Citation :

Paulson Geo Philip, "An Intelligent Hybrid Model for Predicting Building Thermal Load from Geometric and Envelope Attributes," International Journal of Civil Engineering, vol. 13, no. 9, pp. 1-18, 2026. Crossref, https://doi.org/10.14445/23488352/IJCE-V13I9P101

Abstract

Prediction of building thermal loads is vital for improving energy-efficient building design, sustainable infrastructure planning, and intelligent energy management in modern civil engineering applications. Existing thermal load prediction models often struggle to attain nonlinear dependencies and feature interactions in structured building datasets. Several existing studies also focus primarily on component-level thermal analysis or restricted climatic conditions, reducing broader prediction generalization and interpretability. To overcome these limitations, this model introduces a hybrid TabNet-ResMLP approach combined with interaction-based analysis to provide interpretable insights into the influence of geometric and envelope parameters on building thermal load. TabNet was employed for attentive feature selection and contextual feature weighting, while ResMLP performed residual nonlinear regression for modelling complex thermal behavior. The proposed model was validated using the Energy Efficiency dataset. Experimental evaluation achieved Root Mean Square Error (RMSE) values of 1.65 and 1.95, Mean Absolute Error (MAE) values of 1.12 and 1.38, Mean Absolute Percentage Error (MAPE) values of 6.0% and 7.1%, and Coefficient of Determination (R²) values of 0.98 and 0.97 for Heating Load (HL) and Cooling Load (CL), respectively. Feature importance, sensitivity, and interaction analyses further identified relative compactness, Surface Area (SA) and Overall Height (OH) as dominant parameters influencing thermal demand variation. The findings establish the efficacy of the proposed framework for early-stage building energy assessment, sustainable architectural design, and energy-efficient building operation.

Keywords

Building energy, Cooling Load, Deep Learning, TabNet, Sensitivity analysis, Heating Load.

