Research And Application Of Cloud Computing Technology In The Power System Bad Data Processing

In this Big Data era, the traditional pattern of calculation and storage can hardly meet the requirement of power system gradually in the case of the explosive growth of data. As a kind of new technology for processing mass-data, with super-large scale, high-availability as well as high-scalability, Cloud Computing Technology can be used to solve the problem and can supply the solution for the construction of the information platform for the new generation of power system. In this paper, the identification and correction of the bad data in the power system using cloud technology are investigated.This paper overviews the overseas and domestic research status of cloud technology and the bad data processing of power system. This paper studies the core technology and the realization mechanism of batch cloud computing framework Hadoop and real-time cloud computing framework Storm. On this basis, various algorithms for processing bad data using cloud computing in the aspects of data mining technology and topology theory are designed.In the field of data mining, a distributed system clustering algorithm under Hadoop is put forward to pre-process data. Through transforming the identification of bad data into binary classification, a distributed bad data processing algorithm based on SVM, adopting the slide window mechanism, is brought forward. The algorithm can be divided into two patterns, the off-line mode and the online mode. The off-line mode uses the Hadoop for the distributed processing. Clustering is used on the samples of data at first, each reserved class cluster is divided by time interval and use training model of SVM separately. The on-line mode receives the data to be identified through setting the slide window and adopts the corresponding SVM model to identify bad data and modify it with linear interpolation.In the aspect of topology theory, a distributed bad data processing algorithm dealing with active power in the whole power system under Storm is put forward. The algorithm processes data under the condition of circuit constraints, it introduces the redundant information in the edge of subnet so that distributed calculation can be used in each subnet. The integrated computation is adopt in the whole network level so as to identify and modify data in the entire power system. Meanwhile, a distributed state estimation algorithm under Storm is proposed. It introduces virtual nodes in the tie lines, and splits the subnet through virtual node so as to calculate the distributed state estimation and process bad data, the processing of bad data simplifies the realization process of residual search identification method.The paper provides simulation testing for each algorithm, the simulation result expresses that these algorithms possess superior capacity of bad data processing and instantaneity with some practical prospects.