Future proofing a building design using history matching inspired level‐set techniques. Issue 2 (19th December 2020)

• Record Type: Journal Article
• Authors: Baker, Evan; Challenor, Peter; Eames, Matt
• Is Part Of: Journal of the Royal Statistical Society. Volume 70:Issue 2(2021)
• Page Start: 335
• Page End: 350
• Publication Date: 2020-12-19
• Subjects: building performance simulation -- Gaussian processes -- history matching -- level set estimation -- stochastic simulation -- uncertainty quantification; Statistics -- Periodicals; 519.5
• DOI: 10.1111/rssc.12461
• Languages: English
• ISSNs: 0035-9254

Abstract

Abstract: How can one design a building that will be sufficiently protected against overheating and sufficiently energy efficient, whilst considering the expected increases in temperature due to climate change? We successfully manage to address this question—greatly reducing a large set of initial candidate building designs down to a small set of acceptable buildings. We do this using a complex computer model, statistical models of said computer model (emulators), and a modification to the history matching calibration technique. This modification tackles the problem of level‐set estimation (rather than calibration), where the goal is to find input settings which lead to the simulated output being below some threshold. The entire procedure allows us to present a practitioner with a set of acceptable building designs, with the final design chosen based on other requirements (subjective or otherwise).