Aging-Robust and Current Sensor-Free Soft Short-Circuit Fault Diagnosis for Lithium-Ion Batteries

Transportation electrification is deemed an effective means to achieve carbon neutrality and requires the replacement of fossil-fueled vehicles with electric vehicles (EVs). Due to high volumetric energy, power density, and long lifespan, lithium-ion batteries (LIBs) have become the leading power source in EVs. To ensure the safe and reliable operation of LIBs, it is imperative to accurately estimate the real-time battery states and promptly diagnose any battery faults. In this article, we present an adaptive neural network-based observer design approach to achieve coestimation of battery dynamical states and soft short-circuit (SC) faults. Based on a first-order equivalent circuit model (ECM), a simple and parameter-independent state-space representation is developed to describe the electrical behavior of a battery subject to SC faults. Such a battery state-space model has the potential to enhance the robustness of the subsequent fault estimation procedure against different operating conditions (e.g., different temperatures and capacities). Then, a neural network-based adaptive observer is designed to compensate for the adverse effects of battery model nonlinearities and estimate the SC resistance in real-time. Furthermore, the stability of the resulting estimation error system is formally proved by the Lyapunov stability theory. Finally, extensive experiments are conducted to validate the effectiveness of the proposed method with an overall SC resistance estimation error of less than
$2.5 \Omega $
.