| Distribution network is at the end of the power system,It not only can directly reflect the needs of power consumer,but also bring to play the reliability and quality of power supplying.With the application of various modern smart grid technology,the running picture of power system which has large amount of time stamp accumulated in the distribution network,that is the voltage contour of situation map which can provides a new material for power system analysis.These images which storage the times of situation has not been fully developed and utilized,It not only can back play the situation,but also is a big treasure of situation mining.Therefore,it is very important and urgent that how to use these new materials to dig out the valuable and hidden information of power system.The voltage contour map of distribution network centered on substation is a graphics that viewed the uniform wiring map centered on substation as base picture,rendered the voltage contour,saved in SVG format of scalable vector graphics,and can reflect node data.This paper is the first aiming at the clustering analysis of historical trend of SVG image,and mining its hidden information,so as to giving the decision-making personnel of new means to understanding and masting the long history with dynamic rules of power distribution network.The research and innovations of this paper are as follows:The research on feature vector extraction model of voltage contour of distribution network that centered on transformer substation.This part is the data preparation stage,and also the most important step in the process of image clustering analysis.Firstly,this paper introduces the graphics research object,which is the voltage situation picture that based on the SVG technology,and proposes the image based on contour model and the image based on mesh model to obtaining the feature vector of voltage situation map.Aiming at the problem how to selects the extraction method of feature vector,this paper had compared in three aspects,and indicated that the method of image based on contour model can obtain better results in later section.The research on the selection of clustering analysis means based on the sample of voltage situation map.It is an important step in the process of clustering analysis of method choices,and the results of clustering are affects the comprehensive evaluation of the clustering analysis also.According to the characteristics of the feature vector based on contour model of image,It has proposed and given the principle,steps and implementation code of four kinds of clustering analysis algorithms.Then clustering the same data sample is given in the same environment.Aiming at the selection of clustering method,It has maked a comparison between the clustering results and the fast or stability of the algorithm in this paper.Finally,it has selected the K-means algorithm as the best method to the situation picture of clustering analysis.The research on the clustering analysis application of voltage contour map based on K-means algorithm.This aspect is based on the selection of feature vector extraction model and clustering analysis method,which has applied in the small sample and large sample.Firstly,Aiming at the K value selection problem of Kmeans algorithm,It has proposed the method of four determine optimal to confirm K value.and has given the results and comprehensive evaluation of the typical day sample clustering analysis,and has applied in the large sample finally.It is better mining the rules and hidden information of the clustering results of large scale samples,and can be verificated that it is a good help for grid operation to certain its decision of the cluster analysis of reseach distribution network situation picture according to the theory and data of this text. |
