DFG Graduiertenkolleg 2153: "Energiezustandsdaten - Informatik-Methoden zur Erfassung, Analyse und Nutzung"

Vortrag von Philipp Staudt bei der OR2016 Hamburg

Am 1. September 2016, trägt Philipp Staudt zum Thema "Simulation-based Data Selection and Reduction for Battery Degradation Modeling in Electric Vehicles" auf der OR2016 in Hamburg vor.

Englische Zusammenfassung: Simulation-based Data Selection and Reduction for Battery Degradation Modeling in Electric Vehicles

Traction batteries are one of the most cost intensive components of electric vehicles (EV) and therefore limit their driving range. Due to battery degradation and the resulting capacity fade, this phenomenon is amplified. The influence of usage patterns on degradation has not been fully investigated by now and sufficiently detailed field data is not yet available. Consequently, OEMs are facing the challenge of selecting and reducing the relevant features of the potentially vast amount of data created by electric vehicles in the field in order to optimize the ratio of information gained and the data volume transmitted.

To address this challenge, a simulation-based analysis is conducted, which captures the influence of the dynamic user behavior and the resulting degradation on EV batteries. The simulation is based on a realistic aging model and more than 1000 representative driving profiles. By means of variable selection using Lasso regression and dimension reduction techniques, such as PCA, different forecasting models are developed. Furthermore, the trade-off between forecasting ability and the required data volume is evaluated. Additionally, the models are tested for their ability to predict the battery’s end-of-life. We find that by using a small amount of data, which is suitable for mobile transmission, a satisfactory forecasting accuracy can be achieved. All developed models use driving behavior features as input factors. This suggests changes in the warranty design of OEMs. Furthermore, the results of the end-of-life prediction are promising regarding a predictive maintenance strategy.