Simultaneous Fault Isolation and Estimation of Lithium-Ion Batteries via Synthesized Design of Luenberger and Learning Observers

Abstract

Lithium-ion batteries possess high power, energy, and long cycle life. They are best candidates for applications on hybrid and electric vehicles. To ensure reliable operation, one of the functions in a battery management system is health monitoring in terms of fault diagnosis and estimation. The purpose of this brief is to provide a fault-diagnostic scheme that can achieve single fault isolation and estimation for a three-cell battery string subject to uncertainties. Detecting and isolating faults in systems subject to uncertainties is a challenging task due to the difficulty in distinguishing the effects of faults from uncertainties. To facilitate fault isolation, a bank of systems, each corresponding to a particular fault, are formulated by reorganizing the considered system. Each system in the bank is first transformed into two subsystems. Then, a classical Luenberger observer is designed for the first subsystem to generate a fault-detection residual. In this manner, a bank of reduced-order Luenberger observers are designed to locate a specific fault source, and thus fault isolation is achieved. Parallel to the bank of reduced-order Luenberger observers, a bank of learning observers (LOs) are also constructed to provide an estimate of the isolated fault. As a result, the synthesized design of Luenberger observers and LOs can realize simultaneous fault isolation and estimation. Parameters of an A123 battery cell are extracted via experiments, and effectiveness of the proposed design is demonstrated through simulation studies on the model of a three-cell battery string.