Numerical methods for detecting DC arc fault in lithium-ion batteries

In this paper, numerical methods to detect series arc fault in lithium-ion batteries are presented. The arc signals which are required for detecting arc fault are obtained using a test bench on which the Current Interrupt Device (CID) opens dynamically by contact release, associated with a 48 V DC battery pack and a resistor which can deliver a maximum of 1000 A. A comparison between the signals with and without arc is done to detect differences that can be used for arc detection. To isolate the arc signature, several methods are used, among them: the spectral analysis of arc signals using a Fast Fourier Transform (FFT) algorithm and a periodogram, the linear regression (moving average), the arc signals derivative and the filtering techniques. The spectral analysis shows a rise of the signal amplitude at high frequencies while the derivative method and the linear regression, among other things, show the instant when the arcing event occurs. Detection criteria may be set according to the type of method implemented. All of these methods mentioned above can be used to develop an Arc-Fault Circuit Interrupter (AFCI) for lithium-ion batteries.