基于Spark的变压器局部放电模式识别并行化实现  

PARALLEL REALIZATION OF TRANSFORMER PARTIAL DISCHARGE PATTERN RECOGNITION BASED ON SPARK

作　　者：李涛 朱永利[1] Li Tao;Zhu Yongli(School of Control and Computer Engineering,North China Electric Power University,Baoding 071003,Hebei,China)

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003

出　　处：《计算机应用与软件》2023年第1期74-78,145,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(51677072);中央高校基本科研业务费专项资金项目(2018QN078)。

摘　　要：快速高效地识别局部放电类型不仅对电力设备的状况评估具有十分重大的意义,而且能够实现智能电网下对电力设施监测数据的快速诊断。因此,在Spark计算框架的基础上提出基于VPMCD(Variable Predictive Model Based Class Discriminate)的局部放电模式识别的并行化方法。采取对原始放电信号提取其φ-q-n图谱的PRPD(Phase Resolved Partial Discharge)特征构成相关特征向量作为实验输入,采取并行化VPMCD算法对放电类型进行分类。实验结果和分析表明,在Spark计算框架下的分布式处理的计算效率要优于传统单机环境下的计算效率,加速比随着节点数和数据量的增多而显著提升,可以满足智能电网下大数据快速处理的要求。Recognizing the types of partial discharge(PD) quickly and efficiently is not only of great significance to the condition assessment of transformers, but also enables rapid diagnosis of power facility monitoring data under the smart grid. Therefore, based on the Spark computing framework, this paper proposes a parallel method of PD pattern recognition based on VPMCD. The PRPD features of the φ-q-n spectrum extracted from the original discharge signal constituted the relevant feature vector as the experimental input. The parallel VPMCD algorithm was adopted to classify the discharge types. Experimental results and analysis show that the computing efficiency of distributed processing under the Spark computing framework is prior to that of the traditional single-machine environment, and the acceleration ratio is significantly improved with the increase of the number of nodes and the amount of data, which can meet the requirements of fast big data processing under smart grid.

关 键 词：Spark计算框架 VPMCD 局部放电 模式识别 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TM933[电气工程—电力电子与电力传动]