Parallelization of time domain dynamic simulation: Existing tools and new techniques

Enhancing predictability with “faster than real-time” dynamic simulation will enable the dynamic stability margin, proactive real-time control and will improve grid resiliency to fast time-scale phenomena. The existing tools used for dynamic simulation are optimized for single core of a PC. Therefore, they do not take advantage of multiple computational cores and do not meet the real-time need of the industry. The recent advancement in High Performance Computing (HPC) techniques and parallel numerical libraries presents an opportunity to improve the computational speed of dynamic simulation engines. In this context, this paper presents two approaches for parallelization. In the first approach the time domain simulation engine of GE Concorda PSLF has been parallelized using OpenMP under the existing architecture. Improvements in speed have been observed in cases that are large enough to slow down simulation time, such as the US Eastern Interconnect. Further speed gains are dependent on the acceleration of network solution, which is being researched. In the second approach an implicit integration technique with simultaneous solution of multiple time-steps is presented. An initial implementation of this technique on a small test system shows promising results.