A power distribution area topology identification method and system based on intelligent fusion terminal and correlation coefficient

Abstract

The application discloses a power distribution area topology identification method and system based on intelligent fusion terminals and correlation coefficients, the method comprising collecting power distribution area multi-source data through edge fusion terminals, performing data preprocessing and feature extraction on the terminal side, calculating voltage fluctuation Pearson correlation coefficients and load change Kendall rank correlation coefficients in parallel, and constructing correlation coefficient matrices in time periods; calculating adaptive weights based on load fluctuation entropy, and fusing double-mode correlation coefficients; using a graph neural network to generate topology, outputting edge existence probability and node level; using a genetic algorithm to optimize the initial topology under physical constraints; the system comprises an edge fusion terminal cluster, a cloud analysis platform, a topology verification module and a terminal management platform. The application significantly improves topology identification accuracy and dynamic adaptability: innovatively fusing double-mode correlation coefficients and graph neural network spatial modeling, and accurately restoring complex topology structures.

Description

Technical Field

[0001] This invention relates to the field of power system automation technology, and in particular to a method and system for topology identification of distribution substations based on intelligent fusion terminals and correlation coefficients. Background Technology

[0002] With the acceleration of economic development and urbanization, the number of power users in low-voltage distribution areas continues to grow. However, due to imperfect distribution network planning, complex distribution area structures, and lagging information updates, existing distribution area topology data often deviates from the actual situation, leading to reduced power supply reliability and increased difficulty in operation and maintenance management. Therefore, accurately identifying the topology of low-voltage distribution areas is of great significance for optimizing power system dispatch, improving power supply reliability, and achieving refined management. In practical applications, low-voltage distribution area topology identification also faces problems such as data noise and cross-interference between multiple distribution areas. For example, voltage and power data collected by smart meters often contain noise due to equipment failure or environmental interference, resulting in incomplete or distorted time-series data, affecting the accuracy of topology correlation analysis. In power line carrier communication, characteristic signals (such as pulse current) of adjacent distribution areas may interfere with each other, leading to misjudgment of topology affiliation.

[0003] Existing topology identification methods mainly include manual identification and automatic identification based on topology identification instruments. Manual identification relies on on-site inspection and manual recording, which is costly, inefficient, and susceptible to human error. Topology identification instruments use power line carrier signals or pulse current signals to infer the topology structure, but when crossing transformer substations, signal isolation and loss affect the stability of carrier communication, thus impacting identification accuracy. In addition, pulse current signal technology still relies on point-to-point mode, which is labor-intensive, inefficient, and the topology identification instruments are expensive and complex to operate. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a method and system for topology identification of distribution radio stations based on intelligent fusion terminals and correlation coefficients.

[0005] Technical solution: The distribution area topology identification method based on intelligent fusion terminal and correlation coefficient described in this invention includes the following steps:

[0006] Step 1: Collect multi-source data of the distribution substation through the edge fusion terminal, perform data preprocessing and feature extraction on the terminal side, calculate the Pearson correlation coefficient of voltage fluctuation and the Kendall rank correlation coefficient of load change in parallel, and construct the correlation coefficient matrix in different time periods;

[0007] Step 2: Calculate adaptive weights based on load fluctuation entropy, fuse dual-mode correlation coefficients, and introduce time-period difference factors and spatial attenuation factors; construct node features that include voltage statistical features, entropy features, and spatial features, and edge features that fuse correlation coefficients;

[0008] Step 3: Use a graph neural network to generate the topology, outputting the edge existence probability and node level;

[0009] Step 4: Use a genetic algorithm to optimize the physical constraints of the initial topology, including cycle elimination and node degree control; verify the rationality of the topology through theoretical line loss calculation.

[0010] Furthermore, step 1, edge fusion, is performed on the terminal side:

[0011] Real-time acquisition of electricity meter voltage, current, and power data;

[0012] Perform data cleaning and time period segmentation;

[0013] Calculate localization feature metrics;

[0014] Upload preprocessed data via an encrypted channel.

