State-of-Charge Estimation for Lithium-Ion Batteries Using Guided Ultrasonic Waves and Recurrent Neural Network

초록

This study presents a novel methodology for estimating the state of charge (SOC) in lithium-ion batteries using features derived from guided ultrasonic wave (GUW) signals. Accurate SOC estimation is critical for the safe and reliable operation of electric vehicles, yet conventional approaches based solely on electrical signals suffer from limited accuracy and robustness. In contrast, GUW provides a sensitive, non-invasive means of probing internal changes in batteries, offering richer information for data-driven modeling. In this work, GUW signal parameters - including time-of-flight, amplitude, dispersion extent, and asymmetry - were extracted using the Matching Pursuit (MP) algorithm with a dictionary designed for asymmetrically dispersive waves. These parameters served as inputs to a gated recurrent unit-based encoder-decoder (GRU-ED) network, which effectively captures the temporal dynamics of SOC. Experimental validation was performed on commercial lithium iron phosphate (LFP) pouch cells under various charge-discharge conditions. The proposed GUW-based GRU-ED model achieved superior accuracy, with mean absolute error (MAE) and root mean square error (RMSE) values of 0.28% and 0.377%, respectively, compared to 0.737% and 1.044% from conventional electrical-signal-based estimators. These results demonstrate the effectiveness of GUW-informed deep learning models for real-time SOC prediction, highlighting their potential to improve safety, reliability, and efficiency in next-generation battery management systems. © 2025 IEEE.

키워드