Simulation of large-scale electrical power networks on graphics processing units

Abstract

In this paper suitability of applying graphics processing unit (GPU) based computing for electromagnetic transient (EMT) simulation of large scale power systems is demonstrated. As the network size increases there is a corresponding increase in simulation time with conventional central processing unit (CPU) based simulation tools. The paper shows that with a hybrid environment consisting of CPUs and GPUs, simulation time is much less compared to the CPU-only implementations. In this scheme the GPU is mainly used to do the computationally intensive part of the simulation in parallel on its built-in multiple processing cores, and the CPU is assigned for updating history terms and flow control of the simulation. The GPU-based approach is used to simulate a network with 117 nodes, and it is shown that the CPU-GPU based implementation takes less than half of the time taken by the CPU-only implementation of the simulation.