[0015] Furthermore, the voltage fluctuation Pearson correlation coefficient in step 1 Represented as:

[0016]

[0017] in, Represents the total number of samples. Represents a node The x-coordinate value, This represents the mean of the x-axis; Represents a node The ordinate value, This represents the mean of the vertical axis.

[0018] Furthermore, the Kendall rank correlation coefficient of the load change in step 1. Represented as:

[0019]

[0020] in, Represents the total number of samples. and These represent the number of consistent pairs and the number of divergent pairs, respectively.

[0021] Furthermore, the adaptive weights in step 2 are expressed as follows:

[0022]

[0023] in, For nodes i The load fluctuation entropy, The system's average entropy, As a regulating factor;

[0024] Introducing a spatial decay factor, ,in Distance between collection points The feature distance is the final fusion coefficient. Represented as:

[0025]

[0026] in, This represents the Kendall rank correlation coefficient representing load variation; This represents the Pearson correlation coefficient for voltage fluctuations.

[0027] Furthermore, the graph neural network in step 3 includes a terminal feature encoding layer, a feature vector processing layer for the terminal, a spatial attention mechanism, a modeling of electrical distance relationships, a hierarchical prediction module, an output node topology hierarchy, an edge connection predictor, and a generator for generating topology connections.

[0028] Furthermore, the graph neural network in step 3 adopts a course learning strategy: the first stage trains edges with a fusion coefficient > 0.8; the second stage trains edges with a fusion coefficient > 0.6; and the third stage trains all edges.

[0029] Furthermore, the fitness function of the genetic algorithm in step 4 is:

[0030]

[0031] in, Scoring for graph neural networks, Scoring the two-mode correlation coefficient. For the degree penalty term, β=0.8, γ=0.3. , where deg is the number of edges connected to the node.

[0032] The distribution area topology identification system based on intelligent fusion terminal and correlation coefficient of the present invention includes:

[0033] Edge fusion terminal cluster: deployed in the distribution network area, including data acquisition module and preprocessing module;

[0034] Cloud-based analytics platform: includes a correlation coefficient calculation module, a GNN model training module, and a genetic optimization module;

[0035] Topology verification module: used to perform line loss calculation and topology visualization;

[0036] Terminal management platform: used to configure terminal parameters and receive topology update instructions.

[0037] Beneficial Effects: Compared with the prior art, the present invention has the following significant advantages: The present invention significantly improves the accuracy and dynamic adaptability of topology identification by integrating the collaborative architecture of terminal edge computing and cloud intelligent analysis; it innovatively integrates dual-modal correlation coefficient and graph neural network spatial modeling to accurately restore complex topology structures; it introduces a dual guarantee mechanism of genetic algorithm hard constraints and GNN soft constraints to ensure that the topology conforms to the physical laws of the distribution network; and the adaptive fusion strategy based on load entropy effectively copes with the interference of new energy fluctuations. Attached Figure Description

[0038] Figure 1 This is a diagram of the overall architecture of the present invention;

[0039] Figure 2 This is a system processing flowchart of the present invention;

[0040] Figure 3 This is a diagram of the topology update mechanism of the present invention. Detailed Implementation

[0041] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0042] like Figure 1 As shown, the overall architecture of the distribution area topology identification system based on intelligent fusion terminal and correlation coefficient of the present invention includes a distribution management system, a 5G private network, an intelligent fusion terminal, smart meters, transformers and photovoltaic inverters. The top layer is the distribution management system, which is connected to the intelligent fusion terminal in the middle layer through the 5G private network. The intelligent fusion terminal is then connected to the bottom layer of smart meters, transformers and photovoltaic inverters.

