Context
In the era of electric vehicles (EVs), accurately predicting battery states is crucial for both vehicle performance and user experience. Two critical battery states are the State of Charge (SOC), which represents the remaining energy available for use, and the State of Health (SOH), which reflects the long-term degradation of the battery (Luo et al., 2022). SOC indicates the current energy level in the battery as a percentage of its total capacity, directly influencing the vehicle’s remaining range and short-term driving decisions. In contrast, SOH measures how well the battery performs compared to its original state, providing insights into its degradation over time and guiding maintenance or replacement planning. Efficient SOC prediction enables users to estimate the remaining driving range and plan recharging stops, directly influencing energy-efficient driving and short-term decisions. Moreover, this estimation, when correlated with electric vehicle usage, could play a significant role in adapting traffic control strategies to accommodate energy-efficient driving behavior and manage traffic flow based on the real-time energy states of vehicles.
This internship aims to explore advanced deep learning techniques for predicting the State of Charge (SOC) of EV batteries based on historical usage and operational data (Mayemba et al., 2024; Tian et al., 2023). The focus will be on utilizing sequential deep learning algorithms to process time-series data and on integrating recent attention mechanism architectures to enhance the accuracy and scalability of the predictions. Once the SOC is accurately predicted, the approach could be extended to estimate the electric vehicle batteries SOH, leveraging long-term data to analyze battery degradation trends and lifespan.
The project also offers the potential for further development through a PhD opportunity, allowing for in-depth exploration and research in the field.Objectives and Activities
The main goal of this internship is to develop a predictive model for battery health and lifespan in electric vehicles using deep learning techniques, focusing on scalability and accuracy.
