New energy electric vehicle battery health state prediction based on vibration signal characterization and clustering

Abstract

The health status of the battery of new energy electric vehicles is related to the quality of vehicle use, so it is of high practical application value to predict the health status of the battery of electric vehicles. In order to predict the health status of lithium battery, this study proposes to optimize the empirical modal decomposition method and obtain the ensemble empirical modal decomposition algorithm, and use this algorithm to collect the vibration signal of the battery, then use wavelet transform to pre-process the collected signal, and finally combine K-mean clustering and particle swarm algorithm to cluster the signal types to complete the prediction of battery State of Health. The experimental results show that the ensemble empirical modal decomposition algorithm proposed in this study can effectively perform signal acquisition for different state types of batteries, and the K-mean clustering-particle swarm algorithm predicts a 63 % decrease in the health state of the battery at 600 cycles, with a prediction error of 2.6 %. Therefore, the algorithm proposed in this study is feasible in predicting the battery health state.

Keywords

• Vibration signal
• K-mean clustering algorithm
• EEMD
• Prediction
• Lithium battery