基于SARIMA-LSTM组合的机场起降量短时预测方法
Short-Term Forecasting Method for Airport Takeoff and Landing Volume Based on SARIMA-LSTM Combination
杨慧云 1李印凤 1段满珍 1阮昌2
作者信息
- 1. 华北理工大学应急管理与安全工程学院 河北 唐山 063210;唐山市空地智慧交通重点实验室 河北 唐山 063210
- 2. 中国民用航空华北地区空中交通管理局 北京 100621
- 折叠
摘要
机场起降量短时预测方法是根据空中交通流量管理需求,对机场未来24小时时间跨度内起降量情况进行预测.首先,构建了基于季节性差分自回归移动平均(SARIMA)和长短期记忆神经网络(LSTM)的机场起降量预测模型;然后,根据误差倒数法确定组合预测权重以期得到更好的预测效果;最后,使用天津滨海机场进行实例验证,以机场起降量的小时数据建立了SARIMA(0,1,7)×(0,1,1)24和LSTM模型,并分别以0.600和0.400的权重建立了组合预测模型.验证结果显示,组合模型的预测指标R2达到0.904,较反向传播(BP)神经网络等其他单一模型预测性能更佳.
Abstract
Short-term forecasting methods according to the demand of air traffic flow management,predict the airport takeoff and landing volume within a 24-hour time span.Firstly,an airport takeoff and landing volume forecasting models based on seasonal auto regressive integrated moving average(SARIMA)and long short term memory network(LSTM)is constructed.Then,with the error re-ciprocal method,the combined forecasting weights are determined to achieve better prediction results.Finally,using Tianjin Binhai Airport as an example to verify the model,the SARIMA(0,1,7)×(0,1,1)24 and LSTM models based on hourly takeoff and landing volume data are built,and a com-bined prediction model with weights of 0.600 and 0.400 respectively is established.The verification re-sults show that the prediction indicator R2 of the combined model reaches 0.904.It demonstrates better prediction performance than back propagation(BP)neural network and other single models.
关键词
机场起降量/季节性差分自回归移动平均(SARIMA)模型/长短期记忆神经网络(LSTM)模型/误差倒数法Key words
airport takeoff and landing volume/seasonal auto regressive integrated moving average(SARIMA)model/long short term memory network(LSTM)model/error reciprocal method引用本文复制引用
出版年
2024
NETL
NSTL
万方数据