A machine learning based method for lithium-ion battery state of health classification and prediction

Accurate state of health (SOH) prediction for lithium-ion batteries is a key technology in battery applications. However, because of the varied failure modes, complicated electrochemical systems, and production variations, the degradation of lithium-ion batteries frequently exhibits high dispersion, making it challenging to precisely forecast the SOH of the lithium-ion battery. To solve this problem, this paper proposes a machine learning-based method for classifying and predict the SOH of lithium-ion batteries. First, based on the accuracy constraints, the double subgroup optimization algorithm is used to determine the appropriate number of categories and category ranges for the training set data. Then, based on the Softmax classification model, lithium-ion batteries are classified according to the early-cycle data, so that the batteries with a similar degradation trend are divided into one class. To limit the impact of data dispersion and increase forecast accuracy, the SOH prediction model for each kind of battery is built using a backpropagation neural network. Compared with the traditional method, the prediction error of the proposed method is reduced by more than 34%, which verifies the effectiveness and superiority of proposed method. © 2023 Beijing University of Aeronautics and Astronautics (BUAA). All rights reserved.