Emerging power services with massive connections and real-time control have different requirements on delay, rate, reliability, etc. In order to meet the diversified requirements of power services, network slicing can provide guarantee for the wide application of 5G in power. The current research on network slicing adapting power services is still in its infancy. This paper proposes a trade-off analysis method for radio access network (RAN) slice delay and throughput based on dynamic intelligent resource allocation to meet the different QoS requirements of different services in smart grid scenarios. Firstly, clustering analysis is used to classify the communication requirements of smart grid services. A throughput demand model for enhanced mobile broadband (eMBB) slice users based on importance and a delay constraint model for ultra-reliable and low latency communication (uRLLC) slice users are constructed, and a throughput model for slice users based on cell capacity limitation is constructed. Then, the problem model of maximizing the throughput of the eMBB slice by reasonably allocating resource blocks and power under the premise of ensuring the delay of the uRLLC slice is established. The Lyapunov drift plus penalty function method is used to convert the above problem into a tractable form, and then the problem is solved by the active set method. Simulation results show that the different QoS requirements of smart grid services will significantly affect the actual queue length of the uRLLC slice as well as the throughput of the eMBB slice.

新兴的海量连接、实时控制的电力业务在时延、速率、可靠性等方面有着不同的要求，为了满足电力业务多样化的需求，网络切片可以为5G在电力领域的广泛应用提供保障。目前网络切片适配电力服务的研究还处于起步阶段。本文提出了一种基于动态智能资源分配的无线接入网(RAN)切片延迟和吞吐量的权衡分析方法，以满足智能电网场景下不同业务的不同QoS需求。首先，利用聚类分析对智能电网服务的通信需求进行分类。构建了基于重要性的增强型移动宽带（eMBB）切片用户吞吐量需求模型和超可靠低时延通信（uRLLC）切片用户的延迟约束模型，基于小区容量限制的切片用户吞吐量模型为建。然后，建立了在保证uRLLC切片时延的前提下，通过合理分配资源块和功率来最大化eMBB切片吞吐量的问题模型。用李雅普诺夫漂移加罚函数法将上述问题转化为易处理形式，再用主动集法求解。仿真结果表明，智能电网服务的不同QoS要求会显着影响uRLLC切片的实际队列长度以及eMBB切片的吞吐量。