Pages

Predicting The Ageing Of Lithium-Ion Batteries Using AI May Prolong Their Lifespan, New Study Claims

There’s nothing worse than your lithium-ion battery dying without providing a signal that informs the user of what went wrong.

Therefore, we thought it would be interesting to see how such cells are turning into an integral component during the rise of electric mobility, despite such characteristics. And it’s all thanks to researchers who are now working to predict their aging with the help of AI-powered technology as a means to prolong their lifespan.

These cells are very successful in terms of their function. But their major downfall is a decrease in function with time due to degradation. And that is not only restricted to aging but also due to everyday tasks like charging and discharging which many experts deem cycling aging.

Batteries also tend to degrade when users are not making use of them. This process is better known as calendar aging and it takes place during storage or when cells aren’t getting used properly. This is based on three factors that include the resting temperature, rest time duration, and also the resting state of charging.

Knowing that most electric vehicles would spend most of their time in a parked position, estimating their capacity degradation through such means is essential. It’s how users can get a better idea of the battery life and make way for different mechanisms that circumvent the whole idea.

Therefore, so many researchers are busy setting out advanced machine algorithms that correctly estimate the process of calendar aging for these cells.

The news comes after research was published by the European Union’s Horizon program in 2020 where scientists took on new steps by comparing accuracy from two new algorithms on a wider spectrum of commercial chemistries for lithium-ion cells.

This drew attention to aging data from around six different types of chemistries linked to batteries. The latter was aged in the likes of temperature chambers using voltages of various kinds. And to further protect them from the likes of aging, the team was seen investigating the efficiency linked to machine learning algorithms.

But the real question has to do with how such algorithms really end up working in today’s time. It ends up rolling out reliable results but they vary by a huge amount in terms of their operations.

One example is the XGBoost which is supervised through the likes of machine learning that is used for matters like regression or even classification issues.

Then we’ve got the likes of ANN which is adaptive through artificial means and makes use of base elements dubbed connections and neurons. It alters global inputs in a manner that gives rise to predicted outputs.

To better understand the performance here, researchers are using different metrics to determine the severity of errors between the likes of values both measures and predicted. So the smaller the value, the greater the precision or accuracy.

In conclusion, we witnessed testing proving how XGBoost could be used in a successful manner to forecast the calendar aging of batter chemistries with very little error. But algorithms like ANN were not as great and could only produce satisfactory results for certain batteries and thier respective chemistries.


Read next: Toyota Wins the Battle of Being the Longest Lasting Car - Concerning Miles

No comments: