Two-stage model for heavy-duty truck fuel consumption prediction based on incomplete data
-
摘要:
为应对缺失数据下重型卡车油耗预测面临的不确定性与精度下降问题，提出了一种面向缺失数据场景的重型卡车油耗预测端到端双阶段深度学习模型；首先，采用Transformer特征提取与生成对抗网络相结合的持续对抗式数据补全方法，使生成数据分布逼近真实分布；随后，构建融合区域与纯图图注意力机制的油耗预测模型，通过局部与全局协同建模，增强对复杂动态特征的表征能力，从而实现重型卡车油耗的精准补全与预测；在缺失率为0~40%的真实重型卡车油耗数据集上开展试验，对所提模型的有效性进行验证。研究结果表明：在不同缺失率条件下，模型的数据补全均方误差分布于0.873~0.906，且整体性能均优于7种基线补全方法；影响因素分析表明，车速、加速度等局部动态因素对油耗补全与预测具有显著影响，而海拔、里程等全局因素影响相对较弱，说明局部特征对模型性能提升作用更为明显；对比试验结果表明，双阶段补全与预测模型取得最高预测精度时，均方误差达到0.602 0；消融试验进一步验证了各功能模块设计的有效性。该研究可为重型卡车燃油管理、碳排放监控及车队智能调度提供技术支撑，尤其适用于数据不完整条件下的复杂运行工况。
Abstract:To address uncertainty and accuracy loss in heavy-duty truck fuel consumption prediction under missing-data conditions, this study proposed an end-to-end two-stage deep learning model for incomplete data scenarios. First, a continual adversarial imputation method combined Transformer-based feature extraction with a generative adversarial network, so that the generated data distribution closely matched the real distribution. Next, a fuel consumption prediction model integrated regional attention and pure graph attention mechanisms. Joint local and global modeling improved representation of complex dynamic features, enabling accurate fuel consumption imputation and prediction. Experiments on a real-world heavy-duty truck fuel consumption dataset with missing rates from 0 to 40% validated the proposed model. Research results show that, across different missing rates, the imputation mean squared error (MSE) ranges from 0.873 to 0.906, and overall performance surpasses seven baseline imputation methods. Impact analysis reveals that local dynamic variables, such as vehicle speed and acceleration, significantly affect fuel consumption imputation and prediction, whereas global factors, such as altitude and mileage, have a weaker effect. This indicates that local features contribute more to performance improvement. Comparative experiments demonstrate that the proposed two-stage imputation-prediction framework achieves the highest predictive accuracy, with an MSE of 0.602 0. Ablation experiments further confirm the effectiveness of each functional module. This study provides technical support for heavy-duty truck fuel management, carbon emission monitoring, and intelligent fleet scheduling, especially under complex operating conditions with incomplete data.
-
图 2 Trans-GAN判别概率收敛曲线
Figure 2. Convergence curves of Trans-GAN discriminator probability
图 4 30%缺失率下各特征维度补全结果对比
Figure 4. Comparison of imputation results across various feature dimensions under a 30% missing rate
表 1 聚合后数据详情
Table 1. Aggregated data details
发生时间 平均海拔/ m 平均车速/（km·h-1） 平均转速/（r·min-1） 平均加速度/（km·h-2） 刹车次数 平均行驶里程/ km 燃油消耗量/ L 2021-01-01T14：50：00 918.03 10.91 961.82 0.02 15 0.6 0 2021-01-01T14：55：00 860.27 36.05 1 392.67 0.02 6 3.1 1 ︙ ︙ ︙ ︙ ︙ ︙ ︙ ︙ 2021-01-30T18：25：00 698.84 21.91 1 234.55 0.05 0 0.2 0 2021-01-30T18：30：00 698.86 32.08 758.40 -0.02 4 1.0 0 
下载: 导出CSV
表 2 数据补全对比试验结果
Table 2. Data imputation comparison test results
