Пример #1
0
def homotopy(model,
             variables,
             targets,
             max_solver_iterations=50,
             max_solver_time=10,
             step_init=0.1,
             step_cut=0.5,
             iter_target=4,
             step_accel=0.5,
             max_step=1,
             min_step=0.05,
             max_eval=200):
    """
    Homotopy meta-solver routine using Ipopt as the non-linear solver. This
    routine takes a model along with a list of fixed variables in that model
    and a list of target values for those variables. The routine then tries to
    iteratively move the values of the fixed variables to their target values
    using an adaptive step size.

    Args:
        model : model to be solved
        variables : list of Pyomo Var objects to be varied using homotopy.
                    Variables must be fixed.
        targets : list of target values for each variable
        max_solver_iterations : maximum number of solver iterations per
                    homotopy step (default=50)
        max_solver_time : maximum cpu time for the solver per homotopy step
                    (default=10)
        step_init : initial homotopy step size (default=0.1)
        step_cut : factor by which to reduce step size on failed step
                    (default=0.5)
        step_accel : acceleration factor for adjusting step size on successful
                     step (default=0.5)
        iter_target : target number of solver iterations per homotopy step
                    (default=4)
        max_step : maximum homotopy step size (default=1)
        min_step : minimum homotopy step size (default=0.05)
        max_eval : maximum number of homotopy evaluations (both successful and
                   unsuccessful) (default=200)

    Returns:
        Termination Condition : A Pyomo TerminationCondition Enum indicating
            how the meta-solver terminated (see documentation)
        Solver Progress : a fraction indication how far the solver progressed
            from the initial values to the target values
        Number of Iterations : number of homotopy evaluations before solver
            terminated
    """
    eps = 1e-3  # Tolerance for homotopy step convergence to 1

    # Get model logger
    _log = logging.getLogger(__name__)

    # Validate model is an instance of Block
    if not isinstance(model, Block):
        raise TypeError("Model provided was not a valid Pyomo model object "
                        "(instance of Block). Please provide a valid model.")
    if degrees_of_freedom(model) != 0:
        raise ConfigurationError(
            "Degrees of freedom in model are not equal to zero. Homotopy "
            "should not be used on probelms which are not well-defined.")

    # Validate variables and targets
    if len(variables) != len(targets):
        raise ConfigurationError(
            "Number of variables and targets do not match.")
    for i in range(len(variables)):
        v = variables[i]
        t = targets[i]

        if not isinstance(v, _VarData):
            raise TypeError("Variable provided ({}) was not a valid Pyomo Var "
                            "component.".format(v))

        # Check that v is part of model
        parent = v.parent_block()
        while parent != model:
            if parent is None:
                raise ConfigurationError(
                    "Variable {} is not part of model".format(v))
            parent = parent.parent_block()

        # Check that v is fixed
        if not v.fixed:
            raise ConfigurationError(
                "Homotopy metasolver provided with unfixed variable {}."
                "All variables must be fixed.".format(v.name))

        # Check bounds on v (they don't really matter, but check for sanity)
        if v.ub is not None:
            if v.value > v.ub:
                raise ConfigurationError(
                    "Current value for variable {} is greater than the "
                    "upper bound for that variable. Please correct this "
                    "before continuing.".format(v.name))
            if t > v.ub:
                raise ConfigurationError(
                    "Target value for variable {} is greater than the "
                    "upper bound for that variable. Please correct this "
                    "before continuing.".format(v.name))
        if v.lb is not None:
            if v.value < v.lb:
                raise ConfigurationError(
                    "Current value for variable {} is less than the "
                    "lower bound for that variable. Please correct this "
                    "before continuing.".format(v.name))
            if t < v.lb:
                raise ConfigurationError(
                    "Target value for variable {} is less than the "
                    "lower bound for that variable. Please correct this "
                    "before continuing.".format(v.name))

    # TODO : Should we be more restrictive on these values to avoid users
    # TODO : picking numbers that are less likely to solve (but still valid)?
    # Validate homotopy parameter selections
    if not 0.05 <= step_init <= 0.8:
        raise ConfigurationError("Invalid value for step_init ({}). Must lie "
                                 "between 0.05 and 0.8.".format(step_init))
    if not 0.1 <= step_cut <= 0.9:
        raise ConfigurationError("Invalid value for step_cut ({}). Must lie "
                                 "between 0.1 and 0.9.".format(step_cut))
    if step_accel < 0:
        raise ConfigurationError(
            "Invalid value for step_accel ({}). Must be "
            "greater than or equal to 0.".format(step_accel))
    if iter_target < 1:
        raise ConfigurationError(
            "Invalid value for iter_target ({}). Must be "
            "greater than or equal to 1.".format(iter_target))
    if not isinstance(iter_target, int):
        raise ConfigurationError("Invalid value for iter_target ({}). Must be "
                                 "an an integer.".format(iter_target))
    if not 0.05 <= max_step <= 1:
        raise ConfigurationError("Invalid value for max_step ({}). Must lie "
                                 "between 0.05 and 1.".format(max_step))
    if not 0.01 <= min_step <= 0.1:
        raise ConfigurationError("Invalid value for min_step ({}). Must lie "
                                 "between 0.01 and 0.1.".format(min_step))
    if not min_step <= max_step:
        raise ConfigurationError("Invalid argumnets: step_min must be less "
                                 "or equal to step_max.")
    if not min_step <= step_init <= max_step:
        raise ConfigurationError("Invalid arguments: step_init must lie "
                                 "between min_step and max_step.")
    if max_eval < 1:
        raise ConfigurationError(
            "Invalid value for max_eval ({}). Must be "
            "greater than or equal to 1.".format(step_accel))
    if not isinstance(max_eval, int):
        raise ConfigurationError("Invalid value for max_eval ({}). Must be "
                                 "an an integer.".format(iter_target))

