Skip to main content

invert_dense_matrix

Function invert_dense_matrix 

pub fn invert_dense_matrix<T>(
    matrix: &[Vec<T>],
    pivot: DensePivot,
    singular_pivot: f64,
    relative_pivot: f64,
) -> Option<Vec<Vec<T>>>
where T: DenseScalar,
Expand description

The inverse of a small dense square matrix by Gauss-Jordan elimination in the pivot order (DensePivot). None when a pivot’s magnitude is NaN, infinite (under DensePivot::Partial; under DensePivot::OpenDssDiagonal an infinite unscaled magnitude is a pivot), singular_pivot or smaller, or relative_pivot times the largest original magnitude in its column or smaller. Each caller states the pivot its quantity is singular at: an absolute singular_pivot (ohms for a series impedance, a dimensionless coefficient for a potential matrix) and a relative_pivot that measures cancellation within a column against that column’s own scale (the dense PTDF fallback passes n ε, so a column eliminated to rounding noise is singular while a wide but valid spread of magnitudes stays invertible). Zero for both refuses only exact singularity. A one by one matrix inverts as its scaled reciprocal (DenseScalar::reciprocal). An empty matrix inverts to an empty matrix.

The one dense inverse: line and switch series admittances, the open line Schur complement, overhead line potential coefficients (OpenDSS and GridLAB-D geometry), the PMD transformer leakage round trip, Kron reduction (powerio_dist::kron::calc_kron_reduced_impedance), the transformer winding admittance and the dense PTDF fallback of powerio-matrix read it with DensePivot::Partial, as does the transmission bus merge’s zero impedance link flow recovery (powerio-tx, an absolute floor of 1e-13 times the largest entry); the OpenDSS source and line primitive check reads it with DensePivot::OpenDssDiagonal.