数据缺失率/% 均值填充 中位数填充 末次观测值填充 高频值填充 KNN填充 缺失森林填充 多重插补填充 Trans-GAN填充 0 0.880 0.886 0.883 0.890 0.892 0.895 0.894 0.873 5 0.885 0.888 0.890 0.894 0.897 0.902 0.900 0.874 10 0.892 0.890 0.895 0.902 0.902 0.905 0.905 0.875 15 0.895 0.910 0.900 0.906 0.907 0.917 0.911 0.875 20 0.905 0.913 0.905 0.910 0.912 0.919 0.915 0.875 25 0.906 0.916 0.910 0.914 0.917 0.923 0.923 0.906 30 0.913 0.917 0.915 0.920 0.922 0.925 0.927 0.906 35 0.915 0.920 0.918 0.926 0.927 0.935 0.930 0.896 40 0.917 0.925 0.928 0.930 0.932 0.936 0.932 0.896
下载: 导出CSV
表 3 在20%、40%、60%、80%数据缺失率下各模型油耗预测情况
Table 3. Fuel consumption prediction of various models under 20%, 40%, 60%, 80% data missing
模型 EMSE（20%/40%/60%/80%数据缺失率） ERMSE（20%/40%/60%/80%数据缺失率） EMAE（20%/40%/60%/80%数据缺失率） GBM 1.297 0/1.462 8/1.609 8/1.828 8 1.138 8/1.209 4/1.268 8/1.352 3 0.866 0/0.940 2/1.007 1/1.097 0 SVM 1.452 4/1.580 2/1.748 7/1.871 6 1.205 1/1.257 0/1.322 4/1.368 0 0.924 3/0.968 7/1.024 2/1.050 1 LSTM 1.742 7/1.750 3/1.988 4/2.211 8 1.320 1/1.323 0/1.410 1/1.487 2 1.011 6/1.001 0/1.083 0/1.120 6 ANN 1.495 4/1.545 8/1.858 9/1.896 9 1.222 8/1.243 3/1.363 4/1.377 2 0.916 0/0.956 4/1.062 6/1.077 4 TG-RAGAT 0.602 0/0.644 4/0.693 7/0.703 6 0.775 9/0.802 8/0.832 9/0.838 8 0.605 1/0.630 6/0.641 5/0.647 5
下载: 导出CSV
表 4 模型在不同油耗区间的表现效果
Table 4. Model performance across different fuel consumption intervals
分类 EMSE ERMSE EMAE 全部数据 0.602 0 0.775 9 0.605 1 低油耗区间 0.376 0 0.613 2 0.498 6 中等油耗区间 1.076 9 1.037 8 0.890 5 高油耗区间 2.791 3 1.670 7 1.492 0
下载: 导出CSV
表 5 消融试验结果
Table 5. Ablation experiment results
模型 EMSE ERMSE EMAE 完整模型（TG-RAGAT） 0.602 0 0.775 9 0.605 1 对比模型1（RAGAT） 0.634 2 0.796 3 0.615 1 对比模型2（RA） 1.408 7 1.186 9 0.966 8 对比模型3（GAT） 2.630 2 1.621 8 1.391 9
下载: 导出CSV
-
[1] World Health Organization. WHO ambient air quality database, 2022 update: Status report[R]. Geneva: World Health Organization, 2023: 1-44. [2] International Energy Agency. CO2 emissions in 2022[R]. Paris: International Energy Agency, 2023: 1-19. [3] 中华人民共和国交通运输部. 2025年交通运输行业发展统计公报[EB/OL]. (2026-06-22). https://xxgk.mot.gov.cn/jigou/zhghs/202606/t20260618_4207752.html .Ministry of Transport of the People's Republic of China. Statistical bulletin on the development of the transportation sector in 2025[EB/OL]. (2026-06-22).https://xxgk.mot.gov.cn/jigou/zhghs/202606/t20260618_4207752.html .[4] 清华大学社会科学学院经济学研究所. 中国公路货运大数据碳排放报告[R]. 北京: 清华大学社会科学学院经济学研究所, 2024: 1-46.Institute of Economics, School of Social Sciences, Tsinghua University. Big data report on carbon emissions from China's road freight transport[R]. Beijing: Institute of Economics, School of Social Sciences, Tsinghua University, 2024: 1-46. [5] FAN P F, SONG G H, ZHAI Z Q, et al. Fuel consumption estimation in heavy-duty trucks: Integrating vehicle weight into deep-learning frameworks [J]. Transportation Research Part D: Transport and Environment, 2024, 130: 104157. doi: 10.1016/j.trd.2024.104157 [6] WANG J H, RAKHA H A. Fuel consumption model for heavy duty diesel trucks: Model development and testing [J]. Transportation