Struct ObservabilityReport 

pub struct ObservabilityReport {
    pub n_unknowns: usize,
    pub rank: usize,
    pub structurally_unmeasured: Vec<Unknown>,
    pub unobservable: Vec<Unknown>,
    pub skipped_numerical: bool,
    pub de_energized: Vec<Unknown>,
    pub global_current_without_angle_reference: bool,
}

What the measurements do and do not determine.

Fields

n_unknowns: usize

rank: usize

Rank of the gain matrix. Equal to n_unknowns exactly when the state is fully observable.

structurally_unmeasured: Vec<Unknown>

Unknowns no measurement function mentions at all.

unobservable: Vec<Unknown>

Unknowns left undetermined once linear dependence is accounted for.

A superset of structurally_unmeasured in effect — a zero column is also rank-deficient — though the specific variables reported can differ, since which member of a dependent group gets blamed depends on the pivot order. The count is what is meaningful; the names are representatives.

skipped_numerical: bool

Set when the system was too large to analyze densely, in which case only the structural half of the report is filled in.

de_energized: Vec<Unknown>

Unknowns belonging to buses that carry no voltage at all.

These are not a gap in the sensor plan and no sensor would close them. A de-energized bus is reported at exactly zero by every estimator path here — nr::estimate writes it, iterative::estimate skips its rows, linear_start starts it there — so its columns are determined by the network rather than by measurements, and they are zero columns of H by construction. Counting them against the rank would report every CGMES model containing one switched-out node as unobservable, which is the same mistake, in the same function, that ignoring constraint rows was.

They are excluded from structurally_unmeasured, unobservable and is_observable, and reported here instead.

global_current_without_angle_reference: bool

A global-angle current sensor is present with nothing to measure its angle against.

Such a sensor reports I = i·e^{jθ} with θ absolute, which is only meaningful if something else fixes what that reference is. With no voltage angle in the set, StateLayout pins a reference bus instead — and that pin is then a false constraint, rotating the estimate away from the angle the sensor actually measured. The rank check will not catch it, because the system is perfectly determined; it is determined to the wrong thing.

power-grid-model refuses to run at all here, raising NotObservableError with “Global angle current sensors require at least one voltage angle measurement as a reference point”. gridoxide reports rather than refuses, on the grounds that this analysis is the place that says what is wrong and the estimator’s job is to answer the question it was asked.

Implementations

impl ObservabilityReport

pub fn is_observable(&self) -> bool

Whether every state variable is determined.

Deliberately does not fold in global_current_without_angle_reference: that state is fully determined, just to the wrong reference, and calling it unobservable would misname it.

Trait Implementations

impl Clone for ObservabilityReport

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ObservabilityReport

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.