Power distribution network voltage weak point identification method and system

Abstract

The invention discloses a power distribution network voltage weak point identification method and system, and relates to the technical field of power distribution network monitoring. Natural condition data and operation data of a power distribution network are combined, clustering is carried out by considering seasonal factors, and the operation data under similar weather conditions are divided together for feature extraction; the operation states of the power distribution network under different natural conditions can be captured more accurately; the natural condition data after feature extraction and the operation data of the power distribution network are fused to form a positive sample pair and a negative sample pair, the positive sample pair and the negative sample pair are input into the comparative learning model for training, the natural condition data and the operation data are considered during training, the influence of the environment on the judgment of the voltage weak point is reduced, and the comparative learning model has relatively strong generalization ability; operation conditions which are not seen in the training data can be handled; according to the method, the operation state of the power distribution network can be monitored according to the real-time operation condition, and early warning is given out in time when abnormity is found, so that faults or accidents caused by voltage weak points are effectively prevented.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network monitoring, and in particular to a method and system for identifying voltage weak points in a distribution network. Background Art

[0002] In the distribution network structure, voltage weak points are often potential areas where failures occur. By analyzing these weak points, potential safety hazards can be discovered in advance and preventive measures can be taken to reduce the risk of power grid failures. In addition, by analyzing voltage weak points, it is possible to understand whether the structure and layout of the distribution network are reasonable. If it is found that there are many voltage weak points or the problems are serious in certain areas, it can be considered to optimize the power grid structure, such as adding lines, transformers and other equipment to improve the power supply capacity and stability of the power grid.

[0003] After distributed power sources are connected to the distribution network, the original grid structure and power flow distribution will be changed, thereby affecting the voltage. These impacts are mainly reflected in voltage fluctuations, voltage imbalance and voltage stability. The access of distributed power sources will increase the short-circuit capacity of the distribution network, resulting in an increased possibility of voltage fluctuations. At the same time, the output power of some distributed power sources is random, such as being affected by natural conditions such as light intensity and wind speed, which further increases the complexity of identifying voltage weak points.

[0004] In existing technologies, weak points in voltage distribution networks are often identified through association rules and supervised algorithms. Association rules are mainly mined based on static data and cannot capture the dynamic changes of weak points in voltage. They are also greatly affected by parameter settings. Supervised algorithms often have problems such as poor generalization ability and high requirements for the accuracy and consistency of input data.

[0005] Therefore, it is urgent to develop a method for identifying voltage weak points in distribution networks to solve the above problems. Summary of the Invention

[0006] In view of the above-mentioned problems, the present invention is proposed.

[0007] Therefore, the technical problems solved by the present invention are: the existing distribution network voltage weak point identification method has poor dynamic adaptability, does not consider the dynamic coupling influence of natural conditions, has weak model generalization ability, and how to achieve real-time monitoring and precise positioning of abnormal points.

[0008] To solve the above technical problems, the present invention provides the following technical solutions: a method for identifying voltage weak points in a distribution network, comprising obtaining distribution network data and preprocessing it, clustering the natural condition data using a clustering algorithm to obtain data clusters; dividing historical data to obtain data sets, extracting features from the data clusters and data sets using a convolutional neural network, and forming positive and negative sample pairs; training a comparative learning model using the positive and negative sample pairs, outputting positive and negative score values, and judging whether there is an abnormality in the distribution network based on the positive and negative score values.

[0009] As a preferred solution of the method for identifying voltage weak points in a distribution network described in the present invention, the obtaining of distribution network data and preprocessing includes obtaining the topological structure of the distribution network, historical normal operating data, historical abnormal operating data, first natural condition data, and second natural condition data; and performing preprocessing; the first natural condition data is the natural condition data corresponding to the historical normal operating data; the second natural condition data is the natural condition data corresponding to the historical abnormal operating data; the historical normal operating data is the operating voltage at each node in the historical data when the distribution network is in a normal operating state; the historical abnormal operating data is the operating voltage at each node in the historical data when the distribution network is in an abnormal operating state.

