Novel joint algorithm for state-of-charge estimation of rechargeable batteries based on the back propagation neural network combining ultrasonic inspection method

Abstract

State of Charge (SoC) is an essential indicator for energy storage distribution in lithium-ion batteries, which prevents overcharge and over-discharge of the battery by integrating it into the Battery Management System. Common estimation approaches are Extended Kalman Filtering (EKF), Coulomb counting, open-circuit-voltage (OCV), and neural networks (NN). While these methods are able to estimate SoC more accurately, they lack real-time monitoring capability for internal and external battery damage such as air bubbles and foreign objects. This study is the first to propose a novel joint algorithm for real-time monitoring of battery SoC and health with high accuracy, integrating ultrasonic non-destructive testing (NDT). The proposed algorithm predicts the battery overcharge based on the ultrasonic signal and reveals structural changes in commercial batteries during charging/discharging. Experimental results demonstrate that ultrasonic inspection cannot only enhance the SoC estimation accuracy of the backpropagation (BP) neural network, but also detects damaged conditions of the battery.