1. Shifted frequency analysis based, faster-than-real-time simulation of power systems on graphics processing unit.
- Author
-
Zhang, Junjie, Mittenbühler, Marcel, and Benigni, Andrea
- Subjects
- *
GRAPHICS processing units , *ELECTRIC transients , *POWER system simulation , *HETEROGENEOUS computing , *SIMULATION methods & models , *GRAPHIC methods - Abstract
In this paper we present a scalable approach based on the latency-based linear multistep compound method to use graphic processing unit (GPU) to accelerate the power system simulations with electromagnetic transients. To balance between the computational load and accuracy, we modeled the power system using shifted frequency analysis. Then, to efficiently exploit the hardware, our computational approach is designed to utilize both data-parallel and task-parallel execution. In addition, a graph-based thread safety design is introduced to achieve high scalability in the paralleled component computations. Furthermore, our implementation introduces a translation layer for different heterogeneous computing frameworks during compile time, so that a wide range of GPU devices can be supported without losing performance. Finally, benchmark results show that our approach achieves faster-than-real-time capability for systems with hundreds to thousands of nodes. [Display omitted] • Exploits data- and task-parallelism together in an automated way. • A graph-based approach to ensure thread safety to increase scalability. • A compile-time translation layer to allow execution on various platforms. • Faster-than-real-time for system with more than 10,000 dynamic components. • 82 × faster compared to an optimized sequential simulation. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF