A Hybrid Battery Fault Diagnosis Method Based on Internal and External Battery Features

Lithium-ion batteries have become a foundational technology for modern energy storage systems due to their high energy density, long cycle life, and low self-discharge characteristics. Nevertheless, safety issues particularly thermal runaway triggered by internal or external abnormalities remain a critical concern in practical applications. Conventional fault diagnosis methods often rely solely on external electrical features, limiting their ability to detect early-stage faults with high precision. To address this limitation, this study proposes a novel fault diagnosis framework that integrates internal electrochemical parameters and external voltage characteristics. A simplified Single Particle Model is developed to extract internal battery parameters, with model fitting achieved through Particle Swarm Optimization. Simultaneously, voltage inconsistency is quantified using Lempel-Ziv complexity. These features are fused via an attention-based deep neural network, enabling comprehensive fault detection. Experimental validation on datasets collected from 90 electric vehicles demonstrates the effectiveness of the proposed approach, achieving a diagnostic Accuracy of 94.1%. The results highlight the advantages of incorporating multi-source features for robust and reliable battery fault diagnosis.