603
Optimized Discrimination Based on the Data Mining Combined with the Discrete Wavelet Transform for LiFePO4 Cells Consistency

Wednesday, 27 May 2015
Salon C (Hilton Chicago)

ABSTRACT WITHDRAWN

The cell’s discrimination process has been absolutely considered to be a prerequisite for stable configuration of the series/parallel-cell configured battery pack. Specifically, our conventional technique based on the pattern recognition was efficiently implemented to minimize cell-to-cell variation in the LiCoO2 battery pack. However, an identical discharging/charging current profile was only applied to all experimental cells for obtaining each discharging/charging voltage considered as representative patterns in the conventional technique. As a result, the application of the discharging/charging current signal (DCCS) profiles with different time interval, abrupt changes, and current magnitude results in erroneous information for correct discrimination of the cells that have similar electrochemical characteristics when using the conventional technique. Moreover, considered in case of the LiFePO4 cell, because of the flatness and hysteresis effect of the open-circuit voltage (OCV), it has been more difficult to succeed in solving aforementioned problems. Therefore, this research investigates a new approach in order to accomplish an optimized LiFePO4 cell’s discrimination for stable operation of the battery pack, together with the LiCoO2 cell. In this research, two-level basis discrimination based on the data mining combined with the discrete wavelet transform (DWT) for LiFePO4 cells consistency is newly introduced. In the first level, 10 current profiles with different discharging/charging sequence are applied to a LiFePO4 cell for obtaining different DCCS patterns. At this level, low/high frequency components decomposed by DWT-based multi-resolution analysis (MRA) are applied as characteristic parameters in the data mining of an unknown DCCS. In the second level, 15 representative discharging/charging voltage signal (DCVS) patterns considering previously discriminated DCCS in the first level, are applied to other data mining of an arbitrary DCVS. Through proposed work, it can be expected to provide an elaborate identification of the representative DCCS and DCVS that most closely match those of an arbitrary LiFePO4 cell.