A low-voltage power distribution network topology automatic identification method and system
By injecting characteristic current to calculate node characteristic weights, potential parent-child node pairs are established and real parent-child node pairs are screened. This solves the problem of insufficient topology identification accuracy caused by noise in low-voltage distribution networks, and improves identification accuracy and management efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID HENAN ELECTRIC POWER CO TANGHE COUNTY POWER SUPPLY CO
- Filing Date
- 2026-04-01
- Publication Date
- 2026-06-26
Smart Images
Figure CN122286333A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power grid topology identification technology, and specifically to an automatic identification method and system for low-voltage distribution network topology. Background Technology
[0002] In low-voltage distribution networks, transformers typically use three-phase transmission lines for power supply. Each phase line contains multiple branches, and the downstream branches connect to varying numbers of power users, forming a hierarchical power grid topology. The complexity of this structure and the diversity of user electricity consumption characteristics place significant pressure on the operation and management of the distribution network. Improper management can increase line losses and raise the overall energy consumption of the power grid. Therefore, conducting automatic topology identification of low-voltage distribution networks is crucial for facilitating efficient operation and management of the distribution network.
[0003] There are three main types of existing automatic topology identification methods for low-voltage distribution networks: those based on SCADA systems, those based on smart meter data, and those based on characteristic current identification. The first two methods require high data quality, relying on data integrity, accuracy, and high-quality data support, respectively. While the characteristic current identification method can achieve topology identification through different characteristics generated by the connections between users, the data collected from low-voltage distribution networks contains uncertainties and noise. Furthermore, the impact of this noise on characteristic currents at different locations varies, leading to inaccurate characterization of characteristic currents at each node and thus reducing the accuracy of topology identification. Summary of the Invention
[0004] This invention provides an automatic topology identification method and system for low-voltage distribution networks to address the problem that the accuracy of low-voltage distribution network topology identification is insufficient due to the varying degrees of noise affecting the data. The specific technical solution adopted is as follows: In a first aspect, one embodiment of the present invention provides a method for automatic identification of low-voltage distribution network topology, the method comprising the following steps: Three characteristic currents with preset current amplitudes are injected into the low-voltage distribution network at preset injection intervals. Different types of power data are collected at each node within the injection time. The characteristic current injection is repeated three times for each preset current amplitude. For any node in a low-voltage distribution network, based on the trend of fluctuation of the characteristic current received by the node during the injection of a preset current amplitude, and the correlation between the characteristic current received by the node and the injected characteristic current during the injection of a preset current amplitude, the second correlation, fidelity stability and noise stability of the node are calculated respectively. The second correlation is used to evaluate the similarity between the characteristic current received by the node and the injected characteristic current. The fidelity stability and noise stability characterize the signal waveform fidelity of the characteristic current received by the node and the intensity of noise interference, respectively. The feature weight of the node is calculated based on the characterization results. Based on the order of the arrival times of the characteristic currents of each node during characteristic current injection, potential parent-child node pairs are established. Based on the correlation between the power data of different nodes during all characteristic current injections, as well as the difference in the second correlation and feature weight of the nodes in the potential parent-child node pairs, the probability of true point pairs of potential parent-child node pairs is calculated. Based on the probability of true point pairs, true parent-child node pairs are selected. Combining the power data of the two nodes in the true parent-child node pairs, the topological relationship of the low-voltage distribution network is identified.
[0005] Furthermore, the specific methods for determining the second correlation, fidelity stability, and noise stability of the node are as follows: Any node in the low-voltage distribution network is designated as the target node. Any preset current amplitude received by the target node is designated as the target preset current amplitude. An injection characteristic current sequence is established based on the characteristic current injected within the injection time of the target preset current amplitude. A node characteristic current sequence of the target node is established based on the characteristic current received by the target node within the injection time of the target preset current amplitude. The absolute value of the correlation between the injection characteristic current sequence and the node characteristic current sequence of the target node is designated as the first correlation of the target node at the target preset current amplitude. The average value of the first correlation of the three characteristic current injections performed by the target node under all preset current amplitudes is designated as the second correlation of the target node. The fidelity stability of a node is calculated by combining the volatility of the sequence formed by the characteristic current sequence and the node characteristic current sequence. The noise stability of the node is calculated based on the changing trend of the characteristic current received by the node during the injection process of the preset current amplitude, and the second correlation of the node.
