It is still early to give a sure answer, but the laboratories are already experimenting with algorithms that could help develop energy accumulators that are increasingly efficient, safe and quick to recharge.
On the other hand, optimizing the efficiency and duration of lithium batteries is an essential objective, in the context of the transition towards electric mobility and towards the massive use of renewable sources in electricity generation, with the consequent need to install stationary accumulation.
So between 2019 and 2020, a group of scientists from universities and research centers in the United States – including Stanford University, Massachusetts Institute of Technology and Toyota Research Institute – used artificial intelligence to make predictions about battery performance and life.
In practice, they used what a Wired article calls “electrochemical torture chambers” where battery cells are charged / discharged rapidly dozens of times a day.
The goal is to generate a flow of data to be fed to the self-learning algorithms.
And artificial intelligence, by dint of processing huge amounts of data, learns to independently perform the tasks that have been assigned to it in the training phase (learning / training), in this case to predict, based on past experiences, the rate of battery performance during their life cycle.
The point is that experiments of this type normally take several months: in fact, you have to continuously test the batteries until they begin to degrade, to collect data that allow you to predict the future performance of the batteries under certain conditions of use.
With artificial intelligence, however, everything can be speeded up.
A few hours of self-learning with continuous data flows can be enough to formulate reliable predictions.
In a second research, the same scientists used artificial intelligence to define optimal protocols for fast battery charging.
In about a month they obtained results that, without algorithms, would have required a couple of years of work.
The self-learning algorithms, in fact, by grinding data streams, have discovered optimal methods to charge lithium batteries in a few minutes without deteriorating performance in terms of reliability and duration; remember that super-fast charging is a high stress factor for batteries, so the challenge is to find the balance between speed and duration.
With the algorithms, many other aspects of batteries can be tested: use of different materials, chemical recipes, anode / cathode composition, optimal energy density, and so on.
We will see if, at some point, these experiments will come out of the laboratories, to result in new batteries made on the basis of data and forecasts released by the super computers of artificial intelligence.