Meet the authors: Zixin Jiang and Bing Dong

Abstract

What can we do to mitigate climate change and achieve carbon neutrality for buildings? In their recent publication in Patterns, the authors proposed a modularized neural network incorporating physical priors for future building energy modeling, paving the way for scalable and reliable building energy modeling, optimization, retrofit designs, and buildings-to-grid integration. In this interview, the authors talk about incorporating fundamental heat transfer and thermodynamics knowledge into data-driven models.

What can we do to mitigate climate change and achieve carbon neutrality for buildings? In their recent publication in Patterns, the authors proposed a modularized neural network incorporating physical priors for future building energy modeling, paving the way for scalable and reliable building energy modeling, optimization, retrofit designs, and buildings-to-grid integration. In this interview, the authors talk about incorporating fundamental heat transfer and thermodynamics knowledge into data-driven models.

Main text

What would you like to share about your background?

Left: Bing Dong. Right: Zixin Jiang.

Zixin Jiang: Before I started doing research, I dreamed of constructing the tallest building in the world when I was a civil engineering undergraduate student at Tianjin University, China. However, after a 3-month internship at a real estate company, I realized that it wasn’t exactly what I wanted. So, I switched my major to building technology, where I earned my master’s degree working on green building design and simulation. That was my first time participating in a research project, and I realized that was what I truly wanted to do. Now, as a 3^rd year PhD student at Syracuse University, USA, my research interest extends to building-to-grid integration, building energy modeling, building control optimization, building energy resilience, and occupancy behavior. Doing research is cool, let’s explore something fun and crazy.

Bing Dong: I received my PhD in building performance and diagnostics, Carnegie Mellon University, with a focus primarily in intelligent building operations, applied machine learning for smart buildings, and occupant behavior modeling. After that, I spent 2 years in an industrial research center working on field demonstration projects of energy-efficient buildings. I started my tenure track career as an assistant professor at the University of Texas at San Antonio where I extended my work on buildings-to-grid integration, urban mobility, and human performance. Now I am a full professor at Syracuse University, and my current research interest includes modeling and optimization of urban energy systems, grid-interactive efficient buildings, urban scale occupancy modeling, and quantum computing. I have served as the subtask leaders for IEA EBC Annex 66 and 79, the board of directors of the international association of building physics, and I am the director of BEST Lab and associate director for Syracuse Center of Excellence in Environmental and Energy Systems.

What motivated you to become a researcher? Is there anyone or anything that helped guide you on your path?

Zixin Jiang: My dad was my idol when I was a child; he seemed to know everything and loved telling me all interesting stories, from myths and legends to history and science. He taught me how to design circuits and repair computers when I was a kid. We spent a lot of great time together, which greatly enriched my imagination, creativity, and practical abilities. Later on, my curiosity about truth and facts motivated me to pursue a PhD. And I got lucky working with my advisor, Prof. Bing Dong. He is an amazing person who can clearly see the big picture for the future and knows exactly what to do next in detail, step by step. I really enjoy doing research with him, where I receive incredible guidance, abundant experimental resources, and critical perspectives. Of course, I also appreciate my excellent groupmates, from whom we inspire and learn from each other.

Bing Dong: With the growing Internet of Thing (IoT) in buildings, everyday there are lots of data generated from system, sensors, meters, and people, and yet those data are largely under-utilized. I love to explore new ways and methods to investigate data in buildings and provide new insights for future building energy modeling and operation.

What is the role of data science in your field? What advancements do you expect in data science in this field over the next 2–3 years?

Zixin Jiang: With the development of sensor technology, more and more data are available in the building sector, providing ample opportunities to integrate data into building design, construction, retrofitting, maintenance, and operation on a large scale. However, a major challenge is the generalization issue of classical data-driven methods. We need more data scientists with solid domain knowledge dedicated to this field. I am looking forward to a general solution to advance building automation.

Bing Dong: Data science in our field is growing fast. Researchers and scientists are paying more attention to how the latest machine learning and AI methods are applied in building science to tackle energy-efficient building energy modeling and control problems. In the future, I would expect that data science will be fully integrated with the fundamentals of building physics and create a better energy modeling approach.

I have a few questions for the first author. How do you keep up to date with advances in both data science techniques and in your domain? What attributes do you think make a data scientist successful?

Zixin Jiang: Solid math foundations such as calculus, linear algebra and probability, and reading are always the best way to keep ourselves up to date. However, the most important thing is to define the research question. It’s not just about applying a new data science technique but also understanding why and how to incorporate our domain knowledge into the method to fit the specific task more appropriately.

