Driving cycles quantify vehicle energy efficiency and emissions, making their optimization pivotal for advancing sustainable transportation systems and low-carbon urban transitions. Current research predominantly employs clustering-based approaches, which typically segment vehicle driving data into micro-trips. However, this methodology may encounter data imbalance issues. Traditional clustering techniques, commonly used in driving cycle construction, are particularly susceptible to the impact of data imbalance on micro-trip clustering results. Furthermore, the limited sampling of micro-trips during driving cycle construction can introduce sampling errors. To overcome these limitations, we propose a novel driving cycle construction method that integrates contrastive clustering and data reconstruction techniques. Our approach begins with feature extraction from micro-trips, followed by classification using contrastive clustering, with the objective of mitigating data imbalance. This is accomplished by leveraging data augmentation to generate positive and negative sample pairs and learning discriminative similarities. We then implement an adaptive algorithm for segment selection in driving cycle construction, and finally optimize the driving cycle through data reconstruction. To validate our method’s efficacy, we conducted comparative analyses against baseline methods, including standard driving cycles and adaptive algorithms based on spectral clustering and K-means clustering. The results demonstrate the superior performance of our approach, achieving optimal metrics with a feature error rate of 0.71% and a fuel consumption error rate of 1.88%. Thus, with its enhanced precision in driving cycle construction, this method emerges as a robust tool for optimizing emission assessments and energy demand forecasting. This study provides actionable insights for policymakers to refine emission regulations, incentivize low-carbon vehicle adoption, thereby accelerating the transition to sustainable urban mobility systems.
Li et al. (Wed,) studied this question.