The Role of Artificial Intelligence and Optimization Algorithms in Reducing Building Energy Consumption: A Narrative Systematic Review of Machine Learning Models and Genetic Algorithms
The Role of Artificial Intelligence and Optimization Algorithms in Reducing Building Energy Consumption: A Narrative Systematic Review of Machine Learning Models and Genetic Algorithms
Kamran Yeganegi1 Pardis Kokabi Moghadam2
1) Assistant Professor, Department of Industrial Engineering, Zanjan Branch, Islamic Azad University, Zanjan, Iran
2) M.Sc. Student (Energy Engineering), Zanjan Branch, Islamic Azad University, Zanjan, Iran
IOI Code
:
AAHF-ABABEA
Publication :
National Conference on Architecture and Art Technologies, Smart City, Soft and Creative Industries and Technological Culture(icact.ir)
Abstract :
Energy consumption in the building sector is one of the most critical global challenges in the fields of sustainable development and energy management, accounting for approximately 30–40% of total global energy use. Rising energy demand, rapid urbanization, and the impacts of climate change have made the optimization of building energy consumption a vital necessity. Traditional physical simulation methods such as EnergyPlus and TRNSYS, while powerful tools for energy modeling, face serious limitations when dealing with incomplete data, environmental uncertainties, nonlinear occupant behaviors, and the growing complexity of heating, ventilation, and air-conditioning (HVAC) systems.
In recent years, artificial intelligence—particularly machine learning—has emerged as a powerful approach for predicting energy consumption and modeling building behavioral patterns. Models such as artificial neural networks, support vector machines, random forests, XGBoost, and long short-term memory networks have demonstrated a high capability to capture complex energy patterns with very high accuracy. Alongside these advances, evolutionary algorithms such as the genetic algorithm, non-dominated sorting genetic algorithm II (NSGA-II), particle swarm optimization, and multi-objective particle swarm optimization have been widely applied to optimize building design, envelope parameters, shading devices, window-to-wall ratios, materials, and HVAC system components. Many studies report energy savings ranging from 10% to 40%.
Using a hybrid review approach (systematic–narrative review), this study analyzes and compares the performance of machine learning models, genetic algorithms, and hybrid machine learning–genetic algorithm models based on 15 Persian-language and 12 high-quality English-language articles. The results indicate that machine learning models achieve prediction accuracies between 0.85 and 0.99, genetic algorithms enable energy reductions of 10–40%, and hybrid models deliver the best overall performance by improving accuracy and reducing prediction errors.
This review also identifies key research gaps, including the lack of real-time operational data, limited modeling of occupant behavior, computational constraints of genetic algorithms, and a shortage of integrated industrial-scale studies. Future research directions are proposed, including the adoption of digital twins, explainable artificial intelligence, reinforcement learning, and the development of integrated prediction–optimization frameworks.
Keywords : artificial intelligence machine learning genetic algorithm energy optimization