Elektrane (2025) [pp. 545-551]
AUTHOR(S) / AUTOR(I): Mirko Stojiljković, Vladan Jovanović, Dušan Ranđelović, Marko Ignjatović, Goran Vučković
Download Full Pdf 
DOI: https://doi.org/10.46793/EEP25.545S
ABSTRACT / SAŽETAK:
Building energy consumption is related to a large share of the primary energy consumption on the global level. Energy retrofit of buildings, which might involve improving envelopes, as well as installing efficient energy supply systems based on the renewable energy sources, has a potential to considerably reduce the consumption of energy and the environmental impact. An accurate assessment of the energy efficiency measures is essential for careful planning and decision making related to their implementation. Detailed building energy simulations have became widely-used tools for the assessment of these measures. The optimization of energy system design and operation parameters is often a very useful addition. However, this approach is often time-consuming and resource-intensive, and, as such, is being replaced or supplemented with much faster, yet precise enough, so called surrogate models. These models are data-driven and based on machine learning techniques. This paper considers the enhancements of the previously-defined surrogate models for the assessment of the energy retrofit measures, including envelope improvements, and the installation of photovoltaic systems and heat-pump-based heating and cooling. It applies supervised machine learning regression models, which are able to learn from a very limited number of data points, obtained with detailed simulations, and predict the primary energy consumption related to particular measures. The models are based on decision trees and their ensembles: random forest and gradient boosting. They are further improved by implementing with linear regression, which uses modified input features, as well as with the optimization of hyperparameters. The presented approach is tested on a building of a primary school. It yielded the improvement in the prediction performance of the surrogate models of 64–75%, compared to the baseline models. The resulting models, although obtained with relatively small datasets, are accurate enough, so they can partially replace detailed building simulations and energy system optimization. The most precise regression model is based on the combination of linear regression and gradient boosting.
KEYWORDS / KLJUČNE REČI:
machine learning, optimization, primary energy, school building retrofit, simulation surrogate models
ACKNOWLEDGEMENT / PROJEKAT:
This research was financially supported by the Ministry of Science, Technological Development and Innovation of the Republic of Serbia (Contract No. 451-03-137/2025-03/200109).
REFERENCES / LITERATURA:
[1] Thrampoulidis, E. et al., Approximating optimal building retrofit solutions for large-scale retrofit analysis, Applied Energy 333 (2023), 120566.
[2] Życzyńska, A. et al., Energy Effects of Retrofitting the Educational Facilities Located in South-Eastern Poland, Energies 13 (2020), No. 10, 2449.
[3] Ran elović, D. et al., Investigation of d̄ a passive design approach for a building facility: a case study, Energy Sources, Part A: Recovery, Utilization, and Environmental Effects 47 (2021), No. 1., pp. 8890–8908.
[4] Papadakis, N. and Katsaprakakis, D. A., A Review of Energy Efficiency Interventions in Public Buildings, Energies 16 (2023), No. 17, 6329.
[5] ****, The Future of Heat Pumps, International Energy Agency, Paris, France, 2022.
[6] Adebayo, P. et al., Development, modeling, and optimization of ground source heat pump systems for cold climates: A comprehensive review, Energy and Buildings 320 (2024), 114646.
[7] Kossi, P., Rama, M., Improving the accuracy of heat pump feasibility assessment, Thermal Science 28 (2024), No. 5B, pp. 4381–4394.
[8] Li, B. et al., Economy and energy flexibility optimization of the photovoltaic heat pump, system with thermal energy storage, Journal of Energy Storage 100 (2024), 113526.
[9] Pesola, A., Cost-optimization model to design and operate hybrid heating systems – case study of district heating system with decentralized heat pumps in Finland. Energy 281 (2023), 128241.
[10] Krützfeldt H. et al., MILP design optimization of heat pump systems in German residential buildings, Energy and Buildings 249 (2021) 111204.
[11] Stojiljković, M. M. et al., Cost-optimal operation of hybrid heat pump systems with progressive electricity tariffs, Thermal Science 29 (2025), 5A, pp. 3441–3452.
[12] ****, Energy Sector Development Strategy of the Republic of Serbia up to 2040 with Projections up to 2050, Official Gazette of the Republic of Serbia, No. 94 (2024) (In Serbian).
[13] Cruz, A. S. et al., Multi-objective optimization based on surrogate models for sustainable building design: A systematic literature review, Building and Environment 266 (2024), 112147,
[14] Shen, Y. and Pan, Y., BIM-supported automatic energy performance analysis for green building design using explainable machine learning and multi-objective optimization, Applied Energy 333 (2023), 120575.
[15] Zhu, Y. et al., Application of hybrid machine learning algorithm in multi-objective optimization of green building energy efficiency, Energy 316 (2025), 133581.
[16] Shi, Y. and Chen, P, Energy retrofitting of hospital buildings considering climate change: An approach integrating automated machine learning with NSGA-III for multi-objective optimization, Energy & Buildings 319, (2024) 114571.
[17] Sharif, S. A., and Hammad, A., Developing surrogate ANN for selecting near-optimal building energy renovation methods considering energy consumption, LCC and LCA, Journal of Building Engineering 25 (2019), 100790.
[18] Alexakis, K. et al., Genetic algorithm-based multi-objective optimisation for energy-efficient building retrofitting: A systematic review, Energy & Buildings 328 (2025), 115216.
[19] Manmatharasan, P. et al., AI-driven design optimization for sustainable buildings: A systematic review, Energy & Buildings 332 (2025), 115440.
[20] Stojiljković et al., Predicting primary energy savings of building retrofit measures with decision-tree-based ensemble methods, Facta Universitatis, Series: Working and Living Environmental Protection 17 (2020), No. 3, pp. 151–162.
[21] Shen, Y. and Pan, Y., BIM-supported automatic energy performance analysis for green building design using explainable machine learning and multi-objective optimization, Applied Energy 333 (2023), 120575.
[22] Stojiljković, M. M., et al., Assessment of low-energy potential of a school building using operation optimization and surrogate models, Thermal Science 2025, Online First.
[23] ****, EnergyPlus Version 23.2.0 Documentation, Engineering Reference, 2023.