Towards faster Givens rotations based power system state estimator

Abstract

Numerically stable and computationally efficient power system state estimation (PSSE) algorithms are designed using an orthogonalization (QR decomposition) approach. They use Givens rotations for orthogonalization which enables sparsity exploitation during factorization of the large sparse augmented Jacobian. A priori row and column ordering is usually performed to reduce intermediate and and overall fills. Column ordering methods, usually based on minimum degree algorithm (MDA), have matured. However, there exists a significant scope for improving the quality of row ordering. This paper introduces a new row ordering technique for Givens rotations based power system state estimators. The proposed row processing method (VPAIR) requires a shift from conventionally used row oriented QR decomposition implementation to a column oriented QR decomposition implementation. It is demonstrated that, the proposed column oriented QR decomposition algorithm which uses MDA for column ordering and VPAIR for row ordering can lead to a much faster PSSE. These aspects are justified by simulations on large power systems.