Early Faulty Battery Detection in Electric Vehicles Based on Self-Discharge Rate Analysis

Abstract

As the lithium-ion battery technology becomes mature and affordable, it has been widely adopted in transportation equipment and energy storage systems. However, there will always exist defects in the manufacturing process, even though at an extremely small percentage, that would result in the end product performing poorly and in rare cases causing safety issues. Therefore, continuous monitoring of the battery usage and early detection of battery faults become a must. This paper introduces a method to detect self-discharging, a leading phenomenon when batteries are failing, using data analytic algorithm on huge amount of run-time data from electric vehicles. The algorithm focuses on long term trend so that tiny self-discharging could be identified far ahead of it becoming much serious. The experiment on ten electric vehicles shows good results. Three abnormal self-discharging cases are detected in their early stages, ranging from 20 days to 5 months before they became serious enough to cause system malfunctions. It enables the service team to do preventative maintenance at the lowest cost, and most important of all, eliminate potential safety risks, whose value can never be over exaggerated. The method in this research can also be applied to different types of batteries and applications with only parameter adjustment.