Predicting energy consumption of electric two-wheelers under real-driving conditions in urban area: An application of Artificial Neural Network

Abstract

The study employs neural networks to simulate the energy consumption of electric two-wheelers (E2W) under real-world driving conditions to support deploying the E2W eco-system in Vietnam. The test E2W's operation characteristics, consisting of instantaneous speed and energy consumption, were collected on the representative streets of Hanoi, Vietnam. Various Artificial Neural Network-based model architectures and input variables were assessed to determine the optimal one. The developed model consists of three input variables of instant speed, acceleration, and vehicle-specific power, and has the highest predictability with a correlation coefficient R >0.9. A good similarity between the predicted and measured energy consumption was also obtained with an R-value of 0.90 to 0.94 for the independent data not included in the network training step. The model was developed as a valuable tool to assess the energy demand of vehicles under real driving conditions, thereby supporting the construction of the E2W eco-system to meet Vietnam's green energy transition target of Net-zero emissions by 2050.