Parallel solution of large sparse matrix equations and parallel power flow

A very efficient parallel LU factorization and substitution algorithm for solving large sparse network equations on shared memory multi-processor parallel computers is presented. By rearranging the order of the computations, better parallelism is obtained out of the traditionally sequential method. Performance results on an actual parallel computer are presented and discussed. Parallel gains for the power flow solution using Newton´s and fast decoupled methods are presented to demonstrate it´s effectiveness. Better speedup gains are obtained for larger systems and speedup of over 13 for Newton´s power flow on a 20 processor shared memory computer has been obtained