[0006] Furthermore, the specific method for determining the fidelity stability of the node is as follows: The normalized value of the coefficient of variation of the sequence formed by the characteristic current sequence and the node characteristic current sequence of the target node is denoted as the characteristic current change degree of the target node at the target preset current amplitude. The average value of the characteristic current change degree of the target node at all preset current amplitudes is denoted as the average characteristic current change degree of the target node. The fidelity stability of a node is calculated based on the degree of change in the average characteristic current of the node and the second correlation. The fidelity stability is positively correlated with the degree of change in the average characteristic current and negatively correlated with the second correlation.
[0007] Furthermore, the specific method for determining the noise stability of the node is as follows: The characteristic current variation of the target node at all preset current amplitudes is linearly fitted to the preset current amplitude. The normalized value of the sum of squared residuals determined by the linear fitting is denoted as the predictability damage degree of the target node. The noise stability of a node is calculated based on the second correlation and the degree of predictability disruption. The noise stability is negatively correlated with the degree of predictability disruption and positively correlated with the fidelity stability.
[0008] Furthermore, the feature weight of the node is the ratio of the node's fidelity stability to its noise stability.
[0009] Furthermore, the method for establishing the potential parent-child node pair is as follows: For any two different nodes during the same injection of characteristic current, the node with the earliest arrival time of the characteristic current is denoted as the potential parent node, and the node with the latest arrival time of the characteristic current is denoted as the potential child node. The potential parent node and the potential child node form a potential parent-child node pair.
[0010] Furthermore, the method for calculating the probability of a true point pair in the potential parent-child node pair is as follows: For any pair of potential parent-child nodes, the absolute value of the difference between the second relevance of the potential parent node and the potential child node is denoted as the relevance difference of the potential parent-child node pair, and the ratio of the relevance difference of the potential parent-child node pair to the product of the feature weights of the potential parent node and the potential child node in the potential parent-child node pair is denoted as the first probability of the potential parent-child node pair. Based on the correlation coefficients between different types of power data within the injection duration of a single characteristic current, a correlation vector for characteristic current injection is established. The product of the node's feature weight and the correlation vector for characteristic current injection is denoted as the node feature vector for the corresponding characteristic current injection. The mean of the absolute values of the similarity between the node feature vectors of the potential parent node and the potential child node under all preset current amplitudes in three characteristic current injections is denoted as the second probability of the potential parent node pair. The mean of the normalized values of the first and second probabilities of a potential parent-child node pair is denoted as the true probability of the potential parent-child node pair.
[0011] Furthermore, the method for selecting the true parent-child node pairs is as follows: Based on the numerical distribution of the true probability of potential parent-child node pairs, the first judgment threshold and the second judgment threshold are determined respectively. Based on the relationship between the difference in the probability of real points between different potential parent-child node pairs and the second judgment threshold, and the relationship between the probability of real points between potential parent-child node pairs and the first judgment threshold, real parent-child node pairs are selected.
[0012] Furthermore, the method for identifying the topology of the low-voltage distribution network is as follows: The set of nodes contained in all real parent-child node pairs is taken as the node set of the low-voltage distribution network topology; the set of all real parent-child node pairs is taken as the edge set of the low-voltage distribution network topology, and the direction of the edge is defined as from the parent node to the child node; the cosine similarity of the vector composed of the power data of two nodes in the real parent-child node pair is taken as the edge weight of the edge between the real parent and child nodes; based on the node set, edge set, and edge weight, a directed graph representing the topological relationship of the low-voltage distribution network is constructed.
