分布式配电网监测的流计算实时并行处理技术

Real-time Parallel Processing Technique in Stream Computing for Distributed Monitoring of Distribution Network

针对配电网大量实时监测数据先存储后处理造成延迟大的问题,提出一种基于Storm分布式流计算框架的监控信息流快速处理方法。结合流计算和全分布式内存计算优势,利用拓扑组件的并行编程模型和流计算节点集群,设计监测数据流拓扑实例,实现配电网监控信息的流计算集群处理。以铁路10 k V配电网监控系统为算例,对不同喷嘴、螺栓组件构成的拓扑实例进行集群测试,验证了先计算后存储的流计算可以提高调度监控实时处理效率。结果表明:集群环境下,拓扑实例的拓扑组件输出结果正确,在数据存储之前进行内存计算可使监测数据平均处理延时控制在百毫秒级内。

Aiming at the serious latency problem due to the massive data＇ s storage before processing in the real-time monitoring of distribution network, a fast processing method for monitoring the information flow is proposed based on Storm distributed stream computing framework. With the combination of stream computing and full-distributed memory computing, a topology instance for monitoring the data flow is designed to realize cluster processing in the stream computing of monitoring the information from distribution network, by using a parallel programming model of topology components and stream computing of node cluster. A railway 10 kV distribution network monitoring system is taken as an example, and it is used to conduct a cluster test for the topology instance composed of different spouts and bolts, which verifies that stream computing can improve the processing efficiency of real-time monitoring by processing the data before storage. Results show that under a cluster environment, the topology instance correctly outputs the processing results of topology components, and the latency for monitoring data can be controlled to be within hundreds of milliseconds by carrying out memory computing before data storage.