Impact study on continuous overcharging of precycled lithium batteries and control algorithm development using machine learning approach

This research aims to understand the impact of overcharging on precycled Lithium-Nickel Manganese Cobalt (Li-NMC) batteries through continuous charge and discharge cycles. These Li-NMC cells underwent standard charging and discharging, maintaining their state of health (SOH) above 90% even after 200 cycles. Continuous overcharging of these precycled Li-NMC cells caused their SOH to degrade below 20% within the next 180 cycles. To investigate the underlying reasons for such significant differences in SOH degradation, scanning electron microscopy (SEM) and Energy Dispersive Spectroscopy (EDS) analyses were performed on both pristine and overcharged batteries. The structural changes in the batteries due to overcharging provide valuable insights into the degradation mechanisms. In addition, to predict the health degradation of precycled batteries by utilizing a gradient boost (GB) and support vector machine (SVM) regressor with hyperparameters optimized through Bayesian optimization, tuning of these hyperparameters increases accuracy, where mean squared error (MSE) and root mean squared error (RMSE) decrease more than 65% of the conventional algorithm. The GB and SVM regressors were trained using data acquired during the charge and discharge cycles of the batteries. The trained model demonstrated promising performance in predicting SOH degradation, providing valuable guidance for developing control algorithms for battery management strategies.