Outlier mining-based fault diagnosis for multiceli lithium-ion batteries using a low-priced microcontroller

Fault diagnosis for lithium-ion batteries involves detecting faulty cells and identifying types of faults. It is crucial to build safety-critical battery systems, which has not been fully integrated in conventional battery management systems. This paper proposes a novel data mining-based real-time fault diagnosis for multicell lithium-ion batteries using a microcontroller. The proposed fault diagnosis algorithm includes: 1) a model-based battery condition monitoring algorithm that estimates physical model parameters and operational states and 2) an outlier detection algorithm that detects abnormal battery cells based on the outcomes of the condition monitoring and identifies the types of faults such as internally shorted cells and anomaly aged cells. The proposed fault diagnosis method is implemented in a low-priced microcontroller and validated by experiments in a multicell battery simulation testbed.