【作者】 王建斌；

【导师】 瞿逢重； 李虓江；

【作者基本信息】 浙江大学 ， 电子与信息（专业学位）， 2024， 博士

【摘要】 随着国家“21世纪海上丝绸之路”战略的提出,发展海洋经济对于我国而言日益重要。海洋经济的数字化转型以及海洋各类传感器呈指数级的增长对中远海的网络性能提出了更高的要求。目前卫星技术服务单一且资费昂贵,第四代移动通信网络(4th generation mobile networks,4G)已无法满足大连接、高速率、低时延等业务需求。第五代移动通信网络(5th generation mobile networks,5G)在其全生命周期的规划、建设、优化和维护等阶段也都未针对海域规模商用进行系统性的适配和优化。在实际工程中,由于海域场景复杂,海上测试难度大、风险高、成本投入大,5G蜂窝网络在海域的规划、建设、优化和维护的研究相对有限,且主要侧重于基础理论问题的阐述和计算,对于研究方案的测试和验证更是缺乏。5G网络在其规划、建设、优化或维护的生命周期中的任何阶段出现问题,都将无法实现网络的高质量规模化商用。本文针对5G网络在海域超远及深度覆盖场景规模商用所面临的工程和技术难点,进行了系统性的研究。本文中所述的蜂窝网络的超远覆盖和深度覆盖分别是指距离基站15 km至100 km的海域范围或船舱等室内环境需要满足业务需求的有效覆盖,包括覆盖距离、容量和质量等各方面因素。本文基于3GPP标准和已有的成果,针对5G移动网络在海域场景规模化商用过程中遇到的具体痛点和难点问题,进行了系统性的研究分析。首先,在规划阶段,本文深入探讨了海域的信号传播模型和船舱损耗模型,并建立了场景化参数规划方案。网络规划作为无线网络的基础,是保障网络性能的前提。该规划主要包括工程规划和参数规划。虽然陆地场景下的5G网络规划已相对成熟,但海域场景的相关研究仍处于起步阶段,传统的规划方法难以有效满足海域业务对5G网络的需求。针对此问题,本文完成了5G海域场景超远及深度覆盖的工程规划、参数规划和损耗模型的建立。在3GPP规范中提及模型的基础上,本文首次建立了海域3.5 GHz传播模型,并通过仿真与实测对模型进行了校正。此外,针对参数规划,本文首次开展了基于海域特性的物理小区标识(Physical Cell Identifier,PCI)、物理随机接入信道(Physical Random Access Channel,PRACH)和波束等场景化参数的研究,并搭建了5G异帧组网的智能化系统,以最大限度地发挥网络性能。根据船舶中船舱的材料特性,本文校正了穿透损耗模型,并通过现场测试验证了不同距离、不同频段和不同材料的穿透损耗数值。其次,在建设阶段,本研究严格按照规划完成了5G网络的高质量建设,重点解决了以下问题:终端最大发射功率受限导致海域超远场景上行覆盖能力不足,以及大型船舶船舱的厚重钢板和复杂结构严重影响通信性能和用户感知的问题。本文首次提出了一个综合解决方案,包括设计和改造客户前置设备(Customer Premise Equipment,CPE)、场景化适配2.1 GHz+3.5 GHz超级上行聚合技术、研究和开发了一种多维度干扰消除技术,以优化和提升上行覆盖能力。此外,通过结合5G用户平面功能(User Plane Function,UPF)专用网切片技术,本研究创新性地实现了4G S1(4G基站和4G核心网的链路接口)链路通过5G网络的稳定回传。该方案使得在船上低成本开通基站成为可能,从而确保船舱内部能够实现良好的无线信号深度覆盖。第三,在网络优化阶段,本文重点探讨了机器学习算法在海域通信领域的应用可行性。由于海域站点环境复杂、交通不便以及当前的测量报告(Measurement Report,MR)等大数据分析方法在该场景下的局限性,现场测试和优化工作面临诸多困难。为解决这些问题,本文采用了随机森林(Random Forest,RF)、反向传播神经网络(Back Propagation Neural Network,BPNN)和时间序列等机器学习算法,并结合用户的4/5G网络行为进行综合分析。本文首次构建了一个自动化识别系统,能够提前并有效地识别海域问题栅格、贬损用户和质差小区。此外,本文还将该系统与支持向量机(Support Vector Machine,SVM)、决策树(Decision Tree,DT)等其他算法进行了预测准确性对比,验证了所构建模型的预测性能优势。第四,在网络维护阶段,本文重点研究了站点节能技术。由于海域站点动力环境受限制,采用站点节能技术以降低能耗及延长续航能力至关重要。本文首次采用了融合长短期记忆的深度卷积神经网络(Deep Convolutional Neural Network-Long Short Term Memory,DCNN-LSTM)算法模型来预测站点负荷,并对整个5G接入网络的能耗进行动态调控。此外,本文还首次利用基站射频头端模块的温湿度数据和硬件使用寿命设计了一种自主的节能感知控制系统。通过软硬件的结合的方式,实现了海域站点的高效、稳定且可靠的节能,显著提高了在相同能源配套下的站点的续航能力。最后,本文通过仿真和海上实地测试等手段对上述研究内容进行了验证。由于海上风浪对测试环境和测试人员的身体素质要求较高,同时海测成本也较高,这导致了海域场景中超远和深度覆盖参数的验证与测试数据的缺乏。本文首先通过仿真手段验证了所提方案的可行性,并通过采用创新的多测试用例并行测试方法进行海上实地测试,获取了大量的测试数据,进一步证实了本文所提供方案的有效性和实际应用价值。这些验证和分析不仅提升了本文研究的创新性和实用性,同时也为未来相关领域的研究提供了有价值的参考。本文通过上述研究,有效解决了以下一系列实际问题:规划效率和准确性低、海域通信超远覆盖实现困难、船舱内部移动网络深度覆盖质量不佳、海域问题区域与问题用户识别困难、海域站点动力环境受限等。本研究成果可为海域通信和海洋经济的发展提供有力支持。更多还原

【Abstract】 With the proposal of the national “21 Century Maritime Silk Road”strategy,developing the marine economy has become increasingly important for our country.Recently,as the digital and intelligent transformation of the marine economy progresses,the use of sensors is increasing exponentially.This poses higher requirements for the networks in deep sea areas.Currently,satellite technology offers limited and expensive services.The 4th generation mobile networks(4G)cannot meet service requirements such as a large number of connections,high rates,and low latency.Moreover,the 5th generation mobile networks(5G)have not been systematically applied and optimized for large-scale commercial use in sea areas across all phases,including planning,construction,optimization,and maintenance.Since the complexity of sea scenarios has posed great difficulties,more risks,and high costs to the offshore testing,current research mainly focuses on the elaboration and calculation of basic