Data-driven Modeling and Estimation of Li-Ion Battery Properties
Dr. Matthias Preindl
Department of Electrical Engineering,
Columbia University in the City of New York
*The organization and the title are those when awarded
Highly accurate State of Charge (SoC) determination techniques are essential to maximize the use of Li-ion battery pack and design cost-effective Electric Vehicles. Numerous techniques have been proposed to estimate SoC. However, existing techniques still face large uncertainties in general use due to highly varying conditions such as the temperature distribution within battery packs.
In this work, a novel data-driven approach using deep-learning is applied to estimate SoC. It is found to accurately reconstruct the dependencies of SoC at a large range of ambient conditions. The proposed approach reduces uncertainties and achieves a highly accurate estimation with less than 1 % error in a wide temperature range (-25℃ to 45℃). Thus, it has the potential to make laboratory-scale accuracy available to real-world Electrical Vehicles operation.
In future, the technique will be expanded to estimate battery degradation and predict the lifetime of batteries. As a result, the proposed methodology is expected to maximize the use of installed battery capacity or reduce the required battery capacity for a given vehicle range.