A Machine-Learning-Based Approach to Analyse the Feature Importance and Predict the Electrode Mass Loading of a Solid-State Battery

Abstract

Solid-state batteries are currently developing into one of the most promising battery types for both the electrification of transport and for energy storage applications due to their high energy density and safe operating behaviour. The performance of solid-state batteries is largely determined by the manufacturing process, particularly in the production of electrodes. However, efficiently analysing the effects of key manufacturing features and predicting the mass loading of electrodes in the early stages of battery manufacturing remain a major challenge. In this study, a machine-learning-based approach is proposed to effectively analyse the importance of manufacturing features and accurately predict the mass loading of electrodes. Specifically, the importance of four key features during the manufacturing process of solid-state batteries is first quantified and analysed using a machine-learning-based method to analyse the importance of features. Then, four effective machine-learning-based regression methods, including decision tree, boosted decision tree, support vector regression and Gaussian process regression, are used to predict the mass loading of the electrodes in the mixing and coating stages. The comparative results show that the developed machine-learning-based approach is able to provide a satisfactory prediction of the electrode mass loading of a solid-state battery with 0.995 R2 while successfully quantifying the importance of four key features in the early manufacturing stages. Due to the advantages of its data-driven nature, the developed machine-learning-based approach can efficiently assist engineers in monitoring/predicting the electrode mass loading of solid-state batteries and analysing/quantifying the importance of manufacturing features of interest. This could benefit the production of solid-state batteries for further energy storage applications.