Parallel-vector processing synergy and frequency domain transient stability simulations

The use of vector and parallel-vector processing to perform triangular substitution and power system transient stability simulations is explored in this paper. Experimental results obtained from the Cray X/MP-116se and Cray Y/MP-8/8-64 show large gains due to vectorization. Experimental results on the Alliant FX/8 show that the synergistic relationship between parallel and vector processing can be used to achieve gains on the order of 25 for multiple forward and backward substitution problems. The vectorized triangular solver is incorporated into a frequency domain relaxation scheme for performing dynamic simulations. Interfacing time and frequency domain variables is achieved using a vectorized fast Fourier transform routine. A sparse implementation of the frequency domain integrator and lag operators is shown to be highly vectorizable and to require O(N) operations rather than O(N²) required by a dense formulation. Dynamic simulation results are provided for the Cray X/MP and Cray Y/MP showing real-time performance.