1. Fault diagnosis algorithm of electric vehicle based on convolutional neural network and long short-term memory neural network.
- Author
-
Li, Xiaojie, Zhang, Yang, Wang, Haolin, Zhao, HeMing, Cui, XueLiang, Yue, Xikai, and Ma, Zilin
- Subjects
CONVOLUTIONAL neural networks ,PEARSON correlation (Statistics) ,FAULT diagnosis ,FEATURE extraction ,ELECTRIC vehicles - Abstract
Battery fault diagnosis is a key problem to ensure the safe operation of electric vehicles. In order to achieve accurate voltage prediction during the real operation of electric vehicles, this study proposed an electric vehicle real voltage prediction algorithm based on PCC-CNN-LSTM. First, the input and output characteristics were determined by Pearson correlation coefficient method. Considering the sampling time of individual cells in different battery packs, the number of cells and the working range of voltage, the parameters of the algorithm model were determined. Then, the sequential features in the voltage data were mined by CNN for feature extraction and input into LSTM for model training, which effectively solves the problem that the LSTM input feature dimension is too large to grasp the effective information, which makes the prediction accuracy decrease. Through the real vehicle test, for different types of battery packs, the evaluation indexes of the CNN-LSTM-based model are superior to those of other algorithm models (CNN, LSTM, GRU, and CNN-GRU). Finally, the superiority, reliability, and robustness of the proposed algorithm are verified by comparing with other algorithm models. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF