Monitoring method, system, device and storage medium for weak links in distribution network

Abstract

The present invention discloses a monitoring method, system, device and storage medium for weak links in a distribution network. The method includes: obtaining operation-related data of a target distribution network; performing a preliminary state assessment on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix; locating weak links based on the preliminary state matrix to obtain weak link data, and generating an emergency response strategy for the target distribution network according to the weak link data. By locating weak links through the preliminary state matrix to obtain weak link data, it is possible to quickly and accurately find out the parts with problems in the distribution network, thereby improving the efficiency of problem discovery. An emergency response strategy is generated based on the weak link data, so that effective countermeasures can be quickly taken when a fault or abnormality occurs in the distribution network, thereby improving the overall operation reliability of the distribution network and ensuring the stability of power supply. The method can dynamically track changes in the distribution network and adapt to the ever-changing power grid operation environment and needs.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution networks, and particularly to a monitoring method, system, device and storage medium for weak links in a distribution network. Background Art

[0002] In modern power systems, the distribution network, as an important link connecting the transmission network and user terminals, its stable operation is crucial for ensuring the quality and reliability of power supply. With the rapid development of social economy, the power demand continues to grow, the scale of the distribution network is constantly expanding, and the structure is becoming increasingly complex. Traditional distribution network monitoring methods gradually expose many deficiencies when facing such complex and changeable situations.

[0003] Existing distribution network monitoring methods mostly rely on traditional monitoring means and lack in-depth analysis and processing of real-time data. This makes the assessment of the health status of the distribution network often lag behind the occurrence of actual problems, and it is difficult to achieve timely early warning and intervention. The traditional methods have limited data processing capabilities and cannot comprehensively consider multi-dimensional operation data, resulting in incomplete and inaccurate identification of weak links. In addition, the lack of an intelligent decision-making mechanism for weak links makes it difficult to quickly formulate emergency response strategies, affecting the overall safety and stability of the distribution network. Summary of the Invention

[0004] The present invention provides a monitoring method, system, device and storage medium for weak links in a distribution network, which solves the technical problem that the assessment of the health status of the traditional distribution network often lags behind the occurrence of actual problems and it is difficult to achieve timely early warning and intervention.

[0005] According to one aspect of the present invention, there is provided a monitoring method for weak links in a distribution network, the method comprising:

[0006] Obtaining operation-related data of a target distribution network;

[0007] Based on the operation-related data, performing a preliminary status assessment on each node of the target distribution network to generate a preliminary status matrix;

[0008] Based on the preliminary status matrix, locating weak links to obtain weak link data, and generating an emergency response strategy for the target distribution network according to the weak link data.

[0009] Optionally, obtain the operation-related data of the target distribution network, including: collecting data of each node of the target distribution network through a distributed sensor network to obtain the original operation data, where the original operation data includes voltage data, current data, temperature data, and load rate data; performing wavelet transform denoising processing on the original operation data to generate wavelet transform denoised data; performing time-series frequency change identification on the wavelet transform denoised data to generate a frequency change identification result; when the frequency change identification result is not within the preset frequency change range, generating time-series frequency change data according to the frequency change identification result, and performing adaptive sampling on the time-series frequency change data to obtain adaptive sampling data; when the frequency change identification result is within the preset frequency change range, generating data to be reconstructed according to the frequency change identification result; reconstructing the adaptive sampling data and the data to be reconstructed to obtain reconstructed data; performing normalization processing on the reconstructed data to generate operation-related data.

[0010] Optionally, perform a preliminary state assessment on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix, including: performing spectrum analysis on the operation-related data to generate a frequency domain feature set; performing frequency response analysis on each node of the target distribution network based on the frequency domain feature set to obtain a frequency response matrix; performing singular value decomposition on the frequency response matrix to obtain a reduced-dimensional feature vector set; performing trend analysis on the reduced-dimensional feature vector set to obtain a trend analysis result; using a preset long short-term memory network to perform similarity prediction and matching on the trend analysis result to obtain a similarity matrix; performing state correlation analysis on each node of the target distribution network based on the similarity matrix to obtain a state correlation matrix; using a Bayesian network to perform health state probability estimation on each node of the target distribution network based on the state correlation matrix to obtain the state assessment result corresponding to each node degree; constructing a preliminary state matrix according to the state assessment result corresponding to each node.

[0011] Optionally, perform weak link positioning based on the preliminary state matrix to obtain weak link data, including: performing structured processing on the preliminary state matrix through a preset topology analysis algorithm to obtain the topology structure matrix of the target distribution network, where the topology structure matrix includes the connection relationship between each node, the branch impedance parameter corresponding to each node, and the electrical characteristics of each node; performing network modeling on the target distribution network based on the topology structure matrix to obtain a network feature vector; performing weak link analysis on the target distribution network according to the network feature vector to obtain a node vulnerability index; performing comprehensive evaluation on each node of the target distribution network based on the node vulnerability index to obtain the risk score of each node; performing spatial clustering analysis on the target distribution network according to the risk score of each node through a self-organizing mapping neural network algorithm to obtain a risk hot spot map; determining the weak link position in the risk hot spot map through a preset risk assessment algorithm, and performing weak link monitoring at the weak link position to obtain weak link data.

