Meaning
Numerical phenomenon where the sensitivity of a system of linear equations to numerical errors increases dramatically, causing inversion algorithms to produce highly inaccurate results. A condition number explosion typically occurs when modeling electrode impedance or state of charge where matrix columns become nearly co-linear. This happens when the ratio of the largest to smallest singular value grows too large, which amplifies sensor noise during computation.
Numerical Instability
Matrix inversion processes suffer when the coefficients of the system are too close to each other. Under these conditions, the calculation of internal battery states can output physically impossible values like negative concentrations or temperatures above boiling point. Stabilizing the equations is necessary to prevent a condition number explosion from corrupting the monitoring software.
Diagnostic Marker
Real-time algorithms track the ratio of extreme eigenvalues to detect when the computations are becoming unreliable. This monitoring allows the control unit to switch to a fallback state. When a condition number explosion begins, the diagonal elements of the covariance matrix grow rapidly.
Prevention Method
Mathematical constraints can hold the calculated values within physically realistic limits. Applying regularizing terms to the solver prevents the condition number explosion by damping high-frequency noise. These filters modify the original system of equations to ensure stable matrix inversion.