Organisations are actively embracing machine learning techniques to enhance the development of effective batteries for delivering high power to electric vehicles. A robust battery is key to achieve our goal of shifting towards sustainable energy. However, no battery manufacturer can claim supremacy in the highly competitive market.
Every automobile company is striving to blaze their trail and develop a battery that could mitigate numerous challenges associated with it. More notably, Toyota, DOE, ION Energy, and more have been at the forefront of deploying machine learning techniques to enhance the traditional batteries for electric vehicles.
Toyota Research Institute (TRI)
TRI was introduced in 2015 to focus on autonomous driving primarily, and one of them was the research on material science by embracing the latest technologies. And later on, it set aside $35 million for research and development initiative to develop new models and materials for batteries.
This year, MIT, Stanford University, and TRI collaborated and trained a machine learning model with a few hundred million data points for predicting the lifespan of the battery. The model can categorise batteries as long or short life expectancy just by having the data of the first five charge and discharge cycles.
Their effort resulted in obtaining a model that delivers 95% accurate results, thereby opening up the doors to accelerate further research and development. Until recently, companies used to charge and discharge the cells for checking the lifecycle of batteries. With this, firms can now quickly evaluate the battery performance, and in turn, expedite the development processes.
The U.S Department Of Energy (DOE)
The department’s Argonne National Laboratory is leveraging machine learning techniques to make advancement in battery development. They curated a dataset of 133,000 small organic molecules with a computationally intensive model, G4MP2, that has 166 billion large molecules. To find a correlation with the small and large data set, the researchers used machine learning capabilities – taxing modelling framework that is based on density functional theory.
Although the accuracy was lower than the G4MP2, the density functional theory was fruitful in providing the desired approximation. The idea is to improve efficiency further while keeping the computational requirements as low as possible.
Through this, researchers could understand the relationship between atoms and their respective bond. The study is envisioned to help in storing energy more effectively while achieving stability.
The Mumbai-based startup, through its ML-based platform, is helping users to improve the performance of batteries by continually monitoring it. ION Energy offers lithium-ion battery management solution to improve the life and performance. To attain better results, the firm uses Edison Analytics to combine battery data and AI for helping business predict and enhance life by up to 40%.
Various firms are arduously trying to crack the battery’s chemistry to obtain exceptional performance for electric vehicles. However, finding out the chemical reaction will be of no good if the cost of the battery is too high; today, battery accounts for the 40% (a ballpark figure) of the price of electric vehicles. Consequently, the manufacturers will have to innovate while keeping the cost lower.
Manufacturers should use machine learning capabilities to find effective ways of achieving lower-cost. While India has devised various plan to commence Tesla type Gigafactories for EV batteries, it is crucial to integrate ml-based solutions for reducing cost. Such initiatives will assist the country in accomplishing its mission to ban IC engine-based two-wheelers below 150cc by 2025 and three-wheelers by 2023.
Machine learning is playing a crucial role in battery manufacturing, but it is yet to attain a breakthrough for revamping the landscape and proliferate in electric vehicles.