陆怡菲,顾君佳,沈昊骢,吴媖,潘俊姚,邢旭亮
LU Yifei,GU Junjia,SHEN Haocong,WU Ying,PAN Junyao,XING Xuliang

• 1:国网浙江省电力有限公司嘉兴供电公司

摘要(Abstract)：

目前电力通信光缆光路迂回选路的传统方法是人工决策，效率低下，导致电力通信业务恢复较慢。为此，提出了一种基于KSP(K条最短路径)与Dueling DQN（竞争深度Q学习网络）的电力通信光缆光路智能迂回方法。使用电力通信光缆网络拓扑信息以及光缆类型、同沟道情况、光缆长度、光缆芯数等数据，通过KSP算法寻找K条最短迂回路径，然后采用Dueling DQN评估每条迂回路径的风险值并实现路由选择。将所提方法与传统KSP算法、Nature DQN算法的效果进行对比，该算法光缆光路智能迂回准确率达到99.5%，决策时间缩减至秒级。
Currently, the traditional method for optical cable path routing is manually decided and inefficient, leading to the slow recovery of the electric power communication business. Therefore, an intelligent routing of optical cable path based on the combination of KSP(K-shortest paths) and Dueling DQN(deep Q-networks) is proposed.The network topology of the power communication optical cable and data such as the optical cable, the information of optical cables in the same channel, optical cable length, and the number of optical cable cores are used to search the K-shortest path routing under the KSP algorithm. Afterward, the Dueling DQN is used to evaluate the risk value of each routing path and complete the routing selection. The effect of the method above is compared with those of the traditional KSP algorithm and Nature DQN algorithm. The results show that the intelligent routing accuracy of the optical cable path is up to 99.5%, and the decision time is reduced to seconds.

关键词(KeyWords)： 电力通信光缆;光路迂回;Dueling DQN;最短路径
power communication optical cable;optical path routing;Dueling DQN;shortest paths

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司省管产业单位科技项目（2021-KJLH-HC-007）

作者(Author): 陆怡菲,顾君佳,沈昊骢,吴媖,潘俊姚,邢旭亮
LU Yifei,GU Junjia,SHEN Haocong,WU Ying,PAN Junyao,XING Xuliang

DOI: 10.19585/j.zjdl.202212014

参考文献(References)：

• [1]卓华硕.电力通信光传输网的现状与优化[J].贵州电力技术，2014,17(3):61-63.
• [2]金炜.OTN技术在电力通信网中的应用研究[D].北京：华北电力大学，2014.
• [3]赵灿.OTN技术在电力通信传输网中的应用分析[J].河北电力技术，2017,36(3):17-19.
• [4]高会生，王慧芳.基于安全性的继电保护光纤迂回通道路径选择[J].电力系统保护与控制，2014,42(14):25-31.
• [5]林庆达，徐曦，卢琦淇，等.改进Dijktra算法在电力通信光缆恢复业务中的应用[J].广西电力，2016,39(6):53-56.
• [6]国网辽宁省电力有限公司信息通信分公司.一种电力信息通信网络最优迂回路径选择方法：CN108494608A[P].2018-09-04.
• [7]刘全，翟建伟，章宗长，等.深度强化学习综述[J].计算机学报，2018,41(1):1-27.
• [8]魏英姿，赵明扬.一种基于强化学习的作业车间动态调度方法[J].自动化学报，2005,31(5):765-771.
• [9] IPEK E,MUTLU O,MARTFNEZ J F,et al. SelfOptimizing Memory Controllers:A Reinforcement Learning Approach[C]//International Symposium on Computer Architecture.ACM,2008:39-50.
• [10] MNIH V,KAVUKCUOGLU K,SILVER D,et al.Playing Atari with Deep Reinforcement Learning[J]. Computer Science,2013.
• [11] VOLODYMYR M,KORAY K,DAVID S,et al.Humanlevel control through deep reinforcement learning[J].Nature,2015,518(7540):529-533.
• [12] SILVER D,HUANG A,MADDISON C J,et al.Mastering the game of Go with deep neural networks and tree search[J].Nature,2016,529(7587):484-489.
• [13] LILLICRAP T P,HUNT J J,PRITZEL A,et al.Continuous control with deep reinforcement learning[EB/OL]. 2015:arxiv. 1509.02971. https：//arxiv. org/abs/1509.02971.
• [14] HANSEN S.Using deep Q-Learning to control optimization hyperparameters[EB/OL]. 2016:arxiv. 1602.04062.https：//arxiv.org/abs/1602.04062.
• [15]杨明宇.ADSS光缆与OPGW光缆的性能较比及选型[J].通讯世界，2014(20):19-20.
• [16]刁少岚.电力光纤通信线路的安全评估[D].保定：华北电力大学，2007.
• [17] DIJKSTRA E W. A note on two problems in connexion with graphs[J]. Numerische Mathematik,1959,1(1):269-271.
• [18] EPPSTEIN D.Finding the k shortest paths[C]//Symposium on Foundations of Computer Science. IEEE,1994:154-165.
• [19] WANG Z,FREITAS N D,LANCTOT M.Dueling network architectures for deep reinforcement learning[EB/OL]. 2015:arxiv. 1511.06581. https：//arxiv. org/abs/1511.06581.
• [20]冯平，袁亮，钟于，等.基于新型站内引入光缆方案的研究与分析[J].四川电力技术，2020,43(5):62-64.
• [21]卢超，付连宇，李佳泽，等.支撑泛在物联的电力无线专网性能评估[J].内蒙古电力技术，2020,38(1):45-49.
• [22]郑惠文.光传送网在电力系统通信中的应用研究[J].山西电力，2021(4):70-72.
• [23]赵洋，赵晓红，任天成，等.5G通信在电力系统中的应用[J].山东电力技术，2020,47(10):6-10.
• [24]赵晓坤，高洁，郭冰洁，等.基于IoT-G 230 MHz电力无线通信关键技术分析[J].四川电力技术，2021,44(6):58-62.
• [25]熊菲，金翠，陆申，等.内蒙古电网电力无线专网建设方案[J].内蒙古电力技术，2020,38(1):35-39.
• [26]陈兆雁，程紫运，徐慧慧，等.泛在电力物联网通信需求及技术适用性分析[J].内蒙古电力技术，2020,38(1):1-5.