References

1. Fabrizio Ascione et al., “Design of the Building Envelope: A Novel Multi-Objective Approach for the Optimization of Energy Performance and Thermal Comfort,” Sustainability, vol. 7, no. 8, pp. 10809-10836, 2015.
   [CrossRef] [Google Scholar] [Publisher Link]
2. Yiqing Liu, Zhiwen Luo, and Sue Grimmond, “Impact of Building Envelope Design Parameters on Diurnal Building Anthropogenic Heat Emission,” Building and Environment, vol. 234, pp. 1-12, 2023.
   [CrossRef] [Google Scholar] [Publisher Link]
3. Ziyang Hao et al., “Simplified Model of Heat Load Prediction and Its Application in Estimation of Building Envelope Thermal Performance,” Buildings, vol. 13, no. 4, pp. 1-24, 2023.
   [CrossRef] [Google Scholar] [Publisher Link]
4. Meng Wang et al., “A Knowledge Graph-based Framework to Automate the Generation of Building Energy Models using Geometric Relation Checking and HVAC Topology Establishment,” Energy and Buildings, vol. 325, pp. 1-16, 2024.
   [CrossRef] [Google Scholar] [Publisher Link]
5. Rajendra Roka et al., “A Systematic Review of Sensitivity Analysis in Building Energy Modeling: Key Factors Influencing Building Thermal Energy Performance,” Energies, vol. 18, no. 9, pp. 1-27, 2025.
   [CrossRef] [Google Scholar] [Publisher Link]
6. Navid Shirzadi, Dominic Lau, and Meli Stylianou, “Surrogate Modeling for Building Design: Energy and Cost Prediction Compared to Simulation-Based Methods,” Buildings, vol. 15, no. 13, pp. 1-22, 2025.
   [CrossRef] [Google Scholar] [Publisher Link]
7. Youness El Mghouchi, and Mihaela Tinca Udristioiu, “Thermal Load Predictions in Low-Energy Buildings: A Hybrid AI-Based Approach Integrating Integral Feature Selection and Machine Learning Models,” Applied Sciences, vol. 15, no. 11, pp. 1-23, 2025.
   [CrossRef] [Google Scholar] [Publisher Link]
8. Osama Mohsen, “Machine Learning Framework for Predicting Thermal Performance of Building Envelopes in Early-Stage Architectural Design,” Proceedings of International Structural Engineering and Construction, pp. 1-6, 2025.
   [CrossRef] [Google Scholar] [Publisher Link]
9. Xuelong Zhang et al., “Building Cooling Load Prediction based on GWO-CNN-LSTM,” Energies, vol. 19, no. 2, pp. 1-19, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
10. Liliane Magnavaca de Paula, Amr Oloufa, and Omer Tatari, “BIM-based Machine Learning Framework for Early-Stage Building Energy Performance Prediction,” Applied Sciences, vol. 16, no. 1, pp. 1-22, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
11. Yingfeng Wei et al., “An Interpretable Building Air Conditioning Load Forecasting Framework using SSA-Optimized Bi-LSTM and SHAP Analysis,” Frontiers in Environmental Science, vol. 14, pp. 1-14, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
12. Nurlan Zhangabay et al., “Assessment and Numerical Modeling of the Thermophysical Efficiency of Newly Developed Adaptive Building Envelopes Under Variable Climatic Impacts,” Buildings, vol. 16, no. 2, pp. 1-28, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
13. Amjad Baset, and Muhyiddine Jradi, “Prioritizing Building Envelope Retrofits through Data-Driven Heat Flux Prediction,” Journal of Building Engineering, vol. 119, pp. 1-13, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
14. Nurlan Zhangabay et al., “Analysis of the Influence of the Geometry of a Ventilated Facade Subsystem on its Stress–Strain State and the Coefficient of Thermal Nonuniformity,” Results in Engineering, vol. 29, pp. 1-17, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
15. Alireza Karimi, David Moreno-Rangel, and Antonio García-Martínez, “AI-Driven Morphological Optimization for Resilient Retrofitting of Mediterranean Housing Stock under Compounded Urban Heat Island and HEATWAVE Effects,” Building and Environment, vol. 291, pp. 1-21, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
16. Liliane Magnavaca de Paula, Amr Oloufa, and Omer Tatari, “Integrating BIM and Machine Learning for Energy and Carbon Performance Prediction in Office Building Design,” Eng, vol. 7, no. 2, pp. 1-31, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
17. Simon Muhič et al., “Influence of Building Envelope Modeling Parameters on Energy Simulation Results,” Sustainability, vol. 17, no. 12, no. 1-17, 2025.
    [CrossRef] [Google Scholar] [Publisher Link]
18. Dun Cao et al., “Research on the Distribution of the Energy-Saving Benefits of Building Geometric Parameters Under Different Climate Conditions,” Buildings, vol. 15, no. 17, pp. 1-19, 2025.
    [CrossRef] [Google Scholar] [Publisher Link]
19. Yongbao Chen et al., “Machine Learning Approach to Predict Building Thermal Load Considering Feature Variable Dimensions: An Office Building Case Study,” Buildings, vol. 13, no. 2, pp. 1-24, 2023.
    [CrossRef] [Google Scholar] [Publisher Link]
20. Fayez Abdel-Jaber, and Kim N. Dirks, “Thermal Load Prediction in Residential Buildings using Interpretable Classification,” Buildings, vol. 14, no. 7, pp. 1-14, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
21. Adam Slowik, and Dorin Moldovan, “Multi-Objective Plum Tree Algorithm and Machine Learning for Heating and Cooling Load Prediction,” Energies, vol. 17, no. 12, pp. 1-23, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
22. Shahid Mahmood et al., “Active Learning-based Machine Learning Approach for Enhancing Environmental Sustainability in Green Building Energy Consumption,” Scientific Reports, vol. 14, no. 1, pp. 1-18, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
23. Amal H. Alharbi et al., “Forecasting of Energy Efficiency in Buildings using Multilayer Perceptron Regressor with Waterwheel Plant Algorithm Hyperparameter,” Frontiers in Energy Research, vol. 12, pp. 1-16, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
24. Energy Efficiency Dataset, Kaggle. [Online]. Available: https://www.kaggle.com/datasets/elikplim/eergy-efficiency-dataset
25. Kevin McDonnell et al., “Deep Learning in Insurance: Accuracy and Model Interpretability using TabNet,” Expert Systems with Applications, vol. 217, pp. 1-10, 2023.
    [CrossRef] [Google Scholar] [Publisher Link]
26. Xinya Zhou et al., “RMLP-Cap: An End-to-End Parasitic Capacitance Extraction Flow Based on ResMLP,” Electronics, vol. 15, no. 1, pp. 1-23, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]