Accelerating the large-scale simulation of power distribution networks by using the multi-GPU LIM

The GPGPU-based LIM has been proposed as one of the fast circuit simulation methods, which is much faster than the LIM with a single CPU. However, the memory capacity of the GPU memory size is not satisfied for solving large-scale problems. Thus, it is required to expand the memory size by using multiple GPUs in the GPGPU-based method. In this paper, multi-GPU based LIM is proposed for the solution of the large-scale problems. First, the LIM is reviewed briefly. Second, the proposed method is described for the fast circuit simulation. Finally, the efficiency of the proposed method is evaluated by some numerical results.