为解决从铁道配电网监控的海量动态信息中快速筛选和及时处理实时报警信息的问题,提出一种基于发布订阅的铁道供电报警信息的流计算集群处理技术.通过发布订阅方式接入铁道配电网监测数据,结合流计算模式的拓扑并行模型,综合运用多主题消息分类及分区消息缓存技巧,实现报警信息的低延迟和高吞吐量处理.以铁路10 kV动车段配电网调度监控的实测数据为算例,对工程中采集的监测数据进行多机集群测试.结果表明:基于发布订阅的流集群方法可获得数百ms级报警数据平均处理延时,且分布式消息队列的主题分区数并非设置的越大越好,通过对拓扑组件中并行线程数的优化设置,可以有效提高铁道配电网报警信息的实时响应性能和吞吐率.%In order to solve the problem of rapid filtering and timely processing real-time alarm information from the mass dynamic information collected by railway distribution network monitoring system,this paper presented a stream computing clustering technique for the railway power supply alarm information based on publish-subscribe architecture.The low latency and high throughput processing of alarm information can be a-chieved through the access to the railway distribution network monitoring data with publish-subscribe architec-ture,the combination of the topology parallel model of stream computing pattern and comprehensive applica-tion of multi-topic message classification and partition message caching technique.Based on the case study of the remote control monitoring data from the 10 kV high-speed railway distribution network,several multi-ma-chine cluster tests were performed on the monitoring data collected in the proj ect.Experimental results show that the stream clustering method based on publish-subscribe architecture can obtain the hundreds of millisec-ond level average processing delay of alarm data.The larger number of topic partitions of the distributed mes-sage queue is not always the better.This paper also verifies that the optimization of the number of parallel threads in the topology component can effectively improve real-time response performance and throughput rate of railway distribution network alarm information.
展开▼