[0043] like Figure 2 As shown, the distribution area topology identification method based on intelligent fusion terminal and correlation coefficient of the present invention is characterized by including the following steps:

[0044] Step 1: Collect multi-source data of the distribution substation through the edge fusion terminal, perform data preprocessing and feature extraction on the terminal side, calculate the Pearson correlation coefficient of voltage fluctuation and the Kendall rank correlation coefficient of load change in parallel, and construct the correlation coefficient matrix in different time periods;

[0045] Step 2: Calculate adaptive weights based on load fluctuation entropy, fuse dual-mode correlation coefficients, and introduce time-period difference factors and spatial attenuation factors; construct node features that include voltage statistical features, entropy features, and spatial features, and edge features that fuse correlation coefficients;

[0046] Step 3: Use a graph neural network to generate the topology, outputting the edge existence probability and node level;

[0047] Step 4: Use a genetic algorithm to optimize the physical constraints of the initial topology, including cycle elimination and node degree control; verify the rationality of the topology through theoretical line loss calculation.

[0048] This invention achieves dynamic topology updates through multi-stage collaborative processing, such as... Figure 3 The diagram shown illustrates the topology update mechanism. First, the terminal equipment synchronously collects voltage and current time-series data, equipment temperature, and environmental parameters, and performs intelligent data cleaning: it uses a dynamic threshold method to identify abnormal values ​​such as voltage drops exceeding 15%, completes missing data based on spatiotemporal correlation, and divides the daily load curve into four types of operating periods: valley, flat, peak, and peak, extracting key features such as voltage fluctuation rate and load entropy.

[0049] The system then proceeds to the dual-mode correlation analysis phase, where it calculates the Pearson voltage correlation coefficient and the Kendall load rank correlation coefficient in parallel—the former capturing linear voltage fluctuation correlations, and the latter identifying nonlinear load following relationships to resist interference. During dynamic fusion, the system adaptively allocates weights based on load entropy: when node load fluctuations are severe (entropy > 0.8), the Kendall coefficient weight increases to over 70%; simultaneously, a spatial attenuation factor is introduced, with the correlation coefficient decreasing exponentially by 30% for every 200-meter increase in electrical distance. Notably, the system implements differentiated strategies for different time periods: peak periods emphasize voltage correlation analysis, while off-peak periods strengthen load following characteristics.

[0050] Based on the fusion correlation coefficient matrix, the system constructs a topology map that includes electrical statistical characteristics (mean / variance / skewness / kurtosis), entropy characteristics (load entropy / voltage fluctuation entropy), and spatial attributes (coordinates / altitude / topology level).

[0051] Graph neural networks achieve accurate modeling through three layers of processing: the node encoder compresses the 12-dimensional original features into a 64-dimensional feature vector; the graph attention layer calculates the spatial association weights; the edge prediction module outputs the connection probability; and the hierarchical classifier divides the topology into five levels.

[0052] After the initial topology is generated, the system starts the genetic algorithm optimization engine: initializes 50 groups of topology populations, constructs a multi-objective fitness function based on model confidence, fusion correlation coefficient, and physical rule compliance, and iterates and optimizes for 15 generations through crossover and topology mutation operations. During this process, the ring network structure is forcibly eliminated, and the number of node branches is controlled to not exceed 4 to ensure the rigid constraints of the radial network.

[0053] The final stage involves multi-dimensional verification: the difference between the theoretical line loss calculation and the measured value must be controlled within 15%, otherwise topology backtracking correction will be triggered; electrical connectivity testing ensures that the power supply path for each node is unique; load matching verification requires that the total load of the parent node is not less than the sum of the loads of the child nodes. After successful verification, a standardized topology configuration file is generated and distributed to the terminal equipment through a secure channel. When equipment in the distribution area is modified or new nodes are added, the system only initiates incremental identification for the changed area, significantly reducing computational overhead.

[0054] The distribution transformer area topology identification system based on intelligent fusion terminals and correlation coefficients described in this invention includes an edge fusion terminal cluster deployed in the distribution transformer area, comprising a data acquisition module and a preprocessing module; a cloud analysis platform comprising a correlation coefficient calculation module, a GNN model training module, and a genetic optimization module; a topology verification module for performing line loss calculation and topology visualization; and a terminal management platform for configuring terminal parameters and receiving topology update instructions.