[0012] Optionally, a spatial clustering analysis is performed on the target distribution network according to the node risk scores through the self-organizing mapping neural network algorithm to obtain a risk hot spot map, including: performing a non-linear dimensionality reduction mapping on the node risk scores through the self-organizing mapping neural network algorithm to obtain the risk distribution of each node on the mapping grid; using the kernel density estimation method to calculate the relative risk density of each node within the corresponding neighborhood based on the node risk distribution to obtain the node risk density map; performing a local depth analysis on the node density distribution in the node risk density map to obtain a preliminary risk area division map, where the preliminary risk area division map is used to distinguish different risk level areas in the target distribution network; optimizing the risk area boundary of the preliminary risk area division map to obtain the risk hot spot map.

[0013] Optionally, an emergency response strategy for the target distribution network is generated based on the weak link data, including: determining the weak link type corresponding to the weak link data; performing semantic parsing on the weak link type to obtain a semantic feature vector; screening the historical emergency strategies in the preset database based on the semantic feature vector to obtain a candidate strategy set; generating an emergency response strategy according to the candidate strategy set.

[0014] Optionally, determining the weak link type corresponding to the weak link data includes: extracting features from the weak link data to obtain a weak link vector, where the weak link vector includes a voltage fluctuation vector, a harmonic content vector, a power factor vector, and a load imbalance degree vector; generating a time-frequency feature vector according to the weak link vector and performing non-linear dynamic modeling on the time-frequency feature vector to obtain a dynamic modeling vector; performing a fault mode recognition on the weak link location based on the dynamic modeling vector through a preset fuzzy inference algorithm to obtain a fault mode recognition result; performing dynamic classification on the fault mode recognition result to obtain the weak link type.

[0015] According to another aspect of the present invention, a monitoring system for weak links in a distribution network is provided, and the system includes:

[0016] An operation-related data acquisition module for acquiring operation-related data of the target distribution network;

[0017] A preliminary state matrix generation module for performing a preliminary state assessment on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix;

[0018] An emergency response strategy generation module for performing weak link location based on the preliminary state matrix to obtain weak link data, and generating an emergency response strategy for the target distribution network according to the weak link data.

[0019] According to another aspect of the present invention, an electronic device is provided, and the electronic device includes:

[0020] At least one processor;

[0021] and a memory communicatively connected to the at least one processor;

[0022] Wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute a method for monitoring weak links in a distribution network according to any embodiment of the present invention.

[0023] According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement a method for monitoring weak links in a distribution network according to any embodiment of the present invention when executed.

[0024] The technical solution of the embodiment of the present invention locates weak links through a preliminary state matrix to obtain weak link data, which can quickly and accurately find the problematic parts in the distribution network, improving the efficiency of problem discovery. Generating an emergency response strategy based on the weak link data enables effective countermeasures to be taken quickly when a fault or anomaly occurs in the distribution network, enhancing the overall operational reliability of the distribution network and ensuring the stability of power supply. It can dynamically track the changes in the distribution network and adapt to the changing grid operating environment and requirements.

[0025] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0027] Figure 1 is a flowchart of a method for monitoring weak links in a distribution network according to Embodiment 1 of the present invention;

[0028] Figure 2 is a flowchart of another method for monitoring weak links in a distribution network according to Embodiment 2 of the present invention;

[0029] Figure 3 is a schematic structural diagram of a monitoring system for weak links in a distribution network according to Embodiment 3 of the present invention;

[0030] Figure 4It is a schematic structural diagram of an electronic device for implementing a monitoring method for weak links in a distribution network according to an embodiment of the present invention. Detailed implementation manners

[0031] Embodiment 1

[0032] Figure 1 FIG. 9 is a flowchart of a monitoring method for weak links in a distribution network provided in Embodiment 1 of the present invention. This embodiment is applicable to the situation of monitoring weak links in a distribution network. The method can be executed by a monitoring device for weak links in a distribution network. The monitoring device for weak links in a distribution network can be implemented in the form of hardware and / or software, and the monitoring device for weak links in a distribution network can be configured in a computer controller. As Figure 1 shown, the method includes:

[0033] S110. Obtain operation-related data of a target distribution network.

[0034] Among them, a distribution network refers to a power network that receives electric energy from a transmission network or a regional power plant and distributes it locally through distribution facilities or step by step according to voltage levels to various users. The target distribution network can be a certain area of a city, an industrial park, or the part of the distribution network corresponding to a specific user group. The operation-related data refers to data that can reflect various states and characteristics of the distribution network during operation. Each node of the target distribution network can represent different electrical components of the target distribution network.

