Time: Wednesday, 13/Aug/2025: 1:40pm - 3:10pm |
Location: Cripple Creek 1 |
This session explores cutting-edge AI applications in building portfolio management and design, demonstrating how new tools can address persistent challenges in the built environment. The first presentation introduces a machine learning metamodel that leverages few-shot and zero-shot learning to rapidly estimate life cycle metrics for large building portfolios, overcoming traditional data limitations. The second presentation showcases a custom conversational AI assistant that automates early-stage building energy benchmarking, offering accessible and streamlined data analysis for both experts and non-experts. The final presentation examines how AI and traditional methods compare in modeling airflow and exhaust dispersion in complex building designs, discussing AI’s potential to enhance laboratory air quality and system performance while acknowledging critical physics constraints. Together, these talks highlight how AI-driven strategies and metamodels are redefining design workflows, accelerating energy benchmarking, and addressing climate-related demands in the built environment.
AI Agent for Early-Stage Building Energy Benchmarking
Affiliated Engineers, Inc.
This presentation introduces a custom conversational AI assistant developed to automate early-stage building energy benchmarking in the United States. The tool allows both non-experts and experts to quickly retrieve, analyze, and compare data from trusted public energy datasets (such as CBECS, I2SL, and Energy Star) through a user-friendly chatbot interface. By streamlining data access and analysis, it significantly improves the speed and accuracy of benchmarking during early design stages. This innovative approach supports integrated design workflows and helps set realistic energy target goals for projects.
AI in the Air: Transforming Airflow Analysis for Sustainable Buildings
RWDI
AI is emerging as a powerful tool to solve new design challenges, with limitations that must be understood. Additionally, fundamental physics challenges, such as modeling airflow and exhaust dispersion, remain key barriers to energy savings. This presentation explores modeling tools for optimizing building airflow to maximize long-term efficiency, comparing AI with traditional methods like numerical modeling, wind-tunnel testing, and Computational Fluid Dynamics. We will discuss AI applications in early-stage laboratory design, development challenges, and adoption barriers. As climate change drives new building demands, AI, alongside traditional performance-based design tools, can improve building air quality and enhance overall system performance.
Machine Learning Based Metamodel for Faster Life Cycle Assessment of Large Portfolio of Buildings
National Renewable Energy Laboratory (NREL), United States of America
Managing large portfolio of buildings involves decisions on reuse, retrofit, renovation, and new construction, influenced by trade-offs between cost, time, and operational flexibility over their life cycle. Traditional life cycle assessment tools are labor- and compute-intensive. Machine learning-based metamodels (surrogate models) offer faster alternatives but are limited by scarce life cycle data. Recent deep learning techniques, like few-shot and zero-shot learning, enable learning from sparse data. We propose a machine learning metamodel that leverages these techniques for rapid estimation of key life cycle metrics. The presentation will cover the model’s architecture, and potential for rapidly evaluating large portfolio of buildings.