[0013] Secondly, embodiments of the present invention also provide an automatic topology identification system for low-voltage distribution networks, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.
[0014] The beneficial effects of this invention are: The degree of noise influence varies when the characteristic current reaches different nodes, which reduces the accuracy of topology identification. Therefore, this application evaluates the similarity between the characteristic current received by the node and the injected characteristic current to obtain the second correlation of the node. Then, it evaluates the signal waveform fidelity and the intensity of noise interference of the characteristic current received by the node to obtain the fidelity stability and noise stability of the node. The greater the noise stability of the node, the greater the weight of the node features should be when performing distribution network topology identification. The feature weight of the node is calculated based on the characterization results, so that the greater the noise stability and the greater the fidelity stability, the greater the weight of the node, thereby improving the accuracy of low-voltage distribution network topology identification. Considering that upstream nodes will detect the injected characteristic current before downstream nodes when power signals propagate along the distribution network lines, potential parent-child node pairs are established based on the order in which the characteristic current first arrives at each node during the injection. Then, the probability that the potential parent-child node pairs are real point pairs is evaluated to obtain the probability of real point pairs. Based on the probability of real point pairs, real parent-child node pairs are selected. Combining the power data of the two nodes in the real parent-child node pairs, the topological relationship of the low-voltage distribution network is accurately identified. This solves the problem that the accuracy of low-voltage distribution network topology identification is insufficient due to the difference in the degree of noise influence on the data of low-voltage distribution networks. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1This is a flowchart illustrating an automatic topology identification method for low-voltage distribution networks provided in one embodiment of the present invention. Detailed Implementation
[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] Please see Figure 1 The diagram illustrates a flowchart of an automatic low-voltage distribution network topology identification method according to an embodiment of the present invention. The method includes the following steps: Step S001: Inject three characteristic currents with preset current amplitudes into the low-voltage distribution network at preset injection intervals, collect different types of power data for each node within the injection time, and repeat the injection of characteristic currents three times for each preset current amplitude.
[0019] Three different current amplitudes are preset. Under each preset current amplitude, a smart terminal synchronously issues characteristic current injection and reception commands to the low-voltage distribution network. A data device controlling the location of the meter box injects the characteristic current of the preset current amplitude and collects the power data of each node. The power data includes the characteristic current received by the node, the node's current, the node's voltage, the node's active power, and the node's reactive power.
[0020] In this embodiment, the three preset current amplitudes are 0.25 amperes, 0.4 amperes, and 0.6 amperes, and are injected in ascending order of the preset current amplitude. For each preset current amplitude, a characteristic current is injected three times, with an injection interval of 5 seconds, an injection duration of 15 seconds, and an injection frequency of 4 kHz. In practical applications, as other implementation methods, implementers can determine the values of the three preset current amplitudes, as well as the injection interval, injection duration, and injection frequency of the characteristic current according to the actual situation; this application does not impose any special restrictions.
[0021] It is important to note that the preset current amplitude, the injection interval of the characteristic current, the duration of each injection, and the adjustment of the injection frequency must be based on the characteristic current injection standard to avoid the problem of not being able to obtain the characteristic current. For example, when the transmission frequency of a low-voltage distribution network is 50Hz, if the frequency of the injected characteristic current is close to 50Hz, the characteristic current will not be effectively obtained.
[0022] Thus, the power data of the characteristic current injected three times at each node under three preset current amplitude values are obtained.
[0023] Step S002: For any node in the low-voltage distribution network, based on the trend of fluctuation of the characteristic current received by the node during the injection of the preset current amplitude, and the correlation between the characteristic current received by the node and the injected characteristic current during the injection of the preset current amplitude, calculate the second correlation, fidelity stability and noise stability of the node respectively. The second correlation is used to evaluate the similarity between the characteristic current received by the node and the injected characteristic current. The fidelity stability and noise stability characterize the signal waveform fidelity of the characteristic current received by the node and the intensity of noise interference, respectively. Calculate the feature weight of the node based on the characterization results.
