Electric vehicle drivers know the feeling: the dashboard says the battery is 40% charged, but that number doesn’t tell you whether you’ll make it another 100 kilometers over hilly roads with the heater running.
Engineers at the University of California, Riverside want to remove that uncertainty.
They’ve developed a new way of tracking battery health and performance called the “State of Mission,” or SOM, which focuses on what really matters—whether the battery can handle a specific job in real-world conditions.
Unlike the current “percent charged” readouts, SOM uses both battery data and outside factors such as elevation changes, traffic, and temperature to give task-specific predictions.
“SOM fills that gap,” said UCR engineering professor Mihri Ozkan. “It’s a mission-aware measure that combines data and physics to predict whether the battery can complete a planned task under real-world conditions.”
The research, published in iScience, introduces a hybrid approach that sets SOM apart from existing models.
Traditional methods either rely only on physics equations, which can be too rigid for changing conditions, or only on machine learning, which can be flexible but hard to interpret. The UCR team combined both.
Their model learns from how batteries behave—how they charge, discharge, and heat up—but also follows the physical laws of chemistry and thermodynamics.
This dual method produces predictions that are both accurate and trustworthy, even under stress, such as a steep uphill climb or sudden cold weather.
“By combining them, we get the best of both worlds,” explained UCR professor Cengiz Ozkan, who co-led the project. “The model learns flexibly from data but always stays grounded in physical reality. That makes the predictions far more reliable.”
To test their framework, the team used real-world battery datasets from NASA and Oxford University.
These included records of how batteries were charged and discharged over time, how their temperatures changed, and how performance shifted after long use.
When compared with traditional diagnostic methods, the new system reduced prediction errors for voltage, temperature, and charge state significantly, giving drivers and engineers a clearer picture of what the battery could actually do.
Instead of simply reporting how much charge is left, SOM can provide actionable guidance. For example, it could tell a driver they’ll be able to finish a planned trip but will need to recharge halfway. Or it might tell a drone operator that a flight is not possible under current wind conditions.
While the system is still being refined, researchers believe it could be adapted for electric vehicles, drones, grid storage, and even space missions.
They also aim to test it on newer types of batteries, such as solid-state and sodium-ion. “Our approach is designed to be generalizable,” said Cengiz Ozkan. “It can improve safety, reliability, and efficiency across a wide range of energy technologies.”
Source: UC Riverside.
