Generating realistic data for developing artificial neural network based SOC estimators for electric vehicles

Tracking the state of a lithium-ion battery in an electric vehicle (EV) is a challenging task. In order to tackle one aspect of this task, we choose a data-driven approach for estimating the State of Charge (SOC), which is one of the most import parameters. In this context, the quality of the provided data is of utmost importance. Usually, standardized driving profiles are used to generate current profiles which are then applied to battery cells during testing. However, these standardized driving profiles exhibit significant deviation from real-world conditions, which can considerably affect the learning and validation performance of data-driven approaches. In this paper, we first propose a test profile generator which generates realistic current profiles for EV battery testing. Second, to demonstrate the effect of the proposed test profiles a multilayer perceptron (MLP) based SOC estimator is presented. Finally, we compare the results to the standardized driving profiles.