[0024] The topology of a low-voltage distribution network can be abstracted as a directed graph consisting of multiple nodes and their connections. Specifically: nodes correspond to key electrical connection points in the low-voltage distribution network; each node's characteristics are calculated based on power data such as voltage, current, and characteristic current; by analyzing the correlation of node characteristics, the subordinate or parent-child relationship of each node can be determined, and a directed edge representing the dominant direction of the topology hierarchy can be constructed accordingly; the edge weight is calculated by combining the characteristic correlation strength of the nodes at both ends of the directed edge or the node distance; finally, a directed graph representing the topology of the low-voltage distribution network is formed through the node set, node characteristics, directed edge set, and edge weights.
[0025] Among them, key electrical connection points in the distribution network include low-voltage feeder switches, branch box outgoing switches, and meter box incoming switches; correlation of node characteristics such as similarity and synchronization of electrical quantity changes; and dominant direction of topology hierarchy such as parent node pointing to child node.
[0026] The core of topology automatic identification technology based on characteristic current is to inject characteristic current into the power system and utilize its transmission characteristics along the power grid path, combined with the differences in characteristic currents at each node, to achieve topology identification. However, the degree of noise influence varies when the characteristic current reaches different nodes, which can reduce the accuracy of topology identification. For example, if node A in level 1 experiences greater noise interference than node B in level 2, the characteristic current amplitudes of the two nodes may be similar, leading to misclassification of nodes A and B as belonging to the same level.
[0027] Noise in power distribution networks primarily originates from grid loads. Differences in load and load variation at different nodes lead to varying degrees of random noise impact when injected characteristic currents of the same amplitude reach different nodes. This, in turn, results in variations in the characteristic current characteristics collected from the same node at different time periods. When the noise impact on injected characteristic currents of different amplitudes is consistent, higher amplitude characteristic currents are less affected by noise when transmitted to the node. Therefore, theoretically, the variation characteristics of three different amplitude characteristic currents at the same node exhibit a predictable relationship. However, the greater the load variation at a node, the greater the fluctuation in noise impact, disrupting this predictability.
[0028] Any node in the low-voltage distribution network is designated as the target node. Any preset current amplitude received by the target node is designated as the target preset current amplitude. The characteristic currents of the injected preset current amplitudes within the injection time of the target preset current amplitude are arranged chronologically to obtain the injected characteristic current sequence. The characteristic currents received by the node within the injection time of the target preset current amplitude are arranged chronologically to obtain the node characteristic current sequence of the target node. The normalized value of the coefficient of variation of the sequence formed by connecting the first and last parts of the characteristic current sequence of the target node with the node characteristic current sequence of the target node is designated as the degree of change of the characteristic current of the target node at the target preset current amplitude. The characteristic current of the target node at all preset current amplitudes is then considered. The average value of the characteristic current variation is denoted as the average characteristic current variation of the target node. The absolute value of the correlation between the injected characteristic current sequence and the node characteristic current sequence of the target node is denoted as the first correlation of the target node at the target preset current amplitude. The average value of the first correlation of the target node under all preset current amplitudes is denoted as the second correlation of the target node. Based on the average characteristic current variation and the second correlation of the target node, the fidelity stability of the target node is calculated. The fidelity stability of the target node is positively correlated with the average characteristic current variation of the target node, and the fidelity stability of the target node is negatively correlated with the second correlation of the target node.
[0029] It is understood that the positive and negative correlations in this application refer to the relationship between the independent and dependent variables. A positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive or multiplicative relationship. A negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases), and can be an inverse relationship or a subtractive relationship.
[0030] Some embodiments of this application may be that the product of the difference between the number 1 and the second correlation of the target node and the degree of change of the average characteristic current of the target node is recorded as the fidelity stability of the target node.
[0031] This embodiment uses the Pearson correlation coefficient to measure the similarity between different sequences; the calculation of the Pearson correlation coefficient, coefficient of variation and normalized value are all well-known techniques and will not be described in detail here.
