The Exploration Of Big Data Analysis In Distribution Network Statistical Data

With the development of intelligence and informatization of the distribution network, the digital level of the distribution network is improved quickly. The volumes of monitoring and statistical data of the distribution network increase significantly. Unfortunately, there is no mature method to transform the large data generated from the power grid into useful information. It is necessary to probe into the method to analyze the large amount of data of distribution network. The analysis of data can provide the effective information for the operation of the distribution network. Based on the idea of big data, this thesis explored the approach to apply the big data analysis into distribution network. Four modes of data mining: clustering analysis, evolution analysis and correlation analysis are discussed here. The statistical data of distribution network are analyzed in depth and breadth. Three levels, which are data characterization, clustering analysis and evolution analysis, are applied into the 10 kV distribution network outage statistical data in depth. Correlation analysis are applied between devices malfunction statistical data and power quality monitoring data in breadth. The detail of the studies are given as follows.Firstly, outage statistical data are analyzed with data characteristic analysis. Various characteristics of outage in the grid have been presented. By fitting the probability density function of the overhead line outage number, statistical characteristics of overhead line power failure number is extracted. Otherwise, it is found that the data characteristic analysis is not good enough for the information extraction.Secondly, based on the results of data characteristics analysis, the clustering analysis is applied to multi-variables cluster analysis of the unplanned outage statistical data. The distribution characteristics of variables in each cluster are obtained by the cluster analysis. Then the characteristics of the faults in different regions can be described. The extracted features provides the information support of individual distribution network reliability management for each region.Thirdly, the scale and frequency of the power failure numbers and duration in fault outage statistical data are analyzed with the evolution analysis respectively. The mathematical model of the scale and frequency of the power failure numbers and duration testify the self-organized critical characteristics of fault in the distribution network. As consequence, the evolution law in scale and frequency of the distribution network fault outage is obtained.Finally, the power quality monitoring data and device malfunction statistical data are analyzed with the correlation analysis. Pearson correlation coefficients are used to quantify the correlations between the power quality monitoring data and the malfunction data. The coefficients show which power quality problems are strongly associated with device malfunctions. The information provided by the correlation analysis can give the management suggestion to prevent capacitor bank failure.Based on the idea of big data, this thesis analyzed the whole distribution network statistical data and provided the useful information for the distribution network operation. This study is a significant exploration for the application of big data and data mining in power system.