Robust Sensor Fault Diagnosis for Lithium-Ion Battery Systems Using an Electro-Thermal Model

Sensor faults can have a significant negative impact on battery system performance, and if not properly detected, may result in failure or safety issues. Hence, it is important to diagnose these faults in real time. Fault detection and isolation methods in battery systems mainly depend on current and voltage sensor measurements. This paper presents a model-based fault diagnosis approach for a Li-ion battery cell using an electro-thermal model. The proposed method uses the Extended Kalman Filter to estimate the state of charge and terminal voltage in real time. In addition, by incorporating a thermal model, the system not only enhances its fault detection capability by considering temperature effects but also effectively isolates various sensor faults, including current, voltage, and temperature sensor faults. Temperature-adaptive cumulative sum control charts are proposed to detect any small deviation between measured and estimated data, which indicates faults through adaptive process-related parameters. Moreover, the method demonstrates greater robustness against false alarms compared to the threshold method and maintains high fault detection performance even under noisy sensor measurements. The proposed fault diagnosis scheme is then validated by investigating different fault scenarios for each sensor through Li-ion battery cell experimental US06 drive cycle data, which demonstrates the effectiveness of the proposed method in modeling and fault diagnosis for various operating conditions in an electric vehicle battery system. The results demonstrate that the proposed method can be systematically tuned for different fault scenarios and operating temperatures, enabling accurate detection of various sensor faults across a wide temperature range with zero false alarms and minimal detection delay.