Structural Analysis in Reconfigurable Battery Systems for Active Fault Diagnosis

Conventional automotive battery systems consisting of a large number of battery cells pose a variety of challenges in terms of safety, reliability, lifetime, and energy efficiency. Reconfigurable battery systems (RBSs) are a promising solution to these issues of conventional battery systems. However, the large number of components in RBS also increases the fault probability. To meet this challenge on the way to fault tolerance, this article addresses fault isolation in an RBS, which comprises two switches per cell. Based on an electrothermal model, a structural analysis is performed and a sensor set with optimal fault isolation properties is found. Since the system consists of many equations, a novel algorithm is introduced to efficiently calculate minimal structurally overdetermined (MSO) subsystems for fault diagnosis. For each fault, the algorithm allows determining the MSO set that has the least number of equations. A complexity analysis of the algorithm reveals that the proposed algorithm is computationally significantly less expensive for systems with high redundancy, such as RBS, than existing algorithms that compute all MSO sets. Since the algorithm considers the switch states, it is suitable for active fault isolation through switches. The application to the RBS shows that the electrical equations are prioritized over the thermal equations due to the model uncertainties. A video file demonstrating the proposed graph-based algorithm with an example is attached to this article.

Automated summary

This article discusses the differences and advantages between conventional automotive battery systems and reconfigurable battery systems, focusing on fault isolation methods and algorithms in RBS. By experiment and structural analysis, a sensor set with optimal fault isolation properties was found, and a novel algorithm for efficiently calculating minimal structurally overdetermined subsystems for fault diagnosis was proposed.