Umstrukturierung

.gitmodules

@@ -10,3 +10,9 @@
 [submodule "01_Input/Vorlagen/MPC Minimalbeispiele"]
 	path = 01_Input/Vorlagen/MPC Minimalbeispiele
 	url = https://git-ce.rwth-aachen.de/ebc/ebc-general/controls/mpc-tool-examples.git
+[submodule "MA Konrad"]
+	path = MA Konrad
+	url = https://git-ce.rwth-aachen.de/ebc/projects/GGE_EBC0002_BMWi_MPCGeothermie_/data-driven-model-predictive-control.git
+[submodule "01_Input/MA Konrad"]
+	path = 01_Input/MA Konrad
+	url = https://git-ce.rwth-aachen.de/ebc/projects/GGE_EBC0002_BMWi_MPCGeothermie_/data-driven-model-predictive-control.git

MA Konrad @ 95939187

+Subproject commit 959391878e07bdeb38847897bdf8ccbd81665f29

01_Input/Masterarbeit Konrad Beeser/README.md

-# Data Driven Model Predictive Control - DDMPC
-
-Framework for creating and investigating data-driven model predictive controllers.
-
-------------
-
-
-**ann_ddmpc:** Base framework and some examples for ANN based data-driven MPC 
-
-**gpr_ddmpc:** gaussian process based ddmpc, framework needs to be adapted to ann_ddmpc framework
-
-**lsi_ddmpc:** base example for the use of linear system identification
-
-**vdp_example:** proof of concept for ann-based ddmpc for a van der pol oscillator

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/Boptest/test_boptest_base.py

-import requests
-from urllib.parse import urljoin
-import pandas as pd
-import numpy as np
-
-url = "http://134.130.166.203:5000"
-
-""" Get Properties: """
-name = requests.get(url=urljoin(url,'name')).json()
-inputs_meta = requests.get(url=urljoin(url,'inputs')).json()
-measurements_meta = requests.get(url=urljoin(url,'measurements')).json()
-
-
-""" Initialize: """
-init_params = {'start_time': 8*24*3600,
-               'warmup_period': 7*24*3600}
-
-meas_dict = requests.put(url=urljoin(url, 'initialize'), data=init_params).json()
-
-step_size = requests.get(url=urljoin(url, 'step')).json()
-print(step_size)
-
-new_step_size = {'step': 900}
-requests.put(url=urljoin(url,'step'),data=new_step_size)
-
-# Setting a scenario resets the simulation to a given start time,
-# also a predefined duration will be passed
-scenario = {'electricity_price': 'dynamic',
-            'time_period': 'peak_heat_day'}
-scenario_return = requests.put(urljoin(url, 'scenario'), data=scenario).json()
-
-
-""" Set Forecast parameters and get forecast"""
-forecast_param = requests.get(urljoin(url, 'forecast_parameters')).json()
-new_forecast_param = {'horizon': 32*900,
-                      'interval': 900}
-requests.put(urljoin(url, 'forecast_parameters'),data=new_forecast_param)
-print(requests.get(urljoin(url, 'forecast_parameters')).json())
-
-forecast = requests.get(urljoin(url, 'forecast')).json()
-forecast_pd = pd.DataFrame(data=forecast)
-
-""" Set Inputs and Advance"""
-inputs={'oveHeaPumY_u':0.5}
-y=requests.post(urljoin(url, 'advance'),inputs).json()
-
-""" get results and KPIs"""
-res_req = {'point_name': 'oveHeaPumY_u',
-           'start_time': -np.inf,
-           'final_time': np.inf}
-res = requests.put(url=urljoin(url,'results'),data=res_req).json()
-
-kpis = requests.get(url= urljoin(url,'kpi')).json()

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/Boptest/test_boptest_service.py

-import requests
-
-url = "http://134.130.166.203:80"
-
-testcase = 'bestest_hydronic_heat_pump'
-
-testid=requests.post('{0}/testcases/{1}/select'.format(url,testcase)).json()['testid']
-name = requests.get('{0}/name/{1}'.format(url, testid)).json()

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/ashrae140_900_set_point_ctrl_predicting_energy_consumption/2_generate_data.py

-from Examples.ashrae140_900_set_point_ctrl_predicting_energy_consumption.configuration import *
-
-# PID controller for the BKT
-BKT_PID = PID(
-    y=TAirRoom,
-    u=Q_flowTabs,
-    dt=60*15,
-    Kp=1,
-    Ti=600,
-    Td=0,
-)
-
-# PID controller for the AHU
-AHU_PID = PID(
-    y=TAirRoom,
-    u=TsetAHU,
-    dt=60 * 15,
-    Kp=1,
-    Ti=600,
-    Td=0,
-)
-
-ThermalZone_Simulator.plotter.sub_folder_name = 'pid'
-ThermalZone_FMU.setup(start_time=0)
-ThermalZone_DataContainer = ThermalZone_Simulator.run(controllers=(AHU_PID, BKT_PID),
-                                                      duration=60 * 60 * 24 * 6)
-ThermalZone_DataContainers = ThermalZone_DataContainer.split_container(training_share=0.8,testing_share=0.1,validating_share=0.1)
-
-ThermalZone_DataHandler = DataHandler(containers=ThermalZone_DataContainers)
-ThermalZone_DataHandler.save(pid_DataHandler_name, override=True)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/ashrae140_900_set_point_ctrl_predicting_energy_consumption/3_train_anns.py

