Parametric Evaluation of a PCM-Integrated Exterior Wall Across Turkish Climate Zones Using Building Energy Simulation and Machine Learning
Buildings, cilt.16, sa.18, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 16 Sayı: 18
- Basım Tarihi: 2026
- Doi Numarası: 10.3390/buildings16183683
- Dergi Adı: Buildings
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Avery, Compendex, INSPEC, Directory of Open Access Journals, Natural Science Collection (ProQuest), Materials Science & Engineering Collection (ProQuest), Technology Collection (ProQuest)
- Anahtar Kelimeler: building energy efficiency, building energy simulation, DesignBuilder, machine learning, phase change material
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- İstanbul Üniversitesi Adresli: Evet
Özet
This study evaluates the energy and operational carbon performance of phase change material (PCM) integrated into one external wall of a reference office building across 19 Turkish cities representing six TS 825 climate zones. A full-factorial parametric analysis was conducted using DesignBuilder by varying five nominal melting temperatures (21–29 °C), three PCM thicknesses (5, 10, and 20 mm), and four façade orientations. The lowest-energy tested PCM configuration reduced annual total site energy consumption in all investigated cities, with savings ranging from 0.38% in Kayseri to 2.86% in Samsun and average savings of 1.89%. The corresponding CO2 reductions ranged from 0.26% to 1.84%, with an average reduction of 1.25%. Among the investigated thicknesses, a 20 mm PCM layer produced the lowest annual total site energy consumption in all 19 cities. Melting temperatures of 21 °C and 23 °C generally provided the greatest energy savings, while the best-performing façade orientation varied with climate. A machine-learning surrogate model was also developed for rapid PCM screening by comparing Ridge regression, Random Forest, and Gradient Boosting models. Random five-fold cross-validation produced high predictive accuracy. Leave-one-city-out validation showed limited accuracy for absolute annual energy prediction in cities excluded from model training but stronger performance for energy- and CO2-saving percentages. The model is therefore suitable for preliminary screening and ranking PCM configurations, while detailed simulations remain necessary for final design validation.