    # Create solver object
    solver_obj = SolverFactory('ipopt')

    # Perform initial solve of model to confirm feasible initial solution
    results, solved, sol_iter, sol_time, sol_reg = ipopt_solve_with_stats(
        model, solver_obj, max_solver_iterations, max_solver_time)

    if not solved:
        _log.exception("Homotopy Failed - initial solution infeasible.")
        return TerminationCondition.infeasible, 0, 0
    elif sol_reg != "-":
        _log.warning(
            "Homotopy - initial solution converged with regularization.")
        return TerminationCondition.other, 0, 0
    else:
        _log.info("Homotopy - initial point converged")

    # Set up homotopy variables
    # Get initial values and deltas for all variables
    v_init = []
    for i in range(len(variables)):
        v_init.append(variables[i].value)

    n_0 = 0.0  # Homotopy progress variable
    s = step_init  # Set step size to step_init
    iter_count = 0  # Counter for homotopy iterations

    # Save model state to dict
    # TODO : for very large models, it may be necessary to dump this to a file
    current_state = to_json(model, return_dict=True)

    while n_0 < 1.0:
        iter_count += 1  # Increase iter_count regardless of success or failure

        # Calculate next n value given current step size
        if n_0 + s >= 1.0 - eps:
            n_1 = 1.0
        else:
            n_1 = n_0 + s

        _log.info("Homotopy Iteration {}. Next Step: {} (Current: {})".format(
            iter_count, n_1, n_0))

        # Update values for all variables using n_1
        for i in range(len(variables)):
            variables[i].fix(targets[i] * n_1 + v_init[i] * (1 - n_1))

        # Solve model at new state
        results, solved, sol_iter, sol_time, sol_reg = ipopt_solve_with_stats(
            model, solver_obj, max_solver_iterations, max_solver_time)

        # Check solver output for convergence
        if solved:
            # Step succeeded - accept current state
            current_state = to_json(model, return_dict=True)

            # Update n_0 to accept current step
            n_0 = n_1

            # Check solver iterations and calculate next step size
            s_proposed = s * (1 + step_accel * (iter_target / sol_iter - 1))

            if s_proposed > max_step:
                s = max_step
            elif s_proposed < min_step:
                s = min_step
            else:
                s = s_proposed
        else:
            # Step failed - reload old state
            from_json(model, current_state)

            # Try to cut back step size
            if s > min_step:
                # Step size can be cut
                s = max(min_step, s * step_cut)
            else:
                # Step is already at minimum size, terminate homotopy
                _log.exception(
                    "Homotopy failed - could not converge at minimum step "
                    "size. Current progress is {}".format(n_0))
                return TerminationCondition.minStepLength, n_0, iter_count

        if iter_count >= max_eval:  # Use greater than or equal to to be safe
            _log.exception("Homotopy failed - maximum homotopy iterations "
                           "exceeded. Current progress is {}".format(n_0))
            return TerminationCondition.maxEvaluations, n_0, iter_count

    if sol_reg == "-":
        _log.info("Homotopy successful - converged at target values in {} "
                  "iterations.".format(iter_count))
        return TerminationCondition.optimal, n_0, iter_count
    else:
        _log.exception("Homotopy failed - converged at target values with "
                       "regularization in {} iterations.".format(iter_count))
        return TerminationCondition.other, n_0, iter_count
Пример #2
0
    def _Q_at_theta(self, thetavals):
        """
        Return the objective function value with fixed theta values.
        
        Parameters
        ----------
        thetavals: dict
            A dictionary of theta values.

