Estimate e-Golf Battery State Using Diagnostic Data and a Digital Twin

Abstract

Li-ion battery packs are the heart of modern electric vehicles. Due to their perishable
 nature, it is crucial to supervise them closely. In addition to on-board supervision over safety and
 range, insights into the battery’s degradation are also becoming increasingly important, not only
 for the vehicle manufacturers but also for vehicle users. The concept of digital twins has already
 emerged on the field of automotive technology, and can also help to digitalize the vehicle’s battery.
 In this work, we set up a data pipeline and digital battery twin to track the battery state, including
 state of charge (SOC), state of health (Capacity) (SOHc ) and state of health (Resistance) (SOHr ). To
 achieve this goal, we reverse-engineer the diagnostics interface of a 2014 e-Golf to query for Unified
 Diagnostic Services (UDS) messages containing both battery pack and cell-individual data. An
 on-board diagnosis (OBD) logger records the data with edge-processing capability. Pushing this data
 into the cloud twin system using IoT-technology, we can fit battery models to the data and infer
 cell individual internal resistance from them. We find that the resistances of the cells differ by a
 magnitude of two. Furthermore, we propose an architecture for the battery twin in which the twin
 fleet shares resources like models by encapsulating them in Docker containers run on a cloud stack.
 By using web technology, we present the analyzed results on a web interface.