[0035] Optionally, obtaining operation-related data of the target distribution network includes: collecting each node of the target distribution network through a distributed sensor network to obtain original operation data, where the original operation data includes voltage data, current data, temperature data, and load rate data; performing wavelet transform denoising processing on the original operation data to generate wavelet transform denoised data; performing time-series frequency change identification on the wavelet transform denoised data to generate a frequency change identification result; when the frequency change identification result is not within a preset frequency change range, generating time-series frequency change data according to the frequency change identification result, and performing adaptive sampling on the time-series frequency change data to obtain adaptive sampling data; when the frequency change identification result is within the preset frequency change range, generating data to be reconstructed according to the frequency change identification result; reconstructing the adaptive sampling data and the data to be reconstructed to obtain reconstructed data; and performing standardization processing on the reconstructed data to generate operation-related data.

[0036] Specifically, the distributed sensor network can be widely distributed on each node of the target distribution network to monitor key parameters such as voltage, current, temperature, and load rate in real time. Among them, the voltage data reflects the voltage level of each node in the distribution network; the current data provides information about the magnitude and direction of the current in the distribution network; the temperature data represents the temperature change of the device; and the load rate data represents the load degree of the device or line

[0037] Furthermore, when collecting data, a variety of sensors and monitoring devices need to be deployed, and these devices will be distributed on each node of the target distribution network to collect multi-dimensional operation data such as current, voltage, temperature, and load rate in real time. By integrating these sensors, each electrical component of the distribution network can continuously feedback its health status under different operating conditions, thus forming a comprehensive operation data network. For example, in a specific distribution network, intelligent sensors can be installed at key positions such as transformers, switchgear, and distribution lines. These sensors will monitor the load and working status of electrical components in real time and transmit the data to the central processing system through a wireless network. The system can integrate and analyze the collected multi-dimensional data to provide a detailed view of the operation of the distribution network. Such multi-dimensional operation data not only includes the basic electrical parameters of each node but also can combine environmental factors, such as the impact of temperature changes on the performance of electrical equipment, etc., so as to more comprehensively reflect the health status of the distribution network. In an actual application scenario, assuming that during the peak electricity consumption period of a city distribution network, the sensor can monitor in real time that the temperature of a certain transformer rises abnormally. This information will be recorded in time as an important part of the multi-dimensional operation data. By comparing these real-time data with historical data, the system can quickly judge whether the node is in a dangerous state and incorporate the relevant information into the construction of the preliminary state matrix. This process not only improves the efficiency of data collection but also ensures the real-time and accuracy of the data, laying a solid foundation for subsequent risk assessment and monitoring of weak links. Therefore, through effective real-time data collection, the target distribution network can identify potential risks in a timely manner in a dynamically changing environment, ensuring the stability and security of power supply.

[0038] Among them, wavelet transform is a signal processing technology that can decompose a signal into components of different frequencies and time scales. By selecting an appropriate wavelet basis function, the noise in the signal can be effectively removed while retaining useful information. Specifically, the wavelet transform denoising process can be carried out using the following formula (1):

[0039] (1)

[0040] Among them, represents the wavelet transform denoised data, represents the original multi-dimensional operation data, represents the dimension, represents the total number of dimensions, represents the scale of the wavelet transform, represents the maximum scale of the wavelet transform, represents the time shift index of the wavelet transform, represents at the Dimension, the scale, and the wavelet coefficients under translation, represent the scale and the wavelet basis functions under translation, represents the threshold function, which is the threshold function for denoising the wavelet coefficients, represents the threshold parameter for denoising.

[0041] Furthermore, after obtaining the wavelet transform denoised data, a time series analysis method can be used to analyze the wavelet transform denoised data and extract the frequency change characteristics. For example, tools such as Fourier transform and wavelet analysis can be used to identify the frequency change trends and periodicities in the data. Specifically, the following formula (2) can be used to obtain the frequency change recognition result:

[0042] (2)

[0043] where, represents the frequency change recognition result, is the wavelet transform denoised data, represents the frequency index, represents the total number of frequency indices, represents the imaginary unit, satisfying , represents the th frequency value, represents the Dirac delta function, capturing the frequency , represents the threshold function, which is the threshold function for identifying the frequency components, represents the threshold parameter for the identification process.

[0044] Specifically, when the frequency change recognition result is not within the preset frequency change range, according to the frequency change recognition result, determine the frequency change part that exceeds the range, generate the time series frequency change data, and perform adaptive sampling on the time series frequency change data, adjusting the sampling frequency according to the change characteristics of the data to obtain more accurate information. Adaptive sampling can improve the resolution and accuracy of the data, and is helpful for better analyzing and processing abnormal situations. When the frequency change recognition result is within the preset frequency change range, determine the data part within the normal range and generate the data to be reconstructed. These data can be directly used for the subsequent reconstruction process, or reconstructed together with the adaptive sampling data. Specifically, the following formula (3) can be used for adaptive sampling:

[0045] (3)