[0010] As a preferred solution of the method for identifying voltage weak points in a distribution network described in the present invention, the clustering of the natural condition data includes clustering the pre-processed first natural condition data and the second natural condition data according to weather conditions based on different seasons through a clustering algorithm to obtain a first data cluster and a second data cluster.

[0011] As a preferred solution of the method for identifying voltage weak points in a distribution network described in the present invention, the division of historical data to obtain data sets includes clustering the pre-processed historical normal operation data and historical abnormal data according to weather conditions based on different seasons using a clustering algorithm to obtain a first data set and a second data set.

[0012] As a preferred solution of the method for identifying voltage weak points in a distribution network described in the present invention, the formation of positive and negative sample pairs includes fusing the first data cluster after feature extraction and the first data set to obtain first sample data, and introducing noise into the first sample data to form second sample data; the first sample data and the second sample data constitute a positive sample pair; fusing the second data cluster after feature extraction and the second data set to obtain third sample data; the third sample data and the first sample data constitute a negative sample pair.

[0013] As a preferred solution of the method for identifying voltage weak points in a distribution network described in the present invention, the training of the contrastive learning model using positive and negative sample pairs includes inputting positive sample pairs and negative sample pairs into the contrastive learning model for processing to obtain feature representations, respectively calculating feature representation score values ​​of the positive sample pairs and negative sample pairs, calculating loss values ​​based on the score values, backpropagating the loss values ​​using a gradient descent algorithm, updating model parameters, and obtaining a trained contrastive learning model; the contrastive learning model is used to input real-time operating data and output positive score values ​​and negative score values.

[0014] As a preferred embodiment of the method for identifying voltage weak points in a distribution network according to the present invention, the method comprises: determining whether an abnormality exists in the distribution network based on positive and negative score values, based on a trained contrastive learning model, collecting distribution network operation data and natural condition data in real time and inputting the data into the contrastive learning model to obtain positive score values ​​and negative score values; determining whether an abnormality exists based on the positive score values ​​and the negative score values; if an abnormality exists, determining the location of the abnormal point in the distribution network based on the topological structure; the positive score value is a positive sample pair similarity value calculated using a cosine similarity algorithm; and the negative score value is a negative sample pair similarity value calculated using a cosine similarity algorithm.

[0015] Another object of the present invention is to provide a distribution network voltage weak point identification system, which can build a comparative learning model by integrating multi-season natural condition data with grid operation data, and combine feature inversion with gradient sensitivity analysis to solve the technical problems that current static modeling technology is difficult to suppress environmental noise interference and cannot adapt to dynamic scenarios.

[0016] As a preferred solution of the distribution network voltage weak point identification system described in the present invention, it includes: a data clustering module, a feature extraction module, and an abnormality judgment module; the data clustering module is used to obtain distribution network data and preprocess it, clustering the natural condition data through a clustering algorithm to obtain data clusters; the feature extraction module is used to divide historical data to obtain data sets, extract features from the data clusters and data sets through a convolutional neural network, and form positive and negative sample pairs; the abnormality judgment module is used to train a comparative learning model through positive and negative sample pairs, output positive and negative score values, and judge whether there is an abnormality in the distribution network based on the positive and negative score values.

[0017] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a method for identifying voltage weak points in a distribution network.

[0018] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for identifying voltage weak points in a distribution network.

