Just as all of us care about our health and quality of life over time, battery scientists have the identical passion for maximizing the longevity of batteries and closely tracking their health. The health of a battery, just as our own, is a posh product of internal and external aspects that can not be easily summarized by a single number.
A battery is a composite device, where each of the subcomponents, similar to the anode and cathode, ages in other ways. The wide range of conditions during which a battery is used also comes into play: the batteries in a hybrid vehicle driven in Arizona face very different demands than a completely electric vehicle in Alaska. To see an entire picture of battery health, scientists are required to know the varied ways during which batteries can age and methods to link a battery’s degradation to its design and usage.
The complexity of battery aging makes it difficult to review. As scientists, we will use our knowledge of physics and chemistry to develop an understanding of various aging mechanisms, however the wide selection of scenarios and conditions for battery usage makes it difficult for us to achieve a holistic perspective quickly. This makes it essential to attract out the subtle correlations from large battery datasets, which might then inform the judgment of scientists. Just as we use large datasets and advanced analytics to know what helps people to take care of good health with age, we want to do the identical to know battery health.
The usage of artificial intelligence (AI) on data from batteries, a field generally known as battery informatics, has rapidly emerged as a crucial a part of the battery scientist’s toolkit. Previous work from TRI in collaboration with Stanford and MIT has shown that AI has the power to accurately predict the lifetime of batteries and develop optimized algorithms for fast charging. While the tools of conventional AI are undoubtedly powerful for making predictions, they’re often “black box” models, with limited ability to supply explanations for how and why those predictions are made. This will make it difficult for researchers to attract wider conclusions or gain scientific insight, so with a purpose to proceed constructing on the success of our previous results, it requires us to adopt recent approaches – similar to adding explainability to predictive power.
Two novel approaches were introduced to enable interpretability in our recent work here. First, our team used an interpretable machine learning technique, SHAP evaluation, which originates from game theory. SHAP evaluation allows us to quantify the contributions from individual inputs to a machine learning model. For instance, it allows us to deduce whether increasing the charging speed or using a wider range of the battery’s capability is more detrimental to battery health. Second, inputs to the AI models are restricted to features developed by our team’s battery experts which can be physically meaningful and interpretable. Sixteen metrics were developed to explain the state of health of batteries, starting from the cell-level to electrode-specific descriptors. Because of this, the output of our predictive models could be clearly mapped to scientific understanding.
While developing our explainable AI approach, researchers at TRI worked with counterparts at Stanford and MIT over a period of two years to gather a dataset of aging trajectories of 363 business EV battery cells. This collaboration resulted in a comprehensive and diverse dataset, containing a wealth of data concerning the state of health of the batteries tested under 218 distinct aging conditions. The range of aging conditions results in battery lifetimes starting from 4600 cycles for the gentlest conditions to 63 cycles for the harshest. To place that into perspective, for an EV that’s charged every day, that’s the difference between a battery that lasts 12 years to 1 that lasts just two months!
Having the ability to predict and explain the complex state of health of batteries provide tremendous value to battery engineers trying to improve the present battery design, in addition to for policymakers to develop roadmaps of charging and usage guidance for EV drivers. The success of this work exemplifies a recent way of conducting scientific research, where AI acts as a co-worker to scientists. As an alternative of viewing AI as a alternative for humans, here at TRI, we develop AI toolkits that work hand-in-hand with humans to reinforce the capabilities of each.
At TRI, we’re focused on research with the potential to enhance the standard of life for people and society. And we are attempting to resolve one among the largest challenges of our time: how will we power our future in a sustainable, reasonably priced way? TRI’s Energy and Materials team is devoted to researching and developing recent methods to find recent materials and improving batteries to assist humanity achieve these goals. For more on TRI’s approach to AI for science, learn more here (outlined in a previous Medium article).
