Development of a cluster analysis method for solving the problem of identifying static voltage load characteristics

Abstract

Identification of static voltage load characteristics is not a trivial task. It is extremely difficult to develop a single methodology that will allow achieving the required accuracy results for any input data. The authors made an attempt to solve the problem, developed a method based on the use of methods for processing large data sets – BigData. One of the most important stages of this method is the stage of cluster analysis. The problem of choosing a method of cluster analysis also has nuances: it is necessary to maintain a balance between time and accuracy of calculations, while performing calculations with sufficient accuracy on various data. A number of domestic and foreign authors have investigated this issue, but the proposed options cannot be used in the developed methodology. The purpose of this research is to develop a cluster analysis method that will satisfy the above requirements. The author’s method is based on the k-means method. The main feature of the author’s method is the mechanism of initial generation of cluster centroids, which allows to exclude the influence of random generation of two-dimensional coordinates imperfection and to achieve uniform generation of initial clusters centers. The developed method was tested on a model problem with all known solutions, which can be easily verified analytically. Also the method was tested on actual industrial load data. The results obtained the requirements of accuracy and calculation speed and allow us to conclude that the developed method of cluster analysis can be used in the task of identifying static voltage load characteristics.