Concurrent multi-fault diagnosis of lithium-ion battery packs using random convolution kernel transformation and Gaussian process classifier

The timely detection and accurate differentiation of concurrent diverse faults within lithium-ion battery packs are essential for triggering targeted countermeasures by the battery management system, thereby ensuring the safe and stable operation of the battery system. Existing methods for multi-fault diagnosis in lithium-ion battery packs often assume that different types of faults do not occur simultaneously and face difficulties when determining accurate diagnosis thresholds. Therefore, this work proposes a method based on random convolution kernel transformation and Gaussian process classifier to achieve concurrent multi-fault diagnosis of lithium-ion battery packs without establishing battery models or setting diagnosis thresholds. First, an interleaved voltage measurement circuit is employed to capture information about various faults without increasing the number of sensors. Subsequently, a multitude of convolution kernels with random parameters are employed to extract features that effectively represent various fault patterns from voltage measurements. Finally, a diagnosis model based on a Gaussian process classifier is constructed to detect and isolate different types of concurrent faults while integrating the utilized interleaved voltage measurement circuit to pinpoint the specific locations of fault occurrences. The experimental results demonstrate that the proposed method can achieve a diagnosis accuracy of 98.44%, and thus confirms its effectiveness and feasibility.