On the application of high performance computation techniques to nonlinear interior point methods

Abstract

This work focuses on the application of vector processing by levelwise algorithms to the solution of the optimal power flow (OPF) via nonlinear interior point methods. Two optimization algorithms are analyzed: the pure primal-dual and the predictor-corrector. In order to take advantage of the application of vector processing, the concept of factorization path level is used in both the factorization and the forward-back substitution. Results with power systems ranging from 118 to 1700 buses illustrate the performance of the proposed methodology. It is shown that considerable speed-ups in the OPF solution can be achieved if vector processing is suitably used in the solution of the linear system at each iteration of the optimization process.