[0032] It should be noted that this embodiment uses the maximum-minimum normalization method to calculate the normalized value. Specifically, the normalized value of the coefficient of variation is the ratio of the coefficient of variation to the maximum value of all coefficients of variation. In practical applications, implementers may use other methods of existing technology, such as the tanh function or the sigmoid function, to calculate the normalized value, which is not limited here.
[0033] The second correlation of the target node is used to evaluate the similarity between the characteristic current received by the target node and the injected characteristic current.
[0034] The degree of change of the characteristic current of the target node within the target preset current amplitude is used to evaluate the degree of change of the characteristic current collected by the target node within the injection time of the target preset current amplitude.
[0035] The fidelity stability of the target node is used to evaluate the amplitude fluctuation of the characteristic current and the fidelity of the signal waveform under the same preset current amplitude. The greater the fidelity stability of the target node, the more stable the transmission of the characteristic current of the target node, the less transmission distortion, and the less affected by noise.
[0036] In an ideal low-noise environment, the characteristic current response usually exhibits a monotonic trend as the amplitude changes. It is understandable that, during the injection of characteristic current with three preset current amplitude values, if the target node is stable under the influence of noise during the injection process corresponding to the three preset current amplitude values, the signal-to-noise ratio and wave fidelity of the characteristic current at the target node position corresponding to the larger preset current amplitude value should be higher.
[0037] Using the preset current amplitude as the independent variable and the degree of characteristic current change of the target node at all preset current amplitudes as the dependent variable, a linear fit is performed. The normalized value of the sum of squared residuals determined by the linear fit is denoted as the degree of predictability damage to the target node.
[0038] In this embodiment, the least squares method is used for linear fitting. The calculation of the residual sum of squares through linear fitting is a well-known technique and will not be described in detail here. The normalized value of the residual sum of squares is the ratio of the residual sum of squares to the maximum value of the residual sum of squares corresponding to all nodes.
[0039] The noise stability of the target node is calculated based on the second correlation and the degree of predictability disruption of the target node. The noise stability of the target node is negatively correlated with the degree of predictability disruption of the target node, and the noise stability of the target node is positively correlated with the fidelity stability of the target node.
[0040] It is understood that the positive and negative correlations in this application refer to the relationship between the independent and dependent variables. A positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive or multiplicative relationship. A negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases), and can be an inverse relationship or a subtractive relationship.
[0041] Some embodiments of this application may use the normalized value of the ratio of the second correlation of the target node to the degree of predictability disruption as the noise stability of the target node.
[0042] In the process of calculating the ratio, in order to avoid the denominator being zero, a preset value needs to be added to the denominator. In this example, the preset value is 0.001. The normalized value of the ratio is the ratio of the ratio corresponding to the target node to the maximum value of the ratios corresponding to all nodes.
[0043] The noise stability of a target node is used to evaluate the intensity of noise interference to the target node and the stability of the noise environment in which the target node is located. The greater the second correlation of the target node, the smaller the degree of predictability disruption, the greater the fidelity of the characteristic current received by the target node, the smaller the deviation between the characteristic current received by the target node and the predicted characteristic current received by the target node, and the smaller the noise variation during three injections of characteristic currents with different preset current amplitudes. In this case, the noise stability of the target node is greater, and the weight of the target node's features should be increased when performing distribution network topology identification.
[0044] The same method can be used to obtain the noise stability of each node in a low-voltage distribution network.
[0045] The ratio of a node's fidelity stability to its noise stability is denoted as the node's feature weight.
[0046] In the process of calculating the ratio, in order to avoid the denominator being zero, a preset value needs to be added to the denominator. In this example, the preset value is 0.001.
[0047] At this point, the feature weights of each node in the low-voltage distribution network are obtained.
