Joint prediction of SOH and RUL of lithium-ion batteries using single-cycle charging data

Accurately predicting the state of health (SOH) and remaining useful life (RUL) of lithium-ion batteries is crucial for ensuring battery safety and optimizing management strategies. Both SOH and RUL undergo significant changes as the number of charge–discharge cycles increases, and a strong correlation exists between them. However, existing prediction models often overlook the intrinsic relationship between SOH and RUL, failing to achieve joint modeling of both. This limitation leads to error accumulation and impairs the decision-making accuracy of battery management systems (BMS). To address this challenge, we propose an innovative deep learning framework for the joint prediction of SOH and RUL using only a single cycle of charging data. First, a CNN-PCC feature extractor, based on a two-dimensional convolutional neural network (2D-CNN) and the Pearson correlation coefficient (PCC), is designed to extract features related to SOH and RUL. These features are then fused with other handcrafted features to form multi-scale hybrid features, which are processed through joint prediction framework to capture the underlying relationships between SOH and RUL. This method fully utilizes the intrinsic coupling of SOH and RUL in a joint prediction framework and is capable of handling complex fast charging scenarios. Experimental results show that the proposed method reduces the RMSE of SOH and RUL prediction tasks by 35.3% and 37.3%, respectively, compared to existing single-task models. These results validate the effectiveness of the joint learning paradigm and the feasibility of using single-cycle charging data, demonstrating significant improvements in both prediction accuracy and generalization capability.