Topology Recognition Of Low Voltage Distribution Network Based On Data Fusion Of Smart Meter And CTU

The low-voltage distribution network is an important occasion for the ubiquitous power Internet of Things.As a peripheral power supply network directly facing users,its power supply service capacity and power supply reliability directly affect the customer’s power consumption satisfaction.The traditional method of relying on staff to investigate and maintain topological relationships is economically unbearable.In the absence of real-time updated line information,it is easy to cause excessive line loss in the station area,delayed line repairs,and failure to grasp the phenomenon of stealing electricity in time.Therefore,power grid companies have noticed the importance of the level of intelligence in low-voltage stations.In recent years,smart meters have been widely used worldwide,and the coverage of smart measurement equipment in various regions in China has also been greatly improved.A large amount of measured data provides a reliable prerequisite for data-driven topology analysis,and the visibility of the low-voltage distribution network will also provide the basis for the line loss calculation and the power flow calculation of the power grid company.This paper divides the low-voltage distribution network topology identification into the user phase line identification problem,the meter box identification problem,and the low-voltage distribution network real line topology and electrical parameter identification problem.The first two problems are called the Metering Topology Identification in this article;the third problem is called the Electrical Topology&Parameters Identification.First,study the issue of low-voltage user identification,which is one of the content of low-voltage distribution network metering topology identification.After raising the original question as the "user 3 classification problem" and reviewing the existing research,a method for distinguishing low voltage users based on voltage and power data is proposed.First gather users according to the number of line bifurcation points to reduce the number,and then apply nonlinear dimensionality reduction algorithms to form a feature space for further dimensionality reduction;for the voltage waveform and aggregate power in the feature space,a multi-objective optimization problem that considers both power balance and voltage timing waveform dual-objective classification is proposed,in which the solution of the only power balance optimization is used as the initial solution.Two urban residential areas with typical characteristics were selected as pilots,and it was found that only one of the common five types of clustering methods was successfully identified with very different in identification performance;however,the method proposed in this paper achieved results in both cases success which shows its certain degree of stability and versatility.Secondly,it solves the second problem of the identification of the metering topology of the low-voltage distribution network,that is,the identification of the meter box of the user meter.Since the user incoming lines of the same meter box share the same distribution box bus,if the total equivalent single-phase meter box number N is known,the original problem can be converted to N-type classification based on the voltage time waveforms.Based on the analysis of the insufficient success rate of such methods in actual engineering applications,a low-voltage distribution meter box identification method combining known phase and address information is proposed.The first step is to use the clustering method to identify the voltage time waveforms after phase splitting to the distribution box level;The second step is to use the mined and completed address text information to construct constraints based on neighborhood relations,and use the constrained k-medoids algorithm to differentiate users to the meter box.Two low-voltage distribution network cases with typical characteristics prove the effectiveness of the method proposed in this paper.For the identification of electrical topology,this article focuses on solving the real line topology of the distribution network.When the measuring equipment CTU is installed in front of the meter box,the entire station area is divided into the upper and lower parts of the CTU;Use a recognition algorithm based on similar voltage waveforms to construct a tree topology above the CTU;On the premise that the meter box of the user meter is known,the principal component analysis method is used to identify the relationship between the CTU and the meter box.Through the mining of the measurement data,this paper proposes data-driven identification methods for the phase identification and meter box identification of the low-voltage distribution network;Identify the true line topology of the distribution network under the condition that CTU is installed in front of meter boxes.The proposed method will help solve the problem of ambiguity in the topology of low-voltage distribution networks,reduce the manpower requirement for the maintenance of the topology of the station area,and provide a basis for the lean management of the smart grid.