Year: 2025
Type: conference-paper
Abstract: In order to establish a more accurate and stable battery charge state prediction model, a new ultrasonic guided wave signature, signal envelope, is proposed in this paper to predict battery SOC with machine learning. Firstly, ultrasonic characteristic parameters of experimental data are extracted by various algorithms, then the battery state is predicted by several machine learning methods, and finally, these results are analyzed to obtain the most accurate and stable SOC prediction machine learning model. The results show that the ultrasonic envelope signal is closely related to the state of the battery in the ultrasonic guided wave experiment. After the signal envelope is used as the feature, the prediction accuracy of several machine learning methods is greatly improved, and the fitting degree of the Gaussian regression method combined with the signal envelope feature in the charge and discharge prediction is as high as 0.9981 and 0.9971.
DOI: 10.58286/31023
Authors: Huang Hao, Kangpei Meng
Institutions: Chang'an University, Ningbo University of Technology
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4409730738