面向新型电力系统的电力大数据副本管理算法

丁斌; 袁博; 郑焕坤; 邢志坤; 王帆

丁斌,袁博,郑焕坤,邢志坤,王帆.面向新型电力系统的电力大数据副本管理算法[J].电测与仪表,2022,59(1):10-17.
Ding Bin,Yuan Bo,Zheng Huankun,Xing Zhikun,Wang Fan.Research on replica management algorithm of power big data orienting novel power system[J].Electrical Measurement & Instrumentation,2022,59(1):10-17.

面向新型电力系统的电力大数据副本管理算法

Research on replica management algorithm of power big data orienting novel power system

DOI：10.19753/j.issn1001-1390.2022.01.002

中文关键词: 电力大数据副本管理；低时延；流量预测

英文关键词:power big data, replica management, low latency, traffic forecast

基金项目:国家自然科学基金资助项目(61501185);国家电网面向多维信息系统智能运维的关键技术研究(SGTYHT/19-JS-215),中央高校基本科研业务费专项资金资助项目(2014MS87)

作者中文名	作者英文名	单位
丁斌	Ding Bin	国网河北省电力有限公司雄安新区供电公司，河北保定 071700
袁博	Yuan Bo	国网河北省电力有限公司雄安新区供电公司，河北保定 071700
郑焕坤	Zheng Huankun	华北电力大学，河北保定 07100
邢志坤	Xing Zhikun	国网河北省电力有限公司雄安新区供电公司，河北保定 071700
王帆	Wang Fan	国网河北省电力有限公司雄安新区供电公司，河北保定 071700

摘要点击次数: 1926

中文摘要:

随着新型电力系统建设步伐的加快，电网信息化程度不断加深，上万台虚拟机构成的庞大数据中心，仅仅依靠传统任务调度型副本管理策略，已无法满足未来精准负荷控制等新型电力业务对大数据处理时延的需求。对此，文章在充分考虑网络流量和数据中心位置分布的基础上，构建一种基于随机配置网络（SCN）的电力大数据自适应副本管理系统。同时，提出了一种基于C-means聚类的底层设备分类网络流量预测模型，有效地完成数据库网络资源的实时评估。为了有效提升电力大数据副本管理效率降低数据处理延时，提出了一种面向新型电力系统的数据存储和选择的副本管理算法，实现电力大数据副本的灵活存储和最优选择。最后，在相应省公司开展试点验证，实验结果表明该算法能够有效支撑电力大数据存储，降低数据处理延时，跨分布式DCs的现场作业完成时间平均减少12.19%。

英文摘要:

With the acceleration of the construction of new power systems, the degree of informatization of power grids has continued to deepen. Only relying on traditional task scheduling-type replica management strategies, the huge data center composed of tens of thousands of virtual machines can no longer meet the needs of big data processing latency of new power business such as precise load control in the future. In this regard, based on the full consideration of network traffic and data center location distribution, this paper constructs a stochastic configuration network (SCN)-based adaptive replica management system of power big data. Meanwhile, a C-means clustering-based classification of underlying equipment and network traffic prediction is proposed to complete the real-time evaluation of database network resources. On this basis, in order to effectively improve the efficiency of power big data replica management and reduce data processing latency, a novel power system-oriented data storage and selection replica management algorithm is proposed to realize the flexible storage and optimal selection of power big data copies. Finally, a pilot test was carried out in the corresponding provincial companies. The algorithm can effectively handle the storage of power big data, reduce the data processing latency, and reduce the completion time of field operations across distributed DCs by an average of 12.19%.

查看全文查看/发表评论下载PDF阅读器

关闭