Combining adaptive and heat balance models for thermal sensation prediction: A new approach towards a theory and data‐driven adaptive thermal heat balance model

Abstract

The adaptive thermal heat balance (ATHB) framework introduced a method to account for the three adaptive principals, namely physiological, behavioral, and psychological adaptation, individually within existing heat balance models. This work presents a more detailed theoretical framework together with a theory‐driven empirical determination toward a new formulation of the ATHBPMV. The empirical development followed a rigor statistical process known from machine learning approaches including training, validation, and test phase and makes use of a subset (N = 57 084 records) of the ASHRAE Global Thermal Comfort Database. Results show an increased predictive performance among a wide range of outdoor climates, building types, and cooling strategies of the buildings. Furthermore, individual findings question the common believe that psychological adaptation is highest in naturally ventilated buildings. The framework offers further opportunities to include a variety of context‐related variables as well as personal characteristics into thermal prediction models, while keeping mathematical equations limited and enabling further advancements related to the understanding of influences on thermal perception.

Double-Skin Facades for Thermal Comfort and Energy Efficiency in Mediterranean Climate Buildings: Rehabilitating Vulnerable Neighbourhoods

“…The data that has been obtained for analysis from each hypothesis was percentages of hours of comfort temperature and energy demand depending on the set-point temperature established in Table 3. With regard to thermal comfort, the results of the interior temperature in free evolution were analysed according to the adaptive comfort methodology [61,62] defined by the European Standard EN 16798-1:2019 "Energy performance of buildings. Ventilation for buildings.…”

Analysis of outlier detection rules based on the ASHRAE global thermal comfort database

“…To avoid outliers, research [25] and [44] set a minimum sample size for each analyzed group, but extreme values in normal groups were not addressed. The method of inspection was used in research [45] to filter out unexpected values, but it would require too much labor when the sample size is large. The Boxplot rule was used in research [38] and [26] to select outliers, but a more systematic view of different outlier removal approaches in the thermal comfort domain has yet to be discovered.…”

“…In thermal comfort studies, outlier detection is usually involved in data preprocessing to remove the misleading effects of extreme values. The available techniques include stochastic-based methods like the 3-Sigma rule [72], the Boxplot rule [73], and the Hampel rule [74]; distance-based methods such as cook distance [75] and k-nearest neighbour (KNN) [76]; manually inspection [45] or setting fixed ranges [77]; and binning or adjusting variables into specific intervals [78]. The majority of studies in thermal comfort research filtered outliers using stochastic-based methods, and more details can be found in Appendix B.…”