【摘要】 随着人工智能与数字经济的深度融合，传统算网相对独立的架构难以满足计算任务对高性能、实时性及跨域资源共享的需求。将算力网（CPN）定义为以计算为核心、网络为基础、智能为引擎的新型基础设施，系统探讨了其关键技术创新与发展实践。详细阐述了算力路由、高通量数据网、分布式智算组网、智融标识网络、星织网络架构以及算力互联测量感知等六大核心技术体系，并通过现网试点与规模验证，验证了这些技术在提升网络吞吐率、降低端到端时延及优化异构资源调度方面的显著成效。最后，围绕高效基础设施建设、跨域跨平台调度、智能化管理、多样化场景适配以及隐私安全与绿色节能等5个维度，提出了CPN后续研究的重点方向与建议。更多还原

【Abstract】 With the deep integration of artificial intelligence and the digital economy, the traditional architecture—where computing and networks operate relatively independently—struggles to meet the demands of computing tasks for high performance, real-time response, and cross-domain resource sharing. This paper defines the computing power network(CPN) as a new type of infrastructure that is computingcentric, network-based, and intelligence-driven, and systematically explores its key technological innovations and development practices. This paper elaborates on six core technology systems: computing power routing, high-throughput data network, distributed intelligent computing networking, intelligence-converged identification network, star-fabric network architecture, and computing power interconnection measurement and awareness. Through live network pilots and large-scale verifications, the study demonstrates the significant effectiveness of these technologies in improving network throughput, reducing end-to-end latency, and optimizing heterogeneous resource scheduling. Finally, this paper proposes key directions and suggestions for future research on the CPN, focusing on five dimensions: efficient infrastructure construction, cross-domain and cross-platform scheduling, intelligent management, diverse scenario adaptation, and privacy security combined with green energy conservation.更多还原