数字孪生辅助的智慧园区物联终端接入调度方法  

Digital Twin Assisted IoT Device Access Scheduling Method for Smart Park

作　　者：贾泽晗 于子淇 石珵 廖海君 周振宇[1] 孙健[1] 白晖峰 张颉 JIA Zehan;YU Ziqi;SHI Cheng;LIAO Haijun;ZHOU Zhenyu;SUN Jian;BAI Huifeng;ZHANG Jie(State Key Laboratory of Alternate Electrical Power System With Renewable Energy Sources,North China Electric Power University,Beijing 102206,China;Beijing Smart-Chip Microelectronics Technology Co.,Ltd.,Beijing 100192,China;Power Science Research Institute of State Grid,Sichuan Electric Power Company,Chengdu 610041,China)

机构地区：[1]华北电力大学新能源电力系统国家重点实验室,北京102206 [2]北京智芯微电子科技有限公司,北京100192 [3]国网四川省电力公司电力科学研究院,四川成都610041

出　　处：《河北电力技术》2023年第2期12-19,共8页Hebei Electric Power

基　　金：国家重点研发计划资助项目(2020YFB0905900);国家自然科学基金项目(61971189);国家电网有限公司科技项目(52094021N010);国家电网有限公司科技项目(5400-202199541A-0-5-ZN)。

摘　　要：数字孪生通过辅助智慧园区构建能源管理模型实现园区智能能源管理。本文面向数字孪生辅助的智慧园区能源管理场景,基于联邦学习的智慧园区数字孪生框架,构建了园区能源管理模型,在此基础上,提出了一种基于数字孪生和深度Q网络的终端接入调度算法。首先,构建了以能源管理模型损失函数最小化为优化目标,长期接入优先级为约束条件的接入调度优化数学模型;其次,基于虚拟队列将长期接入优先级约束和短期接入调度优化解耦,并结合数字孪生和深度Q网络实现优化问题求解;最后,通过仿真分析验证了所提算法的有效性。Smart park can realize intelligent energy management by constructing an energy management model with the assistance of digital twin(DT).In this paper,a federated learning-based DT framework is proposed to construct an energy management model for smart park.On this basis,a DT and deep Q network(DQN)-based device access scheduling algorithm is proposed.Firstly,an access scheduling optimization mathematical model is constructed with the minimization of the loss function of the energy management model as the optimization objective and the long-term access priority as the constraint condition.Secondly,the long-term access priority constraint and the short-term access scheduling optimization are decoupled based on virtual queue,then DT and DQN are combined to solve the optimization problem.Finally,the effectiveness of the proposed algorithm is verified by simulation analysis.

关 键 词：数字孪生 智慧园区 联邦学习 能源管理模型 接入调度 

分 类 号：TN915.853[电子电信—通信与信息系统] TM734[电子电信—信息与通信工程]