bluemira.equilibria.optimisation.problem._nested_position

OptimisationProblems for coilset design.

New optimisation schemes for the coilset can be provided by subclassing from CoilsetOP, which is an abstract base class for OptimisationProblems that use a coilset as their parameterisation object.

Subclasses must provide an optimise() method that returns an optimised coilset according to a given optimisation objective function. As the exact form of the state vector that is optimised is often specific to each objective function, each subclass of CoilsetOP is generally also specific to a given objective function, since the method used to map the coilset object to the state vector (and additional required arguments) will generally differ in each case.

Classes

`NestedCoilsetPositionCOP`	Coilset OptimisationProblem for coil currents and positions
`PulsedNestedPositionCOP`	Coilset position optimisation problem for multiple sub-optimisation problems.

Module Contents

class bluemira.equilibria.optimisation.problem._nested_position.NestedCoilsetPositionCOP(sub_opt: bluemira.equilibria.optimisation.problem.base.EqCoilsetOptimisationProblem, position_mapper: bluemira.utilities.positioning.PositionMapper, opt_algorithm: bluemira.optimisation.AlgorithmType = Algorithm.SBPLX, opt_conditions: dict[str, float] | None = None, opt_parameters: dict[str, float] | None = None, constraints: list[bluemira.equilibria.optimisation.constraints.UpdateableConstraint] | None = None)

Bases: bluemira.equilibria.optimisation.problem.base.EqCoilsetOptimisationProblem

Coilset OptimisationProblem for coil currents and positions subject to maximum current bounds and positions bounded within a provided region. Performs a nested optimisation for coil currents within each position optimisation function call.

Parameters:

sub_opt (bluemira.equilibria.optimisation.problem.base.EqCoilsetOptimisationProblem) – Coilset OptimisationProblem to use for the optimisation of coil currents at each trial set of coil positions. sub_opt.coilset will be modified during the optimisation.
eq – Equilibrium object used to update magnetic field targets.
position_mapper (bluemira.utilities.positioning.PositionMapper) – position mapper object of the regions to optimise the coil positions within
opt_algorithm (bluemira.optimisation.AlgorithmType) – The optimisation algorithm to use (e.g. SLSQP)
opt_conditions (dict[str, float] | None) – The stopping conditions for the optimiser. for defaults see AlgorithDefaultTolerances along with max_eval=100
opt_parameters (dict[str, float] | None) – Optimiser specific parameters, see https://nlopt.readthedocs.io/en/latest/NLopt_Reference/#algorithm-specific-parameters Otherwise, the parameters can be founded by digging through the source code.
constraints (list[bluemira.equilibria.optimisation.constraints.UpdateableConstraint] | None) – List of optimisation constraints to apply to the optimisation problem

Notes

Setting stopval and maxeval is the most reliable way to stop optimisation at the desired figure of merit and number of iterations respectively. Some NLOpt optimisers display unexpected behaviour when setting xtol and ftol, and may not terminate as expected when those criteria are reached.

position_mapper

bounds

sub_opt

_get_initial_vector() → numpy.typing.NDArray

Returns:: A vector representation of the initial coilset state from the PositionMapper.
Return type:: numpy.typing.NDArray

optimise(x0: numpy.typing.NDArray | None = None, *, keep_history: bool = False, check_constraints: bool = False)

Run the optimisation.

Return type:

Optimised CoilSet object.

Parameters:

x0 (numpy.typing.NDArray | None)
keep_history (bool)
check_constraints (bool)

objective(vector: numpy.typing.NDArray[numpy.float64]) → float

Objective function to minimise.

