TU Berlin

Electrical Energy Storage TechnologyFeBaL

Page Content

to Navigation


Field data based battery diagnosis and remaining life prognosis

  • Collaborative project
  • Partners: Fraunhofer IVI
  • Sponsored by BMBF
  • Project duration: 10/2020- 09/2023
  • Subproject: Prognosis of the knee-behaviour during Ageing and the development of application specific reference cycles


The battery is the last wearing part of an electric vehicle having a significant cost share (up to 30%). Therefore a good management of that asset is needed which leads to the necessity of precise battery dimensioning and evaluation of the utility in case of reuse, resale or switching to another post-automobile application. The system design is often based on datasheets from the manufacturer and accelerated ageing tests, where a remaining capacity measurement is considered adequate for the estimation of the ageing state.  These approaches share one common problem considering real applications: They do not represent the bandwidth of the stress and the interactions. A electric vehicle battery is neither used at a constant current of 1C and 100% DOD (Depth of Discharge) nor used under permanent high or low temperatures (accelerated ageing tests).

Aims and objectives

Based on the limitations of typical ageing tests the aim to learn the following characteristics from field data is designed.

·         Learn the ageing behaviour for identification of stress factors and their interactions

·         Learn the ageing state without additional capacity tests during application

·         Learn the usage patterns for the application based remaining useful life estimation

A new definition for the end of life of the battery could be developed, which describes the end of life as the point in time where the battery does not fulfil the specifications of the application anymore.

The approach of a stress factor based ageing model is used to capture the influence of different stresses on the battery parameters to enable a remaining useful life estimation. A Battery Neural Network (BNN) is deployed to extract ageing parameters from the application data to estimate the parameters of the ageing model.

That leads to the possibility of simulating experiments for any scenario to estimate the actual battery capacity and to predict the remaining useful life. Other factors like inhomogeneities or contact resistances play a decisive role in addition to the battery capacity and the internal resistance.

In addition to the BNN to extend the stress factor based ageing model, measurements are being conducted that should allow the analysis of the knee effect (exponential capacity loss near the end of life), enable the modelling and prediction of the ageing trajectory. The experiments are already running that are extended to reach the knee effect since this effects often appears after the boundary of 80% SOH (State of Health). Therefore, the knee effect prognosis is specifically important for Second-Life applications. The results should be applied to the BNN afterwards.


Innovations and perspectives

The methods described show innovative approaches since the analysis is conducted on field data and the aim is to forgo expensive measurements. That strengthens the main concept of the battery cluster (where this project is part of) because the results could be used to improve cell specifications and production. Second-Life applications profit from the results of the project, because realistic prognosis of the ageing behaviour is possible.


Quick Access

Schnellnavigation zur Seite über Nummerneingabe