Research Part D: Transport and Environment, 2017, 55: 127-141. doi: 10.1016/j.trd.2017.06.011 [7] ZENG X H, JIANG X W, SONG D F, et al. High-precision vehicle engine fuel consumption model based on data augmentation [J]. Energy, 2025, 333: 137400. doi: 10.1016/j.energy.2025.137400 [8] PERROTTA F, PARRY T, NEVES L C. Application of machine learning for fuel consumption modelling of trucks [C]//IEEE. 2017 IEEE International Conference on Big Data (Big Data). New York: IEEE, 2017: 3810-3815. [9] 杜凯, 史晴晴, 李乐天, 等. 融合机器学习与可解释算法的货车燃油效率预测[J]. 公路交通科技, 2025, 42(7): 202-213.DU Kai, SHI Qing-qing, LI Le-Tian, et al. Truck fuel efficiency prediction integrating machine learning and interpretable algorithms [J]. Journal of Highway and Transportation Research and Development, 2025, 42(7): 202-213. [10] KATREDDI S, THIRUVENGADAM A. Trip based modeling of fuel consumption in modern heavy-duty vehicles using artificial intelligence [J]. Energies, 2021, 14(24): 8592. doi: 10.3390/en14248592 [11] 熊中敏, 郭怀宇, 吴月欣. 缺失数据处理方法研究综述[J]. 计算机工程与应用, 2021, 57(14): 27-38.XIONG Zhong-min, GUO Huai-yu, WU Yue-xin. Review of missing data processing methods [J]. Computer Engineering and Applications, 2021, 57(14): 27-38. [12] 马飞, 杨治杰, 王江博, 等. 基于气象-交通多通道数据融合的短时交通流速度预测模型[J]. 交通运输工程学报, 2024, 24(6): 183-196. doi: 10.19818/j.cnki.1671-1637.2024.06.013MA Fei, YANG Zhi-jie, WANG Jiang-bo, et al. Short-term traffic flow velocity prediction model based on multi-channel data fusion of meteorological and traffic [J]. Journal of Traffic and Transportation Engineering, 2024, 24(6): 183-196. doi: 10.19818/j.cnki.1671-1637.2024.06.013 [13] 张淼, 何仪娟, 杨博宇, 等. 基于LSTM-KF模型的高速列车群组追踪运行轨迹预测方法[J]. 交通运输工程学报, 2024, 24(3): 296-310. doi: 10.19818/j.cnki.1671-1637.2024.03.021ZHANG Miao, HE Yi-juan, YANG Bo-yu, et al. Trajectory prediction method for high-speed train group tracking operation based on LSTM-KF model [J]. Journal of Traffic and Transportation Engineering, 2024, 24(3): 296-310. doi: 10.19818/j.cnki.1671-1637.2024.03.021 [14] 陈小波, 陈程, 陈蕾, 等. 基于改进低秩矩阵补全的交通量数据缺失值插补方法[J]. 交通运输工程学报, 2019, 19(5): 180-190. doi: 10.19818/j.cnki.1671-1637.2019.05.018CHEN Xiao-bo, CHEN Cheng, CHEN Lei, et al. Interpolation method of traffic volume missing data based on improved low-rank matrix completion[J]. Journal of Traffic and Transportation Engineering, 2019, 19(5): 180-190. doi: 10.19818/j.cnki.1671-1637.2019.05.018 [15] MIOK K, NGUYEN-DOAN D, ROBNIK-ŠIKONJA M, et al. Multiple imputation for biomedical data using Monte Carlo dropout autoencoders [C]//IEEE. 2019 E-Health and Bioengineering Conference (EHB). New York: IEEE, 2019: 1-4. [16] TASHIRO Y, SONG J M, SONG Y, et al. Csdi: Conditional score-based diffusion models for probabilistic time series imputation[J]. Advances in Neural Information Processing Systems, 2021, 34: 24804-24816. [17] 王新屏, 王曦然, 王运明, 等. 基于LSTM-IMLP-Transformer的列车牵引电机温度预测模型[J/OL]. 