A challenging problem in energy storage systems for electric vehicles is the efficiency use of lithium multicell batteries. Because of production tolerances, unbalanced cells can be overstressed during usage leading to the reduction of the available capacity and premature failure of the battery pack. In order to reduce this problem the Università Politecnica delle Marche and FAAM Group S.p.A. developed a method for the selection and classification of homogenous cells to form uniform battery pack using Self Organizing Maps neural networks. Experimental data are collected from a set of LiFePo4 cells tested in FAAM laboratories. The selection considers both experimental data and identified characteristics: Discharge Voltage, Open Circuit Voltage, Total Capacity and identified parameters from Randle's equivalent circuit modelling. The State of Charge variability within each selected group of cells has been chosen as the clustering criterion to find the method which gives the best results in terms of homogeneity of the battery. Simulation results consider an experimental EV load profile and show a great reduction of the SOC variability and, consequently, in the balance of the battery pack for all the methods presented compared to random selection. Capacity and discharge voltage based methods give the best results over all.
Selection of Lithium cells for EV battery pack using Self Organizing Maps
2010
8 Seiten, 13 Bilder, 1 Tabelle, 20 Quellen
(nicht paginiert)
Aufsatz (Konferenz)
Datenträger
Englisch
Ice Shape Characterization Using Self-Organizing Maps
Online Contents | 2011
|Ice Shape Characterization Using Self-Organizing Maps
AIAA | 2009
|Self Organizing Maps (SOM) for Design Selection in Multi-Objective Optimization Using modeFRONTIER
British Library Conference Proceedings | 2008
|Self Organizing Maps (SOM) for Design Selection in Multi-Objective Optimization using modeFRONTIER
SAE Technical Papers | 2008
|