Periodic steady state determination of power systems using graphics processing units

Modern advances in computer technology will definitely have a great impact on the methodologies used in the expansion and operational planning of power systems. Parallel and distributed systems are the new tendencies in technologies that present a great potential for application in these areas. Graphics processing units (GPUs) have recently become popular in several fields of science due to their capability of performing massively large computational tasks. The use of a GPU for efficiently obtaining the periodic steady state solution in the time domain of large-scale electric power systems with nonlinear components is presented in this paper. Two programming models of the numerical differentiation process are proposed and implemented on a single GPU. The computation time of the simulation performed by co-processing GPU-CPU has been compared with the simulation time required when using a conventional CPU. Several test cases have been analyzed to demonstrate the acceleration of the GPU-CPU simulation and the speed-up achieved.