State of Health Classification for Lead-acid Battery: A Data-driven Approach

Abstract

In general, methods that use a data-driven approach in estimating lead-acid batteries’ State of Health (SoH) rely on measuring variables such as impedance, voltage, current, battery’s life cycle, and temperature. However, these variables only provide limited information about internal changes in the battery and often require sensors for accurate measurements. This study explores ultrasonic wave propagation within a lead-acid battery cell element to gather data and proposes a data-driven approach for classifying the SoH. The results demonstrate that a neural network classifier can effectively distinguish between two classes: 1) batteries in a healthy state with SoH greater than 80%, and 2) batteries in an unhealthy state with SoH less than 80%. The data-driven approach introduced in this study, which uses ultrasonic wave data, provides valuable information relative to the changes in the internal cell of the battery. Conventional external measurements may not capture this information. Consequently, it eliminates the need for additional sensor installations and offers a promising alternative for SoH classification.