







摘" 要: 隨著全球化進程的加快和航空技術的發展,對空中交通流量預測的精度要求也越來越高。為提高空中交通流量預測精度,減輕空中交通管制員的壓力,提出一種增強位置編碼的Transformer模型。利用小波變換對原始空域流量數據進行分析,通過信噪比選出性能最優的小波基函數,再進一步計算出小波系數并將其融入位置編碼,以增強模型對時間序列數據的理解能力。實驗結果表明,所提模型能夠準確捕捉空中交通流量數據中的非平穩性和突變特征,其RMSE和MAPE評估指標較原始Transformer模型分別降低了29.9與2.9%,較LSTM模型分別降低了34.5與3.4%。該模型不僅提升了空域流量預測的準確性,也證實了小波變換在增強模型時間序列數據理解中的有效性,且為交通流量管理提供了一種新的技術方案。
關鍵詞: 空域流量預測; 增強位置編碼; Transformer模型; 小波變換; LSTM模型; 小波基函數
中圖分類號: TN919.3?34" " " " " " " " " " " " " "文獻標識碼: A" " " " " " " " " " " 文章編號: 1004?373X(2025)08?0127?06
Airspace traffic prediction based on wavelet transform enhanced
position coding Transformer
TANG Weizhen, LIU Bo, HUANG Zhousheng, TIAN Qiqi
(Civil Aviation Flight University of China, Deyang 618307, China)
Abstract: With the acceleration of globalization and the development of aviation technology, the accuracy requirements for air traffic flow prediction are also increasing. In order to improve the accuracy of air traffic flow prediction and reduce the pressure of air traffic controllers, a Transformer model with enhanced position coding is proposed. It uses wavelet transform to analyze the original airspace flow data, selects the wavelet basis function with the best performance by means of the signal?to?noise ratio, and further calculates the wavelet coefficient and integrates it into the position coding, so as to enhance the model's understanding ability of time series data. The experimental results show that the proposed model can accurately capture the non?stationary and abrupt characteristics of air traffic flow data. Its RMSE and MAPE evaluation indicators are decreased by 29.9 and 2.9% respectively compared with the original Transformer model, and by 34.5 and 3.4% compared with the long short?term memory (LSTM) model, respectively. This model not only improves the accuracy of airspace flow prediction, but also confirms the effectiveness of wavelet transform in enhancing the understanding of model time series data, and provides a new technical scheme for traffic flow management.
Keywords: airspace flow prediction; enhanced location encoding; Transformer model; wavelet transform; LSTM model; wavelet basis function
0" 引" 言
隨著全球化進程的加速和航空技術的不斷進步,航空運輸需求持續增長,空中交通流量管理與預測的重要性日益凸顯。有效的流量預測在保障空中交通安全、提高航班運營效率、優化航空管理策略等方面起著至關重要的作用。使用傳統預測技術,如差分整合移動平均自回歸模型[1]、季節性指數平滑法[2]以及基于支持向量機[3]的機器學習方法可對空中交通流量進行預測,這些方法通過分析歷史數據中時間序列的特性,或通過建立原始數據與目標變量之間的數學關系來進行預測?!?br>