Framework for emulation and uncertainty quantification of a stochastic building performance simulator

Abstract

A good framework for the quantification and decomposition of uncertainties in dynamic building performance simulation should: (i) simulate the principle deterministic processes influencing heat flows and the stochastic perturbations to them, (ii) quantify and decompose the total uncertainty into its respective sources, and the interactions between them, and (iii) achieve this in a computationally efficient manner. In this paper we introduce a new framework which, for the first time, does just that. We present the detailed development of this framework for emulating the mean and the variance in the response of a stochastic building performance simulator (EnergyPlus co-simulated with a multi agent stochastic simulator called No-MASS), for heating and cooling load predictions. We demonstrate and evaluate the effectiveness of these emulators, applied to a monozone office building. With a range of 25-50 kW h/m 2 , the epistemic uncertainty due to envelope parameters dominates over aleatory uncertainty relating to occupants' interactions, which ranges from 6-8 kW h/m 2 , for heating loads. The converse is observed for cooling loads, which vary by just 3 kW h/m 2 for envelope parameters, compared with 8-22 kW h/m 2 for their aleatory counterparts. This is due to the larger stimuli provoking occupants' interactions. Sensitivity indices corroborate this result, with wall insulation thickness (0.97) and occupants' behaviours (0.83) having the highest impacts on heating and cooling load predictions respectively. This new emulator framework (including training and subsequent deployment) achieves a factor of c.30 reduction in the total computational budget, whilst overwhelmingly maintaining predictions within a 95% confidence interval, and successfully decomposing prediction uncertainties.

Validation of a Computer Code for the Energy Consumption of a Building, with Application to Optimal Electric Bill Pricing

“…We have used a dynamic thermal code implemented with the Dymola software, but other computer codes are available, such as the stochastic building performance simulator (S-BPS) considered in Wate et al (2020), who used a Gaussian process emulator. We have considered a Bayesian approach but others have been proposed in literature; for example, Shamsi et al (2020) took a fuzzy approach.…”

“…This includes: probabilistic modeling of different sources of uncertainty using expert opinion, sensitivity analysis, meta-modeling to accelerate calculations (Sacks et al 1989), and uncertainty propagation using either Monte-Carlo (or quasi Monte-Carlo) methods, or quadratic summation via Taylor's approximation. Other aspects about energy in buildings have been analyzed in, e.g., Wate et al (2020), who considered the effect of insulation thickness and window transmittance on the annual heating and cooling demand per unit floor area.…”

Evaluation of housing stock indoor air quality models: A review of data requirements and model performance

“…The methods used to quantify these uncertainties include the use of clustering techniques, which reflect the variability between groups, and Monte-Carlo sampling methods that account for variability in the descriptive parameters within groups [76]. More recent methods include Gaussian process emulators for uncertainty quantification and sensitivity analyses to perform complex stochastic building performance modelling [77,78].…”

Impact of adjustment strategies on building design process in different climates oriented by multiple performance

“…For example, Pilechiha et al focus on the trade-offs of window design on the quality of views, daylight and building energy loads in an office room [44]. Wate et al presented a framework for the quantification and decomposition of uncertainties in a dynamic building performance (heating and cooling load) simulation of a hypothetical office building [45]. Najjar et al proposed an optimization framework for sustainable building by integrating a building modeling and life cycle assessment [46].…”