A novel Sensor Fault Diagnosis method for Lithium-ion Battery System Using Hybrid System Modeling

Abstract

A novel sensor fault diagnosis method for Li-ion battery system is presented in this paper by using hybrid system modeling. Battery packs are often combined with hundreds of battery cells connected in series and parallel with a request of large number of current and voltage sensors. The fault diagnosis of sensors is essential for battery management system (BMS) to ensure normal operation of battery system. To implement the diagnosis of sensor faults, the battery system is modeled as a hybrid system through stochastic automata in the paper. Several discrete states are defined to describe normal and faulty states of the system. Combining with the prior discrete states transitions, unscented particle filter algorithm is applied for estimating the most likely discrete states of the system so as to output diagnosis results. The experimental verification is conducted through a parallel battery pack with voltage and current sensors to test the algorithm effectiveness and the diagnosis accuracy of the proposed approach under dynamic current cycles.