The use of optimisation in the calibration of building simulation models

Basurra, S.; Jankovic, L.; Huws, H.

Abstract

In this paper, a model-assisted calibration is proposed in order to reduce performance gap of simulation models, being the estimated differences between the predicted performance of the simulation model and the actual performance of the building. We use the nearly unbiased 𝑘 Nearest Neighbour (KNN) algorithm to classify the solution that exhibits the minimum performance gap based on a set of reference points. Density avoidance algorithm is used to further refine the solutions by finding regions in the space of input factors for which the model output is either maximum or minimum to meet optimum criterion, hence fine tuning the model to establish one-to-one relationship between the simulated and actual performance.

• Presentation Conference Type: Conference Paper (published)
• Conference Name: Building Simulation 2015: 14th Conference of IBPSA
• Start Date: Dec 7, 2015
• End Date: Dec 9, 2015
• Publication Date: 2015
• Deposit Date: Oct 14, 2024
• Pages: 2905-2912
• Book Title: Proceedings of Building Simulation 2015: 14th Conference of IBPSA
• ISBN: 9789352301188
• Publisher URL: https://publications.ibpsa.org/conference/?id=bs2015