Logo image
Back
Indoor Temperature and Humidity Prediction in Residential Buildings: A Review of the White Box and Black Box Modeling Techniques
Journal article   Open access   Peer reviewed

Indoor Temperature and Humidity Prediction in Residential Buildings: A Review of the White Box and Black Box Modeling Techniques

Chima Cyril Hampo, Leah Schinasi and Simi T Hoque
Building and environment, v 290, 114125
Feb 2026
DOI: https://doi.org/10.1016/j.buildenv.2025.114125
Featured in Collection :   Research Supported by Drexel Libraries' OA Programs
url
https://doi.org/10.1016/j.buildenv.2025.114125View
Published, Version of Record (VoR) Open Access via Drexel Libraries Read and Publish Program 2025 Open CC BY-NC V4.0

Abstract

Energyplus Physics based modeling Data driven modeling Indoor humidity Indoor temperature Residential Buildings Computational Fluid Dynamics Machine Learning
Residential buildings account for a substantial portion of global energy consumption and are critical for ensuring safe and comfortable indoor environments. Accurate prediction of indoor temperature and humidity is essential for thermal comfort, energy efficiency, and occupant health, yet remains methodologically complex. This review synthesizes 91 peer-reviewed studies published between 2019 and 2025 that applied white box (physics-based) and black box (data-driven) modeling approaches to residential buildings. Studies were identified through Google Scholar and Web of Science, screened using defined inclusion and exclusion criteria, and evaluated based on model type, predictive variables, validation method, and performance metrics. White box models, including nodal, zonal, computational fluid dynamics (CFD), and hybrid physics-based frameworks, capture mechanistic heat and moisture transfer processes and have evolved toward more integrated hygrothermal and airflow coupling since 2021. Black box methods, including shallow and deep neural networks, regression models, and hybrid or ensemble architectures, demonstrate high predictive accuracy for short-term and real-time applications, often achieving sub-degree temperature errors and increasingly predicting both temperature and humidity. Comparative findings show that temperature remains the predominant predictive variable, while humidity, though vital for comfort and health, is less frequently modeled. Geographically, studies are concentrated in Europe, East Asia, and North America, with limited representation of tropical and Global South regions. Seasonally, most research has emphasized heating conditions, though recent efforts address cooling, overheating, and mixed-mode ventilation. Evaluation remains fragmented across metrics such as RMSE, MAE, and correlation coefficients, underscoring the need for standardized reporting practices. Overall, physics-based and data-driven approaches are complementary. The former provide interpretability and physical realism, while the latter offer adaptability and scalability. Future research should prioritize coupled heat–moisture modeling, extend analysis to underrepresented climates and housing types, and develop unified validation benchmarks. Advancing along these directions will strengthen predictive reliability and support healthier, more energy-resilient residential environments.

Metrics

14 Record Views
4 citations in Web of Science

Details

UN Sustainable Development Goals (SDGs)

This publication has contributed to the advancement of the following goals:

#7 Affordable and Clean Energy
#11 Sustainable Cities and Communities
#13 Climate Action

InCites Highlights

Data related to this publication, from InCites Benchmarking & Analytics tool:

Web of Science research areas
Construction & Building Technology
Engineering, Civil
Engineering, Environmental
Logo image