A feature prediction-based method for energy consumption prediction of electric buses

TY - JOUR

T1 - A feature prediction-based method for energy consumption prediction of electric buses

AU - Zhang, Zhaosheng

AU - Wang, Shuai

AU - Ye, Baolin

AU - Ma, Yucheng

N1 - Publisher Copyright: © 2024 Elsevier Ltd

PY - 2025/1/1

Y1 - 2025/1/1

N2 - Due to environmental concerns, electric buses (EBs) are becoming increasingly popular. Predicting the energy consumption of EBs is crucial for the operational management of bus fleets and the development of related infrastructure. This paper proposed a novel method for energy consumption prediction. Firstly, real-world data of 30 EBs covering a time span over one year from 3 bus routes was collected and the features that influence the energy consumption of EBs were selected and analysed. Then, these features were categorized into two groups: directly obtainable and indirectly obtainable, and models were built to predict the latter. Finally, the predicted features were combined with the directly obtainable ones to train an energy consumption prediction model based on Convolutional Neural Network (CNN). Cross validation was applied to evaluate the energy consumption prediction method, which achieved a mean average absolute percentage error (MAPE) of 8.13 %, surpassing other existing researches.

AB - Due to environmental concerns, electric buses (EBs) are becoming increasingly popular. Predicting the energy consumption of EBs is crucial for the operational management of bus fleets and the development of related infrastructure. This paper proposed a novel method for energy consumption prediction. Firstly, real-world data of 30 EBs covering a time span over one year from 3 bus routes was collected and the features that influence the energy consumption of EBs were selected and analysed. Then, these features were categorized into two groups: directly obtainable and indirectly obtainable, and models were built to predict the latter. Finally, the predicted features were combined with the directly obtainable ones to train an energy consumption prediction model based on Convolutional Neural Network (CNN). Cross validation was applied to evaluate the energy consumption prediction method, which achieved a mean average absolute percentage error (MAPE) of 8.13 %, surpassing other existing researches.

KW - CNN

KW - Electric buses

KW - Energy consumption prediction

KW - Feature prediction

KW - Machine learning

KW - XGBoost

UR - https://www.scopus.com/pages/publications/85213709909

U2 - 10.1016/j.energy.2024.134345

DO - 10.1016/j.energy.2024.134345

M3 - Article

AN - SCOPUS:85213709909

SN - 0360-5442

VL - 314

JO - Energy

JF - Energy

M1 - 134345

ER -