Research on a lead-acid battery fault detection method based on LSTM-AE-Cosine
- 1School of Intelligent Connected Vehicle, Hubei University of Automotive Technology, Shiyan, China
- 2Shiyan Key Laboratory of Air-Ground Crowd Cooperation Technology and Application (Hubei University of Automotive Technology), Hubei University of Automotive Technology, Shiyan, China
Abstract
Existing fault detection methods for lead-acid batteries in commercial vehicles are subject to several critical limitations, such as low detection accuracy and a shortage of labeled fault data. To address these issues, this paper presents an unsupervised deep learning framework for anomaly detection. In the framework, a hybrid LSTM-Autoencoder architecture is used to integrate the temporal feature extraction capability of long short-term memory networks with the reconstruction mechanism of Autoencoders, thus markedly improving the temporal modeling of battery data. Since the detection performance is highly dependent on the selection of reconstruction error metrics, multiple such metrics are employed in the model, and are systematically evaluated and compared to determine the most suitable metric for battery anomaly detection. The model is trained separately for static and driving conditions based on corresponding subsets of the battery datasets. With a fixed model structure and experimental conditions, model performance is evaluated using various reconstruction error metrics, and Bayesian optimization is used to identify the optimal detection threshold for each metric. Experimental results show that detection performance changes with different reconstruction error metrics and cosine similarity yields the best performance. This study offers a practical solution for online battery diagnosis and helps lower vehicle maintenance costs.
Related
- Rolling bearing fault diagnosis method based on dual attention multi-scale convolutional network
- Design of a comprehensive protection algorithm for fault detection, classification, and location in high voltage direct current transmission lines connected to wind farms based on local terminal measurements
- Detection of Carbapenemase-Producing Bacteria by Using an Ultra-Performance Liquid Chromatography-Tandem Mass Spectrometry Method
- Frequency modulated semi-continuous wave Doppler ultrasound: A novel method for the detection of scatterer velocity and range using wearable ultrasound instrumentation
- A field observation method to characterize site soundscape and improve detection probability estimation for eastern bat species in North America
BibTeX
@article{liu2026research,
title = {{Research on a lead-acid battery fault detection method based on LSTM-AE-Cosine}},
author = {Le Liu and Hongsheng Xu and Hai Tang and Jianing Cui and Li Feng and Ke Deng},
journal = {PLOS One},
year = {2026},
month = sep,
doi = {10.1371/journal.pone.0357237},
url = {https://doi.org/10.1371/journal.pone.0357237},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Research on a lead-acid battery fault detection method based on LSTM-AE-Cosine
AU - Le Liu
AU - Hongsheng Xu
AU - Hai Tang
AU - Jianing Cui
AU - Li Feng
AU - Ke Deng
PY - 2026
DA - 2026/09/15
JO - PLOS One
DO - 10.1371/journal.pone.0357237
UR - https://doi.org/10.1371/journal.pone.0357237
AB - Existing fault detection methods for lead-acid batteries in commercial vehicles are subject to several critical limitations, such as low detection accuracy and a shortage of labeled fault data. To address these issues, this paper presents an unsupervised deep learning framework for anomaly detection. In the framework, a hybrid LSTM-Autoencoder architecture is used to integrate the temporal feature extraction capability of long short-term memory networks with the reconstruction mechanism of Autoencoders, thus markedly improving the temporal modeling of battery data. Since the detection performance is highly dependent on the selection of reconstruction error metrics, multiple such metrics are employed in the model, and are systematically evaluated and compared to determine the most suitable metric for battery anomaly detection. The model is trained separately for static and driving conditions based on corresponding subsets of the battery datasets. With a fixed model structure and experimental conditions, model performance is evaluated using various reconstruction error metrics, and Bayesian optimization is used to identify the optimal detection threshold for each metric. Experimental results show that detection performance changes with different reconstruction error metrics and cosine similarity yields the best performance. This study offers a practical solution for online battery diagnosis and helps lower vehicle maintenance costs.
SN - 1932-6203
ER -