Parallelizing Methods Analysis for Solving Large Sparse Power Network Equations on Multi-core Processor Platforms
Zhang Jia-an, Zhang Na, Yilang Jiang · 2011
The solution of large sparse network equations is a recurrent problem in almost every algorithm of power system simulation, which is widely used in offline analysis, online stability assessment and control. Nowadays, processor architectures are moving toward the integration of more cores on one chip, whilst traditional serial and parallel software can not fully exploit the performance of multi-core processors. The paper discusses the parallelizing methods for solving large sparse network equations on multi-core processor platforms. The available and feasible parallelizing algorithms such as W-matrix method and BBDF method are investigated. Implicit and explicit parallelization strategies are studied. The parallelizing complexity and computation efficiency are analyzed as well. Refactoring software code to parallel code is a significant ongoing topic of research not only for PCs but also for parallel systems.