ExaPF

ExaPF.jl is a package to solve powerflow problem on exascale architecture. ExaPF.jl aims to implement a reduced method for solving the optimal power flow problem (OPF) fully on GPUs. Reduced methods enforce the constraints, represented here by the power flow's (PF) system of nonlinear equations, separately at each iteration of the optimization in the reduced space. This paper describes the API of ExaPF.jl for solving the power flow's nonlinear equations (NLE) entirely on the GPU. This includes the computation of the derivatives using automatic differentiation, an iterative linear solver with a preconditioner, and a Newton-Raphson implementation. All of these steps allow us to run the main computational loop entirely on the GPU with no transfer from host to device.

This implementation will serve as the basis for the future optimal power flow (OPF) implementation as a nonlinear programming problem (NLP) in the reduced space.

To make our implementation portable to CPU and GPU architectures we leverage two abstractions: arrays and kernels. Both of these abstractions are supported through the packages CUDA.jl and KernelAbstractions.jl Please take a look at the autodiff and linear solver implementations to get a design overview of ExaPF.jl targeted for GPUs.

The user API is separated into three layers:

• First layer: Physical layer, specify the power network topology in powersystem
• Second layer: Interface between power network and NLE or NLP in formulations
• Third layer: Evaluators for non-linear problems

The third layer is for users working in optimization whereas the first layer is for electrical engineers. They meet in the second layer.

Table of contents

Manual

• Quick Start
• AutoDiff
• Linear Solver
• PowerSystem
• Formulations
• Evaluators

Library

• PowerSystem
• Formulations
• Evaluators

Funding

This research was supported by the Exascale Computing Project (17-SC-20-SC), a joint project of the U.S. Department of Energy’s Office of Science and National Nuclear Security Administration, responsible for delivering a capable exascale ecosystem, including software, applications, and hardware technology, to support the nation’s exascale computing imperative.