To satisfy demands of fast calculation in power system analysis, utilizing of the parallel computer is being studied and various parallel algorithms are being developed. System decomposition methods and iterative methods are usually applied to parallel computing due to their parallelism. LU decomposition methods, i. e. direct methods, is also utilized its operating parallelism. However, because of the medium in size or the sparsity of power system matrices, it is difficult to gain a speedup ratio so suficient as the increment of processor number in the region of the large processor number.
Author studied to exploit more parallelism in the LU decomposition, and developed a new parallel algorithm “parallel elimination method” mentioned details in this paper. This method is verified its effectiveness in large number rigion of processors by simulations counting the number of operation units.
J-STAGEがリニューアルされました！https://www.jstage.jst.go.jp/browse/-char/ja/
