TL;DR:
- The data-driven model aims to optimize building energy performance through artificial neural networks and machine learning algorithms.
- Developed by Florida Tech researchers, the model trains on large datasets generated through a Python-based software script and the EnergyPlus simulation tool.
- Optimization is achieved through genetic algorithms and Bayesian optimization, allowing for continuous adaptation to changes in operational parameters through data from building sensors.
- This leads to more accurate predictions of energy consumption, reduced CO2 emissions, and energy savings for building owners.
- The study is part of a larger effort to improve building energy modeling (BEM) and was published in the ASME Journal of Engineering for Sustainable Buildings and Cities in December 2022.
- A separate study explored the impact of COVID-19 on building energy performance and the adaptability of BEMs.
Main AI News:
Florida Tech Researchers Utilize Machine Learning to Reduce Environmental Impact
Building energy use and greenhouse gas emissions in the United States account for over 40% of the country’s total energy consumption, and a recent study from Florida Tech seeks to tackle this problem through the application of machine learning.
The study, featured in the January edition of the Energies journal, presents a novel approach to building energy modeling (BEM) and optimization. Titled “A Novel Approach for Optimizing Building Energy Models Using Machine Learning Algorithms,” the research was authored by Hamidreza Najafi, associate professor of mechanical engineering, and Benjamin Kubwimana, a master’s mechanical engineering graduate from Florida Tech.
Current BEM practices require extensive manual input of various variables, including building materials and operational parameters, which can result in a significant amount of time and effort. Najafi and Kubwimana’s work aims to streamline this process through the development of a Python-based software script that automates data entry into the EnergyPlus physics-based building energy simulation tool. This allows for the creation of large datasets to develop a surrogate energy simulation model using machine learning algorithms, specifically artificial neural networks.
Optimization of the surrogate model is achieved through the use of two approaches: genetic algorithm and Bayesian optimization. This allows for the building design to be optimized and continuously adapted to changes in operational parameters, leading to more accurate predictions of energy consumption and savings, as well as reduced CO2 emissions.
The study is part of a larger effort to enhance BEMs and expand their applications, providing value to building owners and developers throughout the building’s lifespan. The authors’ previous research also explored the impact of COVID-19 on building energy performance and how BEMs can adapt to remain accurate in changing conditions.
Optimizing Building Energy Performance: Automated Approach Utilizes Machine Learning and Digital Twins to Predict Energy Consumption and Reduce CO2 Emissions
Artificial neural networks and machine learning algorithms are at the heart of a new data-driven model aimed at optimizing building energy performance. The approach, developed by Florida Tech researchers Hamidreza Najafi and Benjamin Kubwimana, trains a model using large datasets generated through a Python-based software script and the EnergyPlus physics-based building energy simulation tool.
Optimization of the surrogate model is achieved through the use of genetic algorithms and Bayesian optimization, allowing for the building design to be optimized and continuously adapted to changes in operational parameters. This process is automated, using data from building sensors to facilitate the adaptation of the digital twin to the current operating condition of the building. This provides building owners with the ability to predict energy consumption and budget accordingly, leading to reduced CO2 emissions and energy savings.
The study, part of a broader effort to improve building energy modeling (BEM) and expand its applications, was published in the ASME Journal of Engineering for Sustainable Buildings and Cities in December 2022. In a separate study, Najafi and Ph.D. student Mariana Migliori explored the impact of COVID-19 on building energy performance and the adaptability of BEMs to changing conditions. The researchers collected data from Florida Tech’s Folliard Alumni Center and developed a data-driven model that could adapt to the new operating conditions brought on by the pandemic.
Conlcusion:
The application of machine learning algorithms to building energy modeling (BEM) presents a significant opportunity for the reduction of energy consumption and greenhouse gas emissions in the building sector. Florida Tech researchers have developed a data-driven model that utilizes artificial neural networks and machine learning algorithms to optimize building energy performance.
The model trains on large datasets generated through a Python-based software script and EnergyPlus simulation tool and is optimized through genetic algorithms and Bayesian optimization. The process is automated, using data from building sensors to continuously adapt the digital twin to the current operating condition of the building, providing building owners with the ability to predict energy consumption and budget accordingly, leading to reduced CO2 emissions and energy savings. The study, part of a larger effort to improve BEM and expand its applications, was published in December 2022 in the ASME Journal of Engineering for Sustainable Buildings and Cities.