theoretical problems,not to mention limited research focusing on the planning,construction,optimization,and maintenance of cellular networks in sea areas.There is also a lack of research on testing and verifying proposed solutions.However,any problems encountered during the planning,construction,optimization,or maintenance will prevent achieving highquality large-scale commercial use of the network.This paper presents a systematic study on the engineering and technical challenges associated with achieving large-scale commercial use of networks.In this paper,ultra-long-distance and deep coverage of cellular networks refer to the effective coverage of sea areas within a distance range of 15 km to 100 km from base station to indoor environments such as ship cabins that need to meet business requirements,including coverage distance,capacity,and quality.Based on 3GPP and existing achievements,this paper systematically studies the pain points and difficulties encountered in the process of large-scale commercial use of specific sea scenarios.Firstly,in the network planning stage,this paper studies the signal propagation model and ship cabin loss models in marine environments,and establishes a scenario-based parameter planning scheme.Network planning,which includes engineering and parameter planning,is fundamental to wireless networks and crucial for ensuring network performance.While 5G network planning in land-based scenarios is relatively mature,5G network planning in sea areas remains in the exploratory stage.Additionally,traditional methods cannot effectively meet the requirements of sea area businesses for 5G networks.To address this issue,this paper completes engineering planning,parameter planning,and loss model establishment for 5G communications in sea areas.This paper establishes a novel 3.5 GHz propagation model in marine environments,refined through simulation and field-testing.In addition,this paper studies,for the first time,scenario-based parameter planning such as physical cell identifiers(PCI),physical random access channels(PRACH),and beamforming based on marine characteristics,and develops an intelligent system for 5G heterogeneous frame network architecture to improve network performance.Furthermore,a penetration loss model is established based on the material properties of ship cabins on board,and the penetration losses of different distances,frequency bands,and material types are verified through field tests.Secondly,during the network construction phase,this paper addresses the limited uplink coverage of terminals in offshore ultra-long environments due to restricted transmission power.It also considers the significant impact of thick steel plates and complex cabin structures of large ships on communication performance and user experience.This paper proposes a comprehensive solution.It involves the use of modified customer premise equipment(CPE),2.1GHz + 3.5 GHz super uplink aggregation technology,and multi-dimensional interference cancellation to optimize and enhance uplink coverage capabilities.Furthermore,by integrating 5G user plane function(UPF)and dedicated network slicing technology,stable backhaul through5 G networks for 4G S1 links(the interface between 4G base stations and 4G core networks)is achieved.This approach not only enables the cost-effective establishment of base stations on board but also ensures extensive coverage within ship cabins.Thirdly,during the network optimization phase,this paper examines