Multi-Processor Architectures for Solving Sparse Linear Systems. Application to the Load Flow Problem

Three heuristic algorithms for solving a cluster problem associated with the tearing of a symmetric matrix are presented. Based on these partitioning procedures, a method for the parallel solution of the fast decoupled load flow has been developed, although it is also suitable for the parallel solution of any linear equations system. Experimental results of applying such algorithms on several test systems have been obtained using a multiprocessor architecture.