基于 VMD-GRU 的城市轨道交通短时客流预测

精准的客流预测是轨道交通运输计划编制的基础和依据，为提高城市轨道交通短时客流的预测精准度， 基于城市轨道交通短时客流的动态性、非线性、不确定性、周期性、非平稳性及时序性等特点，提出一种组合 模型预测方法，即 VMD-GRU 神经网络预测模型，由变分模态分解和门控循环单元组合而成。变分模态分解的 作用是分解短时客流，降低数据中的噪声，减少数据波动；门控循环单元的作用是基于分解的短时客流，进行 客流预测。经南京地铁的数据验证，该模型在地铁短时客流预测方面效果良好。与 GRU 相比，VMD-GRU 在 15、30 和 60 min 的时间粒度下，预测准确度分别提升 7.57%，16.93%，18.47%。该模型可为地铁运营管理部 门对车站客流管理、日常行车计划制定等提供有效的数据支撑，从而提升线网总体运营效率以及轨道交通系统 的服务水平。

Forecast of Short-term Passenger Flow of Urban Rail Transit Based on VMD-GRU

Precise passenger flow forecasting is the basis of rail transit planning. In this study, we propose a combined model method to improve the prediction accuracy of short-term passenger flow. The method is based on the dynamic, nonlinear, uncertain, periodic, non-stationary, and sequential characteristics of short-term passenger flow in urban rail transit. The model is a neural network prediction model. This model comprises variational mode decomposition (VMD) and gated recurrent units (GRUs) to forecast the short-term passenger flow of urban rail transit. The role of VMD is to decompose short-time passenger flow and reduce noise and fluctuations in the data. The GRU is based on the breakdown of short-term passenger flow to perform passenger flow prediction. Data verification by Nanjing Metro shows that the model is effective in short-term passenger flow forecasting of urban rail. Compared with the GRU, the prediction accuracy of VMD-GRU at 15 min, 30 min, and 60 min improved by 7.57%, 16.93%, and 18.47%, respectively. The model can provide effective data support for operations and management, such as station passenger flow management and daily train schedules, thereby enhancing the operational efficiency and service level of the rail transit system.