[0048] Step S003: Based on the order of the arrival times of the characteristic currents of each node during the injection of characteristic currents, potential parent-child node pairs are established. Based on the correlation between the power data of different nodes during all characteristic current injections of potential parent-child node pairs, as well as the difference in the second correlation and feature weight of the nodes in the potential parent-child node pairs, the probability of the true point pairs of potential parent-child node pairs is calculated. Based on the probability of the true point pairs, the true parent-child node pairs are selected. Combining the power data of the two nodes in the true parent-child node pairs, the topological relationship of the low-voltage distribution network is identified.
[0049] Calculate the correlation coefficient between different types of power data of a node within the injection duration of a characteristic current, establish a correlation coefficient matrix, denote the vector composed of all data in the upper triangular part of the correlation coefficient matrix as the correlation vector of the corresponding characteristic current injection, and denote the product of the node's feature weight and the correlation vector of the characteristic current injection as the node feature vector of the corresponding characteristic current injection.
[0050] In this embodiment, the Pearson correlation coefficient is used to calculate the correlation coefficient between power data, and the correlation coefficient matrix is a symmetric matrix.
[0051] When power signals propagate along distribution network lines, upstream nodes will detect the injected characteristic current before downstream nodes. Therefore, by comparing the first arrival times of the characteristic current recorded by the data devices of each node, a preliminary hierarchical order of the nodes can be established. Specifically: for the same characteristic current injection process, all nodes are arranged in ascending order according to the time of the first arrival of the characteristic current. For any two different nodes, the node with the earliest first arrival time of the characteristic current is designated as the potential parent node, and the node with the latest first arrival time of the characteristic current is designated as the potential child node. Potential parent nodes and potential child nodes form a potential parent-child node pair.
[0052] Understandably, the process of injecting three characteristic currents under all preset current amplitudes requires determining the corresponding potential parent-child node pairs.
[0053] For any pair of potential parent-child nodes, the absolute value of the difference between the second correlations of the potential parent node and the potential child node is denoted as the correlation difference of the potential parent-child node pair. The ratio of the correlation difference of the potential parent-child node pair to the product of the feature weights of the potential parent node and the potential child node in the potential parent-child node pair is denoted as the first probability of the potential parent-child node pair. The mean of the absolute values of the similarity of the node feature vectors of the potential parent node and the potential child node under all preset current amplitudes and three feature current injections is denoted as the second probability of the potential parent-child node pair. The mean of the normalized values of the first probability and the second probability of the potential parent-child node pair is denoted as the true point-to-point probability of the potential parent-child node pair.
[0054] The normalized value of the first probability of a potential parent-child node pair is the ratio of the first probability of the potential parent-child node pair to the maximum value of the first probability of all potential parent-child node pairs; the normalized value of the second probability of a potential parent-child node pair is the ratio of the second probability of the potential parent-child node pair to the maximum value of the second probability of all potential parent-child node pairs.
[0055] This embodiment uses cosine similarity to measure the similarity of feature vectors of different nodes.
[0056] Clustering algorithms are used to cluster the true point-pair probabilities of all potential parent-child node pairs, resulting in two possible clusters. The cluster with the largest mean of all true point-pair probabilities is designated as the strongly linked cluster, and the cluster with the smallest mean of all true point-pair probabilities is designated as the weakly linked cluster. The mean of all true point-pair probabilities within the strongly linked cluster is designated as the first judgment threshold. In a distribution network, a parent node may correspond to multiple child nodes, and the true point-pair probabilities of the parent-child node pairs should have a high degree of similarity. Therefore, the standard deviation of all true point-pair probabilities within the strongly linked cluster is designated as the second judgment threshold.
[0057] In this embodiment, the K-means algorithm is used to cluster the probability of true point pairs.
[0058] Based on the relationship between the difference in the probability of real points between different potential parent-child node pairs and the second judgment threshold, and the relationship between the probability of real points between potential parent-child node pairs and the first judgment threshold, real parent-child node pairs are selected.
