The voltage abnormal fluctuation is a warning signal of short-circuit, over-voltage and under-voltage. This paper proposes a scheme of three-layer fault detection method for lithium-ion batteries based on statistical analysis. The first layer fault detection is based on the thresholds of over-charge and over-discharge of a battery pack.
The voltage data can also be filtered and decomposed to extract features, and the features can be exploited to visualize the evolution of abnormal cells more intuitively with clustering . The current and temperature data also plays an important role in lithium-ion battery fault diagnosis.
The lithium-ion batteries may experience the abnormal changes of voltages and current, the abrupt rise of temperature during a thermal runaway process , . Therefore, many researchers diagnose faults by using temperature and voltage data. Remarkable endeavors have been dedicated to fault diagnosis of batteries.
The systematic faults of battery pack and possible abnormal state can be diagnosed by one coefficient. For the voltage abnormality, an accurate detection and location algorithm of the abnormal cell voltage are attained by combining the data analysis method and the visualization technique.
From the detection results and the voltage variation trajectories of cells, it can be concluded that the detected abnormality is a rapid descent of voltage caused by the battery pack that is discharged with a high rate current in a low voltage stage.
Root cause 1: High self-discharge, which causes low voltage. Solution: Charge the bare lithium battery directly using the charger with over-voltage protection, but do not use universal charge. It could be quite dangerous. Root cause 2: Uneven current.