Researsh On The Real-time Stream Computing Processing Of Information Of Dispatching And Monitoring System Of Railway Power Supply Based On Storm

With the constant expansion of the scale and capacity of the power grid, Distribution network of modern industry is gradually moving towards a new development direction,including the interaction of energy and information and the combination of massive information and intelligent dispatching. Compared with traditional power network,Dispatching and monitoring of intelligent distribution network is realized based on panoramic real-time state data acquisition of power grid. more acquisition points and higher device status information sampling rate make massive monitoring data need to be processed by dispatching and monitoring system integrated device information. In the future, Interconnection interaction will be realized by application of network equipment, new energy technology, and Electric vehicle access equipment, leading to intensifying massive state information real-time processing pressure of power network monitoring. Take IEC61970 CIM(Common Information Model) of real-time data acquisition and monitoring control system as example,analog telemetry class object such as bus voltage and feeder current is defined as more than50 attributes, and status remote signal class object such as disconnector and circuit breaker is defined as more than 40 attributes. Because the monitoring real-time database usually adopts processing method of resident computer memory, excess real-time monitoring data are placed in memory for millions of levels, tens of millions, or even more, which will eventually exceed memory capacity upper limit of real-time data server, leading to that the access performance and computing capability of conventional real-time database are limited. Facing the pressure of real-time access and interactive processing for massive monitoring information, data server CPU may produce large information processing latency due to communication channel congestion, affecting the real-time performance of dispatching and monitoring. Therefore, it is urgent to carry out the research of rapid processing technology for large amount of real-time monitoring data.Combined with big data status in power field, big data processing technology based on batch processing mode and stream computing mode is introduced to realize the distributed storage and efficient parallel processing of massive monitoring data. However, batch processing mode is limited to general operation process of data’s storage before processing,which makes it difficult to reach subsecond-level real-time processing of large amount of monitoring data. On the contrary, capacity constraint of memory database can be avoided by applying stream computing mode with its full memory computing, high scalability, high faulttolerance and other advantages, it provides a completely new solution for rapid and low latency processing of massive monitoring data in dispatching and monitoring.Considering the particularity of railway power supply distribution system, the reliability of railway power supply and traffic safety will be greatly affected by processing real-time of Dispatching and monitoring information flow. Thus, in this paper, the real-time processing system of railway dispatching and monitoring based on Storm stream computing is constructed, combined with processing requirements of massive real-time monitoring information in practical engineering application. The integrated system takes the Storm stream computing framework as the real-time processing engine for monitoring data, using distributed message system “Metamorphosis” as data source access module, building a distributed storage system of dispatching and monitoring based on HBase. With the massive monitoring data of railway dispatching and monitoring as processing object, Overall computational performance of initial integrated system is evaluated through a series of topology instance cluster tests. Test results show that, stream computing real-time processing system can ensure the accuracy of data processing while obtaining millisecond-level average processing latency; a relatively low average processing latency can be maintained by integrated system when disk message accumulation happens; integrated system has a certain load balancing capacity, when coping with large scale stream computing topology task.In the end, optimization methods of configuring operating parameters and designing parallel sliding window are proposed with the purpose of further improving overall computational performance of the real-time processing system. Compared with test results of before and after optimization, above optimization methods are verified to be effective on improving stream computing parallel processing capacity and bettering system memory consumption. What’s more, in order to provide a more complete and robust big data processing scheme for railway dispatching and monitoring, this paper also considers combining real-time stream computing and batch processing mechanism by using open source project “Storm-Yarn” to deploy stream computing framework on resource management system “YARN” of Hadoop2.0, and cluster test for monitoring data topology instance is finally done to verify the feasibility of this association scheme.