-from Examples.ashrae140_900_set_point_ctrl_predicting_energy_consumption.configuration import *
-ThermalZone_DataHandler = load_DataHandler(pid_DataHandler_name)
-containers = load_DataHandler(pid_DataHandler_name).containers
-
-# ----------------------------------- TAirRoom NetworkTrainer -----------------------------------
-
-TAirRoom_NetworkTrainer = NetworkTrainer(
-    inputs=(
-        # disturbances
-        Input(feature=dry_bul, lag=2),
-        Input(feature=rad_dir, lag=1),
-
-        Input(feature=daily_sin, lag=2),
-        Input(feature=daily_cos, lag=2),
-        Input(feature=weekly_sin, lag=2),
-        Input(feature=weekly_cos, lag=2),
-
-        # Stellgrößen
-        Input(feature=Q_flowTabs, lag=6),
-
-        Input(feature=TsetAHU, lag=2),
-        # T air room
-        Input(feature=TAirRoom, lag=6),
-
-    ),
-    output=Output(feature=d_TAirRoom),
-    step_size=60*15
-)
-
-TAirRoom_NetworkTrainer.load_data(containers=containers)
-TAirRoom_NetworkTrainer.tune(layer_range=(1,), neurons_range=(8, 16), trials=10, epochs=500, batch_size=100)
-TAirRoom_NetworkTrainer.save(d_TAirRoom_NetworkTrainer_name, override=True)
-
-# ----------------------------------- Q_flowAHU  NetworkTrainer -----------------------------------
-
-Q_flowAhu_NetworkTrainer = NetworkTrainer(
-    inputs=(
-
-        # AHU
-        Input(feature=TAirRoom, lag=2),
-        Input(feature=TsetAHU, lag=2),
-        Input(feature=dry_bul, lag=2),
-    ),
-    output=Output(feature=Q_flowAhu),
-    step_size=60*15
-)
-
-Q_flowAhu_NetworkTrainer.clear_data()
-Q_flowAhu_NetworkTrainer.load_data(containers=containers)
-Q_flowAhu_NetworkTrainer.tune(layer_range=(1,), neurons_range=(4, 8), trials=10, epochs=500, batch_size=100)
-Q_flowAhu_NetworkTrainer.save(Q_flowAHU_NetworkTrainer_name, override=True)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/ashrae140_900_set_point_ctrl_predicting_energy_consumption/4_mpc_online.py

-from Examples.ashrae140_900_set_point_ctrl_predicting_energy_consumption.configuration import *
-
-# load ann1
-TAirRoom_NetworkTrainer = load_NetworkTrainer(d_TAirRoom_NetworkTrainer_name)
-TAirRoom_NetworkTrainer.load_best_ann(loc=0)
-
-# load ann2
-Q_flowAhu_NetworkTrainer = load_NetworkTrainer(Q_flowAHU_NetworkTrainer_name)
-Q_flowAhu_NetworkTrainer.load_best_ann(loc=0)
-
-TAirRoom.mode = 'economic'
-
-ThermalZone_MPC = ModelPredictive(
-    dt=60*15,
-    N=32,
-    objectives=(
-        Objective(feature=TAirRoom,     weight=500,                 k_0_offset=1),
-        Objective(feature=Q_flowTabs,   weight=1,     norm=5,       k_N_offset=-1),
-        Objective(feature=d_Q_flowTabs, weight=2,     norm=5,       k_N_offset=-1),
-        Objective(feature=Q_flowAhu,    weight=1,     norm=5,       k_N_offset=-1),
-        Objective(feature=d_TsetAHU,    weight=2,     norm=5,       k_N_offset=-1),
-    ),
-    constraints=(
-        Constraint(features=(Q_flowTabs,),  lb=-5,              ub=5,               k_N_offset=-1),
-        Constraint(features=(TsetAHU,),     lb=273.15 + 18,     ub=273.15 + 25,     k_N_offset=-1),
-    ),
-    model=ThermalZone_Model,
-    disturbances_df=ThermalZone_FMU.load_disturbances(),
-    network_trainers=(TAirRoom_NetworkTrainer, Q_flowAhu_NetworkTrainer),
-    solver_options= {'verbose': False,
-                     'ipopt': {'print_level': 2,
-                               'linear_solver': 'ma57'}}
-)
-
-ThermalZone_Simulator.plotter.sub_folder_name = 'online_mpc'
-ThermalZone_FMU.setup(start_time=60*60*24*0)
-
-ThermalZone_MPC_DataHandler = DataHandler()
-
-# online mpc loop
-for repetition in range(3):
-
-    # run mpc and add to DataHandler
-    ThermalZone_DataContainer = ThermalZone_Simulator.run(
-        controllers=(ThermalZone_MPC,),
-        duration=60 * 60 * 24 * 3,
-    )
-
-    # split data into training, testing and validating
-    ThermalZone_DataContainers = ThermalZone_DataContainer.split_container(testing_share=0.15,training_share=0.7,validating_share=0.15)
-    ThermalZone_MPC_DataHandler.add_containers(ThermalZone_DataContainers)
-
-    # online learning
-    ThermalZone_Simulator.retrain_anns(
-        containers=ThermalZone_DataContainers,
-        network_trainer=TAirRoom_NetworkTrainer,
-        epochs=500,
-        batch_size=250,
-        clear_data=False,
-    )
-    ThermalZone_Simulator.retrain_anns(
-        containers=ThermalZone_DataContainers,
-        network_trainer=Q_flowAhu_NetworkTrainer,
-        epochs=500,
-        batch_size=250,
-        clear_data=False,
-    )
-    ThermalZone_MPC.reinitialize_mpc = True
-
-ThermalZone_MPC_DataHandler.save(MPC_DataHandler_name, override=True)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/ashrae140_900_set_point_ctrl_predicting_energy_consumption/5_analyse_data.py

