Precise State-of-Charge Mapping via Deep Learning on Ultrasonic Transmission Signals for Lithium-Ion Batteries

Abstract

The uneven distribution of state of charge (SoC) in the lithium-ion battery is a key factor to cause fast decay of local electrochemical performance. Here, we report an acoustic method to realize SoC mapping in a pouch cell. A focused ultrasound beam is used to scan the cell, and the transmitted ultrasonic wave is analyzed with a deep learning algorithm based on the feedforward neural network. The deep learning algorithm effectively suppresses the disturbance of structural variation in different cells. As a result, the root mean squared error (RMSE) of the estimated local SoC is reduced to 3.02% when applying to different positions on different pouch cells, which is 11.07% of the RMSE by direct fitting SoC with acoustic time of flight. Combining with the progressive scanning technique, our method can realize non-destructive in situ SoC mapping with 1 mm in-plane resolution on pouch cells.