On-Board Short-Circuit Detection of Li-ion Batteries Undergoing Fixed Charging Profile as in Smartphone Applications

In this paper, an effective and robust algorithm is developed for on-board detection of battery anomaly caused by short circuit (SC). The proposed algorithm uses the battery-terminal voltage and current information measured using battery-management system during standard uses. The proposed method is purely a noninvasive, software-based solution and does not need any additional hardware. It extracts a set of designed features from the recorded data and stores their statistics as normal operating behavior for the initial five charge–discharge cycles. After the initial characterization of the battery, the likelihood of the features of being normal are evaluated for the subsequent cycles. If a feature is likely to be healthy, then the statistics for that feature get updated with the current value. However, if the feature is not likely to be healthy, then the battery is considered to be faulty with respect to that particular feature. If more than
$\text{50}\%$
of the features identify the battery as faulty, then the overall battery status is classified as faulty. A mobile application is developed to log the smartphone's battery data during charging and discharging for testing. The faulty data are generated by connecting SC resistances of various values to the battery terminals, which emulates the SC condition in the battery. The data logged in the smartphones are transferred to a PC for analysis using the proposed algorithm.
$\text{100}\%$
detection accuracy is achieved for SC resistance values less than
$\text{200}\,\Omega$
.