Pipeline multiprocessor algorithm for contingency analysis in power systems

In this paper an efficient multiprocessor algorithm for repetitive sparse linear equations solution with singleton and full right hand side vector is proposed. It features fast analysis of power system based on multiprocessor pipeline, as well as efficient pipeline mapping on hypercube architecture, applied with and without superposition principle in use. A major characteristic of the proposed approach is the optimization of processor load balance (over 90% active time of processors). Achieved speedups are close to the number of processors in the system. The communication overheads are neglectable and the execution times are calculated on the basis of the computing operations. The proposed algorithm has been successfully applied for power mismatch calculations arising in the solution phase of fast decoupled power flow (FDPF), which involves typical repeated sparse power network computations. The experiments supported the approach for suboptimal pipeline mapping on a higher order hypercube, without applying superposition.