def test_clear_dir_k_aug(self):
m = simple_model_1()
sens = SensitivityInterface(m, clone_model=False)
k_aug = K_augInterface()
opt_ipopt.solve(m, tee=True)
m.ptb = pyo.Param(mutable=True, initialize=1.5)
cwd = os.getcwd()
dir_contents = os.listdir(cwd)
sens_param = [m.p]
sens.setup_sensitivity(sens_param)
k_aug.k_aug(m, tee=True)
# We are back in our working directory
self.assertEqual(cwd, os.getcwd())
# The contents of this directory have not changed
self.assertEqual(dir_contents, os.listdir(cwd))
# In particular, the following files do not exist
self.assertFalse(os.path.exists("dsdp_in_.in"))
self.assertFalse(os.path.exists("conorder.txt"))
self.assertFalse(os.path.exists("timings_k_aug_dsdp.txt"))
# But they have been transferred to our k_aug interface's data
# dict as strings.
self.assertIsInstance(k_aug.data["dsdp_in_.in"], str)
self.assertIsInstance(k_aug.data["conorder.txt"], str)
self.assertIsInstance(k_aug.data["timings_k_aug_dsdp.txt"], str)
def test_clear_dir_dot_sens(self):
m = simple_model_1()
sens = SensitivityInterface(m, clone_model=False)
k_aug = K_augInterface()
opt_ipopt.solve(m, tee=True)
m.ptb = pyo.Param(mutable=True, initialize=1.5)
cwd = os.getcwd()
dir_contents = os.listdir(cwd)
sens_param = [m.p]
sens.setup_sensitivity(sens_param)
# Call k_aug
k_aug.k_aug(m, tee=True)
self.assertIsInstance(k_aug.data["dsdp_in_.in"], str)
sens.perturb_parameters([m.ptb])
# Call dot_sens. In the process, we re-write dsdp_in_.in
k_aug.dot_sens(m, tee=True)
# Make sure we get the values we expect. This problem is easy enough
# to solve by hand:
# x = [1, 1, -2] = [v1, v2, dual]
# Sensitivity system:
# | 2 -2 1 |
# |-2 2 1 | dx/dp = -[dL/dxdp, dc/dp]^T = -[0, 0, -1]^T
# | 1 1 0 |
# => dx/dp = [0.5, 0.5, 0]^T
# Here, dp = [0.5]
# => dx = [0.25, 0.25, 0]^T
# => x_new = [1.25, 1.25, -2]
self.assertAlmostEqual(m.v1.value, 1.25, 7)
self.assertAlmostEqual(m.v2.value, 1.25, 7)
# We are back in our working directory
self.assertEqual(cwd, os.getcwd())
# The contents of this directory have not changed
self.assertEqual(dir_contents, os.listdir(cwd))
self.assertFalse(os.path.exists("dsdp_in_.in"))
self.assertFalse(os.path.exists("delta_p.out"))
self.assertFalse(os.path.exists("dot_out.out"))
self.assertFalse(os.path.exists("timings_dot_driver_dsdp.txt"))
# And we have saved strings of the file contents.
self.assertIsInstance(k_aug.data["dsdp_in_.in"], str)
self.assertIsInstance(k_aug.data["delta_p.out"], str)
self.assertIsInstance(k_aug.data["dot_out.out"], str)
self.assertIsInstance(k_aug.data["timings_dot_driver_dsdp.txt"], str)
def get_dsdp(model, theta_names, theta, tee=False):
"""This function calculates gradient vector of the variables
with respect to the parameters (theta_names).
e.g) min f: p1*x1+ p2*(x2^2) + p1*p2
s.t c1: x1 + x2 = p1
c2: x2 + x3 = p2
0 <= x1, x2, x3 <= 10
p1 = 10
p2 = 5
the function retuns dx/dp and dp/dp, and column orders.
The following terms are used to define the output dimensions:
Ncon = number of constraints
Nvar = number of variables (Nx + Ntheta)
Nx = number of decision (primal) variables
Ntheta = number of uncertain parameters.
