Abstract:  As lithium-ion batteries become increasingly popular, the battery life prediction is of crucial importance. Accurate prediction of the remaining useful life (RUL) of lithium-ion batteries is a critical part of their health management. In light of this, this paper proposes an algorithm SCSSA-CNN-BiLSTM, aming to perform RUL prediction for lithium-ion batteries used on electric vehicles. By combining convolutional neural networks (CNN), bidirectional long short-term memory (BiLSTM), and sine-cosine and cauchy mutation sparrow search algorithm (SCSSA), our algorithm forms a novel hybrid neural network that enhances the accuracy and stability of RUL predictions for lithium-ion batteries. CNN is employed for comprehensive extraction of deep features related to the state of health (SOH) of the batteries, while BiLSTM investigates these deep features bidirectionally and generates RUL predictions through dense layers. To validate the effectiveness of the proposed approach, battery data from NASA are compared with multiple commonly used models. Our research results show the hybrid model improves the coefficient of determination (R2) by 4%～23% and reduces the RUL absolute error to 1, demonstrating a higher prediction accuracy. The cyclic experiments are conducted later on vehicles under CLTC dynamic conditions, and predictions are made on battery life degradation. Our results reveal the SCSSA-CNN-BiLSTM model yields a root mean square error (RMSE) of 1.64 A·h and an R2 value of 0.98, delivering strong predictive and generalization performances.