the feasibility of implementing machine learning algorithms within the domain of marine communications.The intricate on-site environments and transportation challenges inherent to maritime settings make on-site optimization and testing particularly daunting tasks.Simultaneously,due to the unique characteristics of these maritime environments,conventional big data analysis methods such as conventional measurement reports(MR)are not fully applicable,further complicating network optimization efforts.Addressing the limitations of conducting tests in maritime settings and the difficulties encountered in complaint management,this paper employs several machine learning algorithms,including random forests(RF),back propagation neural network(BPNN),and time series analysis,to establish an automated identification system.This is the first instance of such a system being used to pinpoint problematic grids,degraded users,and substandard cells in maritime zones.It facilitates the early detection of trouble spots,low-quality cells,and affected users in these areas.Furthermore,when compared with other algorithms such as support vector machines(SVM),decision trees(DT),and so on,our model has demonstrated superiority in terms of prediction accuracy.Fourthly,in the network maintenance stage,since the endurance of power supply and backup power on-site is particularly important,this paper focuses on site energy-saving technology.A deep convolutional neural network-long short term memory(DCNN-LSTM)deep learning algorithm model is used to predict the future load of small cells and to manage the energy consumption of the entire 5G access network dynamically.In addition,an independent energy-saving control system is designed based on temperature and humidity data obtained from the base station radio frequency head module and hardware service life.Such system facilitates the achievement of efficient,stable,and reliable energy saving for sea area sites.Through the combination of software and hardware,the endurance time of sites is significantly increased with the same energy support.Finally,this research is verified through simulations and on-site testing at sea.The transmission model tailored to sea areas,particularly for ultra-long-distance coverage,differs from that on land.Extensive experimental testing is required for the verification and optimization of parameters and technical solutions.Additionally,due to the stringent testing environment requirements,and the high cost of offshore testing,currently there is a lack of validation and testing data for ultra-long and deep coverage parameters in sea areas.This paper,therefore,analyzes the transmission and coverage performance of the aforementioned technical issues through simulations to verify the feasibility of the proposed solutions.Furthermore,through on-site testing at sea,a large amount of test data is gathered to further assess the effectiveness and practical application value of the proposed solutions.In summary,this research effectively solves the problems of low planning efficiency and low accuracy,provides solutions to achieving ultra-far coverage for marine communications and to making deep coverage within ship cabins possible,and enables identifying problematic areas and users within the sea area.Furthermore,it also addresses the limited power environments of sites within the sea area.These contributions are expected to contribute the development of marine communications and marine economy.更多还原