Discrimination of battery characteristics using discharging/charging voltage pattern recognition

It is very important to have methods to determine the battery performances, such as state of charge (SOC) and state of health (SOH). Therefore, there are several methods to determine the performances of a battery in these days. However, these methods have drawbacks - different electrochemical characteristics, serious parameter changes due to temperature and aging. After all, it is limited SOC and SOH estimation for good performances due to unexpected and inconsistent parameters. For verification, few experiments are implemented using Li-Ion batteries. In this paper, battery characteristics are discriminated using discharging/charging voltage (DCV) pattern. In general, the parameters variations are intimately linked with voltage information of a battery. Especially, two resistance - series resistance and diffusion resistance in the lumped parameter battery model affected terminal voltage with the pulse current. When the pulse current is commonly applied to the batteries, the magnitudes of increased or decreased voltages are different, however, the patterns of voltages are similar each other. So, these patterns are used to apply the discrimination of battery characteristics using the hamming network. This network demonstrated proper technique for using a neural network for pattern recognition. The purpose of this is to decide which representative DCV pattern is closest to the current DCV pattern. Through statistical analysis of measured voltages, the proposed method is developed for recognition a current DCV pattern as one of 10 representative DCV patterns. The direct current internal resistance (DCIR) results are used for verification. A total of 10 fresh 1.3 Ah 18650 type Li-Ion batteries are used for DCV patterns and DCIR results at 25degC.