Deep Residual Net-Based Prognosis Method for Lithium-ion Batteries with Information Fusion from Different Scales

Nowadays, prognosis methods based on deep learning have been successfully developed and applied in many industrial fields, such as energy, transportation, aero-space engineering etc. Lithium-ion battery prognosis is very important to indicate the health states of the energy system, which has been a hot topic in the past decades. In this paper, a new method is proposed for battery prognosis. The proposed architecture integrates the traditional DNN and CNN models, and divides the feature graph into two branches for separate computation. Residual network is also used in the prediction model for pursuing better effects. Residual block is implemented by a hidden layer connecting each branch’s inputs and outputs, which improves the model’s generalization ability. The proposed method takes up one-step prediction for CALCE lithium-ion data set. Experimental results show that the proposed method has a better prediction effect. Therefore, it is of great significance to predict the life of lithiumion batteries and become a new basis for a deep learning-based method to predict the life of lithium-ion batteries.