Abstract

Battery fault diagnosis has great significance for guaranteeing the safety and reliability of lithium-ion battery (LIB) systems. Out of many possible failure modes of the series–parallel connected LIB pack, cell open circuit (COC) fault is a significant part of the causes that lead to the strong inconsistency in the pack and the reduction of pack life. Therefore, it is extremely important to diagnose COC faults in real time. Motivated by this fact, we propose Kirchhoff's law based method, short-time Fourier transform based method, the Pearson correlation coefficient based method, dual extended Kalman filter (DEKF) based method, and long short-term memory recurrent neural network based method for diagnosing COC fault. These diagnostic approaches do not rely on other sensor data except pack current and terminal voltages of modules composed of cells in parallel. Furthermore, several experiments on a 4S-3P battery pack are taken under time-varying ambient temperature conditions to evaluate and compare the computation cost, diagnostic delay, and accuracy of these approaches. Test results show that only the DEKF-based approach owns the weakest robustness, and Kirchhoff's law based method with the merit of the lowest diagnostic delay and computation cost is the most appropriate approach for online COC fault diagnosis.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call