Motivation and Goals

Implementing building-energy retrofits is unavoidably part of achieving a sustainable built environment. Currently, the mismatch between predicted and actual energy performance hinders the effectiveness of retrofits and the efficient operation of buildings. This project aims to leverage artificial intelligence to enhance retrofit outcomes by systematically selecting optimal types, times, and locations for measurements, also known as measurement-system-design, and by improving diagnostics of existing-building performance through system identification. Additionally, this project seeks to integrate methods that increase the trustworthiness of AI, enabling reliable and confident retrofit decisions.

Recent Results 

An entropy-based method for measurement-system-design and a modified Bayesian approach for system identification is implemented, demonstrating the potential of these methods in the context of building energy retrofits. Preliminary results have been presented at the 2024 GNI Symposium, with additional journal and conference publications planned for late 2024 and early 2025, respectively. Next steps include further establishing performance and validating our approach through realistic, full-scale case studies.