智能电网监测数据的云存储研究

发布时间：2018-07-25 15:34

【摘要】：智能电网状态监测通过分析电网状态数据,可以实时监控和预测电力系统状况。电网系统中的状态数据数量巨大,格式多样、不统一,有的数据需要实时性处理,这就需要利用云存储技术对海量的电网监测数据进行快速有效地处理与存储。本文利用云计算中的MapReduce并行数据处理编程模型、BigTable和GFS数据存储技术,提出了智能电网监测数据的云存储原型系统,详细介绍了云存储系统整体设计、云存储构架,同时提出了云存储系统的运行流程和云存储构架的故障恢复策略,构建了一个完整、高效、可靠地数据存储和处理的系统。结合聚类算法和一致Hash算法设计了数据均衡分布算法,进行数据分布。首先,综合处理器、内存、网速等因素,进行存储设备聚类,并优先使用性能高的数据服务器；其次,在每个聚类设备内部,利用一致Hash算法均衡地将数据分布在聚类内部的各个服务器上。为了进一步满足数据之间的关联性、数据的访问便利性,寻找高效地进行计算迁移方式的网络环境,需要对已经存储的数据进行数据分布的再优化。本文利用遗传算法,选择出最合理的数据分布的优化方法。经过实验证明,本文提出的数据分布算法具有可行性。数据查询由于查询的顺序不同而造成查询效率的天壤之别,再加上分布式数据的特殊查询流程,使得数据查询效率差距更大。本文比较了不同查询方法,显示了不同查询方法的查询效率的差别。利用代数优化对查询语句进行优化,提高查询效率。进而又证明了分布式数据查询方法的可行性。最后给出了两种多服务器协同查询步骤：迭代查询和递归查询,并做了对比。本文的涉及范围从智能电网监测数据的云存储原型系统,到数据均衡分布和优化再分布,到分布数据的多服务器的分布式协同数据查询,整个从数据存储到数据查询,形成一个完整的体系。
[Abstract]:State monitoring of smart grid can monitor and predict the state of power system in real time by analyzing the state data of power system. The state data in the power system is large in quantity, diverse in format and not uniform. Some of the data need real-time processing, which requires the use of cloud storage technology to quickly and effectively process and store the massive power grid monitoring data. In this paper, using the MapReduce parallel data processing programming model in cloud computing, BigTable and GFS data storage technology, a cloud storage prototype system for smart grid monitoring data is proposed, and the whole design of cloud storage system and cloud storage architecture are introduced in detail. At the same time, the operation flow of cloud storage system and the fault recovery strategy of cloud storage architecture are proposed, and a complete, efficient and reliable data storage and processing system is constructed. Combined with clustering algorithm and uniform Hash algorithm, the data equilibrium distribution algorithm is designed to carry out data distribution. First of all, integrate processor, memory, network speed and other factors to cluster storage devices, and give priority to the use of high-performance data servers; second, within each cluster device, The uniform Hash algorithm is used to distribute the data evenly among the servers within the cluster. In order to further satisfy the relationship between data, the convenience of data access, and to find a network environment that can efficiently compute and migrate, it is necessary to optimize the data distribution of the stored data. In this paper, genetic algorithm is used to select the most reasonable data distribution optimization method. Experimental results show that the proposed data distribution algorithm is feasible. Because the order of data query is different, the query efficiency is greatly different, and the special query flow of distributed data makes the difference of data query efficiency even bigger. This paper compares different query methods and shows the difference of query efficiency of different query methods. The query statements are optimized by algebraic optimization to improve the query efficiency. Furthermore, the feasibility of distributed data query method is proved. Finally, two kinds of multi-server cooperative query steps, iterative query and recursive query, are given and compared. The scope of this paper ranges from cloud storage prototype system of smart grid monitoring data, to data balanced distribution and optimal redistribution, to distributed collaborative data query of multiple servers, from data storage to data query. Form a complete system.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP333;TM769

【参考文献】