Sparse solver application for parallel real-time electromagnetic transient simulations

Abstract

The main purpose of this research is to speed up real-time simulations of electromagnetic transients (EMTs) using sparse linear solver techniques. This paper presents the integration of a direct sparse linear solver (KLU) into a real-time software for EMT simulation. This solver is combined with parallelization of network solution. Fill-in reduction techniques are investigated as well as partial refactorization to speed up computations. The pivoting technique during refactorization is asserted in terms of simulation stability as compared to existing sparse solver based on code generation without pivoting. Performance and validation are studied on practical power system cases with real-time Hardware-In-the-Loop (HIL) simulation. Substantial performance gains, up to 50%, are obtained using fill-in reduction and partial refactorization. Pivoting is necessary to maintain numerical stability.