Fusion of physical domain knowledge with imp AI

image: Estimation of battery capacity and remaining useful life using artificial neural networks based on physical domain knowledge.
see After


Recently, electric vehicles (EVs) have appeared everywhere, from passenger cars to buses to taxis. Electric vehicles have the advantage of being environmentally friendly and having low maintenance costs; but their owners should beware of fatal accidents in case the battery runs out or reaches the end of its life. Therefore, accurate capacity and lifetime predictions of lithium-ion batteries – commonly used in electric vehicles – are essential.

A POSTECH research team led by Prof. Seungchul Lee and Ph.D. Candidate Sung Wook Kim (Department of Mechanical Engineering) collaborated with Prof. Ki-Yong Oh of Hanyang University to develop new intelligence technology artificial intelligence (AI) capable of accurately predicting the capacity and lifetime of lithium-ion batteries. This scientific breakthrough, which dramatically improved the accuracy of predictions by merging knowledge from the physical domain with AI, was recently published in Energy appliedan international academic journal in the field of energy.

There are two methods for predicting battery capacity: a physics-based model, which simplifies the complex internal structure of batteries, and an AI model, which uses the electrical and mechanical responses of batteries. However, the conventional AI model required large amounts of data for training. Moreover, when applied to untrained data, its prediction accuracy was very low, which desperately called for the emergence of next-generation AI technology.

To efficiently predict battery capacity with less training data, the research team combined a feature extraction strategy that differs from conventional methods with neural networks based on physical domain knowledge. As a result, the battery prediction accuracy for testing batteries with different capacities and lifetime distributions improved by up to 20%. Its reliability was ensured by confirming the consistency of the results. These findings should lay the groundwork for the application of highly reliable AI based on physical domain knowledge to various industries.

Professor Lee of POSTECH remarked: “The limitations of data-driven AI have been overcome with knowledge of physics. The difficulty of constructing big data has also been mitigated with the development of the differentiated feature extraction technique.

Professor Oh from Hanyang University added: “Our research is important as it will help spread electric vehicles to the public by enabling accurate predictions of remaining battery life in next-generation electric vehicles. “.

This study was funded by the Institute of Civil-Military Technology Cooperation and the National Research Foundation of Korea.

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of press releases posted on EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.

Donald E. Patel