Fraunhofer Institute for Transportation and Infrastructure Systems IVIThe electric mobility market has grown rapidly in recent years, with the battery remaining the most expensive component in an electric vehicle and thus a valuable asset. To ensure the cost-effective operation of these vehicles and establish a secondary market for used batteries, assessing the battery’s condition is of great importance.
The currently established methods for battery diagnosis are consistently based on complex laboratory measurements of individual battery cells or modules which do not adequately reflect real-world conditions. In an earlier research project (FeBal: Field-Data-Based Battery Diagnosis and Lifespan Prediction), Fraunhofer IVI developed an approach to optimize both capacity estimation and lifespan prediction for vehicle batteries using field data. The model reduces the scope of laboratory measurements and enables the estimation of current battery capacity across different vehicles. This allows various degradation effects to be linked to different usage profiles.
Validation on a larger scale
These results, which are particularly promising in terms of application and utilization, were initially achieved only based on data from a single bus fleet. This fleet consisted of a specific vehicle type, with all vehicles serving the same route, resulting in a relatively homogeneous load cycle. This is where the VALA project comes in, aiming to demonstrate the quality of the analysis results for other usage profiles, vehicle types, and applications as well, thereby validating the concept on a larger scale.
The service life prediction of vehicle batteries is based primarily on the number of cycles achieved to date under reference conditions. However, the reliability of this value under real-world operating conditions is questionable. The actual application scenarios and the resulting inhomogeneous operating conditions for the cells in the battery are not taken into account. It therefore seems obvious to make effective use of field data to improve the accuracy of service life predictions.
Field data for improved prognosis
The VALA project takes the approach of using neural networks to learn battery behavior from field data. The fleet model developed as part of this work
- combines a battery model for voltage prediction with an aging model for internal resistance and capacity,
- enables the determination of battery service life under given scenarios, and
- allows for the adaptation of operational scenarios.
In the process, the focus always remains on significantly extending battery life. Overall, this approach takes the accuracy of battery diagnostics to a new level, thereby significantly increasing both reliability and customer acceptance. As a result, it is possible, for example, to optimize the operational planning processes of logistics and transportation fleets, establish more targeted warranty conditions, and determine residual values with greater precision.
