Technology assessment study of near term computer capabilities and their impact on power flow and stability simulation programs. Technical training study 77-749. Final report

The computer requirements needed for power system simulation in the detail desired today by power system analysts exceeds the capabilities of either: large general-purpose computers shared by several users simultaneously; or medium-scale affordable general-purpose machines dedicated to the simulation task. At the same time, computer manufacturers have responded to the needs of specific groups requiring above all else high-speed arithmetic capability. The result is a host of new machines which are called in this report vector processors and which include large vector machines such as the CRAY-1, parallel processors such as ILLIAC IV and smaller array processors such as the Floating Point Systems AP-120B. The purpose of this project is to assess the applicability of vector processors to power flow and transient stability simulation problems and to indicate how these problems should be organized to run efficiently on these machines. The approach taken was to survey the entire class of vector processors now or shortly to be available, to attempt to raise the reported low efficiency of sparsity-coded programs run on large vector processors by reorganization of their sparse structure, and to show how the most time-consuming parts of a simulation can be vectorized for these machines. Among the conclusionsmore » reached by this project are that: while the supercomputer class of vector processors can be made to process power system simulation problems very rapidly, the efficiency will be far less than for those problems for which these machines were designed; small array processors with sufficient floating-point word length (such as the AP-120B) promise an order of magnitude speed increase on these problems for a very reasonable price; and those portions of simulation programs requiring significant execution time and not concerned with sparse equation solution can be vectorized for any of these machines.« less