As fossil fuel-driven society has brought environmental pollution and climate crisis, environmentally friendly alternative energy has been in the spotlight. In particular, electric vehicles receive attention as a next-generation transportation method. The most core technology for electric vehicles is the lithium-ion battery and the battery management system (BMS). However, due to the long lifespan of lithium-ion batteries, it is very time-consuming to measure and improve performance over time. In addition, since the batteries are driven under various conditions, it is not easy to measure the SOC and SOH in real-time because there are many variables. With the recent development of prediction technology through machine learning, it has become possible to predict with high accuracy without complex physical knowledge through high-quality data.

In this study, machine learning technology was applied by utilizing charge/discharge data of lithium ion batteries. The measured values in the initial charge/discharge cycles were selected as independent variable, and the number of cycles remaining until the end of the life was selected as a dependent variable, and then a regression algorithm was applied. After that, model tuning was performed to increase accuracy by reducing overfitting of the data. This study used machine learning technology to predict the SOH of a lithium ion battery with high accuracy with only the initial cycles, thereby contributing to real-time data of BMS and raising the time efficiency of the study.