Research Of Power Equipment Risk Identification And Early Warning Model Based On Data Stream

With the vigorous promotion of UHV grid and intelligent substation, the increasing size and the automation degree of the power system and continuously has been improved, operation and maintenance of power transmission and distribution equipment come to be more complex, the improvement of the health and life of the equipment can help to ensure the security and reliability of the power system. State Grid Corporation has used lots of modern scientific and technological means and methods to enhance the management level of power equipment, and got lots of real-time data streams. These data streams implies important information relating to equipment failure, the paper means to research the key technology of data stream and the methods of risk identification and early warning of network equipment by using those data stream, the topic has important practical applications..In this paper, the research were put on how to get the overview of data flow model using the sliding window data in data stream, about the data stream sampling and the significant anomaly characteristics analysis of the data stream. By using a sliding window model realizes the data stream data even extraction, while according to the device monitoring data interval abnormalities and abnormal fluctuations of attention, the introduction of the method ratio threshold of fluctuation anomalies were extracted outline model paper proposes generation algorithm except guarantee data uniformity, stability also have the advantage footprint respect, both to improve awareness of important data, but also to enhance the stability of the algorithm.For multi-stream integration issues that this article through the data stream mining algorithms for network equipment risk identification, proposed the idea for a double grid equipment state detection data processing through online micro-clustering multiple devices to achieve fast polyethylene mixed data stream class, and then extract the micro-cluster centroid offline layer classification. While offline classification, the advantages of using traditional algorithms for data multiple scans, made more accurate results. At the same time we consider the importance of multi-device data stream data, in the online micro-clusters of micro clustering does not delete the way to ensure that all data are effectively applied, in the future this algorithm massive equipment on-line monitoring system that can be more effective quickly identify equipment risks, with a high value.In this paper, constantly monitor device status, based on a method of risk status warning equipment, analysis equipment by historical operating data to predict trends device for some time in the future content running state, so a reasonable way to take the appropriate preventive improve equipment operating status. In this paper, the gray of early warning model to predict metabolic filling equipment operating data, and analyze their possible risks for some time based on the predicted value. Algorithm always incoming data stream to the latest data, improve data accuracy, if it is embedded into the online monitoring platform, it can achieve the status of the device always risk warning.