[0059] Specifically, the maximum value among the true point pair probabilities greater than the first judgment threshold is recorded as the feature true point pair probability, and the potential parent-child node pair corresponding to the feature true point pair probability is recorded as the true parent-child node pair. The absolute value of the difference between the true point pair probability greater than the first judgment threshold and the feature true point pair probability is calculated. When the absolute value of the difference is less than or equal to the second judgment threshold, the potential parent-child node pair corresponding to the absolute value of the difference is recorded as the true parent-child node pair.
[0060] It is understandable that in a real parent-child node pair, the potential parent node and the potential child node correspond to the parent node and the child node, respectively.
[0061] The set of nodes contained in all real parent-child node pairs is taken as the node set of the low-voltage distribution network topology; the set of all real parent-child node pairs is taken as the edge set of the low-voltage distribution network topology, and the direction of the edge is defined as from the parent node to the child node. The direction of the edge represents the dominant relationship of the topological hierarchy, thus determining the edge set; the cosine similarity of the vector composed of the power data of two nodes in the real parent-child node pair is taken as the edge weight of the edge between the real parent and child nodes. The edge weight is used to quantify the tightness of the topological association between nodes; using engineering implementation tools, a directed graph representing the topological relationship of the low-voltage distribution network is constructed based on the node set, edge set, and edge weight.
[0062] In this embodiment, Python is used to generate a directed graph.
[0063] This completes the identification of the low-voltage distribution network topology.
[0064] Based on the same inventive concept as the above method, this embodiment of the invention also provides a low-voltage distribution network topology automatic identification system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described low-voltage distribution network topology automatic identification methods.
[0065] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A low-voltage power distribution network topology automatic identification method, characterized in that, The method includes the following steps: Three characteristic currents with preset current amplitudes are injected into the low-voltage distribution network at preset injection intervals. Different types of power data are collected at each node within the injection time. The characteristic current injection is repeated three times for each preset current amplitude. For any node in a low-voltage distribution network, based on the trend of fluctuation of the characteristic current received by the node during the injection of a preset current amplitude, and the correlation between the characteristic current received by the node and the injected characteristic current during the injection of a preset current amplitude, the second correlation, fidelity stability and noise stability of the node are calculated respectively. The second correlation is used to evaluate the similarity between the characteristic current received by the node and the injected characteristic current. The fidelity stability and noise stability characterize the signal waveform fidelity of the characteristic current received by the node and the intensity of noise interference, respectively. The feature weight of the node is calculated based on the characterization results. Based on the order of the arrival times of the characteristic currents of each node during characteristic current injection, potential parent-child node pairs are established. Based on the correlation between the power data of different nodes during all characteristic current injections, as well as the difference in the second correlation and feature weight of the nodes in the potential parent-child node pairs, the probability of true point pairs of potential parent-child node pairs is calculated. Based on the probability of true point pairs, true parent-child node pairs are selected. Combining the power data of the two nodes in the true parent-child node pairs, the topological relationship of the low-voltage distribution network is identified.
2. The method for automatic identification of low-voltage distribution network topology according to claim 1, characterized in that, The specific methods for determining the second correlation, fidelity stability, and noise stability of the node are as follows: Any node in the low-voltage distribution network is designated as the target node. Any preset current amplitude received by the target node is designated as the target preset current amplitude. Based on the characteristic current injected within the injection time of the target preset current amplitude, an injection characteristic current sequence is established. Based on the characteristic current received by the target node within the injection time of the target preset current amplitude, a node characteristic current sequence of the target node is established. The absolute value of the correlation between the injected characteristic current sequence and the node characteristic current sequence of the target node is denoted as the first correlation of the target node at the target preset current amplitude. The average value of the first correlation of the target node under all preset current amplitudes is denoted as the second correlation of the target node. The fidelity stability of a node is calculated by combining the volatility of the sequence formed by the characteristic current sequence and the node characteristic current sequence. The noise stability of the node is calculated based on the changing trend of the characteristic current received by the node during the injection process of the preset current amplitude, and the second correlation of the node.
