基于改进贝叶斯网络的电力大数据存储架构可扩展性优化算法  

Scalability Optimization Algorithm of Power Big Data Storage Architecture Based on Improved Bayesian Network

作　　者：邓志东 刘鲲鹏 DENG Zhidong;LIU Kunpeng(State Grid Customer Service Center,Tianjin 300309,China)

机构地区：[1]国家电网有限公司客户服务中心,天津300309

出　　处：《微型电脑应用》2024年第9期102-106,共5页Microcomputer Applications

基　　金：国网客服中心2023年客户服务数据管理能力成熟度提升技术服务项目(SGKF0000DFJS2310026)。

摘　　要：受到传统贝叶斯网络局部搜索寻优机制的影响,在电力大数据存储框架扩展优化中无法兼顾全局扩展数据间的依赖关系,导致优化出现局部偏差,扩展存储部分性能衰减严重。为了解决这一问题,通过构建电力大数据存储架构的贝叶斯网络,建立电力大数据存储架构可扩展节点,确定其原始框架节点可扩展上限,改进贝叶斯网络结构的扩展性编码计算,以及存储框架扩展性优化适应度评价,由此重新定义优化算法。仿真实验数据表明:提出算法的读取响应和写入响应指标均控制在60 ms以下,且在波动幅度控制方面表现最好;对于不同容量的样本集,其读取性能和写入性能的均值偏差分别为0.2和0.3,接近理想状态,适应度系数值最大为2。提出算法切实可行,具有较高的实际应用价值与市场推广价值。Affected by the traditional local search and optimization mechanism of Bayesian networks,in the expansion and optimization of power big data storage frameworks,it is impossible to balance the dependency relationship among global extended data,resulting in local deviations in optimization and severe performance degradation of the extended storage part.In order to solve this problem,a Bayesian network for power big data storage architecture is constructed.The expandable nodes of the power big data storage architecture are established and the upper limit of the original framework nodes is determined.The scalability coding calculation of the improved Bayesian network structure and the fitness evaluation of the storage framework scalability optimization are redefined as the optimization algorithm.The data from simulation and comparative experiments indicate that the proposed algorithm has both read response and write response indicators controlled below 60 ms,and performs best in controlling fluctuation amplitude.For sample sets with different sizes,the mean deviations of read and write performance are 0.2 and 0.3 respectively,which are close to the ideal state.The maximum fitness coefficient value is 2.The proposed algorithm is practical and feasible,with higher practical application value and market promotion value.

关 键 词：改进贝叶斯网络 电力大数据 存储框架 优化算法 

分 类 号：TM773[电气工程—电力系统及自动化]