Innovative approach to lithium-ion battery health monitoring in electric vehicles: adaptive LAPSVM sensor fault diagnosis

Abstract Accurate current and voltage sensor readings are crucial for battery management, ensuring protection against under-voltage and overcurrent, as well as enabling precise estimation of the battery’s State of Charge (SoC) and State of Health (SoH). To enhance battery reliability, safety, and lifespan, an effective sensor fault diagnosis approach is necessary. This paper presents a structured, model-based method for detecting faults in battery current and voltage sensors. The battery model is formulated using an equivalent circuit approach, while the Adaptive Extended Kalman Filter (AEKF) is employed to estimate the terminal voltage, forming the basis for fault detection. Residuals, obtained from the difference between estimated and measured voltages, serve as indicators of potential faults. A novel adaptive LapSVM method is applied to analyze these residuals and identify sensor anomalies. Additionally, the study examines how sensor faults influence SoC and SoH estimations, underscoring the importance of reliable diagnostics. The proposed method is validated through Matlab/Simulink simulations, demonstrating its capability to accurately detect sensor faults and enhance battery management.