Lithium-Ion Battery Life Estimation Using Machine Learning Algorithm

Accurate estimation of Li-Ion batteries life cycle is important to the battery management system (BMS). Compared to direct measurement and model-based methods, data-driven methods are drawing much attention for online SOH estimation due to their simple structure, flexibility for online application, and independence on battery model. Data-driven methods first extract a strong health indicator (HI) to measure the SOH and then build a machine learning model to map the relationship between them. Based on data analytics on the battery ageing data, this project proposes to use a partial charge and discharge current sequence. Then it is then fed into KNN algorithm, which has fast learning speed and good generalisation property. By selecting charging and discharging capacity, the impact on the estimation accuracy is comprehensively evaluated. The proposed method is tested with an open dataset and the outcomes verify the effectiveness of the proposed method. This project delivers an analysis of the SoH estimation methods for the main battery, give knowledge about their advantages and disadvantages in terms of real time mechanical compatibility and mainly hybrid electric applications.