Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

foxBMS - free and open BMS platform focused on functional safety and AI

: Waldhör, Stefan; Bockrath, Steffen; Wenger, Martin; Schwarz, Radu; Lorentz, Vincent

Postprint urn:nbn:de:0011-n-5967898 (174 KByte PDF)
MD5 Fingerprint: 87f5e374199d0261aae0e9546909f85e
Erstellt am: 24.7.2020

PCIM Europe 2020, International Exhibition and Conference for Power Electronics, Intelligent Motion, Renewable Energy and Energy Management. CD-ROM : 07.-08.07.2020 : proceedings
Berlin: VDE-Verlag, 2020
ISBN: 978-3-8007-5245-4
PCIM Europe Digital Days <2020, Online>
European Commission EC
H2020; 769900; DEMOBASE
DEsign and MOdelling for improved BAttery Safety and Efficiency
European Commission EC
H2020; 826060; AI4DI
Artificial Intelligence for Digitizing Industry
Konferenzbeitrag, Elektronische Publikation
Fraunhofer IISB ()

The last years have shown a strong market demand for lithium-ion battery systems with higher energy densities, longer lifetimes, and lower costs, but at the same time without compromising safety. To help developers, engineers and researchers worldwide, Fraunhofer IISB has established the free and open source Battery Management System(BMS)development platform foxBMS. The foxBMS platform consists of a modular hardware and software architecture and a complete software development toolchain. Based on the experience providing foxBMS-based solutions to customers and the research community, the next generation of foxBMS is strongly focused on functional safety standards. The hardware architecture and the hardware components themselves help to ensure that functional safety standards are met. Additionally, foxBMS supports a workflow for implementing Artificial Intelligence (AI)-based battery state estimators for the BMS. Using foxBMS as a data generator within this workflow a Neural Network (NN) based on a Long Short-Term Memory (LSTM) is trained offline to estimate the state of charge (SOC) based on the current and voltage measurement input. The simulation results, obtained with the trained NN running online on the device, are shown in this paper.