Parameters:: vector (numpy.typing.NDArray[numpy.float64])
Return type:: float

class bluemira.equilibria.optimisation.problem._nested_position.PulsedNestedPositionCOP(coilset: bluemira.equilibria.coils.CoilSet, position_mapper: bluemira.utilities.positioning.PositionMapper, sub_opt_problems: list[bluemira.equilibria.optimisation.problem.base.EqCoilsetOptimisationProblem], opt_algorithm: bluemira.optimisation.AlgorithmType = Algorithm.COBYLA, opt_conditions: dict[str, float] | None = None, opt_parameters: dict[str, float] | None = None, constraints=None, initial_currents=None, current_bounder: bluemira.equilibria.run.MovingCurrentBoundStrategy | None = None, *, debug: bool = False, keep_history: bool = True, check_constraints: bool = True)

Bases: bluemira.equilibria.optimisation.problem.base.CoilsetOptimisationProblem

Coilset position optimisation problem for multiple sub-optimisation problems.

Parameters:

coilset (bluemira.equilibria.coils.CoilSet) – Coilset for which to optimise positions
position_mapper (bluemira.utilities.positioning.PositionMapper) – Position mapper tool to parameterise coil positions
sub_opt_problems (list[bluemira.equilibria.optimisation.problem.base.EqCoilsetOptimisationProblem]) – The list of sub-optimisation problems to solve
opt_algorithm (bluemira.optimisation.AlgorithmType) – The optimisation algorithm to use (e.g. SLSQP)
opt_conditions (dict[str, float] | None) – The stopping conditions for the optimiser. for defaults see AlgorithDefaultTolerances along with max_eval=100
constraints – Constraints to use. Note these should be applicable to the parametric position vector
initial_currents – Initial currents to use when solving the current sub-optimisation problems
debug (bool) – Whether or not to run in debug mode (will affect run-time noticeably)
opt_parameters (dict[str, float] | None)
current_bounder (bluemira.equilibria.run.MovingCurrentBoundStrategy | None)
keep_history (bool)
check_constraints (bool)

position_mapper

sub_opt_problems

debug

iter

bounds

current_bounder = None

keep_history = True

check_constraints = True

static _run_reporting(itern, max_fom, verbose): Keep track of objective function value over iterations.

static _run_diagnostics(debug, sub_opt_prob: bluemira.equilibria.optimisation.problem.base.EqCoilsetOptimisationProblem, opt_result: bluemira.equilibria.optimisation.problem.base.CoilsetOptimiserResult)

In debug mode, store the LCFS at each iteration for each of the sub-optimisation problems.

Notes

This can significantly impact run-time.

Parameters:

sub_opt_prob (bluemira.equilibria.optimisation.problem.base.EqCoilsetOptimisationProblem)
opt_result (bluemira.equilibria.optimisation.problem.base.CoilsetOptimiserResult)

sub_opt_objective(vector: numpy.typing.NDArray[numpy.float64], *, verbose: bool = False) → float

Run the sub-optimisations

Returns:

The largest figure of merit.

Parameters:

vector (numpy.typing.NDArray[numpy.float64])
verbose (bool)

Return type:

float

objective(vector: numpy.typing.NDArray[numpy.float64], *, verbose: bool = False) → float

The objective function of the parent optimisation.

Returns:

Results of the sub-optimisations.

Parameters:

vector (numpy.typing.NDArray[numpy.float64])
verbose (bool)

Return type:

float

_get_initial_vector() → numpy.typing.NDArray[numpy.float64]

Returns:: A vector representation of the initial coilset state from the PositionMapper.
Return type:: numpy.typing.NDArray[numpy.float64]

optimise(x0: numpy.typing.NDArray | None = None, *, verbose: bool = False, keep_history: bool = False, check_constraints: bool = False) → bluemira.equilibria.optimisation.problem.base.CoilsetOptimiserResult

Run the PulsedNestedPositionCOP

Parameters:

x0 (numpy.typing.NDArray | None) – Initial solution vector (parameterised positions)
verbose (bool) – Whether or not to print progress information
keep_history (bool)
check_constraints (bool)

Returns:

Optimised CoilSet

Return type:

coilset