Benchmarking Large-Scale ACOPF Solutions and Optimality Bounds

TL;DR

This paper benchmarks various open-source tools for solving large-scale ACOPF problems, comparing their performance and optimality bounds across extensive instances up to 30,000 nodes.

Contribution

It provides a comprehensive comparison of ACOPF solution methods and introduces state-of-the-art optimality bounds using advanced semidefinite programming techniques.

Findings

Performance differences among tools across network sizes

Identification of the most efficient solvers for large-scale ACOPF

Establishment of new optimality bounds for benchmark instances

Abstract

We present the results of a comprehensive benchmarking effort aimed at evaluating and comparing state-of-the-art open-source tools for solving the Alternating-Current Optimal Power Flow (ACOPF) problem. Our numerical experiments include all instances found in the public library PGLIB with network sizes up to 30,000 nodes. The benchmarked tools span a number of programming languages (Python, Julia, Matlab/Octave, and C$++$), nonlinear optimization solvers (Ipopt, MIPS, and INLP) as well as different mathematical modeling tools (JuMP and Gravity). We also present state-of-the-art optimality bounds obtained using sparsity-exploiting semidefinite programming approaches and corresponding computational times.