With the right training, algorithms can predict Li-ion battery lifetime – with 95% accuracy

Predictive prognosis could boost life, profits

Video Machine-learning algorithms can predict the lifetimes of lithium-ion batteries, and could help scientists develop better battery designs more quickly and at a cheaper cost, according to a paper published yesterdayin Nature Energy.

“The standard way to test new battery designs is to charge and discharge the cells until they fail,” said Peter Attia, co-author of the paper and a PhD student studying materials science and engineering at Stanford University. “Since batteries have a long lifetime, this process can take many months and even years. It’s an expensive bottleneck in battery research.”

Lithium-ion batteries start at peak performance but slowly degrade over time with each charge cycle. Battery designs have to be rigorously tested and performance can vary due to defects in the manufacturing process.

Here’s a vid from the American uni showing off the prediction tech:

Here’s where machine learning comes in: first, researchers at Stanford, MIT, the Toyota Research Institute, and Lawrence Berkeley Lab examined 124 lithium-ion batteries. They repeatedly drained and charged them until the batteries’ performance dropped by about 20 per cent, over typically anywhere from 150 to 2,300 charge cycles.

Second, the eggheads plotted the batteries’ capacity over each cycle to generate a dataset of 96,700 cycles. This was fed into a regression model, a statistical method that examines the relationship between different variables, to predict how long a battery would last given its performance over the first 100 cycles.

The results showed that the model was able to predict the overall lifetime of a battery from its voltage levels and other readings from its first 100 cycles with a 91 per cent accuracy rate. The researchers could also classify the batteries into whether they would have long or short lifetimes after analyzing data from the first five cycles with a 95 per cent accuracy rate.

Using this technique, the performance of individual batteries, after their first few cycles, can be predicted, allowing them to be sorted into bins for different applications and requirements, and presumably, different prices.

“For all of the time and money that gets spent on battery development, progress is still measured in decades,” said Patrick Herring, co-author of the paper and a research scientist at the Toyota Research Institute. “In this work, we are reducing one of the most time-consuming steps – battery testing – by an order of magnitude.”

Source: theregister.co.uk