Abstract: This article predicts the remaining service life of lithium ion batteries by constructing a model, and explores the effects of temperature and network parameters on the prediction accuracy of the constructed model, thereby improving the prediction accuracy of the model. A prediction method for the residual life of lithium ion batteries based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise（CEEMDAN）, One-dimensional Convolutional Neural Network（1D CNN） and Bi-directional Long Short-Term Memory（BiLSTM） neural network was proposed. Select capacity as the health factor, and then use CEEMDAN to decompose complex and non-stationary data to obtain stable components. 1D CNN is used to deeply mine the capacity data of lithium ion batteries.Finally, BiLSTM neural network modeling is used to predict the Remaining Useful Life（RUL） of lithium ion batteries. Using NASA and CALCE datasets for testing, the prediction performance was compared under different temperatures and network parameters, and compared with the BiLSTM model, SVR model, and CNN-BiLSTM model for prediction.