[0046] where, represents the adaptive sampling data, represents the time point of the time-series frequency change data, represents the previous time point, represents the time-series frequency change data, that is, the frequency change recognition result, represents the total number of frequency indices, represents the frequency index, represents the Dirac delta function, capturing time , represents the derivative of the time-series frequency change data, represents the adjustment parameter of the adaptive sampling. Specifically, the following formula (4) can be used to reconstruct the adaptive sampling data and the data to be reconstructed:

[0047] (4)

[0048] where, represents the reconstructed data, is the time point of the reconstructed data, is the total number of interpolation basis functions, is the interpolation coefficient, represents the th interpolation basis function, represents the adaptive sampling data, represents the time-series frequency change data, represents the modulation function, represents the imaginary unit, satisfying represents the frequency change data. Reconstructing the adaptive sampling data and the data to be reconstructed obtains the complete and processed operation-related data. The reconstruction process can restore the integrity and accuracy of the data while removing the influence of noise and outliers.

[0049] Specifically, the normalization process can transform the reconstructed data into a form with a unified scale and distribution, facilitating data analysis and comparison. The normalization process can eliminate the dimensional and unit differences of the data, improving the comparability and generality of the data. The normalization methods can include mean normalization, standard deviation normalization, etc. By calculating the mean and standard deviation of the data, the data is normalized so that the mean of the data is 0 and the standard deviation is 1.

[0050] Exemplarily, in a specific distribution network application scenario, when the system real-time monitors that the current and temperature values of a certain transformer show abnormal fluctuations, these data will be transmitted to the distributed sensor network for collection. After wavelet transform denoising processing, the system can extract clear current and temperature change data and perform frequency change identification. If it is identified that the current frequency change exceeds the preset range, the system will further analyze this data to confirm whether the transformer is in an overloaded state and perform adaptive sampling to obtain relevant high-frequency and low-frequency data. By reconstructing these important data and performing normalization processing, the finally formed multi-dimensional operation data can accurately reflect the operation status of the transformer, providing an important basis for subsequent risk assessment and decision-making. Such a data processing flow ensures the safe and stable operation of the distribution network, improves the system's response ability and intelligent decision-making ability, and thus provides strong support for the reliability of power supply.

[0051] S120. Perform a preliminary state assessment on each node of the target distribution network based on operation-related data to generate a preliminary state matrix.

[0052] Among them, the elements of the preliminary state matrix represent the multi-dimensional operation data of different electrical components at a specific time point. The rows of the preliminary state matrix represent different electrical components of the target distribution network, and the columns represent the categories of multi-dimensional operation data. The preliminary state matrix reflects the health degree of each node at the current moment.

[0053] Optionally, performing a preliminary state assessment on each node of the target distribution network based on operation-related data to generate a preliminary state matrix includes: performing spectrum analysis on the operation-related data to generate a frequency domain feature set; performing frequency response analysis on each node of the target distribution network based on the frequency domain feature set to obtain a frequency response matrix; performing singular value decomposition on the frequency response matrix to obtain a reduced-dimensional feature vector set; performing trend analysis on the reduced-dimensional feature vector set to obtain a trend analysis result; using a preset long short-term memory network to perform similarity prediction and matching on the trend analysis result to obtain a similarity matrix; performing state correlation analysis on each node of the target distribution network based on the similarity matrix to obtain a state correlation matrix; using a Bayesian network to perform health state probability estimation on each node of the target distribution network based on the state correlation matrix to obtain the state assessment result corresponding to each node degree; constructing a preliminary state matrix according to the state assessment result corresponding to each node.

[0054] Among them, spectrum analysis is a method of converting operation-related data in the time domain to the frequency domain for analysis. Through tools such as Fourier transform, the signal can be decomposed into components of different frequencies, thereby revealing the frequency characteristics of the signal. Specifically, the following formula (5) can be used to calculate the frequency domain feature set:

[0055] (5)

[0056] Among them, is an element of the frequency-domain feature set, representing the row and the column of the frequency-domain feature at the frequency . represents the Fourier transform operator, is an element of the multi-dimensional running data matrix, representing the row and the column of the data at the time . is a specific time point, in represents the imaginary unit, satisfying . represents the frequency component of the frequency-domain feature, represents the total number of frequency components, represents the th amplitude coefficient, represents the th angular frequency, represents the th phase shift, represents the th amplitude coefficient, represents the th angular frequency, represents the th phase shift.

[0057] Specifically, frequency response analysis is to study the response characteristics of a system to input signals of different frequencies. Based on the frequency-domain feature set, frequency response analysis is performed on each node of the target distribution network to obtain a frequency response matrix. Specifically, the frequency response matrix can be calculated using the following formula (6):

[0058] (6)

[0059] Among them, is the frequency response matrix, represents an element of the frequency response matrix, representing the frequency response value of the row and the column, represents the index of the node in the target distribution network, represents the index of the frequency-domain feature set, represents the total number of nodes in the target distribution network, represents the total number of the frequency-domain feature set.

