A Multi-Class Classification Based Approach for Remaining Useful Life (RUL) Prediction of Li-Ion Battery

Estimation of Remaining Useful Life of the battery reduces the risk of battery failure and also helps in the optimization of the battery life by proposing its replacement at a proper time. This paper proposes a new RUL prediction technique based on the Multi-Class Classification approach. In order to predict the group membership of data instances, we have utilized the classification approach of the Machine learning technique. To simplify the issues related to classification, a neural network approach is deployed. To estimate the RUL, multiple measurable data features from the battery monitoring system are considered such as capacity, voltage, current, and temperature charging/discharging profiles. This research work utilizes the Li-Ion battery dataset of NASA Prognostics Center of Excellence Data Repository to verify the efficacy of the proposed machine learning model. The proposed method eliminates the need to rely on complicated battery electrochemical principles. This is very crucial as the technique can be deployed for various types of batteries. Accordingly, the RUL estimation tool proposed in this paper may benefit the upcoming automotive industry, particularly to the electric vehicles.