Predicting the health status and remaining service life of lithium batteries is a major problem that limits the wide use of electric vehicles. Over time, the performance of the battery will decline through a series of complex fine chemical processes. Individually, these processes do not have much impact on battery performance, but taken together, they will seriously shorten battery performance and life.
Researchers from the University of Cambridge and the University of Newcastle have devised a new method to monitor the battery by sending an electrical pulse to the battery and measuring its response. Then, they use machine learning algorithms to process the measured data to predict the health status and service life of the battery.
“Safety and reliability are the most important design criteria, because the battery we developed can store a large amount of energy in a small space,” said Dr. alpha Lee of Cavendish Laboratory, University of Cambridge. “By improving the software for monitoring charge and discharge and using data-driven software to control the charging process, I believe we can greatly improve the battery performance.”
The researchers designed a method to monitor the battery by sending electrical pulses to the battery and measuring its response. Then a machine learning model is used to identify the specific characteristics of electrical reaction, which are the signals of battery aging. The researchers conducted more than 20000 experimental measurements to train the model. Importantly, the model learned how to distinguish important signals from irrelevant noises. Their approach is non-invasive and is a simple add-on system.
The researchers also found that machine learning models can provide clues to the physical mechanism of degradation. The model can tell which electrical signals are most relevant to aging, and then allow them to design specific experiments to explore the causes and ways of battery degradation.
“Machine learning is a supplement and enhancement to physical understanding,” said Dr. Yunwei Zhang, one of the first authors and also from Cavendish Laboratory. “The interpretable signals identified by our machine learning model are the starting point for future theoretical and experimental research.”
Responsible editor; zl