Data-driven Automatic Identification Method Of Topology Relationship In Low Voltage Distribution Networks

The topology of the low-voltage distribution network is the basis for application functions for example analysis of loss of voltage for line,fault diagnosis,and three-phase balance,and plays a vital role in improving power supply reliability and power supply service capabilities.However,the low-voltage distribution network is located at the end of the power grid,serving many users,and has a complex and diverse structure.In the process of transformation,the situation of missing information,false alarms and not entering the system of changes in topological relations is common,which seriously affects the lean management of the distribution network.Only relying on manual investigation and topology recognition method not only consumes human resources and increase investment in equipment,but also has low efficiency and low recognition accuracy.Based on the voltage data uplo aded by terminal devices such as smart meters,th e paper proposes the data-driven automatic identification method for the topological relationship of low-voltage distribution stations,which solves the difficult problems in the identification of topological relationships.For the problem of low correct rate of topology identification when voltage data is close,a new method of topology recognition of low-voltage distribution station area based on voltage fluctuation characteristic parameters is proposed.Firstly,the voltage fluctuation characteristic parameters are defined to extract the global and local characteristics of the voltage data,and the voltage fluctuation characteristic matrix is established.Then,the voltage fluctuation characteristic matrix is combined with the K-means clustering algorithm and linear regression algorithm to identify the household-transformer relationship and household-phase relationship.The analysis of calculation examples shows that,compared with the direct use of voltage data,the voltage fluctuation characteristic matrix is m ore effective for topological analysis.Aiming at the problem that the existing methods are not highly recognizable when the voltage data are close to each other,an adaptive segmented cloud model based low-voltage distribution station topology recognition method is proposed.First of all,according to certain rules,adaptively determine the total number of segments for the transformation of the voltage time series of the station transformer and the user into the cloud model,and adaptively determine the st arting time of each segment of the cloud model.Then the similarity of the segmented cloud model is calculated,and the topological relationship identification result is obtained.The analysis of the calculation example shows that the adaptive segmented cl oud model algorithm is more accurate than the overall cloud model algorithm and other similarity algorithms for the situation where the voltage data is close.Finally,from the perspective of different topologies,different user loads,algorithm limitation s,distributed power sources and errors,it shows that the adaptive segmented cloud model algorithm is highly versatile.Aiming at the problems of poor recognition of exist ing methods when the voltage data is close,and missing or abnormal voltage data,a low-voltage power distribution system based on derivative dynamic time warping distance(DDTW)and density-based spatial clustering of application with noise(DBSCAN)is proposed to solve station topology identification.First,the DDTW algorithm is used t o measure the similarity of time series with unequal lengths and different sampling rates to reduce the impact of missing or abnormal voltage data on the recognition of topological relationships.The addition of the analysis of the voltage curve shape characteristics and change trend solves the problem that the identification result is affected by the closeness of the voltage data and the misjudgment rate is high.Based on the calculated DDTW distance,the DBSCAN clustering algorithm is used to cluster the station changes and users.In order to reduce th e influence of the parameter radius and density threshold in the clustering algorithm on the clustering results,the idea of probability is introduced,and the parameters are changed many times for clustering,and the probabilistic result of the topologica l relationship of the station area is obtained.Using the method of combining DDTW and DBSCAN,there is no need to consider missing or abnormal v oltage data,and there is no need to artificially set thresholds,and the recognition accuracy of the topologic al relationship of the station is high.