Data-Driven and Knowledge-Driven Deep Learning for Battery Safety Modeling

Dr. Anika Tabassum

Abstract: Li-ion Batteries, one of the most efficient energy storage devices, are widely adopted in many industrial applications, e.g., laptops, mobile devices, electric cars, etc.  Hence, battery safety is extremely important as several designs are considered during integration into energy storage systems. E.g., Imaging data of battery electrodes obtained from X-ray tomography can explain the distribution of material constituents and allow reconstructions to study electron transport pathways.  Therefore, it can eventually help quantify various associated properties of electrodes (e.g., volume-specific surface area, porosity) which determines the performance of batteries.  These images often suffer from low contrast and resolution, making it difficult for humans to distinguish and characterize the material constituents.  Another extreme event is the internal short circuit that leads to the thermal runaway of the battery.  This short circuit event is dictated by chemical reactions of various electrode/electrolyte materials, which vary significantly between different device manufacturers.  For safety modeling, this situation must be mitigated under all conditions.

In this talk, I will talk about how we incorporate domain knowledge in data-driven deep-learning models to mitigate such extreme conditions and ensure safety.  Our models can work well with noisy data, generalizable for different battery materials.  In a broader aspect, our model can also be adaptable and extended for other applications like electron microscopy images of material science.

Speaker’s Bio: Anika Tabassum is a Postdoctoral research associate at Discrete Algorithms groups, Computer Science and Mathematics Division at the Oak Ridge National Laboratory, where she is contributing towards developing deep Learning for multi-scale and multimodal data evolved from energy research, e.g.., battery, plasma simulations, and spike neural network. Her research interest broadly lies in robust and generalizable deep learning models for scientific applications.  She recently received an outstanding postdoctoral award from Computational Science and Mathematics Division.  She received her Ph.D. from the Department of Computer Science at Virginia Tech where she worked on bringing domain-guided machine learning to address multiple challenges to prepare and mitigate power system failures and disaster vulnerabilities.  She was a National Science Foundation Urban Computing fellow during her Ph.D. and earned urban computing certification.  She won 1st prize in designing a COVID-19 forecasting model for the Facebook-CDC challenge.  With her team, she developed a COVID-19 forecasting model, which leveraged a deep learning model for the first time.  She has published in multiple venues, such as the Association for Computing Machinery (ACM) Special Interest Group Knowledge and Data Mining, NeuRIPS, the Association for the Advancement of Artificial Intelligence, the Conference on Information and Knowledge Management, the Institute of Electrical and Electronics Engineers- BigData, the Innovative Application of Artificial Intelligence, and journals such as ACM Transactions on Intelligent Systems and Technology and Elsevier.

Last Updated: January 17, 2023 - 9:50 am