Optimal parametrization of electrodynamical battery model using model selection criteria

This paper describes the mathematical parametrization of an electrodynamical battery model using different model selection criteria. A good modeling technique is needed by the battery management units in order to increase battery lifetime. The elements of battery models can be mathematically parametrized to enhance their implementation in simulation environments. In this work, the best mathematical parametrizations are selected using three model selection criteria: the coefficient of determination (R-2), the Akaike Information Criterion (AIC) and the Bayes Information Criterion (BIC). The R-2 criterion only takes into account the error of the mathematical parametrizations, whereas AIC and BIC consider complexity. A commercial 40 Ah lithium iron phosphate (LiFePO4) battery is modeled and then simulated for contrasting. The OpenModelica open-source modeling and simulation environment is used for doing the battery simulations. The mean percent error of the simulations is 0.0985% for the models parametrized with R-2, 0.2300% for the AIC ones, and 03756% for the BIC ones. As expected, the R-2 selected the most precise, complex and slowest mathematical parametrizations. The MC criterion chose parametrizations with similar accuracy, but simpler and faster than the R-2 ones. (C) 2015 Elsevier B.V. All rights reserved.