3. The method for automatic topology identification of a low-voltage distribution network according to claim 2, characterized in that, The specific method for determining the fidelity stability of the node is as follows: The normalized value of the coefficient of variation of the sequence formed by the characteristic current sequence and the node characteristic current sequence of the target node is denoted as the characteristic current change degree of the target node at the target preset current amplitude. The average value of the characteristic current change degree of the target node at all preset current amplitudes is denoted as the average characteristic current change degree of the target node. The fidelity stability of a node is calculated based on the degree of change in the average characteristic current of the node and the second correlation. The fidelity stability is positively correlated with the degree of change in the average characteristic current and negatively correlated with the second correlation.
4. The method for automatic identification of low-voltage distribution network topology according to claim 2, characterized in that, The specific method for determining the noise stability of the node is as follows: The characteristic current variation of the target node at all preset current amplitudes is linearly fitted to the preset current amplitude. The normalized value of the sum of squared residuals determined by the linear fitting is denoted as the predictability damage degree of the target node. The noise stability of a node is calculated based on the second correlation and the degree of predictability disruption. The noise stability is negatively correlated with the degree of predictability disruption and positively correlated with the fidelity stability.
5. The method for automatic identification of low-voltage distribution network topology according to claim 1, characterized in that, The feature weight of the node is the ratio of the node's fidelity stability to its noise stability.
6. The method for automatic identification of low-voltage distribution network topology according to claim 1, characterized in that, The method for establishing the potential parent-child node pairs is as follows: For any two different nodes during the same injection of characteristic current, the node with the earliest arrival time of the characteristic current is denoted as the potential parent node, and the node with the latest arrival time of the characteristic current is denoted as the potential child node. The potential parent node and the potential child node form a potential parent-child node pair.
7. The method for automatic identification of low-voltage distribution network topology according to claim 6, characterized in that, The method for calculating the probability of a true point pair of potential parent-child node pairs is as follows: For any pair of potential parent-child nodes, the absolute value of the difference between the second relevance of the potential parent node and the potential child node is denoted as the relevance difference of the potential parent-child node pair, and the ratio of the relevance difference of the potential parent-child node pair to the product of the feature weights of the potential parent node and the potential child node in the potential parent-child node pair is denoted as the first probability of the potential parent-child node pair. Based on the correlation coefficients between different types of power data within the injection duration of a single characteristic current, a correlation vector for characteristic current injection is established. The product of the node's feature weight and the correlation vector for characteristic current injection is denoted as the node feature vector for the corresponding characteristic current injection. The mean of the absolute values of the similarity between the node feature vectors of the potential parent node and the potential child node under all preset current amplitudes in three characteristic current injections is denoted as the second probability of the potential parent node pair. The mean of the normalized values of the first and second probabilities of a potential parent-child node pair is denoted as the true probability of the potential parent-child node pair.
8. The method for automatic identification of low-voltage distribution network topology according to claim 1, characterized in that, The method for selecting the real parent-child node pairs is as follows: Based on the numerical distribution of the true probability of potential parent-child node pairs, the first judgment threshold and the second judgment threshold are determined respectively. Based on the relationship between the difference in the probability of real points between different potential parent-child node pairs and the second judgment threshold, and the relationship between the probability of real points between potential parent-child node pairs and the first judgment threshold, real parent-child node pairs are selected.
9. The method for automatic topology identification of a low-voltage distribution network according to claim 1, characterized in that, The method for identifying the topology of the low-voltage distribution network is as follows: The set of nodes contained in all real parent-child node pairs is taken as the node set; the set of all real parent-child node pairs is taken as the edge set, and the direction of the edge is defined as from the parent node to the child node; the cosine similarity of the vector composed of the power data of two nodes in the real parent-child node pair is taken as the edge weight of the edge between the real parent and child nodes; based on the node set, edge set and edge weight, a directed graph representing the topology of the low-voltage distribution network is constructed.
10. A low-voltage distribution network topology automatic identification system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as claimed in any one of claims 1-9.