[0019] Beneficial effects of the present invention: The distribution network voltage weak point identification method provided by the present invention combines natural condition data and distribution network operation data, clusters them considering seasonal factors, and divides the operation data under similar weather conditions together for feature extraction, which can more accurately capture the operation status of the distribution network under different natural conditions; the natural condition data and distribution network operation data after feature extraction are fused to form positive sample and negative sample pairs, which are input into the contrast learning model for training. During training, the natural condition data and operation data are comprehensively considered, which reduces the impact of the environment on the judgment of voltage weak points, so that the contrast learning model has strong generalization ability and can cope with newly emerging operation conditions that do not appear in the training data; according to the topological structure, the position of the abnormal point in the distribution network is obtained by feature inversion, thereby realizing accurate fault location; the present invention can monitor the operation status of the distribution network through real-time operation conditions, and issue early warnings in time when abnormalities are found, thereby effectively preventing power grid failures or accidents caused by voltage weak points. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 This is an overall flow chart of a method for identifying voltage weak points in a distribution network provided by the first embodiment of the present invention.

[0022] Figure 2 This is an overall flow chart of a distribution network voltage weak point identification system provided by the third embodiment of the present invention. DETAILED DESCRIPTION

[0023] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0024] Example 1, reference Figure 1 , as one embodiment of the present invention, provides a method for identifying voltage weak points in a distribution network, comprising:

[0025] S1: Obtain distribution network data and preprocess it, cluster the natural condition data using a clustering algorithm to obtain data clusters.

[0026] Furthermore, obtaining distribution network data and performing preprocessing includes obtaining the topological structure of the distribution network, historical normal operation data, historical abnormal operation data, first natural condition data, and second natural condition data; and performing preprocessing; the first natural condition data is the natural condition data corresponding to the historical normal operation data; the second natural condition data is the natural condition data corresponding to the historical abnormal operation data; the historical normal operation data is the operating voltage at each node when the distribution network operation status is normal in the historical data; the historical abnormal operation data is the operating voltage at each node when the distribution network operation status is abnormal in the historical data.

[0027] Distributed energy is often affected by natural environmental conditions, which reduces its operating efficiency. Therefore, the first natural condition data and the second natural condition data include but are not limited to sunshine intensity, wind speed, temperature, precipitation, snowfall and air quality, and various influencing factors are selected according to actual conditions.

[0028] For the same distribution network, there are consistent operating voltages at each node at different times. For the same distribution network, each node of the distribution network is numbered, and historical normal operation data and historical abnormal operation data of the distribution network at different times are obtained. The data are arranged in the order of the numbers of each node of the distribution network to form a sequence.

[0029] The first natural condition data matches the timestamp of the historical normal operating data; the second natural condition data matches the timestamp of the historical abnormal operating data.

[0030] Preprocessing includes but is not limited to filling missing values, correcting outliers, and data standardization to ensure data quality and analysis accuracy.

[0031] Methods used to fill missing values ​​include but are not limited to interpolation, regression prediction, and mean imputation.

[0032] Correcting outliers includes, but is not limited to, deleting outliers, replacing outliers, scaling outliers, and outlier detection.

[0033] Methods used for data normalization include, but are not limited to, min-max normalization, Z-score normalization, logarithmic transformation, and exponential transformation.

[0034] It should be noted that clustering the natural condition data includes clustering the pre-processed first natural condition data and the second natural condition data according to weather conditions using a clustering algorithm based on different seasons to obtain a first data cluster and a second data cluster.

[0035] Based on the fact that distributed energy is affected by natural conditions, and considering the characteristics of natural conditions changing with the four seasons of spring, summer, autumn and winter, dividing the first natural condition data and the second natural condition data by season helps to capture the differences in natural conditions in different seasons, and splitting the data into several categories for separate processing can reduce the complexity of data processing.

[0036] When dividing by season, ensure that the distribution of each seasonal data set is relatively balanced to avoid deviations during model training. The seasonal changes in different regions may vary, and different adjustments should be made based on the actual conditions of the specific region.

[0037] In different seasons, the preprocessed first natural condition data and the second natural condition data are clustered using a clustering algorithm, and data with similar weather conditions are divided into the same cluster.

[0038] Clustering algorithms include but are not limited to K-means and DBSCAN.