        Returns
        -------
        objectiveval: float
            The objective function value.
        thetavals: dict
            A dictionary of all values for theta that were input.
        solvertermination: Pyomo TerminationCondition
            Tries to return the "worst" solver status across the scenarios.
            pyo.TerminationCondition.optimal is the best and 
            pyo.TerminationCondition.infeasible is the worst.
        """

        optimizer = pyo.SolverFactory('ipopt')
        dummy_tree = lambda: None  # empty object (we don't need a tree)
        dummy_tree.CallbackModule = None
        dummy_tree.CallbackFunction = self._instance_creation_callback
        dummy_tree.ThetaVals = thetavals
        dummy_tree.cb_data = self.callback_data

        if self.diagnostic_mode:
            print('    Compute objective at theta = ', str(thetavals))

        # start block of code to deal with models with no constraints
        # (ipopt will crash or complain on such problems without special care)
        instance = _pysp_instance_creation_callback(dummy_tree, "FOO1", None)
        try:  # deal with special problems so Ipopt will not crash
            first = next(
                instance.component_objects(pyo.Constraint, active=True))
        except:
            sillylittle = True
        else:
            sillylittle = False
        # end block of code to deal with models with no constraints

        WorstStatus = pyo.TerminationCondition.optimal
        totobj = 0
        for snum in self._numbers_list:
            sname = "scenario_NODE" + str(snum)
            instance = _pysp_instance_creation_callback(
                dummy_tree, sname, None)
            if not sillylittle:
                if self.diagnostic_mode:
                    print('      Experiment = ', snum)
                    print(
                        '     First solve with with special diagnostics wrapper'
                    )
                    status_obj, solved, iters, time, regu \
                        = ipopt_solver_wrapper.ipopt_solve_with_stats(instance, optimizer, max_iter=500, max_cpu_time=120)
                    print(
                        "   status_obj, solved, iters, time, regularization_stat = ",
                        str(status_obj), str(solved), str(iters), str(time),
                        str(regu))

                results = optimizer.solve(instance)
                if self.diagnostic_mode:
                    print('standard solve solver termination condition=',
                          str(results.solver.termination_condition))

                if results.solver.termination_condition \
                   != pyo.TerminationCondition.optimal :
                    # DLW: Aug2018: not distinguishing "middlish" conditions
                    if WorstStatus != pyo.TerminationCondition.infeasible:
                        WorstStatus = results.solver.termination_condition

            objobject = getattr(instance, self._second_stage_cost_exp)
            objval = pyo.value(objobject)
            totobj += objval
        retval = totobj / len(self._numbers_list)  # -1??

        return retval, thetavals, WorstStatus
Пример #3
0
    def _Q_at_theta(self, thetavals):
        """
        Return the objective function value with fixed theta values.
        
        Parameters
        ----------
        thetavals: dict
            A dictionary of theta values.

        Returns
        -------
        objectiveval: float
            The objective function value.
        thetavals: dict
            A dictionary of all values for theta that were input.
        solvertermination: Pyomo TerminationCondition
            Tries to return the "worst" solver status across the scenarios.
            pyo.TerminationCondition.optimal is the best and 
            pyo.TerminationCondition.infeasible is the worst.
        """

        optimizer = pyo.SolverFactory('ipopt')
        dummy_tree = lambda: None # empty object (we don't need a tree)
        dummy_tree.CallbackModule = None
        dummy_tree.CallbackFunction = self._instance_creation_callback
        dummy_tree.ThetaVals = thetavals
        dummy_tree.cb_data = self.callback_data
        
        if self.diagnostic_mode:
            print('    Compute objective at theta = ',str(thetavals))

        # start block of code to deal with models with no constraints
        # (ipopt will crash or complain on such problems without special care)
        instance = _pysp_instance_creation_callback(dummy_tree, "FOO1", None)    
        try: # deal with special problems so Ipopt will not crash
            first = next(instance.component_objects(pyo.Constraint, active=True))
        except:
            sillylittle = True 
        else:
            sillylittle = False
        # end block of code to deal with models with no constraints

        WorstStatus = pyo.TerminationCondition.optimal
        totobj = 0
        for snum in self._numbers_list:
            sname = "scenario_NODE"+str(snum)
            instance = _pysp_instance_creation_callback(dummy_tree,
                                                        sname, None)
            if not sillylittle:
                if self.diagnostic_mode:
                    print('      Experiment = ',snum)
                    print('     First solve with with special diagnostics wrapper')
                    status_obj, solved, iters, time, regu \
                        = ipopt_solver_wrapper.ipopt_solve_with_stats(instance, optimizer, max_iter=500, max_cpu_time=120)
                    print("   status_obj, solved, iters, time, regularization_stat = ",
                           str(status_obj), str(solved), str(iters), str(time), str(regu))

                results = optimizer.solve(instance)
                if self.diagnostic_mode:
                    print('standard solve solver termination condition=',
                            str(results.solver.termination_condition))

                if results.solver.termination_condition \
                   != pyo.TerminationCondition.optimal :
                    # DLW: Aug2018: not distinguishing "middlish" conditions
                    if WorstStatus != pyo.TerminationCondition.infeasible:
                        WorstStatus = results.solver.termination_condition
                    
            objobject = getattr(instance, self._second_stage_cost_exp)
            objval = pyo.value(objobject)
            totobj += objval
        retval = totobj / len(self._numbers_list) # -1??
        return retval, thetavals, WorstStatus