Title :
An automatic clustering algorithm using factorisation tree for parallel power system simulation
Author :
Bialek, J. ; Grey, D.J.
Author_Institution :
Sch. of Eng. & Comput. Sci., Durham Univ., UK
Abstract :
Parallel simulation of power systems requires the system to be partitioned into subnetworks which are processed on individual processors. Maximum computational efficiency is achieved when the network is split such that each processor has an equal computational load. This paper proposes an automatic method of network partitioning which gives well balanced network splits, based upon an analysis of the factorisation tree for the system. The method also predicts the expected parallel speed-up for the split and allows the visualisation of large networks. A modified minimum degree minimum length node ordering algorithm is also presented which gives well balanced factorisation trees
Keywords :
digital simulation; parallel algorithms; power system analysis computing; trees (mathematics); automatic clustering algorithm; balanced network splits; factorisation tree; large networks visualisation; minimum degree algorithm; minimum length node ordering algorithm; modified algorithm; parallel simulation; power system partitioning; power system simulation; Clustering algorithms; Computer networks; Equations; Partitioning algorithms; Power system analysis computing; Power system dynamics; Power system reliability; Power system security; Power system simulation; Sparse matrices;
Conference_Titel :
Electrotechnical Conference, 1994. Proceedings., 7th Mediterranean
Conference_Location :
Antalya
Print_ISBN :
0-7803-1772-6
DOI :
10.1109/MELCON.1994.380937