A major safety concern for electric vehicles (EVs) is managing battery temperature, as overheating can lead to hazardous situations, including fire. A University of Arizona research team, led by doctoral student Goswami, has developed a new method to predict and prevent these temperature spikes in lithium-ion batteries—the primary power source for most EVs. Supported by a $599,808 grant from the Department of Defense’s Defense Established Program to Stimulate Competitive Research, the team is pioneering a framework that combines multiphysics and machine learning models to detect and anticipate overheating, also known as thermal runaway.
The goal is to integrate this predictive system into electric vehicles’ battery management systems, offering drivers an additional layer of protection against battery malfunctions. “We need to move to green energy, but there are safety concerns associated with lithium-ion batteries,” said Goswami.
Thermal runaway occurs when the temperature in a battery increases exponentially, potentially leading to fire. EV batteries are composed of numerous interconnected “cells,” and modern vehicles can contain over 1,000 cells in a single battery pack. If one cell overheats, it can trigger a chain reaction, causing nearby cells to heat up as well. This domino effect can result in the entire battery pack exploding, posing a severe risk to the vehicle’s occupants.
To address this issue, the researchers have developed thermal sensors that wrap around the battery cells. These sensors collect historical temperature data, which is then processed by a machine-learning algorithm to predict future temperature trends and potential thermal runaway events. The algorithm identifies “hotspots” where thermal runaway is likely to begin, allowing for early intervention. “If we know the location of the hotspot, we can implement solutions to stop the battery before it reaches a critical stage,” Goswami explained.
The framework developed by Goswami and his adviser, aerospace and mechanical engineering professor Vitaliy Yurkiv, has already demonstrated impressive accuracy. Prior to this research, machine learning had not been applied to predict thermal runaway. “We didn’t expect machine learning to be so superior in predicting thermocouple temperature and the exact location of hotspots,” Yurkiv said. “No human would ever be able to do that.”
This recent breakthrough builds on earlier work by Goswami and Yurkiv, published in January, which explored the use of thermal imaging to predict overheating. While effective, that approach required constant photo reviews using heavy imaging equipment. The new machine-learning solution is more lightweight and cost-effective.
Goswami’s research arrives at a pivotal moment for the electric vehicle industry. In July, the same month the team’s findings were published, the Biden administration announced a $1.7 billion investment in EV manufacturing across eight U.S. states. With global EV sales up 35% in 2023 compared to the previous year, ensuring the safety of these vehicles is more important than ever.
“Many people are still hesitant to embrace batteries due to various safety concerns,” Goswami said. “To gain widespread acceptance, it’s crucial for the public to know that ongoing research is actively addressing these critical safety issues.” The team’s work is a significant step towards making electric vehicles safer, thus encouraging wider adoption of green energy transportation.
By Impact Lab