Contribution type: Lightning talk and poster
The Open Energy Benchmark
Presenters: Siddharth Krishna
Description:
The Open Energy Benchmark is an open and reproducible platform for evaluating optimization solvers on real-world energy system planning problems. It compares four open-source
solvers (CBC, GLPK, HiGHS, and SCIP) and a proprietary baseline (Gurobi) across 213 linear and mixed-integer problems from 13 energy modelling frameworks. Solvers are assessed under standardized computational settings using metrics such as runtime, memory consumption, solve success, and other established solution quality indicators. Among the recent findings, the new HiPO method in HiGHS increased the number of problems that can be solved by open-source solvers by 7%, completing complex models with millions of variables up to 15 times faster. The lightning talk will introduce the benchmark, highlight its main findings, and outline opportunities for researchers and modelling teams to contribute new problems and evaluate solver performance on their applications.
Background:
Optimization solvers are central to energy system modelling, but their performance can vary substantially depending on model formulation, problem size, temporal and spatial resolution, and solver configuration. Comparisons based on small or synthetic test cases may therefore provide limited guidance for practical modelling applications. The Open Energy Benchmark addresses this gap by testing solvers on a transparent and continuously expanding collection of problems derived from established energy modelling frameworks. The current benchmark includes linear and mixed-integer problems spanning a wide range of sizes and formulations. Rather than identifying a universal “best” solver, the project aims to improve understanding of solver performance, scaling choices, solver options, and what is feasible to solve with open-source and proprietary solvers.
Optional links