Parallel computation of transient stability using symplectic Gauss method and GPU

This study presents a new parallel algorithm for power system transient stability computation based on symplectic Gauss method. The s-stage 2s-order symplectic Gauss method is used to convert the differential-algebraic system simultaneously at s time points into a set of non-linear algebraic equatio...

Full description

Saved in:
Bibliographic Details
Published in:IET generation, transmission & distribution transmission & distribution, 2016-11, Vol.10 (15), p.3727-3735
Main Authors: Liao, Xiaobing, Wang, Fangzong
Format: Article
Language:English
Subjects:
Algebra
Algorithms
block matrix characteristics
central processing unit computing
Central processing units
Computation
cooperative computing
CPU‐single graphics processing unit
differential algebraic equations
differential‐algebraic system
GPU
Graphics processing units
linear equations
Mathematical analysis
Mathematical models
matrix algebra
microprocessor chips
Newton method
nonlinear algebraic equations
nonlinear differential equations
parallel algorithm
power system transient stability
power system transient stability computation
preconditioned GMRES method
programming models
s‐stage 2s‐order symplectic Gauss method
Transient stability
V‐transformation
Citations: Items that this one cites
Items that cite this one
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study presents a new parallel algorithm for power system transient stability computation based on symplectic Gauss method. The s-stage 2s-order symplectic Gauss method is used to convert the differential-algebraic system simultaneously at s time points into a set of non-linear algebraic equations, and the algebraic system is then solved by Newton method. Based on the block matrix characteristics, the solution of the linear equations involved in Newton method's process can be divided into two parts: the first part is fully parallelisable in time, and the second part is solved by preconditioned GMRES method while an efficient preconditioner has been proposed based on the V-transformation. For test, the proposed algorithm is implemented on three programming models, the first is the traditional central processing unit (CPU) computing, the second is CPU-single graphics processing unit (GPU) cooperative computing, and the last one is CPU-multiple GPUs cooperative computing. The results show that, the proposed parallel algorithm is accurate, has good convergence, and has good scalability both in the problem size and in the number of used GPU.
ISSN:1751-8687
1751-8695
1751-8695
DOI:10.1049/iet-gtd.2016.0033