State of Health and Remaining Useful Life Prediction of Lithium-Ion Batteries Based on a Hybrid Network

Accurate estimation of the state of health (SOH) and remaining useful life (RUL) of lithium-ion batteries is critical for enhancing the safety and reliability of electric vehicles. Nonetheless, accurately assessing the SOH is quite challenging due to the intricate dependencies on factors such as temperature, voltage, and current. This study presents a hybrid model that enhances SOH estimation by integrating multihead attention (MHA), a bidirectional gated recurrent unit (BiGRU), and a bidirectional temporal convolution network (BiTCN). The main goal of the model is to improve the accuracy of SOH predictions throughout the battery’s lifecycle. To achieve this, the model initially selects relevant health indicators through the use of the Pearson correlation coefficient, followed by a verification process to confirm the reliability of these selected features. The BiTCN-MHA-BiGRU model then incorporates these validated health indicators and employs leave-one-out cross-validation to generate accurate SOH predictions throughout the entire battery lifecycle. Experimental results highlight the effectiveness of this hybrid model, as it achieves a mean squared error of less than 0.1% when trained on N-1 datasets and tested on the remaining dataset. This showcases the model’s superior accuracy and robustness compared to existing SOH estimation methods. Furthermore, the model ensures that the prediction error remains within 10% when applied to the NASA dataset by merging the predicted SOH with actual SOH values for joint RUL estimation. This confirms the model’s efficacy in accurately predicting both SOH and RUL, providing a comprehensive and reliable approach for battery management systems in electric vehicles. The findings highlight the model’s ability to boost the performance and extend the lifespan of lithium-ion batteries in real-world applications.