Could AI Accurately Predict Battery Health for EVs and Consumer Electronics?
British researchers seem to think so—they claim that artificial intelligence (AI) and machine learning could be used to predict the health of a battery with 10 times higher accuracy than current techniques.
The researchers, from the Universities of Cambridge and Newcastle in the UK, also claim that their new machine learning model could also help the development of safer and more reliable batteries for electric vehicles (EVs) and consumer electronics like smartphones and wearable devices.
If this is true, one of the most prominent problems facing the widespread adoption of technologies like EVs could be solved: Predicting the state of health and remaining useful lifespan of lithium-ion batteries.
Monitoring Battery Degradation With AI
With time, batteries degrade. This is no secret. This happens via a complex network of subtle chemical processes that collectively shorten performance and lifespan. And being able to monitor this is just as tricky as it is important.
However, current methods of assessing a battery’s condition rely on the measurement of current and voltage during charge cycles. This provides us with some information about its current state and how much it has degraded, but not the processes involved.
Using AI and Machine Learning Methods
To take battery monitoring to the next level, the non-invasive and retrofittable system developed is based on AI and employs a machine learning method to monitor batteries by sending electrical pulses into them and measuring the responses received from them. An AI algorithm then assesses the different features of the responses to discover specific features that are the signs of a battery degrading.
The team performed over 20,000 experimental measurements to train their machine learning model, providing the underlying AI with a huge data set to compare new information to. According to the researchers, this technology is capable of not only accurately monitoring but also predicting battery aging.
Three graphs created by researchers that display estimations of measuring battery capacity. Image used courtesy of Zhang Y., Tang, Q., Zhang, Y. et al
Investigating How and Why Batteries Degrade
By using AI, vast amounts of data can be very quickly and efficiently analyzed. It can recognize processes taking place in a battery and compare them to those seen in previous data sets. Over time, this produces increasingly more reliable results and new information which could be used to make significant improvements in battery technology.
The researchers also demonstrated how their machine learning model can provide hints about the physical mechanism of degradation. It can inform which electrical signals correlate most with aging, allowing engineers to investigate why and how batteries degrade.
Dr. Yunwei Zhang, co-author of the study, said: “Machine learning complements and augments physical understanding. The interpretable signals identified by our machine learning model are a starting point for future theoretical and experimental studies”.
Controlling the Battery Charging Process
The machine learning model is currently being used to monitor processes in different battery systems. The researchers behind it want to look at how degradation happens and how it can be stopped or slowed down.
Dr. Alpha Lee from Cambridge’s Cavendish Laboratory said: “By improving the software that monitors charging and discharging, and using data-driven software to control the charging process, I believe we can power a big improvement in battery performance.”