Parameters
----------
model: Pyomo ConcreteModel
model should include an objective function
theta_names: list of strings
List of Var names
theta: dict
Estimated parameters e.g) from parmest
tee: bool, optional
Indicates that ef solver output should be teed
Returns
-------
dsdp: scipy.sparse.csr.csr_matrix
Ntheta by Nvar size sparse matrix. A Jacobian matrix of the
(decision variables, parameters) with respect to parameters
(theta_names). number of rows = len(theta_name), number of
columns = len(col)
col: list
List of variable names
"""
# Get parameters from names. In SensitivityInterface, we expect
# these to be parameters on the original model.
param_list = []
for name in theta_names:
comp = model.find_component(name)
if comp is None:
raise RuntimeError("Cannot find component %s on model" % name)
if comp.ctype is Var:
# If theta_names correspond to Vars in the model, these vars
# need to be fixed.
comp.fix()
param_list.append(comp)
sens = SensitivityInterface(model, clone_model=True)
m = sens.model_instance
# Setup model and calculate sensitivity matrix with k_aug
sens.setup_sensitivity(param_list)
k_aug = K_augInterface()
k_aug.k_aug(m, tee=tee)
# Write row and col files in a temp dir, then immediately
# read into a Python data structure.
nl_data = {}
with InTempDir():
base_fname = "col_row"
nl_file = ".".join((base_fname, "nl"))
row_file = ".".join((base_fname, "row"))
col_file = ".".join((base_fname, "col"))
m.write(nl_file, io_options={"symbolic_solver_labels": True})
for fname in [nl_file, row_file, col_file]:
with open(fname, "r") as fp:
nl_data[fname] = fp.read()
# Create more useful data structures from strings
dsdp = np.fromstring(k_aug.data["dsdp_in_.in"], sep="\n\t")
col = nl_data[col_file].strip("\n").split("\n")
row = nl_data[row_file].strip("\n").split("\n")
dsdp = dsdp.reshape((len(theta_names), int(len(dsdp) / len(theta_names))))
dsdp = dsdp[:len(theta_names), :len(col)]
col = [i for i in col if sens.get_default_block_name() not in i]
dsdp_out = np.zeros((len(theta_names), len(col)))
for i in range(len(theta_names)):
for j in range(len(col)):
if sens.get_default_block_name() not in col[j]:
dsdp_out[
i,
j] = -dsdp[i, j] # e.g) k_aug dsdp returns -dx1/dx1 = -1.0
return scipy.sparse.csr_matrix(dsdp_out), col
def get_dfds_dcds(model, theta_names, tee=False, solver_options=None):
"""This function calculates gradient vector of the objective function
and constraints with respect to the variables and parameters.
e.g) min f: p1*x1+ p2*(x2^2) + p1*p2
s.t c1: x1 + x2 = p1
c2: x2 + x3 = p2
0 <= x1, x2, x3 <= 10
p1 = 10
p2 = 5
- Variables = (x1, x2, x3, p1, p2)
- Fix p1 and p2 with estimated values
The following terms are used to define the output dimensions:
Ncon = number of constraints
Nvar = number of variables (Nx + Ntheta)
Nx = number of decision (primal) variables
Ntheta = number of uncertain parameters.
Parameters
----------
model: Pyomo ConcreteModel
model should include an objective function
theta_names: list of strings
List of Var names
tee: bool, optional
Indicates that ef solver output should be teed
solver_options: dict, optional
Provides options to the solver (also the name of an attribute)
Returns
-------
gradient_f: numpy.ndarray
Length Nvar array. A gradient vector of the objective function
with respect to the (decision variables, parameters) at the optimal
solution
gradient_c: scipy.sparse.csr.csr_matrix
Ncon by Nvar size sparse matrix. A Jacobian matrix of the
constraints with respect to the (decision variables, parameters)
at the optimal solution. Each row contains [column number,
row number, and value], column order follows variable order in col
and index starts from 1. Note that it follows k_aug.
If no constraint exists, return []
col: list
Size Nvar list of variable names
row: list
Size Ncon+1 list of constraints and objective function names.
The final element is the objective function name.
line_dic: dict
column numbers of the theta_names in the model. Index starts from 1
Raises
------
RuntimeError
When ipopt or k_aug or dotsens is not available
Exception
When ipopt fails
"""
# Create the solver plugin using the ASL interface
ipopt = SolverFactory('ipopt', solver_io='nl')
if solver_options is not None:
ipopt.options = solver_options
k_aug = SolverFactory('k_aug', solver_io='nl')
if not ipopt.available(False):
raise RuntimeError('ipopt is not available')
if not k_aug.available(False):
raise RuntimeError('k_aug is not available')
# Declare Suffixes
_add_sensitivity_suffixes(model)
# K_AUG SUFFIXES
model.dof_v = Suffix(direction=Suffix.EXPORT) #: SUFFIX FOR K_AUG
model.rh_name = Suffix(
direction=Suffix.IMPORT) #: SUFFIX FOR K_AUG AS WELL
k_aug.options["print_kkt"] = ""
results = ipopt.solve(model, tee=tee)
# Raise exception if ipopt fails
if (results.solver.status == SolverStatus.warning):
raise Exception(results.solver.Message)
for o in model.component_objects(Objective, active=True):
f_mean = value(o)
model.ipopt_zL_in.update(model.ipopt_zL_out)
model.ipopt_zU_in.update(model.ipopt_zU_out)
# run k_aug
k_aug_interface = K_augInterface(k_aug=k_aug)
k_aug_interface.k_aug(model, tee=tee) #: always call k_aug AFTER ipopt.
