def test_optimisation_recovers_optimal_position(self):
problem = self.default_problem()
farm = problem.parameters.tidal_farm
solver = DummySolver(problem)
functional = PowerFunctional(problem)
control = TurbineFarmControl(farm)
rf_params = ReducedFunctionalParameters()
rf_params.automatic_scaling = 5.
rf = ReducedFunctional(functional, control, solver, rf_params)
m0 = farm.control_array
p = numpy.random.rand(len(m0))
minconv = helpers.test_gradient_array(rf.__call__,
rf.derivative,
m0,
seed=0.005,
perturbation_direction=p)
assert minconv > 1.9
bounds = [[Constant(0), Constant(0)], [Constant(3000), Constant(1000)]]
maximize(rf, bounds=bounds, method="SLSQP")
m = farm._parameters["position"][0]
log(INFO, "Solution of the primal variables: m=" + repr(m) + "\n")
assert abs(m[0] - 1500) < 40
assert abs(m[1] - 500) < 0.4
def test_optimisation_recovers_optimal_position(self):
problem = self.default_problem()
farm = problem.parameters.tidal_farm
solver = DummySolver(problem)
functional = PowerFunctional(problem)
control = TurbineFarmControl(farm)
rf_params = ReducedFunctionalParameters()
rf_params.automatic_scaling = 5.
rf = ReducedFunctional(functional, control, solver, rf_params)
m0 = farm.control_array
p = numpy.random.rand(len(m0))
minconv = helpers.test_gradient_array(rf.__call__, rf.derivative, m0, seed=0.005,
perturbation_direction=p)
assert minconv > 1.9
bounds = [[Constant(0), Constant(0)], [Constant(3000), Constant(1000)]]
maximize(rf, bounds=bounds, method="SLSQP")
m = farm._parameters["position"][0]
log(INFO, "Solution of the primal variables: m=" + repr(m) + "\n")
assert abs(m[0]-1500) < 40
assert abs(m[1]-500) < 0.4
# Compute the derivatives with respect to the turbine position
for n in range(len(config.params["turbine_pos"])):
for var in ('turbine_pos_x', 'turbine_pos_y'):
tfd = turbines(derivative_index_selector=n,
derivative_var_selector=var)
dj.append(2 * v.inner(tfd.vector()))
dj = numpy.array(dj)
return j, dj
else:
return j, None
j = lambda m, forward_only=False: j_and_dj(m, forward_only)[0]
dj = lambda m, forget: j_and_dj(m, forward_only=False)[1]
# run the taylor remainder test
config = default_config()
m0 = ReducedFunctional(config).initial_control()
# We set the perturbation_direction with a constant seed, so that it is consistent in a parallel environment.
p = numpy.random.rand(len(m0))
# Run with a functional that does not depend on m directly
minconv = test_gradient_array(j, dj, m0, 0.001, perturbation_direction=p)
if minconv < 1.99:
info_red("The turbine test failed")
sys.exit(1)
info_green("Test passed")
config.params["turbine_x"] = 800
config.params["turbine_y"] = 800
config.params["controls"] = ['turbine_pos']
config.params["initial_condition"] = BumpInitialCondition(config)
config.params["automatic_scaling"] = True
return config
config = default_config()
rf = ReducedFunctional(config, forward_model=mini_model_solve)
m0 = rf.initial_control()
config.info()
p = numpy.random.rand(len(m0))
minconv = test_gradient_array(rf.j, rf.dj, m0, seed=0.005, perturbation_direction=p)
if minconv < 1.9:
info_red("The gradient taylor remainder test failed.")
sys.exit(1)
# If this option does not produce any ipopt outputs, delete the ipopt.opt file
g = lambda m: []
dg = lambda m: []
bounds = [[Constant(0), Constant(0)], [Constant(3000), Constant(1000)]]
m = maximize(rf, bounds = bounds, method = "SLSQP")
info("Solution of the primal variables: m=" + repr(m) + "\n")
exit_code = 1
# Compute the derivatives with respect to the turbine position
for n in range(len(config.params["turbine_pos"])):
for var in ('turbine_pos_x', 'turbine_pos_y'):
tfd = turbines(derivative_index_selector=n, derivative_var_selector=var)
dj.append( 2 * v.inner(tfd.vector()) )
dj = numpy.array(dj)
return j, dj
else:
return j, None
j = lambda m, forward_only = False: j_and_dj(m, forward_only)[0]
dj = lambda m, forget: j_and_dj(m, forward_only = False)[1]
# run the taylor remainder test
config = default_config()
m0 = ReducedFunctional(config).initial_control()
# We set the perturbation_direction with a constant seed, so that it is consistent in a parallel environment.
p = numpy.random.rand(len(m0))
# Run with a functional that does not depend on m directly
minconv = test_gradient_array(j, dj, m0, 0.001, perturbation_direction=p)
if minconv < 1.99:
info_red("The turbine test failed")
sys.exit(1)
info_green("Test passed")
