人工智能在飞机颠簸预报中的应用进展及未来趋势展望  

Progress and future trend of artificial intelligence application in aircraft turbulence prediction

作　　者：朱玉祥[1] 刘海文[2] 万文龙 姜晓飞 尕藏程林 ZHU Yuxiang;LIU Haiwen;WAN Wenlong;JIANG Xiaofei;GAZANG Chenglin(CMA Training Center,China Meteorological Administration,Beijing 100081,China;Civil Aviation University of China,Tianjin 300300,China;Dongying Meteorological Bureau,Dongying 257091,China)

机构地区：[1]中国气象局气象干部培训学院,北京100081 [2]中国民航大学,天津300300 [3]东营市气象局,山东东营257091

出　　处：《大气科学学报》2023年第6期825-836,共12页Transactions of Atmospheric Sciences

基　　金：国家自然科学基金委员会-中国民用航空局民航联合研究重点基金项目(U2033207);中国气象局气象干部培训学院重点项目(2022CMATCZD10);中国气象局创新发展专项(CXFZ2023J027)。

摘　　要：飞机颠簸对飞机运行安全及旅客舒适度具有很大危害,提高飞机颠簸的预报准确率对于减轻人员伤害和财产损失具有重要意义,且一直是航空气象研究的重点领域。随着观测手段的丰富和科学技术的进步,飞机颠簸预报也取得了长足发展。本文首先从定性预报和定量预报的角度对国内外飞机颠簸预报的主要方法,特别是人工智能(AI)方法在其中的最新应用进行了综述。在此基础上,归纳总结出AI用于飞机湍流预报存在的主要问题和未来的重点研究方向:(1)飞机实况数据的开放共享以及多源湍流数据的融合构建问题;(2)基于人工智能构建飞机颠簸预报模型的可解释性及物理机制问题;(3)基于人工智能构建飞机颠簸的集合预报问题。最后,提出了气象领域未来第三代人工智能的发展思路。Aircraft turbulence poses great danger to safe aircraft operation and affects the comfort of passengers.Therefore,improving the accuracy of aircraft turbulence forecast is of great significance to reducing injury and property loss,and for many years has been the focus of aviation meteorological research.With the enrichment of observation means and progress of science and technology,great progress has been made in the prediction of aircraft turbulence.In this paper,the main methods of aircraft turbulence prediction used throughout the world,especially the latest application of artificial intelligence(AI),are reviewed from the perspectives of both qualitative and quantitative forecast.Then,on this basis,the main problems present in the application of AI to aircraft turbulence prediction and key research directions in the future are summarized as follows:(1)open sharing of aircraft real data,and integration and construction of multi-source turbulence data;(2)interpretability and physical mechanism of aircraft turbulence prediction model built based on AI;and(3)ensemble forecasting of aircraft turbulence based on AI.Finally,the development ideas of the third generation of AI in the field of meteorology are proposed.The future third generation of AI in meteorology may be based on deep learning,integrating mathematical logic reasoning algorithms,and based on“knowledge”(physical mechanisms,experience of forecasters,or basic principles of weather and climatology)and“data”(model data and observation data).That is,it may combine the advantages of the schools of connectionism and symbolism to carry out weather forecasting and climate prediction work.Of course,the third generation of meteorological AI must have characteristics such as higher prediction accuracy,self-learning,and self-adaptation(such as adapting to changes in the rule of meteorological data caused by climate change),and strong explanation in its mechanism.

关 键 词：航空气象 飞机颠簸 第三代人工智能 概率预报 机器学习 

分 类 号：V328[航空宇航科学与技术—人机与环境工程]