, lambda x: x, self.data_manager.add_solar_element)
process_elements(temper, lambda x: x, self.data_manager.add_load_element) process_elements(temper, lambda x: x, self.data_manager.add_temper_element)
process_elements(energy, lambda x: x, self.data_manager.add_irradiance_element) process_elements(energy, lambda x: x, self.data_manager.add_electricity_element)
process_elements(water, lambda x: x, self.data_manager.add_temperature_element) process_elements(heat, lambda x: x, self.data_manager.add_heat_element)
process_elements(people, lambda x: x, self.data_manager.add_wind_element) process_elements(people, lambda x: x, self.data_manager.add_people_element)
process_elements(price, lambda x: x, self.data_manager.add_price_element) process_elements(price, lambda x: x, self.data_manager.add_price_element)

28
models/module.py
View File

@ -7,7 +7,7 @@ class EC:
self.ramp = params['ramp'] self.ramp = params['ramp']
self.lifetime = params['lifetime'] self.lifetime = params['lifetime']
self.equipment_cost = params['equipment_cost'] self.equipment_cost = params['equipment_cost']
self.electricity_efficiency = params['electricity_efficiency'] self.electrolysis_efficiency = params['electrolysis_efficiency']
self.carbon_reduce = params['carbon_reduce'] self.carbon_reduce = params['carbon_reduce']
def step(self, action_ec): def step(self, action_ec):
@ -20,17 +20,14 @@ class EC:
return self.equipment_cost / self.lifetime * price * self.current_power / self.power_max return self.equipment_cost / self.lifetime * price * self.current_power / self.power_max
def get_hydrogen(self): def get_hydrogen(self):
return self.current_power * self.electricity_efficiency * self.hydrogen_produce return self.current_power * self.electrolysis_efficiency * self.hydrogen_produce
def get_heat(self): def get_heat(self):
return self.current_power * (1 - self.electricity_efficiency) return self.current_power * (1 - self.electrolysis_efficiency)
def less_carbon(self): def less_carbon(self):
return self.current_power * self.carbon_reduce return self.current_power * self.carbon_reduce
def get_power_max(self):
return self.power_max
def reset(self): def reset(self):
self.current_power = 0 self.current_power = 0
@ -42,7 +39,6 @@ class HST:
self.capacity = params['capacity'] self.capacity = params['capacity']
self.min_soc = params['min_soc'] self.min_soc = params['min_soc']
self.max_soc = params['max_soc'] self.max_soc = params['max_soc']
self.ramp = params['ramp']
self.lifetime = params['lifetime'] self.lifetime = params['lifetime']
self.equipment_cost = params['equipment_cost'] self.equipment_cost = params['equipment_cost']
self.charge_efficiency = params['charge_efficiency'] self.charge_efficiency = params['charge_efficiency']
@ -51,12 +47,12 @@ class HST:
''' '''
储氢罐的充气速率 = 电解水制氢速率 （电解水制氢放的热会满足热水需求?） 储氢罐的充气速率 = 电解水制氢速率 （电解水制氢放的热会满足热水需求?）
如何控制上述待补充
储氢罐的放气速率 = 供电 （电价低时多电解，电价高时释放） 储氢罐的放气速率 = 供电 （电价低时多电解，电价高时释放）
''' '''
def step(self, action_hst): def step(self, action_hst):
energy = action_hst * self.ramp energy = action_hst * self.capacity
updated_soc = max(self.min_soc, min(self.max_soc, (self.current_soc * self.capacity + energy) / self.capacity)) updated_soc = max(self.min_soc, min(self.max_soc, (self.current_soc * self.capacity + energy) / self.capacity))
self.hydrogen_charge = (updated_soc - self.current_soc) * self.capacity self.hydrogen_charge = (updated_soc - self.current_soc) * self.capacity
self.current_soc = updated_soc self.current_soc = updated_soc
@ -85,16 +81,16 @@ class Grid:
def __init__(self): def __init__(self):
self.delta = 1 self.delta = 1
self.carbon_increace = 0.9 self.carbon_increace = 0.9
self.trade_energy = None # self.trade_energy = None
def step(self, action_grid): # def step(self, action_grid, ec_power_max):
self.trade_energy = (action_grid + 1) / 2 * EC.get_power_max() # 反标准化 # self.trade_energy = (action_grid + 1) / 2 * ec_power_max # 反标准化
def get_cost(self, price): def get_cost(self, price, trade_energy):
return price * self.trade_energy * self.delta return price * trade_energy * self.delta
def get_carbon(self): def get_carbon(self, trade_energy):
return self.trade_energy * self.carbon_increace return trade_energy * self.carbon_increace
def retrieve_past_price(self): def retrieve_past_price(self):
result = [] result = []

6
models/tools.py
View File

@ -15,12 +15,10 @@ def test_one_episode(env, act, device):
action = a_tensor.detach().cpu().numpy()[0] action = a_tensor.detach().cpu().numpy()[0]
state, next_state, reward, done = env.step(action) state, next_state, reward, done = env.step(action)
record_system_info.append([state[1], state[2], env.HST.current_soc(), env.HST.get_power(), record_system_info.append([state[1], state[2], env.HST.current_soc(), env.HST.get_power(),
env.EC.current_power, env.grid.current_power, action, reward]) env.EC.current_power, action, reward])
state = next_state state = next_state
# add information of last step EC, HST.current_soc, HST.power, grid # add information of last step EC, HST.current_soc, HST.power, grid
final_step_outputs = [self.HST.current_soc, self.HST.get_power(), self.EC.current_power, self.grid.current_power] record_system_info[-1][2:5] = [env.final_step_outputs[0], env.final_step_outputs[1], env.final_step_outputs[2]]
record_system_info[-1][2:6] = [env.final_step_outputs[0], env.final_step_outputs[1], env.final_step_outputs[2],
env.final_step_outputs[3]]
record = {'init_info': record_init_info, 'system_info': record_system_info} record = {'init_info': record_init_info, 'system_info': record_system_info}
return record return record

5
train.py
View File

@ -3,11 +3,12 @@ import pickle
os.environ['OMP_WAIT_POLICY'] = 'PASSIVE' # 确保在pytorch前设置 os.environ['OMP_WAIT_POLICY'] = 'PASSIVE' # 确保在pytorch前设置
from copy import deepcopy from copy import deepcopy
import pandas as pd import numpy as np
import torch
import torch.nn.functional as F import torch.nn.functional as F
from models.env import WgzGym from models.env import WgzGym
from models.net import ActorPPO, CriticAdv from models.net import ActorPPO, CriticAdv
from models.tools import get_episode_return, test_one_episode from models.tools import get_episode_return
def smooth_rewards(rewards, window=10): def smooth_rewards(rewards, window=10):