Detection and isolation of faults in a lithium-ion battery pack using a switched architecture of equivalent cell diagnosers

Abstract

Lithium-ion battery packs are typically built as a series network of Parallel Cell Modules (PCM). A fault can occur within a specific cell of a PCM, in the sensors, or the numerous connection joints and bus conductors. This paper presents a method of detecting a single occurrence of various common faults in a Lithium-ion battery pack and isolating the fault to the faulty PCM, its connecting conductors, and joints, or to the sensor in the pack using a Diagnostic Automata of configurable Equivalent Cell Diagnosers. This is achieved by activating a sequence of diagnosers that generate residuals for various sub-networks of the pack and test hypotheses on them to search for the smallest sub-network and the associated sensors that contain the source of the fault. The sequence of activation of the diagnosers follows the state transition of the Diagnostic Automata triggered by the outcomes of the hypotheses tests. Once the faulty PCM sub-network is identified, the fault type is classified based on selected features of the residuals. For robust detection, events arising from hypothesis tests in the diagnosers and the classifiers are used to compute the probability of occurrence of a specific fault in a specific PCM. The performance of the diagnostic framework is tested offline with experimentally obtained data from an automotive battery pack. Finally, a Hardware-in-the-Loop simulation test demonstrates that the proposed method can be implemented on standard Battery Management System hardware to avoid extensive damage to the pack, enhance safety, and facilitate quick cell-level replacements to repair expensive Lithium-ion battery packs.