-from Examples.ashrae140_900_set_point_ctrl_predicting_energy_consumption.configuration import *
-
-TAirRoom_DataHandler_online = load_DataHandler(MPC_DataHandler_name)
-complete_df_online = DataContainer.merge_containers(TAirRoom_DataHandler_online.containers)
-
-# comparison of online and offline learning
-complete_df_online = abs(complete_df_online.rolling(window=4*6*24*7, center=False).mean())
-plt.plot(complete_df_online['TAirRoom error_bounds'], color=fmt.red, label='TAirRoom error_bounds mean online')
-plt.legend()
-plt.show()
-
-# exit()
-TAirRoom_NetworkTrainer = load_NetworkTrainer('d_TAirRoom')
-TAirRoom_NetworkTrainer.evaluate_data(containers=TAirRoom_DataHandler_online.containers)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/ashrae140_900_set_point_ctrl_predicting_energy_consumption/configuration.py

-import numpy as np
-import pandas as pd
-from ddmpc.controller import *
-from ddmpc.modeling import *
-from ddmpc.utils import *
-
-
-# ---------------------------------------------- names -----------------------------------------------
-pid_DataHandler_name = 'ThermalZone_PID_Data'
-d_TAirRoom_NetworkTrainer_name = 'd_TAirRoom_NetworkTrainer'
-Q_flowAHU_NetworkTrainer_name = 'Q_flowAHU_NetworkTrainer'
-MPC_DataHandler_name = "Online_Learning_MPC"
-# -------------------------------------------- FEATURES --------------------------------------------
-
-time = Time(dis_name='Simulation Time', var_name='SimTime')
-
-dry_bul =        Temperature(dis_name='Außentemperatur',               var_name='weaBus.TDryBul',       plt_opts=PlotOptions(color=blue,         line=line_solid,    label='Außentemperatur'))
-rad_dif =        Radiation(dis_name='Horizontale diffuse Strahlung',   var_name='weaBus.HDifHor',       plt_opts=PlotOptions(color=light_red,    line=line_solid,    label='Strahlung'))
-rad_dir =        Radiation(dis_name='Direkte normale Strahlung',       var_name='weaBus.HDirNor',       plt_opts=PlotOptions(color=red,          line=line_dashdot,  label=''))
-rad_glo =        Radiation(dis_name='Horizontale globale Strahlung',   var_name='weaBus.HGloHor',       plt_opts=PlotOptions(color=dark_red,     line=line_dotted,   label=''))
-int_gain_human = Percentage(dis_name='Personen',                       var_name='internalGains.y[1]',   plt_opts=PlotOptions(color=light_grey,   line=line_solid,    label='Interne Gewinne'))
-int_gain_elect = Percentage(dis_name='Verbraucher',                    var_name='internalGains.y[2]',   plt_opts=PlotOptions(color=grey,         line=line_dashdot,  label=''))
-int_gain_light = Percentage(dis_name='Licht',                          var_name='internalGains.y[3]',   plt_opts=PlotOptions(color=dark_grey,    line=line_dotted,   label=''))
-
-Q_flowTabs =            HeatFlowSetPoint(dis_name='Wärmestrom', var_name='QFlowTabsSet', lb=-5, ub=5,     plt_opts=PlotOptions(color=red, line=fmt.line_solid))
-d_Q_flowTabs =          Change(base_feature=Q_flowTabs)
-d_d_Q_flowTabs =        Change(base_feature=d_Q_flowTabs)
-
-Q_flowCold =            HeatFlowMeas(dis_name='Wärmestrom_kalt', var_name='QFlowCold', plt_opts=PlotOptions(color=red, line=fmt.line_solid))
-Q_flowHot =             HeatFlowMeas(dis_name='Wärmestrom_warm', var_name='QFlowHeat', plt_opts=PlotOptions(color=red, line=fmt.line_solid))
-
-TsetAHU =               TemperatureSetPoint(dis_name='Tsoll AHU',  var_name='TAhuSet', lb=290.15, ub=299.15,default=290, plt_opts=PlotOptions(color=red, line=fmt.line_solid))
-d_TsetAHU =             Change(base_feature=TsetAHU)
-d_d_TsetAHU =           Change(base_feature=d_TsetAHU)
-
-TsetAHU_movsum =        MovingSum(base_feature=TsetAHU, n=3)
-d_TsetAHU_movsum =      MovingSum(base_feature=TsetAHU, n=2)
-
-TAirRoom = Controlled(
-    dis_name='Raumlufttemperatur',
-    var_name='TAirRoom',
-    mode='random',
-    plt_opts=PlotOptions(color=red, line=line_solid),
-    day_start=8,
-    day_end=18,
-    day_target=295.15,
-    day_lb=294.15,
