基于数字孪生的算力网络自优化技术研究

摘要：将数字孪生网络技术引入算力网络，可以建立算力网络的虚拟映射网络。数字孪生网络系统通过高保真的虚实实时交互，实现对算力网络的高效分析、诊断和控制。以算力网络的自优化为例，提出了一种数字孪生算力网络的结构自优化模型，实现了数字孪生算力网络中网络自学习、自验证、自演进的实时闭环控制。与传统自组织网络（SON）不同的是，将物理网络基础设施与SON模块分离，将SON自优化的过程迁移到虚拟网络中，降低了算力网络运维的复杂度，提高了网络的灵活性和适应性。仿真实验证明，引入数字孪生网络技术后，可以迅速地处理算力网络服务超时问题，降低网络整体服务时延。

关键词：数字孪生；算力网络；SON；遗传算法

Abstract: By introducing the digital twin network technology into the computing power network, the virtual mapping network of the computing power network can be established. Digital twin network system realizes efficient analysis, diagnosis, and control of computing power network through high-fidelity real-time interaction between virtual and real. Taking the self-optimization of computing power network as an example, a structural self-optimization model of digital twin computing power network is proposed, which realizes the real-time closed-loop control of network self-learning, self-verification, and self-evolution in digital twin computing power network. Different from the traditional self-organizing network (SON), the physical network infrastructure is separated from the SON module, and the SON self-optimization process is migrated to the virtual network, which reduces the complexity of computing power network operation and maintenance, and improves the flexibility and adaptability of the network. Simulation results show that the introduction of digital Siamese network technology can quickly deal with the problem of computing power network service timeout and reduce the overall service delay of the network.

Keywords: digital twin; computational network; SON; genetic algorithm