An Efficient Parallel Krylov-Schur Method for Eigen-Analysis of Large-Scale Power Systems

An efficient parallel Krylov-Schur method is proposed for computing eigenvalues and eigenvectors related with oscillating modes of low damping for large-scale power systems. Improved restarting techniques are explained and demonstrated in details, which focus on a refined selection of kept subspace in contraction and a reliable mechanism of no missing target eigenvalues. Cayley and shift-invert transforms are used to de-couple the computation of eigen-analysis and enable the proposed parallelization with a master-slave scheme. Based on the improved restarting techniques, the strategy for adaptive allocation of shifts and the coordination of parallel computing tasks, the proposed method is capable of computing a large number of eigen-pairs with satisfactory accuracy, convergence rate and parallel acceleration. The computational efficiency is validated by three test cases from real-world large-scale power systems on a symmetric multi-processing (SMP) computer.