-    day_ub=296.15,
-    night_lb=290.15,
-    night_ub=300.15,
-    night_target=295.15
-)
-TAirIn = Temperature(dis_name='Einlasstemperatur',  var_name='Bus.ahuBus.TSupAirMea')
-
-TsetAHU_minus_dry_bul = Subtraction(base_feature_one=TsetAHU, base_feature_two=dry_bul)
-d_TsetAHU_minus_dry_bul = Change(base_feature=TsetAHU_minus_dry_bul)
-TsetAHU_minus_TAirRoom = Subtraction(base_feature_one=TsetAHU, base_feature_two=TAirRoom)
-d_TsetAHU_minus_TAirRoom = Change(base_feature=TsetAHU_minus_TAirRoom)
-dry_bul_minus_TAirRoom = Subtraction(base_feature_one=dry_bul, base_feature_two=TAirRoom)
-d_dry_bul_minus_TAirRoom = Change(base_feature=dry_bul_minus_TAirRoom)
-
-d_dry_bul =        Change(base_feature=dry_bul)
-d_rad_dir =        Change(base_feature=rad_dir)
-d_TAirRoom =       Change(base_feature=TAirRoom)
-d_TAirIn =         Change(base_feature=TAirIn)
-
-daily_sin = PeriodicTime(name='Daily Sin',     time=time, function=np.sin, frequency=60 * 60 * 24)
-daily_cos = PeriodicTime(name='Daily Cos',     time=time, function=np.cos, frequency=60 * 60 * 24)
-weekly_sin = PeriodicTime(name='Weekly Sin',    time=time, function=np.sin, frequency=60 * 60 * 24 * 7)
-weekly_cos = PeriodicTime(name='Weekly Cos',    time=time, function=np.cos, frequency=60 * 60 * 24 * 7)
-
-t_1 = Temperature(dis_name='AHU_hot_in',   var_name='Bus.ahuBus.heaterBus.hydraulicBus.TFwrdInMea')
-t_2 = Temperature(dis_name='AHU_hot_out',  var_name='Bus.ahuBus.heaterBus.hydraulicBus.TRtrnOutMea')
-t_3 = Temperature(dis_name='AHU_cold_in',  var_name='Bus.ahuBus.coolerBus.hydraulicBus.TFwrdInMea')
-t_4 = Temperature(dis_name='AHU_cold_out', var_name='Bus.ahuBus.coolerBus.hydraulicBus.TRtrnOutMea')
-
-f_1 = MassFlow(dis_name='AHU_hot',  var_name='Bus.ahuBus.heaterBus.hydraulicBus.VFlowInMea',  temperature_in=t_1, temperature_out=t_2)
-f_2 = MassFlow(dis_name='AHU_cold', var_name='Bus.ahuBus.coolerBus.hydraulicBus.VFlowInMea',  temperature_in=t_3, temperature_out=t_4)
-
-Q_flowAhu = EnergyBalance(name='Q_flow_Ahu', flows=[f_1, f_2])
-
-# ------------------------------------------- Model -------------------------------------------
-
-ThermalZone_Model = Model(
-    t=time,
-    X=(
-        TAirRoom,
-        Q_flowAhu,
-    ),
-    U=(
-        Q_flowTabs,
-        TsetAHU
-    ),
-    D=(
-        dry_bul,
-        rad_dir,
-        daily_sin,
-        daily_cos,
-        weekly_sin,
-        weekly_cos,
-    ),
-    C=(
-        d_TAirRoom,
-        TsetAHU_minus_dry_bul,
-        TsetAHU_minus_TAirRoom,
-        d_TsetAHU,
-        d_Q_flowTabs,
-    ),
-    T=(
-        Q_flowCold,
-        Q_flowHot,
-        t_1,
-        t_2,
-        t_3,
-        t_4,
-        f_1,
-        f_2
-    ),
-)
-
-# ------------------------------------------- Plotter -------------------------------------------
-
-ThermalZone_Plotter = Plotter(
-    SubPlot(features=(TAirRoom,),       y_label='Raumlufttemperatur in °C', shift=273.15),
-    SubPlot(features=(Q_flowTabs,),     y_label='Sollwert Wärmestrom BKT', step=True),
-    SubPlot(features=(TsetAHU,),        y_label='Sollwert Einlasstemperatur', shift=273.15, lb=14, ub=31, step=True),
-    SubPlot(features=(Q_flowAhu,),      y_label='Ist Wärmestrom AHU', step=True),
-    SubPlot(features=(dry_bul, rad_dir), y_label='Störgrößen in %', normalize=True),
-    size=(8, 8)
-)
-
-
-# ---------------------------------------------- FMU -----------------------------------------------
-
-ThermalZone_FMU = FMUSystem(model= ThermalZone_Model,
-                            step_size=60 * 15,
-                            fmu_name='ashrae140_900_set_point_fmu.fmu')
-# ------------------------------------------ Simulator ------------------------------------------
-ThermalZone_FMU.load_disturbances()
-ThermalZone_Simulator = Simulator(
-    system=ThermalZone_FMU,
-    plotter=ThermalZone_Plotter
-)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/ashrae140_900_set_point_ctrl_predicting_energy_consumption/learn_mpc.py