[0060] Specifically, singular value decomposition is a matrix decomposition method that can decompose a matrix into the product of three matrices. Through singular value decomposition, the main feature information of the matrix can be extracted while realizing the dimensionality reduction of the data. The dimensionality-reduced eigenvector set can be calculated specifically using the following formula (7):

[0061] (7)

[0062] where, represents the frequency response matrix, represents the left singular matrix, represents the singular value matrix, represents the right singular matrix, represents the rank after dimensionality reduction, represents the first columns of the left singular matrix, represents the first columns of the right singular matrix, represents the diagonal matrix composed of the first singular values of the singular value matrix, represents the frequency response matrix after dimensionality reduction.

[0063] Specifically, trend analysis is to analyze the time series of data to find the changing trends and patterns of the data. In the state assessment of the distribution network, trend analysis can help discover the changing trends of node states and predict possible problems. Using a preset long short-term memory network to perform similarity prediction and matching on the trend analysis results to obtain a similarity matrix, the similarity matrix can be calculated specifically using the following formula (8):

[0064] (8)

[0065] where, represents the similarity matrix, represents the activation function (such as sigmoid), represents the total number of time steps, represents the time step weight, tanh represents the hyperbolic tangent activation function, represents the hidden state of the previous time step, represents the weight matrix of the hidden state, represents the input of the current time step, represents the weight matrix of the input, b represents the bias vector, A represents the first trend analysis result matrix, B represents the second trend analysis result matrix, C represents the third trend analysis result matrix, D represents the fourth trend analysis result matrix, d represents the normalization factor, and o represents the element-wise multiplication (Hadamard product).

[0066] Specifically, state correlation analysis is to find the correlation relationships between nodes by analyzing the correlations between the states of each node. The state correlation matrix can be calculated using the following formula (9):

[0067] (9)

[0068] Among them, represents the state correlation matrix, represents the element in the th row and th column of the similarity matrix, represents the element in the th row and th column of the state influence factor matrix, represents the element in the th row and th column of the state correlation calculation matrix, represents the element in the th row and th column of the state correlation calculation matrix, represents the normalization factor, corresponding to the th column, represents the total number of matching operations, represents the total number of nodes in the target distribution network, represents the total number of categories of multi-dimensional operation data.

[0069] Specifically, a Bayesian network is a probabilistic graphical model that can estimate the probabilities of unknown variables based on known variable relationships and prior knowledge. In the state assessment of a distribution network, the Bayesian network can use the information in the state correlation matrix, combined with prior knowledge and observed data, to estimate the probabilities of the health states of each node. Then, the state assessment results of each node are arranged in the order of the nodes to form a matrix. This matrix is the preliminary state matrix, which reflects the preliminary state assessment of each node in the target distribution network. The following formula (10) shows an example of the preliminary state matrix:

[0070] (10)

[0071] Among them, is an element of the preliminary state matrix, representing the multi-dimensional operation data at the th row and th column at time , represents the index of the electrical components in the target distribution network, represents the category index of the multi-dimensional operation data, represents a specific time point, represents the total number of nodes in the target distribution network, represents the total number of categories of multi-dimensional operation data.

[0072] Specific application scenario: Based on the operation-related data, a preliminary state assessment is carried out for each node of the target distribution network. First, spectrum analysis is required to extract the frequency-domain feature set. This process can reveal the characteristics of the nodes at different frequencies by transforming the time-domain signal. For example, if the spectrum analysis of a certain transformer shows abnormal frequency components, this may imply potential faults. After obtaining the frequency-domain features, frequency response analysis is carried out to form a frequency response matrix. This matrix provides detailed information on the responses of each node to different frequencies and can help identify which nodes perform abnormally at specific frequencies. Then, by performing singular value decomposition on the frequency response matrix, a set of reduced-dimensional feature vectors can be obtained, which helps simplify the data and highlight important changes in the operating state. Subsequently, trend analysis is carried out on the set of reduced-dimensional feature vectors to obtain the trend analysis results, so as to reveal the changing trends of the operating states of each node at different time periods. For example, if the operating state of a certain node shows an upward trend, this may indicate that it is gradually approaching the overload state, and vice versa. At this time, a preset long short-term memory network is used to perform similarity prediction and matching on the trend analysis results to form a similarity matrix, which can capture the relationship between the state changes of nodes. Finally, based on the similarity matrix, state correlation analysis is carried out on each node of the target distribution network to obtain a state correlation matrix. This matrix helps identify which nodes are correlated in terms of health status and whether other nodes will be affected if a certain node fails. Using a Bayesian network, based on the state correlation matrix, probability estimation of the health status of each node is carried out to obtain the corresponding state assessment results for each node. This comprehensive analysis method ensures the accuracy and timeliness of distribution network monitoring and provides a scientific basis for operation and maintenance decisions. In this way, potential risks of the distribution network can be effectively warned and prevented, ensuring its stable operation.

[0073] S130. Locate weak links based on the preliminary state matrix to obtain weak link data, and generate an emergency response strategy for the target distribution network according to the weak link data.