[0039] In an embodiment of the present application, the preprocessed first natural condition data is clustered by the K-means clustering algorithm. Taking heavy rain weather as an example, the data under heavy rain weather conditions are clustered into the same cluster. Specifically, the data is divided according to the season of operation, and summer data is selected as the analysis object. The K value is selected according to the distribution of Summer data, and the K value is determined by the silhouette coefficient. K data points are randomly selected as initial cluster centers, and the distance from each data point to the K cluster centers is calculated by Euclidean distance. Each data point is assigned to the cluster to which the nearest cluster center belongs, and the mean of all data points in each cluster is calculated. The mean is used as the new cluster center for iteration until the cluster center converges, that is, the change is less than the prediction threshold or the maximum number of iterations is reached.

[0040] In an optional embodiment, the preprocessed first natural condition data is clustered using the DBSCAN clustering algorithm. Taking high temperature weather as an example, the high temperature data is clustered into the same cluster. Specifically, the eps and minPts parameters of the DBSCAN algorithm are set based on the high temperature characteristics, where eps represents the maximum distance between two data points. If the distance is less than eps, the two data points are considered to be in the same cluster. minPts represents the minimum data point data that must be included in a cluster. The DBSCAN algorithm is run using the Euclidean distance and set parameters to cluster the data into different clusters. The characteristics of each cluster are analyzed to determine the high temperature weather cluster, and the cluster representing high temperature weather is marked.

[0041] S2: Divide the historical data into data sets, extract features from data clusters and data sets through convolutional neural networks, and form positive and negative sample pairs.

[0042] Furthermore, dividing the historical data to obtain data sets includes clustering the pre-processed historical normal operation data and historical abnormal data according to weather conditions using a clustering algorithm based on different seasons to obtain a first data set and a second data set.

[0043] The distribution network operation data is divided according to environmental factors, so that there are sufficient samples of historical normal operation data and historical abnormal operation data under various environments, which is helpful for the subsequent analysis of the operation status of the distribution network under different natural conditions.

[0044] In an embodiment of the present application, for the preprocessed historical normal operation data, the operation data corresponding to the rainstorm weather condition is divided into one cluster in the first data set, and the operation data corresponding to the high temperature condition is divided into another cluster in the first data set.

[0045] It should be noted that extracting features from data clusters and data sets through a convolutional neural network includes extracting features from the first data cluster, the second data cluster, the first data set, and the second data set, respectively.

[0046] By setting up convolutional neural network learning features, the bias of manually selected features is reduced, and the extracted features are merged into the original data.

[0047] It should also be noted that the formation of positive and negative sample pairs includes fusing the first data cluster after feature extraction and the first data set to obtain first sample data, and introducing noise into the first sample data to form second sample data; the first sample data and the second sample data constitute a positive sample pair; fusing the second data cluster after feature extraction and the second data set to obtain third sample data; the third sample data and the first sample data constitute a negative sample pair.

[0048] Fusion can directly splice the dataset after feature extraction into the data cluster to achieve data fusion and improve the robustness of the model.

[0049] In order to enhance the generalization ability of the model, noise is introduced into the first sample data to form the second sample data; the noise can be a small disturbance to the natural condition characteristics or voltage data, or a small disturbance to the natural condition characteristics and voltage data to simulate the uncertainty in the actual environment.

[0050] S3: Train the contrastive learning model through positive and negative sample pairs, output positive and negative score values, and judge whether there is an abnormality in the distribution network based on the positive and negative score values.

[0051] Furthermore, training the contrastive learning model through positive and negative sample pairs includes inputting positive sample pairs and negative sample pairs into the contrastive learning model for processing to obtain feature representations, calculating feature representation score values ​​of the positive sample pairs and negative sample pairs respectively, calculating loss values ​​based on the score values, using the gradient descent algorithm to backpropagate the loss values, updating the model parameters, and obtaining a trained contrastive learning model; the contrastive learning model is used to input real-time running data and output positive score values ​​and negative score values.

