Machine Learning Based State Prognosis and Performance Assessment of Lithium-Ion Battery

The rising demand for lithium-ion batteries in electric vehicle application, necessitates precise estimation of their state parameters and enhanced controllability. While conventional methods have long been used for this purpose, the emergence of data science has introduced machine learning techniques for prognosis and monitoring of lithium-ion batteries (LIBs. This paper implements two distinct machine learning algorithms, Gradient Boost Regression (GBR) and Support Vector Machine (SVM), to prognose state parameter of LiFeP04 cell. The extensive data set is generated with the predefined empirical findings based on experimental study using the programming in python to train these ML models. The performance assessment of battery cell in electric vehicle application is done to analyse the effect of Equivalent Circuit Parameters (ECM) and temperature on battery's state parameters. Radial Basis Function (RBF) kernel is used in SVM for capturing the non-linear behaviour of ECM to evaluate the SoC Prognosis. Finally, the performance of these models is validated with the experimental data and compared for the selection of better model to be integrated for the real-time applications. The comparative analysis of these models in terms of scatter plot, feature importance plot and accuracy in terms of root mean square error is done. Furthermore, a feature correlational-based C-map is employed for the SVM model to evaluate the impact of ECM parameters and temperature on the battery's state.