Electric vehicle powertrain sizing through deep learning analysis of multiple drive cycle profiles for evaluating the energy consumption

Abstract

Accurate estimation of the power requirements of an electric vehicle (EV) under various driving conditions is critical for efficient component design and optimal performance. In this paper, an integrated framework is proposed that leverages physics based vehicle dynamics simulation and deep learning based predictive modeling for the analysis of EV energy consumption on various standardized global drive cycles such as Urban Dynamometer Driving Schedule (UDDS), US06, Highway Fuel Economy Driving Schedule (HWFET), Worldwide Harmonized Light Vehicle Test Procedure (WLTP) Class 2 and Class 3. Simulation analysis reveals that the maximum traction force required ranges from 2800 N to 6200 N, and the maximum torque force ranges from 950 Nm to 2145 Nm depending on the driving conditions. The average energy consumption is observed to vary between 660 Wh/km to 730 Wh/km, with the US06 aggressive cycle showing the maximum energy consumption. Deep learning based comparative analysis reveals that the Long Short-Term Memory (LSTM) network offers better prediction accuracy for the transient torque, force, and power outputs of the EV compared to the standard feedforward neural network due to its ability to model the temporal dynamics of driving patterns. The hybrid modeling framework presented in this paper facilitates the accurate estimation of EV power requirements and aids in the design of practical battery and motor sizing.