Identification of load dependent cell voltage model parameters from sparse input data using the Mixed Integer Distributed Ant Colony Optimization solver

Abstract

This work focuses on an approach in mixed integer nonlinear programming for parameter identification based on the implementation of an extended ant colony optimization (eACO) algorithm combined with a universal penalty method. The eACO algorithm is based on multi-kernel Gaussian probability density functions, which can be discretized for integer and mixed integer problems. Based on an equivalent circuit model, the battery parameter identification with the Mixed Integer Distributed Ant Colony Optimization (MIDACO) solver is proposed to estimate the load dependent voltage model parameters using only constant current charge and discharge characteristics of the cell. The parameter identification takes the influence of the charge or discharge rate into account as well as the state of charge of the battery. Additional weightings are introduced to set the focus on the peripheral regions of the voltage curves in order to achieve the best possible match between catalog and modeled voltage values. The MIDACO solver is used to extract parameters of six different battery cells. The results demonstrate that the MIDACO solver can extract optimal parameters in a robust, fast, and reliable process. A validation and thus, the general validity of the methodology described is presented based on the identified parameters.