How much charge is left in a battery, and how much useful life remains? These questions drive Braden Holst’s research with the Carolina Institute for Battery Innovation (CIBI). Using mathematical models and sensor data, he is working to improve how engineers estimate battery health and performance. In this Q&A, Holst discusses his research, recent progress in battery pack modeling, and the everyday connections that help explain his work.
Q: Can you explain your battery research and what you hope to accomplish through your work?
My research is specifically focused on reduced-order modeling of lithium-ion cells and battery packs for the electrification of different transportation modes, including applications in the United States defense sector and private industry.
My work focuses on mathematical modeling and what we call state estimation. A battery is, in essence, a closed system while it's operating. There are certain things, like the charge of the battery or how much life is left, that we need to know but can't directly observe as it operates. Our job as the mathematicians of the discipline is to give good predictions, with a certain amount of confidence, of what those states are. These are called the state of charge and the state of health: how much charge is left and how much useful life remains in the battery.
What I hope to accomplish through my work is to develop a very robust battery management system. It would estimate these states using physics-based battery models, which provide predictions of what's going on inside the battery, and combine those predictions with sensor information, such as voltage, current, and temperature. We can use that to make state-of-the-art estimates of the state of charge and state of health.
By the end, I want to give CIBI state-of-the-art state estimation and battery management system software that I can use on my resume to say, “Look at what CIBI can provide. That is something I was on the frontier of making.”
Q: What has been your biggest success or breakthrough in your research so far?
I would say three things. One was serving as a teaching assistant for my advisor, Dr. Ralph White. He's been with the university for a very long time. He was the dean of the School of Engineering back around 2000 or 2005, I think. I was a teaching assistant for one of his courses, and I love teaching. I love taking what might seem like complex topics and breaking them down so people can understand them and have a good, productive conversation.
I loved being a teaching assistant for his course. His courses are challenging, and not everyone appreciates them. What I loved was that students would come from his class confused and then see me, and we could break things down by principle. I got good feedback from the students during the fall and spring courses last year, so I think that was a big success for me.
My second success involves connecting electrochemistry with electrical engineering. Most people in the department are experimentalists working on the leading edge of making batteries better, making them last longer, or developing materials to prevent them from catching fire or make batteries that don't depend on lithium.
Our group was on an island a little bit. They're now hiring more faculty to bring in more of the computational side, but for a long time, we were on an island, communicating with electrical engineers about how to couple batteries to their power systems.
They're not chemical engineers or electrochemists. How can I bring the language of electrochemistry over and marry the two? That's a big area that's missing: how can we incorporate good electrochemical predictions into an overall power system?
I recently wrote a paper with an electrical engineering team at USC on this subject, involving a DC-to-DC power converter. A battery is a DC voltage source, and we need to couple these two for the power system to work properly. For a DC load, such as a pulse load from a laser or other digital technologies, you need a DC-to-DC converter for the battery.
Electrical engineers generally work on that, rather than electrochemists, so they often treat the battery in a rudimentary way. We combined our electrochemistry knowledge with their expertise, and I think we made a wonderful paper. It's now going through the clearance process. I was very proud of that cooperation because we're on two sides of the same coin.
The third success is that I've recently made good strides in battery pack modeling. A lot of people in the department are building good battery cells or modeling cells to make predictions about what happens inside them. Not as many people focus on the battery packs themselves.
Battery packs are used in electric vehicles and other applications. They can be dangerous because, when cells are connected in parallel, their terminals are connected by a bus bar. If one cell were to degrade or a short circuit were to happen, charge could flow into that cell and contribute to thermal runaway.
You can think of one cell at a high voltage and another at a low voltage like a ball on a high mountain and another on a low mountain. The charge is going to move toward equilibrium. Charge flowing into the battery with the low voltage can propagate thermal runaway.
Battery packs are also very difficult to model because modeling individual cells is already intricate and difficult. You also can't have sensors on every single cell in the battery pack. Think of Tesla: they have at least six hundred battery cells in a pack, if not more. Six hundred sensors can run up the cost, and that's a lot of information for the car's computer to process. Modeling battery packs is a lot more difficult than it sounds.
I've recently made great strides using a physics-based model and state estimation, and I'm very proud of that work. It's still in progress, so I won't go into too much detail, but soon enough it will be on the market. I'm very proud of what I've done so far and excited to showcase it soon.
Q: How can heat affect batteries and contribute to thermal runaway?
Think about leaving your phone in the sun. If it gets really hot, it shuts down. That's happened to me while playing basketball outside and leaving my phone in the sun.
Heat can overcome certain barriers that would otherwise keep reactions from happening, especially side reactions on the anode, where the solid electrolyte interphase, or SEI, forms. Higher temperatures can allow side reactions that might not otherwise happen. Thermal runaway can happen if a battery or battery pack gets really hot in its environment.
Q: How could this research eventually help engineers monitor battery health, predict degradation, and make better operating decisions in real-world battery packs?
My research provides a better way to make initial predictions about cells in battery packs. With better physics-based predictions, two benefits stand out to me.
First, physics-based models give us knowledge about what's going on inside a battery. Instead of using an equivalent circuit model or other rudimentary ways of treating cells, we're making predictions about things like porosity, solid electrolyte interphase thickness, and electrolyte transport.
These predictions are about what's actually happening inside the cell, rather than just parameters that help us fit a curve. One benefit of using a physics-based model over an empirical model is that we have meaningful parameters that change as the battery degrades.
Second, with a better prediction, we have less reliance on noisy sensors. This is a major part of state estimation. We combine our model predictions with sensors measuring voltage, current, and temperature to estimate what's happening inside the battery in real time.
For example, say a battery is operating and its voltage is going down over time. We have what the model says the battery should be doing, including what's happening inside, and we have the sensor measurements. If the model is highly inaccurate, we have to rely very heavily on the information we're getting from the sensors.
But what if we have a faulty sensor that deteriorates over time, or sparse data from a battery pack? As I mentioned, we can't have six hundred different sensors. If we can rely less on a noisy sensor that might give us faulty data or get worse over time, we can have more confidence in our overall state estimation than we would using a more traditional, rudimentary cell or pack model.
My research gives us a way to have more confidence in the accuracy of our estimates of what's really going on inside a battery. With that increased confidence, we can give an operator, whether on an electric ship or in an electric vehicle, better insight into how much charge remains or when the battery next needs to be charged.
With physics-based models, we can make better predictions with greater confidence.