考虑航班计划的机场短时停车需求预测

为克服现有短时停车需求模型无法直接利用于机场停车需求预测这一问题，利用停车数据、航班计划和气象信息，建立了面向机场停车场的短时停车需求预测模型。首先使用机场停车数据分析了停车场短时车辆到达与离去特性，然后考虑到航班计划对机场停车场短时停车需求的影响，将其与气象状况同时引入到短时停车需求影响因素中，建立了基于Conv1D-LSTM神经网络结构的机场短时停车需求模型。以上海虹桥机场停车场为实例，Conv1D-LSTM模型实验结果的平均绝对误差和均方根误差分别为12.057 辆和14.237 辆；对比多个其他模型实验结果，本文构建的Conv1D-LSTM模型预测效果更优，能有效应用于机场停车场短时停车需求预测。

Short-term Parking Demand Prediction Model for Airport Parking Lots Considering Flight Schedules

To overcome the problem that the existing short-term parking demand model cannot be directly used for airport parking demand prediction, a short-term parking demand prediction model for airport parking lots has been proposed based on parking data, flight schedules, and weather data. Firstly, the parking data was used to analyze the short-term vehicle arrival and departure characteristics. Then taking the impact of flight schedules and the weather on the short-term parking demand of airport parking lots, an airport short-term parking demand prediction model using the Conv1D-LSTM neural network was established. Taking the Shanghai Hongqiao Airport parking lots as an example, the mean absolute error and root mean square error of the Conv1D-LSTM model are 12.057 and 14.237 vehicles, respectively. Compared with several other models, the Conv1D-LSTM model built in this study has a better prediction performance and can be effectively used for short-time parking demand prediction in airport parking lots.