def test_rp_example3():
m = rp_example3()
with pytest.raises(RuntimeError):
petsc.petsc_dae_by_time_element(
m,
time=m.time,
)
def test_petsc_read_trajectory():
"""
Check that the PETSc DAE solver works.
"""
m, y1, y2, y3, y4, y5, y6 = dae_with_non_time_indexed_constraint()
m.scaling_factor = pyo.Suffix(direction=pyo.Suffix.EXPORT)
m.scaling_factor[m.y[180, 1]] = 10 # make sure unscale works
petsc.petsc_dae_by_time_element(
m,
time=m.t,
ts_options={
"--ts_type": "cn", # Crank–Nicolson
"--ts_adapt_type": "basic",
"--ts_dt": 0.01,
"--ts_save_trajectory": 1,
"--ts_trajectory_type": "visualization",
},
vars_stub="tj_random_junk_123",
trajectory_save_prefix="tj_random_junk_123",
)
assert pytest.approx(y1, rel=1e-3) == pyo.value(m.y[m.t.last(), 1])
assert pytest.approx(y2, rel=1e-3) == pyo.value(m.y[m.t.last(), 2])
assert pytest.approx(y3, rel=1e-3) == pyo.value(m.y[m.t.last(), 3])
assert pytest.approx(y4, rel=1e-3) == pyo.value(m.y[m.t.last(), 4])
assert pytest.approx(y5, rel=1e-3) == pyo.value(m.y[m.t.last(), 5])
assert pytest.approx(y6, rel=1e-3) == pyo.value(m.y[m.t.last(), 6])
tj = petsc.PetscTrajectory(json="tj_random_junk_123_1.json.gz")
assert tj.get_dt()[0] == pytest.approx(
0.01) # if small enough shouldn't be cut
assert tj.get_vec(m.y[180, 1])[-1] == pytest.approx(y1, rel=1e-3)
assert tj.get_vec("_time")[-1] == pytest.approx(180)
times = np.linspace(0, 180, 181)
vecs = tj.interpolate_vecs(times)
assert vecs[str(m.y[180, 1])][180] == pytest.approx(y1, rel=1e-3)
assert vecs["_time"][180] == pytest.approx(180)
tj.to_json("some_testy_json.json")
with open("some_testy_json.json", "r") as fp:
vecs = json.load(fp)
assert vecs[str(m.y[180, 1])][-1] == pytest.approx(y1, rel=1e-3)
assert vecs["_time"][-1] == pytest.approx(180)
os.remove("some_testy_json.json")
tj.to_json("some_testy_json.json.gz")
tj2 = petsc.PetscTrajectory(json="some_testy_json.json.gz")
assert tj2.vecs[str(m.y[180, 1])][-1] == pytest.approx(y1, rel=1e-3)
assert tj2.vecs["_time"][-1] == pytest.approx(180)
os.remove("some_testy_json.json.gz")
tj2 = petsc.PetscTrajectory(vecs=vecs)
assert tj2.vecs[str(m.y[180, 1])][-1] == pytest.approx(y1, rel=1e-3)
assert tj2.vecs["_time"][-1] == pytest.approx(180)
def test_rp_example4():
m = rp_example4()
petsc.petsc_dae_by_time_element(m,
time=m.time,
ts_options={
"--ts_dt": 1,
"--ts_adapt_type": "none",
})
assert pyo.value(m.u[10]) == pytest.approx(398)
assert pyo.value(m.x[10]) == pytest.approx(20)
def test_car():
m = car_example()
m.a.fix(1.0)
m.tf.fix(16.56)
# solve
petsc.petsc_dae_by_time_element(
m,
time=m.tau,
ts_options={
"--ts_type": "cn", # Crank–Nicolson
"--ts_adapt_type": "basic",
"--ts_dt": 0.01,
},
)
assert pyo.value(m.x[1]) == pytest.approx(131.273, rel=1e-2)
def test_petsc_dae_idaes_solve():
"""
Check that the PETSc DAE solver works.
"""
m, y1, y2, y3, y4, y5, y6 = dae()
petsc.petsc_dae_by_time_element(
m,
time=m.t,
ts_options={
"--ts_type": "cn", # Crank–Nicolson
"--ts_adapt_type": "basic",
"--ts_dt": 0.1,
},
)
assert pytest.approx(y1, rel=1e-3) == pyo.value(m.y[m.t.last(), 1])
assert pytest.approx(y2, rel=1e-3) == pyo.value(m.y[m.t.last(), 2])
assert pytest.approx(y3, rel=1e-3) == pyo.value(m.y[m.t.last(), 3])
assert pytest.approx(y4, rel=1e-3) == pyo.value(m.y[m.t.last(), 4])
assert pytest.approx(y5, rel=1e-3) == pyo.value(m.y[m.t.last(), 5])
assert pytest.approx(y6, rel=1e-3) == pyo.value(m.y6[m.t.last()])
def test_petsc_with_pid_model():
m = create_model(
time_set=[0, 24],
nfe=1,
calc_integ=True,
)
# time_var will be an explicit time variable we can use in constraints.
m.fs.time_var = pyo.Var(m.fs.time)
# We'll add a constraint to calculate the inlet pressure based on time,
# so we need to unfix pressure.
m.fs.valve_1.control_volume.properties_in[0].pressure.unfix()
m.fs.valve_1.control_volume.properties_in[24].pressure.unfix()
# The solver will directly set the time variable for the DAE solve, but
# solving the initial conditions is just a system of nonlinear equations,
# so we need to fix the initial time.
m.fs.time_var[0].fix(m.fs.time.first())
# We could break up the time domain and solve this in pieces, but creative use
# of min and max will let us create the ramping function we want.
# From 10s to 12s ramp inlet pressure from 500,000 Pa to 600,000 Pa
@m.fs.Constraint(m.fs.time)
def inlet_pressure_eqn(b, t):
return b.valve_1.control_volume.properties_in[t].pressure == \
smooth_min(600000, smooth_max(500000, 50000*(b.time_var[t] - 10) + 500000))
res = petsc.petsc_dae_by_time_element(
m,
time=m.fs.time,
timevar=m.fs.time_var,
ts_options={
"--ts_type": "beuler",
"--ts_dt": 0.1,
"--ts_monitor": "", # set initial step to 0.1
"--ts_save_trajectory": 1,
"--ts_trajectory_type": "visualization",
},
vars_stub="tj_vars")
# read the trajectory data, and make it easy by interpolating a time point
# every second
tj = petsc.PetscTrajectory(stub="tj_vars", delete_on_read=True)
vecs = tj.interpolate_vecs(np.linspace(0, 24, 25))
# For more details about the problem behavior see the PETSc examples in the
# examples repo.
# make sure the inlet pressure is initially 5e5 Pa
assert (pyo.value(vecs[str(
m.fs.valve_1.control_volume.properties_in[24].pressure)][5]) ==
pytest.approx(5e5))
# make sure the inlet pressure ramped up to 6e5 Pa
assert (pyo.value(vecs[str(
m.fs.valve_1.control_volume.properties_in[24].pressure)][20]) ==
pytest.approx(6e5))
# make sure after the controller comes on the presure goes to the set point
assert (pyo.value(vecs[str(
m.fs.tank.control_volume.properties_out[24].pressure)][9]) ==
pytest.approx(3e5))
# make sure after ramping inlet pressure the tank pressure gets back to the
# setpoint
assert (pyo.value(vecs[str(
m.fs.tank.control_volume.properties_out[24].pressure)][22]) ==
pytest.approx(3e5))