电力电网中台架构的数据统一模型构建

Construction of universal data model for power grid middle office architecture

针对传统电网数据分析平台计算效率低、数据误差大等问题,提出了一种基于Apache-Spark(AS)平台的集成数据统一模型(UDM),该模型结合了故障数据预测的优化线性回归模型、皮尔逊相关系数提取的新数据融合以及基于二叉树优化的支持向量机的故障类型分类等技术,提高了对电网故障数据的预测精度和分类精度。实验结果表明,UDM框架处理电力电网数据具有较低的均方误差,当测试样本数为50000个时,AS平台的响应时间小于5秒,体现出极佳的性能优势。

Aiming at the first mock exam of traditional power grid data analysis platform,such as low computation efficiency and large data error,a unified data model(UDM) based on Apache-Spark(AS) platform is proposed.This model combines the optimized linear regression(OLR) model and Pearson correlation coefficient.New data fusion extracted by persons correlation coefficient(PCC) and binary tree support vector machine(BTSVM) based on binary tree optimization The experimental results show that the UDM framework has low mean square error in processing power grid data.When the number of test samples is as high as 50000,the response time of as platform is less than 5 seconds,reflecting excellent performance advantages.