karaktus / beobenchmark

Building Energy Optimization (BEO) problem test bed for benchmarking black-box optimization algorithms

BEOBenchmark

This repository aims to provide a Building Energy Optimization (BEO) problem test bed for benchmarking black-box optimization algorithms. Previous benchmarking studies in BEO usually included very few problems, thus inhibiting a fair and traceable performance comparison of optimization algorithms. By providing a larger set of BEO problems, we can facilitate reproducible future benchmarking studies, where researchers can focus on improving and designing algorithms suited to solve building energy design problems. Furthermore, it is crucial that benchmarks are conducted on the same problem set in order to make generalizable conclusions.

The first test bed version BEOBenchmark_V1 includes 15 single-objective BEO problems from the literature, all using EnergyPlus as simulator. Original files are updated to EnergyPlus V 8.5.0 and implemented in Rhinoceros 3D Grasshopper as common software platform. An easy interface for connecting any optimization algorithm (library) to Grasshopper is given with the FrOG tool (Framework for Optimization in Grasshopper). This test bed is used in a forthcoming journal publication.

Contributions and collaborations are more than welcome to extend the test bed! Further developments could include more EnergyPlus problems, problems using other simulators (such as Modelica, CitySim, or BRCM), implementations in other software platforms such as GenOpt, or multi-objective problems.

Contact us under: [email protected]