Machine Learning for Sustainable Building Design: Energy, Emissions & Comfort
This study investigates the application of six machine learning regression models to predict building performance in a residential unit located in Sari, Iran. Using a calibrated EnergyPlus model and three years of utility data, the research evaluates primary energy consumption, emissions, indoor air quality, thermal comfort, and visual discomfort. The aim is to enhance sustainable design decision-making by comparing the efficiency and accuracy of different models.
Machine Learning Models in Building Performance
The research evaluates Random Forest, K-Nearest Neighbors, Support Vector Regression, Artificial Neural Network, Extreme Gradient Boosting, and Linear Regression. Each model is tested against real-world performance indicators, highlighting their predictive strength and weaknesses in handling multidimensional building datasets.
Dataset and Methodology
A synthetic dataset of 1,826 configurations with 25 input variables was developed using EnergyPlus. The dataset was split into training and testing sets, ensuring robust validation of model performance. Grid Search and Bayesian Optimization were used for hyperparameter tuning, enhancing predictive reliability.
Model Performance and Evaluation Metrics
Performance was assessed using RMSE, R², and MAPE across five metrics. Random Forest and XGBoost emerged as the most accurate models, achieving R² values above 0.91. In contrast, Linear Regression underperformed with R² values between 0.35 and 0.50, confirming the superiority of advanced ensemble methods.
Sensitivity, SHAP, and Scenario Analysis
Sensitivity and SHAP analyses revealed that ventilation strategies and HVAC configurations significantly influence building performance. Scenario analysis with 1,000 bootstrap iterations highlighted trade-offs between energy efficiency and air quality, offering valuable insights into sustainable design choices.
Optimization for Sustainable Building Design
Optimization results demonstrated that energy consumption and CO₂-equivalent emissions can be reduced by over 30% while improving indoor comfort. The study concludes that Random Forest and XGBoost are highly robust models for supporting sustainable building design optimization and policy-making.
Architecture Engineers Awards
🔗 Nominate now! 👉 https://architectureengineers.com/award-nomination/?ecategory=Awards&rcategory=Awardee 🌐 Visit: architectureengineers.com 📩 Contact: contact@architectureengineers.com Get Connected Here: ***************** Instagram : https://www.instagram.com/architecture_engineers_awards/ Facebook : https://www.facebook.com/profile.php?id=61576995475934 Tumblr : https://www.tumblr.com/blog/architectureengineers Pinterest : https://in.pinterest.com/ Blogger : https://architectureengineers.blogspot.com/ Twitter : https://twitter.com/Architectu54920 YouTube : https://www.youtube.com/@Architechtureengineer Linkedin : https://www.linkedin.com/in/architecture-engineer-01a044361/
#MachineLearning
#BuildingPerformance
#EnergyEfficiency
#SustainableDesign
#ArtificialIntelligence
#RandomForest
#XGBoost
#NeuralNetworks
#DataDrivenDesign
#EnergyModeling
#SmartBuildings
#SustainableArchitecture
#CO2Reduction
#IndoorAirQuality
#ThermalComfort
#VisualComfort
#GreenBuildings
#ClimateAction
#EnergyOptimization
#BuildingScience
Comments
Post a Comment