多基准时钟同步技术研究

The accuracy of the current clock reference source is difficult to meet the ultra-high precision of time synchronization demands in future communications, such as 5G, internet of things, quantum communications. Under the existing problems such as limited accuracy of multi-references clock sources, synchronous operations of multiple references, difficulty to realize networking in different places, the subject of the reference clock synchronization technology mainly contains two aspects: First, putting forward the proposal of synthesis with weighting based on machine learning method, multi-reference synchronization error depth control, and ultimately improving one level of precision time reference clock; After solving the local synthesis of different sources , the second is synthesis system of different nodes in network combined with the idea of introducing a full synchronization network, with machine learning method, putting forward in-depth study of remote multi-node multi-reference clock synthesis model, fully synchronized network strategy as well as off-site transfer and synchronization errors, ultimately making a proposal of full synchronization implementation under multi-reference clocks with ultra-high precision. The research results provide a novel method for ultra-high precision multi reference clock synchronization technology, putting forwad the necessary theoretical basis for the future construction of fully synchronous timing network, a strong improving to innovation and steady of multi-reference clock technology and full synchronization network applications evolution.

现有时钟基准源的精度难以满足5G、物联网、量子通信等未来通信领域的超高精度时间同步需求，主要存在现有单时钟源精度受限、多基准时钟源准同步运行、异地时钟源难以组网等技术难题。针对如何解决上述问题，本课题围绕多基准时钟时间精度提升和全同步组网两个方面开展研究：一是基于机器学习提出提升多基准时钟时间精度的加权合成方法，对多基准源同步误差进行深度控制，拟提高本地多基准时钟组的时间精度一个量级；二是在本地多基准加权合成系统基础上结合全同步网的特性，深入研究异地多节点多基准时钟合成模型、全同步组网策略以及异地同步传送的误差机制等关键技术，提出超高精度多基准时钟全同步组网方案。本项目研究工作及成果为我国地面提供超高精度授时信号、建设全同步授时网络提供关键基础，有力推动我国多基准时钟同步技术和全同步网应用的创新及平稳演进。

现有时钟基准源的精度难以满足5G、物联网、量子通信等未来通信领域的超高精度时间同步需求，主要存在现有单时钟源精度受限、多基准时钟源准同步运行、异地时钟源难以组网等技术难题。基于此问题，本项目开展了两方面研究，一是在本地多基准时钟同步方面，提出基于神经网络的多基准原子钟加权合成模型，仿真性能优于高品质单原子钟指标，理论上可达到ITU-T G.811.1标准要求；二是在异地多基准同步研究方面，结合所提出的组网架构，探讨了异地同步比对关键技术，修正了多基准加权合成模型，并搭建了异地多基准同步误差回归仿真平台与实验验证系统，在频率准确度和稳定度方面优于单铯钟性能指标，在时间同步精度方面，性能优于ITU-T G.8272.1 ePRTC超高精度时间服务器标准要求，理论上可满足5G超高精度时间服务器部署应用要求。本项目研究成果验证了在单原子钟物理性能受限的约束下，基于时空大数据+机器学习进行同步特征建模的方法，可优于高品质单原子钟的性能指标，提供了超高精度同步基准源技术实现的新路径，一定程度上可摆脱国家对超高精度原子钟设备进口的依赖，对于国家战略基础设施的自主建设具有一定的战略意义，对于国内通信领域同步技术升级演进也具有一定的应用推广意义。