[0074] Among them, weak link location refers to finding the relatively vulnerable parts that are prone to problems in the distribution network based on the preliminary state matrix. The emergency response strategy refers to the countermeasures formulated for possible emergencies in the distribution network. It includes isolation measures during a fault, load transfer plans, repair plans, emergency power supply arrangements, etc. The purpose is to restore the normal operation of the distribution network in the shortest time and reduce the impact on users.

[0075] Optionally, weak link positioning is performed based on the preliminary state matrix to obtain weak link data, including: structuring the preliminary state matrix through a preset topology analysis algorithm to obtain the topology structure matrix of the target distribution network, where the topology structure matrix includes the connection relationships between nodes, the branch impedance parameters corresponding to each node, and the electrical characteristics of each node; performing network modeling on the target distribution network based on the topology structure matrix to obtain network feature vectors; performing weak link analysis on the target distribution network according to the network feature vectors to obtain node vulnerability indicators; comprehensively evaluating each node of the target distribution network based on the node vulnerability indicators to obtain the risk scores of each node; performing spatial clustering analysis on the target distribution network according to the risk scores of each node through a self-organizing mapping neural network algorithm to obtain a risk hot spot map; determining the weak link positions in the risk hot spot map through a preset risk assessment algorithm, and monitoring the weak links at the weak link positions to obtain weak link data.

[0076] Specifically, the topology analysis algorithm can determine the connection relationships between nodes, the impedance parameters of branches, and the electrical characteristics of nodes by traversing each node and branch of the distribution network. Specifically, the following formula (11) can be used to calculate the topology structure matrix of the target distribution network:

[0077] (11)

[0078] Among them, T is the topology structure matrix, Zij represents the branch impedance between the th node and the th node, represents the current of the th node, represents the voltage of the th node, represents the power of the th node, represents the connection relationship between the th node and the th node, represents the total number of nodes.

[0079] Specifically, network modeling is to abstract the distribution network into a mathematical model for analysis and calculation. Specifically, through graph spectrum theory technology, network modeling can be performed on the target distribution network based on the topology structure matrix, as shown in the following formula (12):

[0080] (12)

[0081] Among them, represents the network feature vector, H represents the feature matrix, the total number of nodes, is the The element in the th row and th column represents the connection relationship between node and node and represents the element in the th row and th column of the degree matrix, indicating the degree of node . Represents the eigenvector of node . Represents the regularization parameter, which controls the difference between the feature matrix H and the node eigenvector . Represents the regularization parameter, which controls the difference between the feature matrix H and the network eigenvector .

[0082] Specifically, the node vulnerability index can be calculated using the following formula (13):

[0083] (13)

[0084] where represents the vulnerability index of node , is the element in the th row and th column of the topological structure matrix, indicating the connection relationship between node and node , represents the network eigenvector of node , represents the total number of nodes, represents the regularization parameter for controlling the power impact, indicates the power of node on the th path, indicates the influence coefficient of path on node , represents the regularization parameter for controlling the impact of current and voltage, Ii The current of node , represents 's voltage, L represents the total number of normalization factors, represents the element in the th row and th column of the normalization matrix, represents the element in the th column of the l-th row of the normalization matrix.

[0085] Specifically, the risk scores of each node can be calculated using the following formula (14):

[0086] (14)

[0087] Among them, represents the risk score of node i, represents the adjacency matrix and the vulnerability coordinate weight parameter, the power path and the influence coefficient weight parameter, represents the current and voltage weight parameters, represents the temperature weight parameter, the element of the adjacency matrix, indicating the connection relationship between node and node ; is the vulnerability index of node ; the total number of nodes, is the power of node on the th path, represents the path 's influence coefficient on node ; represents the total number of power paths, represents the element in the l-th column of the th row of the normalization matrix, represents the element in the th row and the th column of the normalization matrix. L represents the total number of normalization factors, Ii represents the current of node ; represents the voltage of node ; represents the geographical location feature vector of node ; is the influence coefficient of the communication feature on node ; represents the total number of temperature feature vectors.

[0088] Among them, the self-organizing map neural network is an unsupervised learning algorithm that can map high-dimensional data to a low-dimensional space and maintain the topological structure of the data. In the risk analysis of the distribution network, the self-organizing map neural network can perform spatial clustering analysis based on the risk scores of each node, clustering nodes with similar risk levels together.

[0089] Specifically, the risk scores of each node are used as input data and input into the self-organizing mapping neural network. Through competitive learning and weight adjustment, the network gradually forms clusters of the input data. Each cluster represents a risk level or area. Visualize the clustering results to obtain a risk hot spot map. The risk hot spot map is usually presented in the form of a map, with different colors or patterns representing different risk levels. High-risk areas appear as hot spots in the map, that is, the risk hot spot map. The risk hot spot map includes high-risk areas, medium-risk areas, and low-risk areas. Finally, by real-time positioning the high-risk areas and medium-risk areas, the location of weak links is obtained, using the following formula (15):

[0090] (15)

[0091] where represents the location of weak links, represents the weight parameter of the overall risk assessment, and represent the weight parameters of high-risk and medium-risk areas respectively, and represent the weight parameters of power path and temperature characteristics respectively, represents node 's high-risk score, represents node 's medium-risk score, represents node 's vulnerability index, represents the total number of nodes, represents node 's power, represents node 's influence coefficient on the risk assessment algorithm, represents node 's normalization factor, represents node 's normalization factor, represents node 's geographical location feature vector, represents node 's communication feature influence coefficient.

