Machine Learning Techniques Suitability to Estimate the Retained Capacity in Lithium-ion Batteries from Partial Charge/Discharge Curves

IEEE Transactions on Transportation Electrification, DOI: 10.1109/TTE.2023.3252169.

Machine Learning Techniques Suitability to Estimate the Retained Capacity in Lithium-ion Batteries from Partial Charge/Discharge Curves

Hector Beltran¹, Emilio Sansano¹ and Michael Pecht²
¹Department of Industrial Engineering Systems and Design, Universitat Jaume I, Castelló de la Plana 12003, Spain
²Center for Advanced Life Cycle Engineering, University of Maryland, College Park, MD 20742, USA

For more information about this article and related research, please contact Prof. Michael Pecht

Abstract:

The accurate estimation of the retained capacity in a lithium-ion battery is an essential requirement for the electric vehicles. The aging of the batteries depends on parameters and factors that are not easily monitored by the battery management system. This paper analyzes the ability of various machine learning algorithms to deal with the data generated by the battery management system during the partial charging/discharging process to instantly diagnose and estimate the retained capacity of the battery. Experimental data from an online dataset containing thousands of battery cycles are used for training and validation of the different models. Results demonstrate that the developed convolutional neural network outperforms the rest of the machine learning algorithms implemented, regardless of the portion of the cycle registered by the battery management system. The estimates obtained outperform most previous references. However, the estimation error values registered when analyzing partial cycles with depths lower than 50 % (above 1.5 %) remain too high to validate any of the analyzed algorithms as a solution for commercial systems.

This article is available online here.