def run(fname):
circuit = FileOpen(fname).open()
options = PowerFlowOptions(
solver_type=SolverType.NR,
retry_with_other_methods=False,
verbose=False,
initialize_with_existing_solution=False,
tolerance=1e-4,
max_iter=5,
max_outer_loop_iter=10,
control_q=ReactivePowerControlMode.NoControl,
control_taps=TapsControlMode.NoControl,
multi_core=False,
dispatch_storage=False,
control_p=False,
apply_temperature_correction=False,
branch_impedance_tolerance_mode=BranchImpedanceMode.Specified,
q_steepness_factor=30,
distributed_slack=False,
ignore_single_node_islands=False,
correction_parameter=1e-4)
driver = SampledTimeSeries(grid=circuit,
options=options,
number_of_steps=100)
driver.run()
# compose the train set
nc = circuit.compile_time_series()
Pbus = nc.get_power_injections().real.T
model = KNeighborsRegressor(n_neighbors=2)
model.fit(driver.results.S.real, np.abs(driver.results.voltage))
V_pred = model.predict(Pbus)
model = KNeighborsRegressor(n_neighbors=2)
model.fit(driver.results.S.real, driver.results.Sbranch.real)
Sbr_pred = model.predict(Pbus)
return V_pred, Sbr_pred
def run(fname):
circuit = FileOpen(fname).open()
pf_options = PowerFlowOptions(
solver_type=SolverType.NR,
retry_with_other_methods=False,
verbose=False,
initialize_with_existing_solution=False,
tolerance=1e-6,
max_iter=5,
max_outer_loop_iter=10,
control_q=ReactivePowerControlMode.NoControl,
control_taps=TapsControlMode.NoControl,
multi_core=False,
dispatch_storage=False,
control_p=False,
apply_temperature_correction=False,
branch_impedance_tolerance_mode=BranchImpedanceMode.Specified,
q_steepness_factor=30,
distributed_slack=False,
ignore_single_node_islands=False,
correction_parameter=1e-4)
nc = circuit.compile_time_series()
ts_driver = TimeSeries(circuit, pf_options)
ts_driver.run()
ptdf_driver = PtdfTimeSeries(circuit, pf_options, power_delta=10)
ptdf_driver.run()
npoints = int(len(circuit.time_profile) * 1)
lhs_driver = LatinHypercubeSampling(circuit,
pf_options,
sampling_points=npoints)
lhs_driver.run()
P = nc.get_power_injections().real.T
Q = nc.get_power_injections().imag.T
Pbr_ts = ts_driver.results.Sbranch.real
Pbr_ptdf = ptdf_driver.results.Sbranch.real
P_lhs = lhs_driver.results.S_points.real
Q_lhs = lhs_driver.results.S_points.imag
Pbr_lhs = lhs_driver.results.Sbr_points.real
# KNN
n_neighbors = 3
model = neighbors.KNeighborsRegressor(n_neighbors)
# model.fit(P[:40], Pbr_ts[:40])
# model.fit(P_lhs, Pbr_lhs) # just the LHS for training
# X = np.r_[np.c_[P_lhs, Q], np.c_[P, Q]]
# Y = np.r_[Pbr_lhs, Pbr_ts]
X = np.c_[P, Q][:60]
Y = Pbr_ts[:60]
model.fit(X, Y) # LHS + TS for training ("dreaming")
Pbr_knn = model.predict(np.c_[P, Q])
fig = plt.figure(figsize=(12, 8))
ax = fig.add_subplot(111)
i = 10 # branch index
ax.plot(Pbr_ts[i, :], label='Real flow', linewidth=5, c='orange')
ax.plot(Pbr_ptdf[i, :], label='PTDF', c='b', linestyle='--')
ax.plot(Pbr_knn[i, :], label='KNN', c='k', linestyle=':')
ax.set_xlabel('Time')
ax.set_ylabel('MW')
fig.legend()
plt.show()