[0052] In an embodiment of the present application, the contrastive learning model adopts a Siamese network architecture, which includes two neural network branches with shared weights, respectively used to process two samples in a positive sample pair and a negative sample pair, and output feature representations respectively.

[0053] The loss function of the contrastive learning model can use contrastive loss, triplet loss, or other appropriate loss functions as the training target.

[0054] It should be noted that the loss value is the specific value of the loss function under specific parameter settings. By iteratively optimizing the parameters of the contrastive learning model, the loss value is gradually reduced until the model converges, and a trained contrastive learning model is obtained. During the training process, data enhancement technology can be used to further enhance sample diversity and improve the model generalization ability.

[0055] Data augmentation techniques include but are not limited to rotation, scaling, and translation.

[0056] It should also be noted that judging whether there is an abnormality in the distribution network based on positive and negative score values ​​includes, based on a trained contrastive learning model, collecting distribution network operation data and natural condition data in real time and inputting them into the contrastive learning model to obtain positive score values ​​and negative score values; judging whether there is an abnormality based on the positive score values ​​and negative score values; if there is an abnormality, determining the position of the abnormal point in the distribution network based on the topological structure; the positive score value is the similarity value of the positive sample pair calculated using the cosine similarity algorithm; the negative score value is the similarity value of the negative sample pair calculated using the cosine similarity algorithm.

[0057] It should also be noted that judging whether there is an abnormality based on the positive score value and the negative score value includes calculating the difference between the positive score value and the negative score value to obtain a difference value; setting a threshold value, if the difference value is greater than the threshold value, it is judged that there is no abnormality; if the difference value is less than or equal to the threshold value, it is judged that there is an abnormality.

[0058] The threshold can be adjusted according to actual needs to balance the false positive rate and the false negative rate.

[0059] It should also be noted that determining the position of the abnormal point in the distribution network according to the topological structure includes obtaining the position of the abnormal point in the distribution network by using feature inversion according to the topological structure.

[0060] In an embodiment of the present application, the location of the abnormal point in the distribution network is determined based on the topological structure, specifically, the mapping relationship between the known input features and the output score value; the input features include real-time operation data and real-time natural condition data; when there is a weak point, an abnormality location map is initialized. The abnormality location map is based on the topological structure of the distribution network, and each node or line segment is marked as a potential abnormal point, and the initial abnormality degree is set to the same or preset weight; the real-time input real-time operation data is slightly disturbed, that is, the voltage value of a certain node is adjusted, and then it is input into the comparative learning model again to observe the change in the score value; the sensitivity of the score value to the slight disturbance, that is, the gradient, is calculated through gradient descent or other optimization algorithms; the size and direction of the gradient represent the feature that has the greatest impact on the score value; the obtained gradient is used to infer the abnormal feature that causes the abnormality; in the topological map of the distribution network, the corresponding position and abnormality degree mark on the abnormality location map are adjusted according to the node or line segment corresponding to the feature.

[0061] Embodiment 2 is the second embodiment of the present invention, which differs from the first two embodiments in that:

[0062] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0063] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0064] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0065] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0066] Example 3, reference Figure 2 , which is the third embodiment of the present invention, provides a system for a computing platform load balancing method based on a particle swarm genetic algorithm, including a data clustering module, a feature extraction module, and an anomaly judgment module.

[0067] Among them, the data clustering module is used to obtain distribution network data and perform preprocessing, clustering the natural condition data through the clustering algorithm to obtain data clusters; the feature extraction module is used to divide the historical data to obtain data sets, extract features from the data clusters and data sets through the convolutional neural network, and form positive and negative sample pairs; the anomaly judgment module is used to train the comparative learning model through positive and negative sample pairs, output positive and negative score values, and judge whether there is an anomaly in the distribution network based on the positive and negative score values.

