Currently, the remaining useful life (RUL) prediction of lithium-ion batteries through data-driven methods undergoes extensive investigation. However, due to the influence of individual battery differences and dynamic environmental conditions, traditional data-driven modeling is difficult to accurately represent the actual working status of lithium batteries. This paper proposes a method for predicting the RUL of lithium batteries by dynamically updating the degradation model. It utilizes a hybrid data-driven approach combining convolutional neural networks (CNN) and long short-term memory networks (LSTM) to build a predictive model for lithium batteries. Additionally, it incorporates mechanisms for concept drift detection and dynamic model updates, enabling real-time monitoring and pinpointing of concept drifts within the actual data streams. By dynamically updating the pre-trained static models, we aim to enhance the accuracy and adaptability of predictions. Through a comparative analysis of prediction outcomes before and after the updates, we validate the effectiveness of our proposed method in addressing data drifts and improving the RUL prediction performance. This innovative technique offers a new approach for lithium battery management and maintenance, contributing significantly to the realm of monitoring battery health.