-from Examples.ashrae140_900_set_point_ctrl_predicting_energy_consumption.configuration import *
-
-# ----------------------------------- TsetAHU NetworkTrainer -----------------------------------
-
-TAirRoom_NetworkTrainer = NetworkTrainer(
-    inputs=(
-
-        # disturbances
-        Input(feature=dry_bul, lag=2),
-        Input(feature=rad_dir, lag=1),
-        Input(feature=d_dry_bul, lag=2),
-        Input(feature=d_rad_dir, lag=1),
-
-        Input(feature=daily_sin, lag=1),
-        Input(feature=daily_cos, lag=1),
-        Input(feature=weekly_sin, lag=1),
-        Input(feature=weekly_cos, lag=1),
-
-        # Stellgrößen
-        Input(feature=Q_flowTabs, lag=6),
-        Input(feature=TsetAHU, lag=3),
-
-        Input(feature=d_Q_flowTabs, lag=5),
-        Input(feature=d_TsetAHU, lag=2),
-
-        # T air room
-        Input(feature=TAirRoom, lag=1),
-        Input(feature=d_TAirRoom, lag=1),
-    ),
-    output=Output(feature=TsetAHU),
-    step_size=60*15
-)
-
-containers = load_DataHandler('ThermalZone_MPC_online').containers
-TAirRoom_NetworkTrainer.load_data(containers=containers)
-TAirRoom_NetworkTrainer.tune(layer_range=(1,), neurons_range=(8, 16), trials=5, epochs=500, batch_size=100)
-TAirRoom_NetworkTrainer.save('TsetAHU', override=True)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/ashrae140_900_set_point_ctrl_predicting_energy_consumption/load_res.py

-from ddmpc import *
-import pandas as pd
-
-ThermalZone_DataHandler = read_pkl(r"C:\Users\Administrator\Documents\data-driven-model-predictive-control\DataDrivenMPC\ashrae140_900_set_point_ctrl\stored_data\Res_6\MPC_one_year_8h_no_online_learning_reduced_features.pkl")
-
-training = ThermalZone_DataHandler[0].df
-training.index = training['SimTime']
-for i in range(1,12):
-    container = ThermalZone_DataHandler[i].df
-    container.index = container['SimTime']
-    container = container[1:]
-    training = training.append(container)
-
-mpc_res = ThermalZone_DataHandler[12].df
-mpc_res.index = mpc_res['SimTime']
-for i in range(13,49):
-    container = ThermalZone_DataHandler[i].df
-    container.index = container['SimTime']
-    container = container[1:]
-    mpc_res = mpc_res.append(container)
-
-write_pkl(data={'Training':training,'MPC':mpc_res},filename="DDMPC_one_year_8h_online_learning_reduced_features")
-mpc_res.to_csv("data_driven_mpc_one_year_8h_prediction_online_learning_reduced_features.csv")
-training.to_csv("data_driven_mpc_one_year_8h_prediction_online_learning_reduced_features_training.csv")
\ No newline at end of file

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/boptest_hydronic_heat_pump_cost_based/2_generate_data_boptest.py

-from Examples.boptest_hydronic_heat_pump_cost_based.configuration_boptest import *
-
-# PID controller for the HP
-HP_PID = PID(
-    y=TAirRoom,
-    u=u_hp_rel,
-    dt=900,
-    Kp=0.05,
-    Ti=300,
-    Td=0,
-)
-
-
-TAirRoom.mode = 'random'
-
-hydronic_hp_Simulator.plotter.sub_folder_name = 'pid'
-hydronic_hp_system.setup(start_time=0,warmup_period=7*24*3600)
-hydronic_hp_system.activate_ctrl_layer({'oveHeaPumY_activate':1})
-hydronic_hp_DataContainer_random = hydronic_hp_Simulator.run(controllers=(HP_PID,),
-                                                      duration=60 * 60 * 24 * 14)
-TAirRoom.mode = 'steady'
-hydronic_hp_DataContainer_steady = hydronic_hp_Simulator.run(controllers=(HP_PID,),
-                                                      duration=60 * 60 * 24 * 14)
-hydronic_hp_DataContainers_random = hydronic_hp_DataContainer_random.split_container(training_share=0.7,testing_share=0.15,validating_share=0.15)
-hydronic_hp_DataContainers_steady = hydronic_hp_DataContainer_steady.split_container(training_share=0.7,testing_share=0.15,validating_share=0.15)
-
-hydronic_hp_DataHandler = DataHandler()
-hydronic_hp_DataHandler.add_containers(*hydronic_hp_DataContainers_random)
-hydronic_hp_DataHandler.add_containers(*hydronic_hp_DataContainers_steady)
-
-hydronic_hp_DataHandler.save(pid_DataHandler_name, override=True)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/boptest_hydronic_heat_pump_cost_based/3_train_anns_boptest.py

