Electric Vehicle Range Estimation based on Driving Behaviour Employing Long Short-term Memory Neural Network

• Published in: International Symposium on Power Electronics, Electrical Drives, Automation and Motion pp. 551 - 555
• Main Authors: Tiwary, Anyuti, Mishra, Sukumar, Kumar, Utkarsh
• Format: Conference Proceeding
• Language: English
• Published: IEEE 19.06.2024
• Subjects: Batteries; Driving behaviour; Electric vehicles; Estimation; Long Short-term Memory; Range estimation; Recording; Recurrent neural networks; Robustness; Urban areas
• ISSN: 2835-8457
• DOI: 10.1109/SPEEDAM61530.2024.10609085

Although, internal combustion engine (ICE) vehicles are being substituted with electric vehicles (EVs), however, the range anxiety still prohibits widespread adoption of EVs. Therefore, it is crucial to comprehend how a driver's driving style affects the EV's range when covering a trip over different routes. Thus, this paper focuses on providing the EV user with a driving behaviour-based, range estimation using a Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN). It uses three crucial variables to estimate the range travelled: velocity, acceleration/deceleration, and EV's battery State of Charge (SoC). Here, LSTM RNN is used because it comprehends the long-term dependencies of EV drivers' driving characteristics. Moreover, the paper utilizes Adam optimizer and a dropout technique that adaptively optimize the LSTM RNN to address the overfitting issue. Lastly, the algorithm has been verified for four distinct trips such as Campus, Market, Ring, and City routes recording an error of less than 1%.