Research on Thermal Insulation for EV Batteries with Latent Heat Storage System

Tatsuyuki Ohashi･Haruyuki Iijima (F.C.C.)･Takahiro Nomura･Tomokazu Nakamura (Hokkaido University)

Towards achieving carbon neutrality by 2050, the widespread use of EVs requires battery insulation technology suitable for cold climates. We are researching an EV battery insulation system with latent heat storage and high insulation technology. This presentation reports on the latent heat storage technology for EV battery insulation.

Development of Battery Data Generation Technology
Using Time Series Wasserstein GAN based on Deep Learning.

HYUNJUN JANG･TAEKYU KANG･WOOSUNG KIM (Hyundai Motor)

This paper presents a GAN-based deep learning approach for generating synthetic battery data. While lithium-ion batteries in xEVs are widely studied using deep learning to analyze their electrochemical characteristics under various conditions (current load, SOC, temperature), obtaining real data is challenging. Battery degradation testing is time-consuming, and extreme condition testing (like internal short circuits) is dangerous. The proposed solution uses a GAN architecture where Generator and Discriminator networks compete to create new, realistic battery data that matches the patterns of existing data, enabling testless data generation for battery research.

Realizing an energy and resource saving society by Dynamic Wireless Power Transfer for EVs

Naoya Kato･Tetsuji Mitsuda･Keisuke Tani･Takuya Osugi･Koji Yamaguchi･Yuki Konno･Mitsuhiro Ishihara･Toshiki Nagamatsu･Atsuki Ito･Keiichi Oshima (DENSO)

In order to achieve carbon neutrality, it is important to reduce the carbon generated during the production and the disposal of electric vehicles. The authors examined the specifications of Dynamic Wireless Power Transfer system while vehicle running, which has the potential to significantly reduce the on-board battery capacity, and estimated the effects of energy and resource savings.

Early fault detection in lithium-ion batteries using machine learning

Maximilian Kloock･Ethelbert Ezemobi･Seyedmehdi Hosseininasab･Lennart Bauer (FEV Europe)

Major battery faults include soft and hard short circuits, abnormal aging, over-charging, and over-discharging. Early detection of these faults can help to mitigate against thermal runaway. This work demonstrates a machine learning algorithm designed to detect these faults in an early stage and, consequently, prevent severe damage through thermal runaway. The machine learning algorithm combines the advantages of data-driven and model-based approaches. The algorithm is validated under dynamic load conditions using standard driving profiles of 100 simulated battery packs - each pack consisting of 80 cells, each including one faulty cell.