[0068] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for identifying voltage weak points in a distribution network, characterized in that: include: Obtain distribution network data and preprocess it, cluster the natural condition data using a clustering algorithm to obtain data clusters; Divide historical data into data sets, extract features from data clusters and data sets through convolutional neural networks, and form positive and negative sample pairs; The contrastive learning model is trained by positive and negative sample pairs, and positive and negative score values ​​are output. The positive and negative score values ​​are used to determine whether there is an abnormality in the distribution network.

2. The method for identifying voltage weak points in a distribution network according to claim 1, wherein: The obtaining and preprocessing of the distribution network data includes obtaining the topological structure of the distribution network, historical normal operation data, historical abnormal operation data, first natural condition data, and second natural condition data; And pre-processing; The first natural condition data is the natural condition data corresponding to the historical normal operating data; The second natural condition data is the natural condition data corresponding to the historical abnormal operating data; The historical normal operation data is the operating voltage at each node when the distribution network is operating normally in the historical data; The historical abnormal operation data is the operating voltage at each node when the distribution network operation state is abnormal in the historical data.

3. The method for identifying voltage weak points in a distribution network according to claim 2, wherein: The clustering of the natural condition data includes clustering the pre-processed first natural condition data and the second natural condition data according to weather conditions using a clustering algorithm based on different seasons to obtain a first data cluster and a second data cluster.

4. The method for identifying voltage weak points in a distribution network according to claim 3, wherein: The dividing of the historical data into data sets includes clustering the pre-processed historical normal operation data and historical abnormal data according to weather conditions using a clustering algorithm based on different seasons to obtain a first data set and a second data set.

5. The method for identifying voltage weak points in a distribution network according to claim 4, wherein: The forming of the positive and negative sample pairs includes fusing the first data cluster after feature extraction with the first data set to obtain first sample data, and introducing noise into the first sample data to form second sample data; The first sample data and the second sample data constitute a positive sample pair; Fusing the second data cluster after feature extraction and the second data set to obtain third sample data; The third sample data and the first sample data constitute a negative sample pair.

6. The method for identifying voltage weak points in a distribution network according to claim 5, wherein: The training of the contrastive learning model using positive and negative sample pairs includes inputting the positive sample pairs and the negative sample pairs into the contrastive learning model for processing to obtain feature representations, respectively calculating feature representation scores of the positive sample pairs and the negative sample pairs, calculating loss values ​​based on the score values, backpropagating the loss values ​​using a gradient descent algorithm, updating model parameters, and obtaining a trained contrastive learning model; The contrastive learning model is used to input real-time operation data and output positive score values ​​and negative score values.

7. The method for identifying voltage weak points in a distribution network according to claim 6, wherein: The determining whether there is an abnormality in the distribution network according to the positive and negative score values ​​includes collecting distribution network operation data and natural condition data in real time based on a trained comparative learning model and inputting them into the comparative learning model to obtain positive and negative score values; Determine whether there is an abnormality based on the positive and negative score values; If there is an anomaly, the location of the abnormal point in the distribution network is determined based on the topological structure; The positive score value is the similarity value of the positive sample pair calculated using the cosine similarity algorithm; The negative score value is a similarity value of the negative sample pair calculated using the cosine similarity algorithm.

8. A system using the method for identifying voltage weak points in a distribution network according to any one of claims 1 to 7, characterized in that: Including data clustering module, feature extraction module, and anomaly judgment module; The data clustering module is used to obtain and pre-process the distribution network data, and cluster the natural condition data using a clustering algorithm to obtain data clusters; The feature extraction module is used to divide the historical data into data sets, extract features from the data clusters and data sets through a convolutional neural network, and form positive and negative sample pairs; The abnormality judgment module is used to train the contrast learning model through positive and negative sample pairs, output positive and negative score values, and judge whether there is an abnormality in the distribution network based on the positive and negative score values.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for identifying a voltage weak point in a distribution network according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for identifying voltage weak points in a distribution network according to any one of claims 1 to 7 are implemented.

Images