-from Examples.boptest_hydronic_heat_pump_cost_based.configuration_boptest import *
-
-containers = load_DataHandler(pid_DataHandler_name).containers
-
-
-# ------------------ P_el Networktrainer
-
-P_el_NetworkTrainer = NetworkTrainer(
-    inputs=(
-        # disturbances
-        Input(feature=t_amb, lag=3),
-
-        # Stellgrößen
-        Input(feature=u_hp_rel, lag=3),
-
-        # T air room
-        Input(feature=TAirRoom, lag=3),
-
-    ),
-    output=Output(feature=power_hp),
-    step_size=60*15,
-)
-
-P_el_NetworkTrainer.load_data(containers=containers)
-P_el_NetworkTrainer.tune(layer_range=(1,),
-                         neurons_range=(8, 16),
-                         trials=5,
-                         epochs=1000,
-                         batch_size=100,
-                         output_scale=6000,
-                         loss='mse')
-P_el_NetworkTrainer.save(P_hp_el_NetworkTrainer_name, override=True)
-
-# ----------------------------------- TAirRoom NetworkTrainer -----------------------------------
-
-TAirRoom_NetworkTrainer = NetworkTrainer(
-    inputs=(
-        # disturbances
-        Input(feature=t_amb, lag=2),
-        Input(feature=rad_dir, lag=1),
-
-        Input(feature=daily_sin, lag=2),
-        Input(feature=daily_cos, lag=2),
-        Input(feature=weekly_sin, lag=2),
-        Input(feature=weekly_cos, lag=2),
-
-        # Stellgrößen
-        Input(feature=u_hp_rel, lag=4),
-
-        # T air room
-        Input(feature=TAirRoom, lag=4),
-
-    ),
-    output=Output(feature=d_TAirRoom),
-    step_size=60*15
-)
-
-TAirRoom_NetworkTrainer.load_data(containers=containers)
-TAirRoom_NetworkTrainer.tune(layer_range=(1,), neurons_range=(8, 16), trials=5, epochs=1000, batch_size=100)
-TAirRoom_NetworkTrainer.save(d_TAir_NetworkTrainer_name, override=True)
-
-

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/boptest_hydronic_heat_pump_cost_based/4_mpc_online_boptest.py

-from Examples.boptest_hydronic_heat_pump_cost_based.configuration_boptest import *
-
-MPC_DataHandler_name = 'mpc_online_2'
-# load ann1
-TAirRoom_NetworkTrainer = load_NetworkTrainer(d_TAir_NetworkTrainer_name+'online_learning')
-TAirRoom_NetworkTrainer.load_best_ann(loc=0)
-
-# load ann2
-P_el_NetworkTrainer = load_NetworkTrainer(P_hp_el_NetworkTrainer_name+'online_learning')
-P_el_NetworkTrainer.load_best_ann(loc=0)
-TAirRoom.mode = 'economic'
-
-hydronic_MPC = ModelPredictive(
-    dt=60*15,
-    N=96,
-    objectives=(
-        Objective(feature=TAirRoom,     weight=25,                 k_0_offset=1),
-        Objective(feature=costs_el,   weight=10,     norm=1000,       k_N_offset=-1, cost_order=CostOrder.linear),
-        Objective(feature=d_u_hp_rel, weight=1,  norm=1 ,k_N_offset=-1),
-    ),
-    constraints=(
-        Constraint(features=(u_hp_rel,),  lb=0,              ub=1,               k_N_offset=-1),
-        ),
-
-    model=hydronic_hp_model,
-    disturbances_df=pd.DataFrame(),
-    network_trainers=(TAirRoom_NetworkTrainer, P_el_NetworkTrainer,),
-    solver_options={'verbose': False,
-                    'ipopt': {'print_level': 2,
-                              'linear_solver': 'ma27'}}
-)
-
-hydronic_hp_Simulator.plotter.sub_folder_name = 'online_mpc'
-hydronic_hp_system.setup(start_time=0,warmup_period=7*24*3600)
-hydronic_hp_system.activate_ctrl_layer({'oveHeaPumY_activate': 1})
-
-hydronic_hp_MPC_DataHandler = DataHandler()
-
-# online mpc loop
-for repetition in range(12):
-
-    # run mpc and add to DataHandler
-    hydronic_DataContainer = hydronic_hp_Simulator.run(
-        controllers=(hydronic_MPC,),
-        duration=60 * 60 * 24 * 6,
-    )
-
-    # split data into training, testing and validating
-    hydronic_DataContainers = hydronic_DataContainer.split_container(testing_share=0.2,training_share=0.6,validating_share=0.2)
-    hydronic_hp_MPC_DataHandler.add_containers(*hydronic_DataContainers)
-    hydronic_MPC.reinitialize_mpc = True
-
-    # online learning
-    hydronic_hp_Simulator.retrain_anns(
-        containers=hydronic_DataContainers,
-        network_trainer=TAirRoom_NetworkTrainer,
-        epochs=500,
-        batch_size=250,
-        clear_data=False,
-    )
-    hydronic_hp_Simulator.retrain_anns(
-        containers=hydronic_DataContainers,
-        network_trainer=P_el_NetworkTrainer,
-        epochs=500,
-        batch_size=250,
-        clear_data=False,
-   )
-
-
-hydronic_hp_MPC_DataHandler.save(MPC_DataHandler_name, override=True)
-TAirRoom_NetworkTrainer.save_ann()
-P_el_NetworkTrainer.save_ann()
-TAirRoom_NetworkTrainer.save(d_TAir_NetworkTrainer_name+'_online_learning_2', override=True)
-P_el_NetworkTrainer.save(P_hp_el_NetworkTrainer_name+'_online_learning_2', override=True)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/boptest_hydronic_heat_pump_cost_based/configuration_boptest.py

