Support-based Neural Network Ensemble Method for Predicting the SoH of Lithium-ion Battery

Abstract

Lithium-ion batteries are widely used in industrial and domestic applications because of their high energy ratio and low self-discharge rate. It is important to accurately predict the State of Health (SoH) of lithium-ion batteries as they degrade during use, which can lead to serious safety hazards. We propose a support-based neural network ensemble method, which incorporates the prediction results of several basic neural network models. First, a set of better initial integration weights is calculated and the initial integration result is obtained, then the support degree between this result and the prediction result of each basic neural network is calculated, and the final integration weights are calculated by the weight iterative update ensemble algorithm and the integration prediction result of lithium-ion batteries SoH is obtained. This method avoids the risk of the "majority principle" which does not guarantee that most models perform better, and removes the constraint of positive integration weights, which can further reduce the adverse effects of poorly performing models on the integration results. We demonstrate the effectiveness of the proposed ensemble method for the lithium-ion batteries SoH prediction problem through a 5-fold cross-validation experiment on two datasets.