Improved Numerical Methodologies on Power System Dynamic Simulation Using GPU Implementation

This paper proposes an improved numerical methodology for faster than real time (FTRT) power system dynamic simulation. It is implemented on a graphics processing unit (GPU) with parallel programming, and can handle large-scale power systems. In power system dynamic simulation, the biggest challenge for FTRT simulation is the matrix solver for the network equation. State of the art using the analog approach or parallel high-performance computing (HPC) can barely meet the real-time requirement on a system with approximately 10,000 buses and 2,000 generator-buses. We use the Sherman-Morrison-Woodbury formula to Gauss Block Elimination to take advantage of the low-rank change in the admittance matrix while lowering the computational dimension. With the proposed numerical methodology, we implemented the dynamic simulation on GPUs with classical generator models. To validate its FTRT capability of our methodology, we demonstrate this with the example of Western Electricity Coordinating Council (WECC) systems with over 20,000 buses and over 2,000 generator-buses being simulated, which is very capable of running over four times faster than real-time. The promising results exhibit outstanding scalable FTRT performance.