-import numpy as np
-import pandas as pd
-from ddmpc.controller import *
-from ddmpc.modeling import *
-from ddmpc.utils import *
-
-
-# ---------------------------------------------- names -----------------------------------------------
-pid_DataHandler_name = 'hydronic_PID_Data'
-d_TAir_NetworkTrainer_name = 'd_TAir_NetworkTrainer'
-P_hp_el_NetworkTrainer_name = 'P_hp_el_NetworkTrainer'
-MPC_DataHandler_name = "Online_Learning_MPC"
-# -------------------------------------------- FEATURES --------------------------------------------
-time = Time(dis_name='Simulation Time', var_name='SimTime')
-
-t_amb = Temperature(dis_name='Ambient temperature_',
-                           var_name='TDryBul',
-                           plt_opts=PlotOptions(color=blue,
-                                                line=line_solid,
-                                                label='Ambient Temperature'))
-rad_dir = Radiation(dis_name='direct radiation',
-                      var_name='HDirNor',
-                      plt_opts=PlotOptions(color=light_red, line=line_solid, label='Radiation'))
-
-TAirRoom = Controlled(
-    dis_name='TAir',
-    var_name='reaTZon_y',
-    mode='random',
-    plt_opts=PlotOptions(color=red, line=line_solid),
-    day_start=8,
-    day_end=18,
-    day_target=295.15,
-    day_lb=294.15,
-    day_ub=296.15,
-    night_lb=290.15,
-    night_ub=300.15,
-    night_target=295.15,
-    random_step_size=3600*4
-)
-
-d_TAirRoom = Change(base_feature=TAirRoom)
-
-u_hp_rel = RelPower(var_name='oveHeaPumY_u',
-                    dis_name='u_hp',
-                    plt_opts=PlotOptions(color=blue, line=line_solid, label='u_hp'),
-                    lb=0,
-                    ub=1,
-                    default=0)
-d_u_hp_rel = Change(base_feature=u_hp_rel)
-
-daily_sin = PeriodicTime(name='Daily Sin',     time=time, function=np.sin, frequency=60 * 60 * 24)
-daily_cos = PeriodicTime(name='Daily Cos',     time=time, function=np.cos, frequency=60 * 60 * 24)
-weekly_sin = PeriodicTime(name='Weekly Sin',    time=time, function=np.sin, frequency=60 * 60 * 24 * 7)
-weekly_cos = PeriodicTime(name='Weekly Cos',    time=time, function=np.cos, frequency=60 * 60 * 24 * 7)
-
-power_fan = PowerMeas(var_name='reaPFan_y', dis_name='el. Power Fan', plt_opts=PlotOptions(color=blue, line=line_dotted, label='P_fan'),)
-power_hp = PowerMeas(var_name='reaPHeaPum_y', dis_name='el. Power HP', plt_opts=PlotOptions(color=red, line=line_solid, label='P_hp'),)
-power_ec = PowerMeas(var_name='reaPPumEmi_y', dis_name='el. Power emission circuit', plt_opts=PlotOptions(color=grey, line=line_dashdot, label='P_ec'),)
-
-price_el = Price(var_name='PriceElectricPowerHighlyDynamic', dis_name='el. Power Price', plt_opts=PlotOptions(color=black, line=line_solid, label='P_ec'))
-
-costs_el = Product(base_feature_one=price_el, base_feature_two=power_hp,plt_opts=PlotOptions(color=black, line=line_solid, label='P_ec'))
-# ------------------------------------------- Model -------------------------------------------
-
-hydronic_hp_model = Model(
-    t=time,
-    X=(
-        TAirRoom,
-        power_hp,
-    ),
-    U=(
-        u_hp_rel,
-    ),
-    D=(
-        t_amb,
-        rad_dir,
-        daily_sin,
-        daily_cos,
-        weekly_sin,
-        weekly_cos,
-        price_el,
-    ),
-    C=(
-        d_TAirRoom,
-        d_u_hp_rel,
-        costs_el,
-
-    ),
-    T=(
-        power_fan,
-        power_ec,
-    ),
-)
-
-# ------------------------------------------- Plotter -------------------------------------------
-
-plotter_hydronic = Plotter(
-    SubPlot(features=(TAirRoom,),       y_label='Room temperature in °C', shift=273.15),
-    SubPlot(features=(u_hp_rel,),     y_label='Modulation hp', step=True),
-    SubPlot(features=(t_amb,), y_label='Ambient temperature in °C', normalize=False, shift=273.15),
-    SubPlot(features=(power_fan, power_hp, power_ec), y_label='el. Power in W', normalize=False),
-    SubPlot(features=(price_el,), y_label='el. Price in €', normalize=False),
-    size=(8, 8)
-)
-
-
-# ---------------------------------------------- FMU -----------------------------------------------
-
-hydronic_hp_system = BOPTESTSystem(model=hydronic_hp_model,
-                                   step_size=60 * 15,
-                                   url='http://134.130.166.203:5000',
-                                   name='simple_test')
-
-# ------------------------------------------ Simulator ------------------------------------------
-
-hydronic_hp_Simulator = Simulator(
-    system=hydronic_hp_system,
-    plotter=plotter_hydronic
-)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/boptest_hydronic_heat_pump_hp_ctrl/2_generate_data_boptest.py

