【摘要】 随着新能源汽车的快速发展，纯电动公交车在城市交通中的广泛应用使其能耗管理与运行效率问题日益受到关注。为实现对车辆运行过程中能耗的精准评估与预测，提出一种融合物理机制建模与深度学习方法的纯电动公交车能耗预测方法，以车辆纵向动力学模型为理论基础，构建能量消耗与运行参数之间的关系，并通过自然驾驶数据挖掘关键物理特征变量，设计多层长短期记忆（LSTM）网络结构，构建数据驱动的时序预测模型。预测模型在4辆不同公交车辆自然驾驶数据上进行预测实验，平均R~2为0.85，平均绝对百分比误差（MAPE）为8.34%，显示出良好的预测精度与泛化性能。该方法能够有效捕捉复杂运行工况下纯电动公交车的能耗特征，具备较高的拟合能力与适应性，为后续节能驾驶策略优化及能量管理系统设计提供了理论依据与方法支撑。更多还原

【Abstract】 With the rapid development of new energy vehicles, the widespread application of pure electric buses in urban transportation has brought increasing attention to the challenges of energy consumption management and operational efficiency. To achieve accurate evaluation and prediction of energy consumption during vehicle operation, a hybrid prediction method is proposed that integrates physical mechanism modeling with deep learning techniques. Taking the vehicle longitudinal dynamics model as the theoretical foundation, the method establishes the relationship between energy consumption and operational parameters. Key physical features are extracted from naturalistic driving data. To effectively capture the temporal characteristics of energy consumption, a multi-layer long short term memory(LSTM) network architecture is designed to construct a datadriven sequence prediction model. The proposed model is validated using real-world driving data from four different electric buses. Experimental results show that the model achieves an average R2of 0.85 and a mean absolute percentage error(MAPE) of 8.34%, demonstrating strong predictive accuracy and generalization capability. This method proves effective in characterizing the energy consumption behavior of pure electric buses under complex driving conditions, exhibiting high fitting ability and adaptability. It provides a solid theoretical basis and methodological support for the optimization of energy-efficient driving strategies and the design of energy management systems.更多还原