Research on SOH Prediction Method of New Energy Vehicle Power Battery

Abstract

The battery state of health (SOH) prediction is an important part of the new energy vehicle battery management system (BMS). Accurately predicting the SOH of the lithium-ion battery is of great significance for evaluating the health of the new energy vehicle power system and the remaining service life. The existing models for estimating the SOH of lithium-ion batteries have much room for improvement in terms of prediction accuracy and applicability. This article addresses the general accuracy and generalization problems of the existing lithium battery SOH prediction models. This paper proposes a lithium battery SOH prediction model based on the Temporal Convolutional Network, and uses particle swarm algorithm to optimize the model's hyper parameters. The model has high prediction accuracy on a variety of battery datasets. Subsequently, the transfer learning method is used to transfer the Temporal Convolutional Network model to the actual working condition data set, and the training set size is effectively reduced under the condition that the model prediction accuracy remains unchanged. Combined with the wavelet decomposition method, the Temporal Convolutional Network model is improved to achieve a fast and accurate estimation of the SOH of lithium batteries with fewer cycles.