As we know, battery performance can determine the electric vehicle experience, no one wants an electric vehicle with a short range or requiring long periods of charge. However, artificial intelligence could make it possible for the dream of recharging an electric vehicle to become a reality in the same time it would take to stop at a gas station, even helping to improve other aspects related to rechargeable vehicle battery technology.
For years the development and implementation of electric vehicles has been associated with battery development. A typical goal in this type of industry is to maximize the life of the batteries as their autonomy to improve the driving experience of this type of vehicles; however, performing a single experiment to evaluate the life can take from months to years, and the great variability of sampling requires a large number of experiments. But now, a team led by Stanford professors has developed a method that uses machine learning to reduce these test times by 98 percent. According to the researchers, with the AI algorithm they are able to quickly identify the most promising approaches and eliminate a lot of unnecessary experiments.
The study was published in the February issue of Nature and is a broader collaboration between scientists at Stanford, MIT and the Toyota Research Institute. The goal is to find the best method to charge an electric vehicle battery in the shortest time possible that maximizes battery life. You can read in the article that they have developed a machine learning methodology to efficiently optimize a parameter space that specifies current and voltage profiles in the battery's rapid charge protocols, which can alleviate waiting anxiety for electric vehicle users. By also reducing the duration as well as the number of experiments, the researchers reduced the testing process from nearly two years to 16 days.
By learning the machines, the number of methods to be tested was reduced, so instead of trying all possible charging methods equally, or relying on the researcher's expertise, the computer learned from their experiences to quickly find the best protocols to test. So, by testing fewer methods for fewer cycles, the study authors quickly found an optimal ultra-fast charging protocol for their battery, dramatically accelerating the testing process.
This approach could speed up almost all stages of the battery development process: even from the design of the battery chemistry itself, to the configuration of its size, shape and materials. So these results could be derived for other types of energy storage, a key requirement for less dependence on fossil fuels.
Not only that, the potential of these study methods can be extended beyond battery development, as other data testing problems could be revolutionized by the use of machine learning optimization as well.
Thank you for reading me. Until the next post!