Non-linear robust identification of a lead acid battery model using multiobjective evolutionary algorithms

Abstract

This paper presents a complete (charging, discharging and thermal characteristics) non-linear model identification of a lead acid battery by using a multi-objective (MO) approach. Batteries are a very important part for a lot of processes, hybrid electric vehicles (HEV) are an example. In an HEV, the battery model is critical because defines the electrical behaviour of the system and directly influences the control strategy. The batteries can be charged and discharged at any time, so a complete battery model to simulate the vehicle behaviour and its control is needed. This paper presents a solution to this problem, taking into a account accuracy (the best fits batteries dynamics behaviour) and the robustness (non-modelled dynamics and manufacturing tolerance) properties of the model are considered simultaneously under a MO approach. Thanks to the MO approach the performance of each one of the Pareto optimal models can be analysed and a preferred model selected. In addition, an evaluation phase of the model has been done where it is observed that the model performance is accurate.