Machine Learning Approaches for Enhancing the SoH Estimation of LTO Batteries

Lithium titanate oxide (LTO) batteries' practical application in modern technologies depends on accurately predicting their state of health (SoH). Using advanced machine learning (ML) techniques, our study examined how to estimate LTO batteries' SoH. For this purpose, we aged rechargeable LTO batteries for 3500 cycles with a battery analyzer and performed differential voltage analysis (DVA). To estimate SoH as a regression problem, we used three machine learn-ing methods: Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Gaussi-an Process Regressions (GPR). As a novel approach to SoH estimation, our research uses a feedforward neural network to solve the categorization problem. In analyzing and comparing the performance of all methods, we found that this categorization-based neural network ap-proach improved computational efficiency by 60.89% while achieving SoH estimation accura-cy of 93.18%. By advancing the field of battery health monitoring, these findings contribute to more reliable and efficient battery management algorithms. In addition to improving battery management systems' accuracy and computational efficiency, the categorization approach demonstrated here could also be used to extend the life and reliability of LTO batteries, includ-ing those used in electric vehicles and renewable energy storage systems. The results of this study illustrate the importance of applying innovative machine learning applications to en-hance battery SoH estimations, providing important implications for future research and prac-tice.