For a successful data scientist, I would say intuition is very important. When we start a new task, good intuition can always help us understand the problem. For example, what is the input and target, how does the mapping look like, what model structure should we use, and how can we draw inspiration from our domain knowledge? Furthermore, during the debugging process, good intuition can help you locate and fix problems.

What advice would you have given yourself at the start of the project? Is there anything you would have done differently?

Zixin Jiang: It was my first time working on such a high-quality journal paper, and I was scared and nervous at the beginning. I was unfamiliar with the paper structure, unsure of how to display my results with representative images, and even didn’t know how to divide the main content and supplementary results. I worked inefficiently, trying to get everything ready and perfect before starting to draft the paper, which is not possible. Thanks to the encouragement and help from my advisor, I took the first step, and then everything became easier. I would say that no research is perfect. We should be brave and start first, face challenges, seek help when stuck, pay attention to details, work hard, and eventually, maybe in half a year or longer, you will achieve your goals.

Was there a particular element like a paper, collaboration, talk, conference, key experiment, idea, or result that motivated you to start this project?

Zixin Jiang: 7 years ago, when I was a master student and read my first journal article in our field. Most papers began with the statement “building accounts for 30% of energy consumption in the world,” however, this number has not decreased until now. I was always curious about the reasons behind this and realized that scalability is one of the major barriers. Buildings are complex, highly nonlinear systems, and building energy modeling is challenging. It requires detailed metadata, substantial modeling efforts, and case-by-case calibrations, which limit large-scale real-world applications. This led me to explore how to listen to real-world data and learn from them to make building energy modeling easier and scalable. More importantly, how to drive physical priors into data-driven models to make it reliable and robust. I discussed this idea with my advisor, Prof. Bing Dong, who encouraged me to pursue it. After completing the energy modeling and optimization part, he inspired me to think bigger and aim higher. This led to another inspiration about modularization—by sharing and inheriting modules, people can achieve various tasks flexibly through module connections.

Dr. Dong, how did you introduce your team to the community? Were conferences important in this? What support did you get from the community?

Bing Dong: I often reach out to our community by hosting workshops, visiting students and professors, attending conferences regularly, and conducting research webinars. The community provides me with a lot of positive feedback while I present our research work. I really appreciate the very good research community that I’m in.

Who were the driving forces behind the project? Was there a particular result that surprised you, or did you have a eureka moment? How did you react?

Bing Dong: Building energy modeling, which is known as BEM, is fundamental for achieving optimized energy control, resilient retrofit designs, and sustainable urbanization to mitigate climate change. However, traditional BEM requires detailed building information, expert knowledge, substantial modeling efforts, and customized case-by-case calibrations. This process must be repeated for every building, thereby limiting its scalability. This process has been seen for the last three decades, and there is no efficient solution.

The validation results for all four cases we presented in the paper¹ are above our expectations, including load prediction, indoor environment modeling, building retrofitting, and energy optimization. We immediately think that this is a promising method that potentially transforms the traditional energy modeling field.

What drew you to this area of research? How has your team’s research focus evolved over the years?

Bing Dong: Our group has been doing energy modeling, control, and optimization for the last 15 years. We have published more than 120 papers in this field. However, for every energy model and control optimization case study, we have to spend numerous time doing modeling tuning and validation. There are many existing approaches including machine learning ways to help for the model tuning but still there is not a scalable way that can easily transfer from one case study to another. The arising of data science, in particular physics-informed neural network models, inspires our group into this research direction. We have developed and implemented physics-inspired neural-network-driven building controls in real buildings.

Biographies

About the authors

Zixin Jiang: Zixin is a PhD student in Syracuse University, department of Mechanical and Aerospace Engineering. His research focuses on building-to-grid integration, building energy modeling and control optimization, building energy resilience, and occupancy behavior.

Bing Dong: Dr. Dong is a full professor at Syracuse University. He also serves as associate director of Syracuse Center of Excellence in Environmental and Energy Systems. Dr. Dong’s research and education are in the broad area of smart buildings and cities, including data-driven building energy modeling and controls, modeling occupant behavior at both building and urban scales, energy policy, quantum computing, and buildings-to-grid integration. His research has over 10,000 citations on Google Scholar and over 120 peer-reviewed papers published. He is the recipient of the 2019 NSF CAREER Award and the 2018 IBPSA-USA Emerging Contributor Award. He is included as a 2023 IBPSA world Fellow.