GM 22 Tutorial: GridPACK - Framework and library for accelerating HPC in grid applications (Slides)

This tutorial will introduce GridPACK (Parallel Advanced Computing Kernels for Power Grid), an open-source platform for developing power grid applications designed to run on high-performance computing (HPC) architectures, ranging from a few cores to many cores and medium- and large-scale clusters. The objective of GridPACK is to lower the threshold of parallel programming to allow power grid application developers to focus on algorithms, instead of parallel computing, to address the emerging challenges of renewables, decarbonization, and extreme weather by significantly increasing the computational speed of grid simulations.

The GridPACK framework provides strong support for a wide variety of parallel solvers and parallels linear algebra that can be tailored for different simulation problems, with a low threshold for HPC implementation. The framework can be used to develop new power grid applications with prebuilt modules. Multiple case studies will be discussed to show that GridPACK can significantly increase computational speed for applications such as transient analysis, contingency analysis, and dynamic security assessment, using both task-level parallelism and multi-level parallelism via the task manager framework. Improved computational performance can enable larger and more comprehensive calculations to improve grid reliability, efficiency, and resilience.