[0092] Specific application scenario: First, it is necessary to use a preset topology analysis algorithm to structurally process the preliminary state matrix to obtain the topology structure matrix of the target distribution network. This topology structure matrix contains the connection relationships between nodes, the branch impedance parameters corresponding to each node, and electrical characteristics, which lay an important foundation for subsequent risk assessment. By analyzing these structural features, the system can clearly identify the overall layout of the distribution network and the interactions between nodes, thereby providing a basis for locating weak links. After forming the topology structure matrix, the system will then apply graph theory technology to perform network modeling on the target distribution network to obtain network feature vectors. At this time, the network feature vectors not only describe the basic features of the network but also reveal the connection strength and interdependence between nodes. Based on these network feature vectors, the system will conduct a weak link analysis on the target distribution network to obtain node vulnerability indicators. The calculation of node vulnerability indicators will reflect the ability of each node to withstand stress in the network. The higher the vulnerability indicator, the more likely the node is to become a weak link in the system. Based on the node vulnerability indicators, the system will comprehensively evaluate each node of the target distribution network to obtain the risk scores of each node. This process combines multiple evaluation dimensions to ensure the accuracy and reliability of the final risk scores. After obtaining the risk scores of each node, the system will use the self-organizing mapping neural network algorithm to perform spatial clustering analysis on these scores to generate a risk hot spot map. The risk hot spot map divides the distribution network into high-risk areas, medium-risk areas, and low-risk areas, enabling managers to intuitively understand which areas may have potential problems. After that, the system will once again use the preset risk assessment algorithm to perform real-time positioning on the high-risk areas and medium-risk areas to obtain the locations of the weak links. This process ensures that in actual operation, managers can quickly identify the areas that need to be monitored closely. After confirming the locations of the weak links, the system will monitor these locations for weak links to collect more detailed data. Through monitoring, managers can obtain weak link data in a timely manner, thereby taking corresponding measures to avoid potential risks affecting the safety and stability of the distribution network. For example, in the distribution network of a city, if the nodes in certain areas show high vulnerability indicators due to aging, through the above steps, the system can quickly identify these high-risk nodes and conduct real-time monitoring to ensure that these weak links do not cause large-scale power outages or other safety problems under high load or extreme weather conditions. This method not only improves the management efficiency of the distribution network but also enhances its reliability, providing effective support for practical applications.

[0093] Optionally, perform spatial clustering analysis on the target distribution network according to the risk scores of each node through the self-organizing map neural network algorithm to obtain a risk hot spot map, including: performing non-linear dimensionality reduction mapping on the node risk scores through the self-organizing map neural network algorithm to obtain the risk distribution of each node on the mapping grid; using the kernel density estimation method to calculate the relative risk density of each node within the corresponding neighborhood based on the node risk distribution to obtain the node risk density map; performing local depth analysis on the node density distribution in the node risk density map to obtain a preliminary risk area division map, where the preliminary risk area division map is used to distinguish different risk level areas in the target distribution network; optimizing the risk area boundary of the preliminary risk area division map to obtain the risk hot spot map.

[0094] Specifically, the self-organizing map (SOM) neural network is an unsupervised learning algorithm that can map high-dimensional data to a low-dimensional space (usually a two-dimensional grid) while preserving the topological structure of the data. The controller can input the risk scores of each node as input vectors into the input layer of the SOM. Then, each node in the output layer is compared with the input vector, and the distance between them (usually using the Euclidean distance) is calculated. The output node with the smallest distance (referred to as the winning node) is activated. The weights of the winning node and its neighborhood nodes are adjusted according to the input vector so that they are closer to the input vector. This process is repeated continuously until the node weights in the output layer stabilize. Finally, each input vector (node risk score) is mapped to a node in the output layer, thereby obtaining the risk distribution of each node on the mapping grid, as shown in the following formula (16):

[0095] (16)

[0096] Where, represents the risk distribution of node and node on the mapping grid, represents the non-linear activation function, is the total number of node features, is the activation function in the self-organizing map neural network, is the dimension of the input features, is the weight parameter of the self-organizing map neural network, is node 's l-th risk score, is an element of the adjacency matrix, indicating the connection relationship between node and node , is node 's vulnerability index.

