【作者】 张婷；

【导师】 张民；

【作者基本信息】 北京邮电大学 ， 电子科学与技术， 2022， 硕士

【摘要】 现今,随着信息量飞速增长,互联网、大数据的卓越发展,人们对于通信质量与通信容量的高需求迫在眉睫。卫星通信由于覆盖范围广、受地形影响小、波束覆盖范围大等优点进入人们的视野。而性能监测技术是地球同步轨道卫星通信系统安全可靠运行的重要基础保障。为了保障系统的高效运转,如何有效地对复杂多变的地球同步轨道卫星通信系统进行监测、对系统风险进行预判、对系统故障进行诊断定位成为急需迫切解决的关键问题。性能监测可以防范星地之间通信系统故障,加快故障定位与系统恢复,避免造成重大损失。但是传统的性能监测借助频谱检测仪,依赖大量人工经验进行监测,难以实现系统内物理损伤等信息获取,而且人力成本高,效率低下,容易造成不可避免的人为误差。最近,深度学习技术作为人工智能领域的核心技术,已在多种场景的性能监测方面被广泛使用,但是在卫星通信中应用较少。基于深度学习强大的自适应学习能力,本论文围绕地球同步轨道卫星通信系统的性能监测技术开展研究,主要研究内容如下:第一,针对地球同步轨道卫星通信系统信号的传输过程,搭建了Ka频段地球同步轨道卫星通信仿真系统,该系统可以模拟真实大气信道晴天、乌云、冰霜、小雨、大雨、暴雨六种不同的雨衰。仿真实现十种不同调制格式下,系统非线性损伤、不同天气状况、I\Q不平衡以及接收端热噪声这4种常见卫星系统的损伤。第二,针对传统性能监测依赖人工经验且准确性不足的问题,提出了利用卷积神经网络(CNNs)进行智能损伤监测的方案。在地球同步轨道卫星通信系统传输的正交相移键控(Quadrature Phase Shift Keying,QPSK)信号的星座图损伤监测中,对4种不同损伤类型下多种损伤程度监测准确率为99.14%。第三,针对调制格式识别传统算法无法实现特征自动提取的问题。本论文提出了基于CNNs的调制格式监测方案。在对地球同步轨道卫星通信中的十种常见信号的星座图调制格式监测中,在信噪比(Signal Noise Ratio,SNR)为8dB的情况下实现了对其调制格式100%正确的识别,即使在SNR低至-4dB的情况下,也能实现对其调制格式90.5%正确的识别。从而验证了借助CNNs能够快速高效的实现智能调制格式监测。更多还原

【Abstract】 Nowadays,with the rapid growth of the amount of information and the excellent development of the Internet and big data,people’s high demand for communication quality and communication capacity is imminent.Satellite communication has entered people’s field of vision due to its advantages of wide coverage,little influence by terrain,and large beam coverage.The performance monitoring technology is an important basic guarantee for the safe and reliable operation of the geosynchronous orbit satellite communication system.In order to ensure the efficient operation of the system,how to effectively monitor the complex and changeable geosynchronous orbit satellite communication system,predict system risks,and diagnose and locate system faults have become key issues that need to be urgently solved.Performance monitoring can prevent communication system failures between satellite and ground,speed up fault location and system recovery,and avoid major losses.However,traditional performance monitoring relies on a large amount of manual experience for monitoring with the help of spectrum detectors.It is difficult to obtain information such as physical damage in the system,and the labor cost is high,the efficiency is low,and it is easy to cause inevitable human errors.Recently,deep learning,one of the core technologies in the field of artificial intelligence,has been widely used in performance monitoring in various scenarios,but it is rarely studied in satellite communications.Based on the powerful adaptive learning ability of deep learning,this paper focuses on the performance monitoring technology of geosynchronous orbit satellite communication system.The main research contents are as follows:Firstly,for the signal transmission process of the geosynchronous orbit satellite communication system,a Ka-band geosynchronous orbit satellite communication simulation system is built,which can simulate six different types of rain in the real atmospheric channel:sunny,dark clouds,frost,light rain,heavy rain,and heavy rain.decline.Four common satellite system damages,including system nonlinear damage,different weather conditions,I\Q imbalance and receiver thermal noise,are simulated under ten different modulation formats.Secondly,to address the problem that traditional performance monitoring relies on human experience and lacks accuracy,a scheme for intelligent damage monitoring using convolutional neural networks(CNNs)is proposed.In the constellation damage monitoring of Quadrature Phase Shift Keying(QPSK)signals transmitted by the geosynchronous orbit satellite communication system,the monitoring accuracy of multiple damage degrees under four different damage types is 99.14%.Thirdly,the traditional algorithm for modulation format recognition cannot realize the problem of automatic feature extraction.This paper proposes a modulation format monitoring scheme based on CNNs.In the monitoring of the constellation pattern modulation format of ten common signals in the geostationary orbit satellite communication,100%correct identification of the modulation format is achieved when the Signal Noise Ratio(SNR)is 8dB.In the case of SNR as low as-4dB,it can also achieve 90.5%correct identification of its modulation format modulation.Thus,it is verified that intelligent modulation format monitoring can be realized quickly and efficiently with the help of CNNs.更多还原