Template Class Evaluator

Inheritance Relationships

public GridKit::PhasorDynamics::BusBase< scalar_type, index_type > (Template Class BusBase)
public GridKit::PhasorDynamics::Component< scalar_type, index_type > (Template Class Component)

template<class scalar_type, typename index_type> class Evaluator

Abstract class describing a model.

Public Functions

virtual int allocate() = 0

Allocate all internal memory used by the model.

Post:: Changing the model by changing its internal topology (i.e. size(), nnz(), sparsity pattern in getCsrJacobian(), etc.) may invalidate the model until allocate() is called again. Calling any function that requires allocate() to be called in between modifying the model and reallocating may invoke undefined behaviour.

virtual int tagDifferentiable() = 0

Fill the tag() vector with proper indicators.

Note

For some models, this may be a no-op, and the vector returned by tag() will still be empty.

virtual int setAbsoluteTolerance(RealT rel_tol) = 0

Compute the absolute tolerance for each variable in the model.

This represents a “noise” level close to zero for which pure relative error cannot be used.

Parameters:: rel_tol – The relative tolerance which can be used to pick the absolute tolerance.
Returns:: int 0 if successful, non-zero otherwise.

virtual int evaluateResidual() = 0

Evaluate the model’s residual function.

For models which can be appropriately modeled in semi-explicit form, this will evaluate

\[F(t, y, \dot{y}) = f(t,y) - M \dot{y},\]

where

\[M\]

is the mass matrix. For models which can be modeled in Hessenberg form,

\[M\]

should be a simple indicator function for if the variable is differential (if

\[\dot{y}\]

appears in its equation). This indicator should be given by tag().

virtual IdxT nnz() = 0

How many non-zero elements are in this model’s Jacobian.

This knowledge may be needed to allocate the buffers for the model’s Jacobian, so it it should not depend on allocate() having been called.