def env():
net = Network.load('net/net_test.yaml')
env = gym.make('openmodelica_microgrid_gym:ModelicaEnv_test-v1',
viz_mode=None,
model_path='omg_grid/test.fmu',
max_episode_steps=100,
net=net)
return env, net.in_vars(), flatten(net.out_vars())
def test_load():
Network.load('net/net_valid.yaml')
assert True
def test_load_dup_inputs():
with pytest.raises(ValueError):
Network.load('net/net_dupinputs.yaml')
from matplotlib import pyplot as plt
from openmodelica_microgrid_gym import Runner
from openmodelica_microgrid_gym.agents import StaticControlAgent
from openmodelica_microgrid_gym.aux_ctl import PI_params, DroopParams, MultiPhaseDQ0PIPIController, \
MultiPhaseDQCurrentController, InverseDroopParams, PLLParams
from openmodelica_microgrid_gym.env import PlotTmpl
from openmodelica_microgrid_gym.net import Network
# Simulation definitions
max_episode_steps = 3000 # number of simulation steps per episode
num_episodes = 1 # number of simulation episodes
# (here, only 1 episode makes sense since simulation conditions don't change in this example)
DroopGain = 40000.0 # virtual droop gain for active power / W/Hz
QDroopGain = 1000.0 # virtual droop gain for reactive power / VAR/V
net = Network.load('../net/net.yaml')
delta_t = net.ts # simulation time step size / s
freq_nom = net.freq_nom # nominal grid frequency / Hz
v_nom = net.v_nom # nominal grid voltage / V
v_DC = net[
'inverter1'].v_DC # DC-link voltage / V; will be set as model parameter in the FMU
i_lim = net['inverter1'].i_lim # inverter current limit / A
i_nom = net['inverter1'].i_nom # nominal inverter current / A
logging.basicConfig()
def load_step(t, gain):
"""
Doubles the load parameters
:param t:
from openmodelica_microgrid_gym.util import FullHistory
include_simulate = True
show_plots = False
balanced_load = False
do_measurement = False
save_results = True
current_directory = os.getcwd()
save_folder = os.path.join(current_directory, r'4Dtest1')
os.makedirs(save_folder, exist_ok=True)
np.random.seed(1)
# Simulation definitions
net = Network.load('../../net/net_single-inv-Paper_Loadstep.yaml')
delta_t = 1e-4 # simulation time step size / s
undersample = 1
max_episode_steps = 2000 # number of simulation steps per episode
num_episodes = 1 # number of simulation episodes (i.e. SafeOpt iterations)
n_MC = 1 # number of Monte-Carlo samples for simulation - samples device parameters (e.g. L,R, noise) from
v_DC = 600 # DC-link voltage / V; will be set as model parameter in the FMU
nomFreq = 60 # nominal grid frequency / Hz
nomVoltPeak = 169.7 # 230 * 1.414 # nominal grid voltage / V
iLimit = 16 # inverter current limit / A
iNominal = 12 # nominal inverter current / A
vNominal = 190 # nominal inverter current / A
vLimit = vNominal * 1.5 # inverter current limit / A
funnelFactor = 0.02
vFunnel = np.array([
vNominal * funnelFactor, vNominal * funnelFactor, vNominal * funnelFactor
from openmodelica_microgrid_gym import Runner
from openmodelica_microgrid_gym.agents import SafeOptAgent
from openmodelica_microgrid_gym.agents.util import MutableFloat
from openmodelica_microgrid_gym.aux_ctl import PI_params, DroopParams, MultiPhaseDQ0PIPIController
from openmodelica_microgrid_gym.env import PlotTmpl
from openmodelica_microgrid_gym.net import Network
from openmodelica_microgrid_gym.util import dq0_to_abc, nested_map, FullHistory
# Simulation definitions
max_episode_steps = 300 # number of simulation steps per episode
num_episodes = 2 # number of simulation episodes (i.e. SafeOpt iterations)
mu = 2 # factor for barrier function (see below)
DroopGain = 40000.0 # virtual droop gain for active power / W/Hz
QDroopGain = 1000.0 # virtual droop gain for reactive power / VAR/V
net = Network.load('../net/net_single-inv-volt.yaml')
i_lim = net['inverter1'].i_lim # inverter current limit / A
i_nom = net['inverter1'].i_nom # nominal inverter current / A
# Files saves results and resulting plots to the folder saves_VI_control_safeopt in the current directory
current_directory = os.getcwd()
save_folder = os.path.join(current_directory, r'saves_VI_control_safeopt')
os.makedirs(save_folder, exist_ok=True)
class Reward:
def __init__(self):
self._idx = None
def set_idx(self, obs):
if self._idx is None:
# - Ki: 1D example: Only the integral gain Ki of the PI controller is adjusted
# - Kpi: 2D example: Kp and Ki are adjusted simultaneously
adjust = 'Kpi'
# Check if really only one simulation scenario was selected
if adjust not in {'Kp', 'Ki', 'Kpi'}:
raise ValueError(
"Please set 'adjust' to one of the following values: 'Kp', 'Ki', 'Kpi'"
)
# Simulation definitions
max_episode_steps = 600 # number of simulation steps per episode
num_episodes = 10 # number of simulation episodes (i.e. SafeOpt iterations)
mu = 2 # factor for barrier function (see below)
net = Network.load('../net/net_single-inv-curr.yaml')
i_lim = net['inverter1'].i_lim # inverter current limit / A
i_nom = net['inverter1'].i_nom # nominal inverter current / A
i_ref = np.array(net['inverter1'].i_ref
) # exemplary set point i.e. id = 15, iq = 0, i0 = 0 / A
class Reward:
def __init__(self):
self._idx = None
def set_idx(self, obs):
if self._idx is None:
self._idx = nested_map(lambda n: obs.index(n),
[[f'lc1.inductor{k}.i'
for k in '123'], 'master.phase',
include_simulate = True
show_plots = True
balanced_load = True
do_measurement = False
save_results = False
# Files saves results and resulting plots to the folder saves_VI_control_safeopt in the current directory
current_directory = os.getcwd()
save_folder = os.path.join(current_directory, r'../1Test')
os.makedirs(save_folder, exist_ok=True)
np.random.seed(1)
# Simulation definitions
net = Network.load('../../net/net_single-inv-curr_Paper_SC.yaml')
delta_t = 1e-4 # simulation time step size / s
undersample = 1 # undersampling of controller
max_episode_steps = 1000 # number of simulation steps per episode
num_episodes = 1 # number of simulation episodes (i.e. SafeOpt iterations)
n_MC = 1 # number of Monte-Carlo samples for simulation - samples device parameters (e.g. L,R, noise) from
iLimit = 16 # inverter current limit / A
iNominal = 12 # nominal inverter current / A
mu = 80 # factor for barrier function (see below)
i_ref1 = np.array([10, 0, 0]) # exemplary set point i.e. id = 10, iq = 0, i0 = 0 / A
i_ref2 = np.array([5, 0, 0]) # exemplary set point i.e. id = 15, iq = 0, i0 = 0 / A
# plant
L = 2.3e-3 # / H
R = 400e-3 # / Ohm