Article Preview
Top1. Introduction
With the vigorous promotion of smart power grid and the application and popularization of data acquisition and transmission technology and intelligent data processing technology in power system, various data acquisition systems in the power grid produce a large amount of data all the time (Peng Wu & Jing Tan, 2019). Such data gradually show the characteristics of big data flow, such as fast data flow speed and large data volume (B. Rakhmetulayeva et al., 2018). And it hides a lot of data information about abnormal state of power grid equipment. Therefore, it is possible to make efficient use of such data flow to timely discover the hidden risks in power grid equipment. However, how to build a low-latency and high-precision data flow processing model for power grid equipment is a key problem faced by smart power grid at present (Honarvar, Ali Reza, et al., 2016).
From the perspective of distributed processing of big data flow, this paper, based on the data flow cleaning method for power grid equipment, solves the problem of local node mining in distributed processing of online monitoring data flow for power grid equipment. First need to select to clear representation of equipment state and easy access to the data sequence, and then with the help of the existing concept of distributed data stream to complete power equipment on-line monitoring data formatting of the abstract, distributed processing technology, based on large data flow is presented based on the large distributed data stream classification model of the basic definitions, and puts forward the efficient expression of the local node mining mode structure, and the block of data to the data block in the local node local incremental mining strategy (Zhixiang Wang.etal.,2018). In order to realize the local node to its local mining mode of real-time maintenance, and the grid equipment all sequences of the data in the data stream clustering, coarse to reduce the impact on power grid equipment data stream classification concept drift, and ensure that the local node and the center of information interaction between efficiency, in order to improve the integrated classifier in the center node of data stream classification accuracy and efficiency, thus to achieve the requirements of real-time risk identification for grid equipment (Bouhafs, Faycal, et al., 2014).