More than 40% of all US energy use and greenhouse gas emissions are related to the building sector. A study by researchers at the Florida Institute of Technology is investigating whether machine learning can help reduce this environmental impact.
The study was featured in a paper published in the January issue of the journal energy The title is “A New Approach to Optimizing Building Energy Models Using Machine Learning Algorithms”. Authored by his Hamidreza Najafi, an associate professor of mechanical engineering at the Florida Institute of Technology, and his Benjamin Kubwimana, who holds a master’s degree in mechanical engineering, the research focuses on both building energy modeling (BEM) and building energy model optimization. I am using a new approach to
Current practice for building energy simulation software tools requires manual entry of a vast list of detailed inputs. This includes design and operational variables, including building properties such as walls, building envelopes, window materials, or operational parameters such as set temperature. for different thermal zones.
“When it comes to optimizing BEM, it is very difficult because of the large number of variables involved in BEM and the potential to develop thousands or millions of BEMs with different combinations of these variables,” Najafi said. said. “Achieving a truly optimal design for a building requires evaluating all of these possible design/operational parameters, which is very computationally expensive and often impossible.”
Najafi and Kubwimana’s work includes developing software scripts in the Python language that enable automated data entry into a physics-based building energy simulation tool called EnergyPlus. Using a set of variables as input via this Python script covers large variations in multiple parameters and creates a large data set that can be used to develop a surrogate energy simulation model.
Machine learning algorithms, especially data-driven models using artificial neural networks, are trained using these datasets. Two optimization approaches, genetic algorithm and Bayesian optimization, are applied to the surrogate model to achieve optimal building design. This approach can be easily adjusted for different design or operational parameters.
“This process can be automated, making it easier to feed data from sensors in the building into a computer model to continuously adapt the digital twin to the current operational state of the building,” said Najafi. I’m here. “This helps building owners predict how much energy they will consume based on changes that may occur in operating parameters. It also enables the prediction of energy consumption and energy production.As a reduction of CO2 Power generation related to energy savings. ”
This research was part of a broader effort to improve BEM and expand its applications. An enhanced BEM can be used as a digital twin of a building, providing value to owners and developers not only before construction, but also during the building’s lifetime. The research ASME Journal of Engineering for Sustainable Buildings and Cities December 2022.
“This study with Mariana Migliori, one of my PhD students, explores the impact of COVID-19 on the energy performance of buildings and how BEM can be adapted to maintain accuracy when operating conditions change. We researched ways to make it happen,” Najafi said. “We were able to run a case study based on the data we collected from the Florida Tech Folliard Alumni Center and the physics-based model we had previously developed for our building, developing a data-driven model that could adapt to the new environment.” Operating conditions with extended HVAC system operating hours and changes to occupancy schedules in light of the COVID-19 situation.”
For more information:
Benjamin Kubwimana et al., A New Approach to Optimizing Building Energy Models Using Machine Learning Algorithms, energy (2023). DOI: 10.3390/en16031033