Claims

1. A power distribution district topology identification method based on intelligent fusion terminal and correlation coefficient, characterized in that, Includes the following steps: Step 1: Collect multi-source data of the distribution substation through the edge fusion terminal, perform data preprocessing and feature extraction on the terminal side, calculate the Pearson correlation coefficient of voltage fluctuation and the Kendall rank correlation coefficient of load change in parallel, and construct the correlation coefficient matrix in different time periods; Step 2: Calculate adaptive weights based on load fluctuation entropy, fuse dual-mode correlation coefficients, and introduce time-period difference factors and spatial attenuation factors; construct node features that include voltage statistical features, entropy features, and spatial features, and edge features that fuse correlation coefficients; Step 3: Use a graph neural network to generate the topology, outputting the edge existence probability and node level; Step 4: Use a genetic algorithm to optimize the initial topology based on physical constraints, including cycle elimination and node degree control; verify the rationality of the topology through theoretical line loss calculation; The adaptive weights in step 2 are represented as follows: ， in, For nodes i The load fluctuation entropy, The system's average entropy, As a regulating factor; introducing a spatial decay factor, , wherein is the distance between the acquisition points, is the characteristic distance, the final fusion coefficient is expressed as: ， in, This represents the Kendall rank correlation coefficient representing load variation; This represents the Pearson correlation coefficient for voltage fluctuations.

2. The distribution area topology identification method based on intelligent fusion terminal and correlation coefficient according to claim 1, characterized in that, Step 1, edge fusion, is performed on the terminal side: Real-time acquisition of electricity meter voltage, current, and power data; Perform data cleaning and time period segmentation; Calculate localization feature metrics; Upload preprocessed data via an encrypted channel. 3.The power distribution feeder topology identification method based on smart converged terminal and correlation coefficient according to claim 1, characterized in that, The step 1 voltage fluctuation Pearson correlation coefficient is represented as: ， in, Represents the total number of samples. Represents a node The x-coordinate value, This represents the mean of the x-axis; Represents a node The ordinate value, This represents the mean of the vertical axis. 4.The power distribution feeder topology identification method based on intelligent fusion terminal and correlation coefficient according to claim 1, characterized in that, The step 1 load change Kendall rank correlation coefficient is represented as: ， wherein, denotes the total number of samples, and denotes the number of consistent pairs and the number of inconsistent pairs, respectively. 5.The power distribution feeder topology identification method based on smart converged terminal and correlation coefficient according to claim 1, characterized in that, The graph neural network in step 3 includes a terminal feature encoding layer, which processes the feature vectors uploaded by the terminal, a spatial attention mechanism, a modeling electrical distance relationship, a hierarchical prediction module, an output node topology hierarchy, an edge connection predictor, and a generator for generating topology connection relationships.

6. The distribution area topology identification method based on intelligent fusion terminal and correlation coefficient according to claim 1, characterized in that, The graph neural network in step 3 adopts a course learning strategy: in the first stage, edges with a fusion coefficient > 0.8 are trained; The second stage trains edges with a fusion coefficient > 0.6; the third stage trains all edges. 7.The smart fusion terminal based power distribution feeder topology identification method and correlation coefficient related method according to claim 1, characterized in that, The fitness function of the genetic algorithm in step 4 is: ， in, Scoring for graph neural networks, Scoring the two-mode correlation coefficient. For the degree penalty term, β=0.8, γ=0.

3. , where deg is the number of edges connected to the node.

8. A power distribution district topology identification system based on intelligent fusion terminal and correlation coefficient, which is implemented by the method of any one of claims 1-7, characterized in that, include: Edge fusion terminal cluster: deployed in the distribution network area, including data acquisition module and preprocessing module; Cloud-based analytics platform: includes a correlation coefficient calculation module, a GNN model training module, and a genetic optimization module; Topology verification module: used to perform line loss calculation and topology visualization; Terminal management platform: used to configure terminal parameters and receive topology update instructions.

Images