Deep reinforcement learning framework for power 5G slice resource allocation
电力5G切片资源分配的深度强化学习框架

摘要

电力5G切片资源分配需满足不同电力业务的差异化需求，而传统分配方法难以应对动态变化的网络环境与复杂业务场景。深度强化学习框架通过智能体与环境的持续交互，可实现资源分配策略的自主优化。该框架将电力业务需求、网络负载状态作为状态输入，以资源利用率最大化与业务QoS保障为目标，通过深度神经网络拟合价值函数，动态调整计算、存储、带宽等资源的分配方案。其具备较强的环境适应性与决策时效性，能有效解决切片资源竞争、负载不均衡等问题，为电力5G网络的高效运行提供技术支撑，对推动电力系统数字化转型具有重要意义。

Abstract

The resource allocation of power 5G slices needs to meet the differentiated requirements of various power services, while traditional allocation methods struggle to cope with dynamically changing network environments and complex service scenarios. The deep reinforcement learning framework can realize the autonomous optimization of resource allocation strategies through continuous interaction between agents and the environment. This framework takes power service requirements and network load status as state inputs, with the goals of maximizing resource utilization and ensuring service QoS. It dynamically adjusts the allocation schemes of computing, storage, bandwidth and other resources by fitting the value function through deep neural networks. It has strong environmental adaptability and decision-making timeliness, which can effectively solve problems such as slice resource competition and load imbalance. It provides technical support for the efficient operation of power 5G networks and is of great significance for promoting the digital transformation of power systems.