In the quest for sustainable mobility, electric vehicles (EVs) have emerged as a promising solution, but their widespread adoption hinges on overcoming several challenges, one of which is battery longevity. A team of researchers from the Korea Advanced Institute of Science and Technology (KAIST), including Yonggeon Lee, Jibin Hwang, Alfred Malengo Kondoro, Juhyun Song, and Youngtae Noh, has developed a novel approach to extend the life of EV batteries by optimizing the charging process. Their work was recently published in the journal Nature Energy.
The team’s research focuses on the fact that lithium-ion batteries (LIBs), the most common type of battery used in EVs, degrade more rapidly when kept at high states of charge (SOC) for prolonged periods. To mitigate this issue, they propose delaying full charging until just before the user’s departure. However, this strategy requires accurate prediction of departure times, which has been a challenging task due to the irregularity of individual routines.
The researchers developed a Transformer-based real-time-to-event (TTE) model that predicts EV departure times with high accuracy. Unlike previous methods that rely heavily on historical patterns, their model leverages streaming contextual information to capture irregular departure patterns within individual routines. The model represents each day as a TTE sequence by discretizing time into grid-based tokens, allowing it to process and learn from real-time data effectively.
To validate their approach, the team conducted a real-world study involving 93 users and passive smartphone data. The results demonstrated that their model outperformed baseline models in predicting departure times, highlighting its potential for practical deployment. By enabling delayed-full charging, the TTE model can help extend the life of EV batteries, reducing the environmental impact and cost of ownership associated with frequent battery replacements.
The practical applications of this research are significant for the energy sector, particularly in the realm of EV charging infrastructure. By integrating the TTE model into charging algorithms, EV charging stations can optimize charging schedules to minimize battery degradation, ultimately enhancing the sustainability and affordability of electric mobility. Furthermore, this approach can contribute to the development of smarter energy management systems that balance the demand for electricity with the availability of renewable energy sources.
In conclusion, the work of Yonggeon Lee and his colleagues at KAIST presents a promising solution to one of the key challenges facing the widespread adoption of EVs. By leveraging advanced machine learning techniques, their research offers a practical and effective way to extend battery life, paving the way for a more sustainable and efficient transportation system.
This article is based on research available at arXiv.