A genetic algorithm applied to the validation of building thermal models

TL;DR

This paper introduces a method combining genetic algorithms with thermal simulation to identify faulty components in building models, validated on numerical and real building cases, demonstrating its effectiveness.

Contribution

It presents a novel coupling of genetic algorithms with thermal models to detect defective sub-models, improving model validation processes.

Findings

Effective identification of faulty sub-models

Validated on numerical and real building data

Potential for improved model validation

Abstract

This paper presents the coupling of a building thermal simulation code with genetic algorithms (GAs). GAs are randomized search algorithms that are based on the mechanisms of natural selection and genetics. We show that this coupling allows the location of defective sub-models of a building thermal model i.e. parts of model that are responsible for the disagreements between measurements and model predictions. The method first of all is checked and validated on the basis of a numerical model of a building taken as reference. It is then applied to a real building case. The results show that the method could constitute an efficient tool when checking the model validity.