User-Specific Energy Management for Urban Lightweight Electric Fleet Vehicles with Hybrid Energy Storage Systems

This research introduces an innovative approach to enhance energy management in urban lightweight electric fleet vehicles by leveraging Hybrid Energy Storage Systems (HESS) based on user driving velocity profiles. Long Short-Term Memory (LSTM) network is used to predict vehicle energy consumption by forecasting velocity profiles based on historical data. The LSTM-derived energy predictions are then utilized as inputs for the Energy Management System (EMS) of the HESS. A HESS, combining batteries and supercapacitors, offers a promising solution to urban lightweight electric fleet vehicles due to its ability to harness the strengths of both high energy and high power density storage technologies. Through accurate vehicle power demand anticipation enabled by LSTM's temporal analysis of urban driving conditions, the HESS model optimizes energy flow within the system, contributing to improved energy efficiency, reduced range anxiety, smoother user driving experience and improved energy recovery from braking. The created LSTM neural network is able to predict the velocity of 21 drive cycles with good accuracy, having a maximum Root Mean Square Error (RMSE) of 1.94. This prediction capability is then used for forecasting the energy consumption of the vehicle. The developed HESS model was able to split the predicted energy consumption signal into low-frequency components for batteries and high-frequency components for supercapacitors, with an error of 3\% between the required power and the delivered power. This research lays the foundation for a user-centric solution that holds significant potential to enhance energy management in lightweight electric fleet vehicles employing hybrid energy storage systems.