铁道标准设计, 2025-10-27. https://doi.org/10.13238/j.issn.1004-2954.202508170001 .WANG Xin-ping, WANG Xi-ran, WANG Yun-ming, et al. Temperature prediction model for train traction motors using LSTM-IMLP-Transformer[J/OL]. Railway Standard Design, 2025-10-27.https://doi.org/10.13238/j.issn.1004-2954.202508170001 .[18] LIU J, PASUMARTHI S, DUFFY B, et al. One model to synthesize them all: Multi-contrast multi-scale transformer for missing data imputation [J]. IEEE Transactions on Medical Imaging, 2023, 42(9): 2577-2591. doi: 10.1109/TMI.2023.3261707 [19] 梁俊杰, 韦舰晶, 蒋正锋. 生成对抗网络GAN综述[J]. 计算机科学与探索, 2020, 14(1): 1-17.LIANG Jun-jie, WEI Jian-jing, JIANG Zheng-feng. Generative adversarial networks GAN overview[J]. Journal of Frontiers of Computer Science and Technology, 2020, 14(1): 1-17. [20] QIU H, ZHENG Q K, MSAHLI M, et al. Topological graph convolutional network-based urban traffic flow and density prediction [J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 22(7): 4560-4569. doi: 10.1109/TITS.2020.3032882 [21] GENG Z Q, ZHANG Y H, LI C F, et al. Energy optimization and prediction modeling of petrochemical industries: An improved convolutional neural network based on cross-feature [J]. Energy, 2020, 194: 116851. doi: 10.1016/j.energy.2019.116851 [22] ZHANG Z, WU L, LIU Q, et al. Understanding and improving fairness in cognitive diagnosis [J]. Science China Information Sciences, 2024, 67(5): 152106. doi: 10.1007/s11432-022-3852-0 [23] WANG X H, LI C, SUN Z G, et al. Review of GAN-based research on Chinese character font generation [J]. Chinese Journal of Electronics, 2024, 33(3): 584-600. doi: 10.23919/cje.2022.00.402 [24] SINGH D, SINGH B. Feature wise normalization: An effective way of normalizing data [J]. Pattern Recognition, 2022, 122: 108307. doi: 10.1016/j.patcog.2021.108307 [25] SUN J, FAN Y, YEH C C M, et al. Revealing the power of masked autoencoders in traffic forecasting[C]//SERRA E, SPEZZANO F. Proceedings of the 33rd ACM International Conference on Information and Knowledge Management. New York: Association for Computing Machinery, 2024: 4071-4075. [26] DU K, GUO X P, LI L T, et al. Traffic prediction in time series, spatialtemporal, and OD data: A systematic survey [J]. Journal of Traffic and Transportation Engineering (English Edition), 2025, 12(3): 666-700. doi: 10.1016/j.jtte.2025.03.001 [27] ZHU X Y, SHEN X Y, CHEN K L, et al. Research on the prediction and influencing factors of heavy duty truck fuel consumption based on LightGBM [J]. Energy, 2024, 296: 131221. doi: 10.1016/j.energy.2024.131221 [28] 程颖, 张佳乐, 张少君, 等. 大型货运车辆生态驾驶及节油潜力评估[J]. 交通运输系统工程与信息, 2020, 20(6): 253-258.CHENG Ying, ZHANG Jia-le, ZHANG Shao-jun, et al. Evaluation of eco-driving behavior and fuel-saving potential of large freight vehicles [J]. Journal of Transportation Systems Engineering and Information Technology, 2020, 20(6): 253-258. [29] MORADI E, MIRANDA-MORENO L. Vehicular fuel consumption estimation using real-world measures through cascaded machine learning modeling [J]. Transportation Research Part D: Transport and Environment, 2020, 88: 102576. doi: 10.1016/j.trd.2020.102576 -
计量
- 文章访问数: 12
- HTML全文浏览量: 8
- PDF下载量: 1
- 被引次数: 0
