GPU platforms for highly parallel EMT simulation

A parallelized electromagnetic transient (EMT) simulation tool for power system transients will be developed in this research to accelerate the internal computation process. An EMT simulator uses a highly detailed representation for the power systems components. Conventional EMT simulators typically execute sequentially on a single processor; and computational effort increases significantly with network size. Graphics Processing Units (GPUs) have massively parallel architecture and can accelerate EMT simulation. The investigator’s recent Ph.D. research, was of a preliminary nature, and showed promising performance gain for GPU based EMT simulations. The main focus of this project is to explore the suitability of deploying multiple EMT instances on GPUs to increase the simulation speed for cases requiring concurrent processing of the EMT sub-systems. Also a larger class of parallelized models will be developed for GPU implementation for complex power systems components such as high-voltage dc (HVDC) transmission systems.

Faculty Supervisor: 
Aniruddha Gole
Partner University: