def test_uguess_outside_bounds(self):
"""Test initial guess for u outsidethe feasible region
(should be adjusted)"""
# Bounds
ubounds = [(0., 4000.)]
xbounds = [(21., 23.), (0., 50.), (0., 50.)]
# Initial u
uguess = np.full((self.inp.index.size - 1, 1), 9999.)
# Initial state
x0 = [21., 21., 21.]
# Instantiate multiple shooting
optc = mshoot.MShoot(cfun=cfun)
# Optimize
udf, xdf = optc.optimize(model=self.model,
inp=self.inp,
free=['qhvac'],
ubounds=ubounds,
xbounds=xbounds,
x0=x0,
uguess=uguess,
ynominal=[20., 20., 20., 4000.],
join=1,
maxiter=2)
# Heating power should not exceed 4000 W
self.assertTrue((udf['qhvac'] < 4000.).all())
def test_normalized_uy(self):
"""Test the right solution is achieved with nominal u and y"""
# Bounds
ubounds = [(0., 4000.)]
xbounds = [(21., 23.), (0., 50.), (0., 50.)]
# Initial state
x0 = [21., 21., 21.]
# Instantiate multiple shooting
optc = mshoot.MShoot(cfun=cfun)
# Optimize
udf, xdf = optc.optimize(model=self.model,
inp=self.inp,
free=['qhvac'],
ubounds=ubounds,
xbounds=xbounds,
x0=x0,
uguess=None,
unominal=[4000.],
ynominal=[20., 20., 20., 4000.],
join=3)
# Final room temperature should be equal to lower bound
self.assertLess(np.abs(xdf.iloc[-1][0] - xbounds[0][0]), 1e-4,
"Final state different than lower bound")
# Heating should be provided
self.assertGreater(udf['qhvac'].sum(), 0, "qhvac should be positive")
def test_infeasible_x0(self):
"""Test if good solution is found even with infeasible x0"""
# Bounds
ubounds = [(0., 4000.)]
xbounds = [(21., 23.), (0., 50.), (0., 50.)]
# Initial state
x0 = [15., 21., 21.] # <- infeasible x0
# Instantiate multiple shooting
optc = mshoot.MShoot(cfun=cfun)
# Optimize
udf, xdf = optc.optimize(model=self.model,
inp=self.inp,
free=['qhvac'],
ubounds=ubounds,
xbounds=xbounds,
ynominal=[20., 20., 20., 4000.],
x0=x0,
uguess=None,
join=3)
# Final room temperature should be equal to lower bound
self.assertLess(np.abs(xdf.iloc[-1][0] - xbounds[0][0]), 1e-4,
"Final state different than lower bound")
# Heating should be provided
self.assertGreater(udf['qhvac'].sum(), 0, "qhvac should be positive")
def test_optimize_step_bounds(self):
"""Test preheating (step increase in state lower bound)"""
# Bounds
Tlo = np.full(self.inp.index.size, 16.)
Tlo[-2:] = 19.
plt.plot(Tlo)
ubounds = [(0., 5000.)]
xbounds = [(Tlo, 30.), (0., 50.), (0., 50.)]
# Initial state
x0 = [16., 16., 16.]
# Instantiate multiple shooting
optc = mshoot.MShoot(cfun=cfun)
# Optimize
udf, xdf = optc.optimize(model=self.model,
inp=self.inp,
free=['qhvac'],
ubounds=ubounds,
xbounds=xbounds,
ynominal=[20., 20., 20., 4000.],
x0=x0,
uguess=None,
join=1)
# Final room temperature should be equal to lower bound
self.assertLess(np.abs(xdf.iloc[-1][0] - xbounds[0][0][-1]), 1e-4,
"Final state different than lower bound")
# Room temperature should be always higher or equal to the lower bound
self.assertTrue((xdf['x0'].values >= Tlo).all(),
"Temperature bound violated")
# Heating should be provided
self.assertGreater(udf['qhvac'].sum(), 0, "qhvac should be positive")
def test_optimal_control(self):
print('** Test FMU optimal control **')
# Inputs
t = np.arange(0, 3600 * 10, 3600)
inp = pd.DataFrame(index=pd.Index(t, name='time'), columns=['q', 'Tout'])
inp['q'] = np.full(t.size, 0)
inp['Tout'] = np.full(t.size, 273.15)
# Bounds
ubounds = [(0., 4000.)]
xbounds = [(293.15, 296.15)]
# Initial state
x0 = [293.65]
# Optimization
mpc = mshoot.MShoot(cfun=cfun)
udf, xdf = mpc.optimize(
model=self.model,
inp=inp,
free=['q'],
ubounds=ubounds,
xbounds=xbounds,
x0=x0,
uguess=None,
unominal=[4000.],
ynominal=[4000., 293.15],
)
# ax = udf.plot(title='udf')
# ax.set_ylim(0, 4000)
# ax = xdf.plot(title='xdf')
# ax.set_ylim(293.15, 296.15)
# plt.show()
self.assertLess(abs(xdf['x0'].iloc[-1] - 293.15), 0.01)
self.assertLess(abs(udf['q'].iloc[-1] - 2000.), 0.01)
# Inputs
t = np.arange(0, 3600 * 10, 3600)
inp = pd.DataFrame(index=pd.Index(t, name='time'), columns=['q', 'Tout'])
inp['q'] = np.full(t.size, 0.)
inp['Tout'] = np.full(t.size, 273.15)
# Bounds
ubounds = [(0., 4000.)]
xbounds = [(293.15, 296.15), (293.15, 296.15)]
# Initial state
x0 = [293.65, 293.35]
# Optimization
mpc = mshoot.MShoot(cfun=cfun)
udf, xdf = mpc.optimize(
model=get_model(),
inp=inp,
free=['q'],
ubounds=ubounds,
xbounds=xbounds,
x0=x0,
uguess=None,
ynominal=[4000, 295.],
join=1,
maxiter=30
)
# Show results
print(udf)