nl_data = {}
with InTempDir():
base_fname = "col_row"
nl_file = ".".join((base_fname, "nl"))
row_file = ".".join((base_fname, "row"))
col_file = ".".join((base_fname, "col"))
model.write(nl_file, io_options={"symbolic_solver_labels": True})
for fname in [nl_file, row_file, col_file]:
with open(fname, "r") as fp:
nl_data[fname] = fp.read()
col = nl_data[col_file].strip("\n").split("\n")
row = nl_data[row_file].strip("\n").split("\n")
# get the column numbers of "parameters"
line_dic = {name: col.index(name) for name in theta_names}
grad_f_file = os.path.join("GJH", "gradient_f_print.txt")
grad_f_string = k_aug_interface.data[grad_f_file]
gradient_f = np.fromstring(grad_f_string, sep="\n\t")
col = [
i for i in col
if SensitivityInterface.get_default_block_name() not in i
]
grad_c_file = os.path.join("GJH", "A_print.txt")
grad_c_string = k_aug_interface.data[grad_c_file]
gradient_c = np.fromstring(grad_c_string, sep="\n\t")
# Jacobian file is in "COO format," i.e. an nnz-by-3 array.
# Reshape to a numpy array that matches this format.
gradient_c = gradient_c.reshape((-1, 3))
num_constraints = len(row) - 1 # Objective is included as a row
if num_constraints > 0:
row_idx = gradient_c[:, 1] - 1
col_idx = gradient_c[:, 0] - 1
data = gradient_c[:, 2]
gradient_c = scipy.sparse.csr_matrix((data, (row_idx, col_idx)),
shape=(num_constraints, len(col)))
else:
gradient_c = np.array([])
return gradient_f, gradient_c, col, row, line_dic
def sensitivity_calculation(method,
instance,
paramList,
perturbList,
cloneModel=True,
tee=False,
keepfiles=False,
solver_options=None):
"""This function accepts a Pyomo ConcreteModel, a list of parameters, and
their corresponding perturbation list. The model is then augmented with
dummy constraints required to call sipopt or k_aug to get an approximation
of the perturbed solution.
Parameters
----------
method: string
'sipopt' or 'k_aug'
instance: Block
pyomo block or model object
paramSubList: list
list of mutable parameters or fixed variables
perturbList: list
list of perturbed parameter values
cloneModel: bool, optional
indicator to clone the model. If set to False, the original
model will be altered
tee: bool, optional
indicator to stream solver log
keepfiles: bool, optional
preserve solver interface files
solver_options: dict, optional
Provides options to the solver (also the name of an attribute)
Returns
-------
The model that was manipulated by the sensitivity interface
"""
sens = SensitivityInterface(instance, clone_model=cloneModel)
sens.setup_sensitivity(paramList)
m = sens.model_instance
if method not in {"k_aug", "sipopt"}:
raise ValueError("Only methods 'k_aug' and 'sipopt' are supported'")
if method == 'k_aug':
k_aug = SolverFactory('k_aug', solver_io='nl')
dot_sens = SolverFactory('dot_sens', solver_io='nl')
ipopt = SolverFactory('ipopt', solver_io='nl')
k_aug_interface = K_augInterface(k_aug=k_aug, dot_sens=dot_sens)
ipopt.solve(m, tee=tee)
m.ipopt_zL_in.update(m.ipopt_zL_out) #: important!
m.ipopt_zU_in.update(m.ipopt_zU_out) #: important!
k_aug.options['dsdp_mode'] = "" #: sensitivity mode!
k_aug_interface.k_aug(m, tee=tee)
sens.perturb_parameters(perturbList)
if method == 'sipopt':
ipopt_sens = SolverFactory('ipopt_sens', solver_io='nl')
ipopt_sens.options['run_sens'] = 'yes'
# Send the model to ipopt_sens and collect the solution
results = ipopt_sens.solve(m, keepfiles=keepfiles, tee=tee)
elif method == 'k_aug':
dot_sens.options["dsdp_mode"] = ""
k_aug_interface.dot_sens(m, tee=tee)
return m