Dual-Powered Trolley Bus Inter-Stop Travel Energy Consumption Prediction Method: Temporal Feature Transformer Framework

Abstract

Dual-powered trolley buses (DTBs), with their low vehicle weight, extended range, high passenger capacity, and low investment costs, play a significant role in reducing infrastructure construction expenses, decreasing energy consumption, and promoting lightweight construction of public transit vehicles. Reliable prediction of the inter-stop travel energy consumption of DTBs is a crucial prerequisite for flexibly adjusting routes in the case of operational anomalies and developing new routes without the need for overhead cables. Although many scholars have conducted extensive research on the operational energy consumption prediction of electric buses and achieved substantial theoretical and practical results, studies on inter-stop travel energy consumption prediction for DTBs, particularly those that integrate regression prediction and temporal dependency, remain relatively uncommon. To address this gap, this paper presents a DTB inter-stop travel energy consumption prediction method, the temporal feature transformer model architecture. This approach, based on the self-attention mechanism, models both temporal dependency and the correlation among various variables. Experiments conducted on real driving data from two DTBs operating within the Zurich public transport network from 2019 to 2022 demonstrate that the proposed model outperforms traditional baseline models. Specifically, the model achieves a mean absolute percentage error of 8.76%, reflecting a reduction of more than 26.4% compared to the best-performing baseline. The findings of this study contribute significantly to the dynamic scheduling and effective management of DTBs, enhancing public transportation service levels and delivering environmental benefits through energy savings and reduced emissions. These promising results underscore the practical utility of the model.