Phase Line Loss Calculation And Analysis Based On Improved Nearest Neighbor Clustering Theory

The line loss is an important indicator to examine the production, operation andmanagement level of power supply enterprises. At the present time, the line loss of data ofa time section or a typical day which is calculated via power flow algorithm is tooone-sided, and fails to describe long-term operational performance of power grid, butobtain phase loss through calculating in each section via power flow algorithm is limitedby collecting data integrity and time. The phase loss has the characteristic of long time,multi section, and big date. Therefore, clustering technology can effectively avoid theabove problems, is a kind of effective method.In this paper, the phase loss calculation methods were studied based on the clusteringtechnology; in order to enhance the result’s accuracy of comparative analysis ontheoretical line losses and statistical line losses, provide powerful reference for line lossassessment and management of power supply enterprise. The main contents of this paperare as follows:Firstly, the clustering analysis theory is studied, including the definition, process andthe commonly used methods of clustering. At the same time, an improved nearestneighbor clustering algorithm is proposed for such problems as the unreasonable centerselection, inflexible selection of threshold and the imperfect clustering results.Secondly, the cause effect of section loss fluctuation is analyzed by using theprinciple of superposition. The section feature vector is defined, and in the case of missingdata, feature vector selection via core reducing of rough set theory is used to obtain thecore feature vector. After Weighted Euclidean distance is be used to measure the similaritybetween sections, and the weight coefficient vector of The clustering center is calculatedaccording to quantification of nodal injection power fluctuation and incremental losscaused by network parameters changes.Finally, the improved nearest neighbor clustering calculation method is used to calculatethe phase loss, and the processing method which through the correlation coefficient andthe similarity matching rule to clustering is given for the data deletion serious section. In order to simplify the calculation, on the basis of above methods, a simplified method ofsection clustering based on information fusion theory is proposed. Based on the analysisof the load curve and time course the calculation accuracy was not affected by preservingcharacteristic section load changes. The simulation analysis verified feasibility andcorrectness of the proposed method.