-from Examples.boptest_hydronic_heat_pump_hp_ctrl.configuration_boptest import *
-
-# PID controller for the HP
-HP_PID = PID(
-    y=TAirRoom,
-    u=u_hp_rel,
-    dt=60*30,
-    Kp=0.05,
-    Ti=300,
-    Td=0,
-)
-
-
-TAirRoom.mode = 'random'
-
-hydronic_hp_Simulator.plotter.sub_folder_name = 'pid'
-hydronic_hp_system.setup(start_time=0,warmup_period=7*24*3600)
-hydronic_hp_system.activate_ctrl_layer({'oveHeaPumY_activate':1})
-hydronic_hp_DataContainer = hydronic_hp_Simulator.run(controllers=(HP_PID,),
-                                                      duration=60 * 60 * 24 * 7)
-hydronic_hp_DataContainers = hydronic_hp_DataContainer.split_container(training_share=0.7,testing_share=0.15,validating_share=0.15)
-
-hydronic_hp_DataHandler = DataHandler(containers=hydronic_hp_DataContainers)
-hydronic_hp_DataHandler.save(pid_DataHandler_name, override=True)

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/boptest_hydronic_heat_pump_hp_ctrl/3_train_anns_boptest.py

-from Examples.boptest_hydronic_heat_pump_hp_ctrl.configuration_boptest import *
-hydronic_DataHandler = load_DataHandler(pid_DataHandler_name)
-containers = load_DataHandler(pid_DataHandler_name).containers
-
-# ----------------------------------- TAirRoom NetworkTrainer -----------------------------------
-
-TAirRoom_NetworkTrainer = NetworkTrainer(
-    inputs=(
-        # disturbances
-        Input(feature=t_amb, lag=2),
-        Input(feature=rad_dir, lag=1),
-
-        Input(feature=daily_sin, lag=2),
-        Input(feature=daily_cos, lag=2),
-        Input(feature=weekly_sin, lag=2),
-        Input(feature=weekly_cos, lag=2),
-
-        # Stellgrößen
-        Input(feature=u_hp_rel, lag=6),
-
-        # T air room
-        Input(feature=TAirRoom, lag=6),
-
-    ),
-    output=Output(feature=d_TAirRoom),
-    step_size=60*30
-)
-
-TAirRoom_NetworkTrainer.load_data(containers=containers)
-TAirRoom_NetworkTrainer.tune(layer_range=(1,), neurons_range=(8, 16), trials=10, epochs=1000, batch_size=100)
-TAirRoom_NetworkTrainer.save(d_TAir_NetworkTrainer_name, override=True)
-

01_Input/Masterarbeit Konrad Beeser/ann_ddmpc/Examples/boptest_hydronic_heat_pump_hp_ctrl/4_mpc_online_boptest.py

-from Examples.boptest_hydronic_heat_pump_hp_ctrl.configuration_boptest import *
-
-# load ann1
-TAirRoom_NetworkTrainer = load_NetworkTrainer(d_TAir_NetworkTrainer_name)
-TAirRoom_NetworkTrainer.load_best_ann(loc=0)
-
-TAirRoom.mode = 'economic'
-
-hydronic_MPC = ModelPredictive(
-    dt=60*30,
-    N=32,
-    objectives=(
-        Objective(feature=TAirRoom,     weight=50,                k_0_offset=1),
-        Objective(feature=u_hp_rel,   weight=1,     norm=1,       k_N_offset=-1, cost_order=CostOrder.linear),
-        Objective(feature=d_u_hp_rel, weight=0.01,  norm=1 ,      k_N_offset=-1),
-    ),
-    constraints=(
-        Constraint(features=(u_hp_rel,),  lb=0,              ub=1,               k_N_offset=-1),
-        ),
-
-    model=hydronic_hp_model,
-    disturbances_df=pd.DataFrame(),
-    network_trainers=(TAirRoom_NetworkTrainer,),
-)
-
-hydronic_hp_Simulator.plotter.sub_folder_name = 'online_mpc'
-hydronic_hp_system.setup(start_time=0,warmup_period=7*24*3600)
-hydronic_hp_system.activate_ctrl_layer({'oveHeaPumY_activate':1})
-
-hydronic_hp_MPC_DataHandler = DataHandler()
-
-# online mpc loop
-for repetition in range(1):
-
-    # run mpc and add to DataHandler
-    hydronic_DataContainer = hydronic_hp_Simulator.run(
-        controllers=(hydronic_MPC,),
-        duration=60 * 60 * 24 * 3,
-    )
-
-    # split data into training, testing and validating
-    hydronic_DataContainers = hydronic_DataContainer.split_container(testing_share=0.15,training_share=0.7,validating_share=0.15)
-    hydronic_hp_MPC_DataHandler.add_containers(hydronic_DataContainers)
-
-    # online learning
-    hydronic_hp_Simulator.retrain_anns(
-        containers=hydronic_DataContainers,
-        network_trainer=TAirRoom_NetworkTrainer,
-        epochs=500,
-        batch_size=250,
-        clear_data=False,
-    )
-    hydronic_MPC.reinitialize_mpc = True
-
-
-hydronic_hp_MPC_DataHandler.save(MPC_DataHandler_name, override=True)