[0097] Specifically, kernel density estimation is a non-parametric estimation method used to estimate the probability density function of a random variable. Kernel density estimation estimates the probability density by placing a kernel function around each data point and summing all the kernel functions. When calculating the relative risk density, for each node on the mapped grid (corresponding to a risk score), its neighborhood range is determined. The neighborhood range can be a fixed radius or an adjacent area. Within the neighborhood of each node, the kernel density estimation method is used to calculate the risk density of that node. Specifically, for each node within the neighborhood, the distance between it and the current node is calculated, and its contribution to the risk density of the current node is calculated according to the kernel function. The contributions of all nodes within the neighborhood are added together to obtain the relative risk density of the current node. This density value reflects the relative risk magnitude of the node within its neighborhood. The above process is repeated for all nodes on the mapped grid to obtain the risk density map of each node. The risk density map shows the relative risk density of each node, and the areas with darker colors represent the areas with higher risk density. Specifically, the risk density map of each node can be calculated using the following formula (17):

[0098] (17)

[0099] Wherein, represents the risk density of node at position , represents the total number of nodes, h represents the bandwidth of kernel density estimation, represents the kernel function, represents the position of node , represents node and node the risk relationship between them.

[0100] Specifically, local depth analysis is a data analysis method used to explore the local structure and distribution of data. In the risk density map, local depth analysis can help determine the boundary between the areas with higher risk density and the areas with lower risk density. The local depth can be determined by calculating the distance from the node to other nodes within its neighborhood. Nodes with larger distances have higher local depths, meaning they are relatively isolated within the local area. According to the local depth of the nodes, the risk density map is divided into different regions. Generally, nodes with larger local depths are considered to be in the areas with lower risk density, while nodes with smaller local depths are considered to be in the areas with higher risk density.

[0101] However, the preliminary risk area division map may have problems such as unclear or inaccurate boundaries, so it is necessary to refine and adjust the boundaries in the preliminary risk area division map. Image processing techniques, such as edge detection and morphological operations, can be used to improve the quality of the boundaries. The boundary of the risk area can be optimized specifically using the following formula (18):

[0102] (18)

[0103] Wherein, represents the risk hot spot map, represents the boundary of the risk area, is the risk value of the th row and th column in the preliminary risk area division map, is the boundary optimization function, based on the boundary and the distance function , is the boundary condition of the th row and th column, represents the th row and th column of the distance function, , are the number of rows and columns of the preliminary risk area division map.

[0104] Specific application scenario: First, the controller can use the self-organizing mapping neural network algorithm to perform non-linear dimensionality reduction mapping on the node risk scores, thereby generating the risk distribution of each node on the mapping grid. This mapping process can transform high-dimensional risk score data into a more understandable two-dimensional or three-dimensional form, making the position of each node in the risk distribution more explicit. This result will provide an intuitive basis for subsequent risk analysis. After obtaining the risk distribution of each node, the system will use the kernel density estimation method to calculate the relative risk density of each node within the corresponding neighborhood based on these risk distributions, thereby forming a risk density map of each node. Through this process, the system can identify the concentration degree of risks in different regions, and then reflect the risk levels that certain nodes may face in the overall network. For example, in a certain distribution network area, if the node density map shows a high-risk aggregation, it indicates that there may be potential problems such as equipment aging and overloading in this area. Then, the system will perform a local in-depth analysis on the node density distribution in the risk density map of each node to obtain a preliminary risk area division map. This division map aims to distinguish different risk level areas in the target distribution network, enabling operation and maintenance personnel to quickly identify high-risk, medium-risk, and low-risk areas. In this process, the formation of the preliminary risk area division map is based on the comprehensive evaluation of the risk density and node status of each area, thus ensuring the accuracy of the division. After obtaining the preliminary risk area division map, the system will optimize the boundaries of the risk areas to obtain the final risk hot spot map. This step will further improve the accuracy of the area division, ensure that the boundaries of each risk level area are more in line with the actual situation, and enable more effective monitoring and management in actual operation and maintenance. For example, when operation and maintenance personnel find that a certain area is marked as a high-risk area according to the risk hot spot map, they can immediately take measures such as increasing the detection frequency or performing equipment maintenance to prevent potential failures. In summary, through the self-organizing mapping neural network algorithm, kernel density estimation, and local in-depth analysis, the system can achieve a comprehensive risk assessment and hot spot monitoring of the distribution network, ensure the safe and stable operation of the distribution network, and provide a scientific basis for operation and maintenance decisions.

[0105] The technical solution of the embodiment of the present invention can quickly and accurately find the problematic parts in the distribution network by locating weak links through the preliminary state matrix, improving the efficiency of problem discovery. Generate an emergency response strategy based on the weak link data, enabling effective countermeasures to be taken quickly when a failure or abnormality occurs in the distribution network, enhancing the overall operation reliability of the distribution network, and ensuring the stability of power supply. It can dynamically track the changes in the distribution network and adapt to the changing power grid operation environment and requirements.

[0106] Embodiment 2

[0107] Figure 2The flowchart of a monitoring method for weak links in a distribution network provided in the second embodiment of the present invention. In this embodiment, on the basis of the above-mentioned first embodiment, the specific process of generating an emergency response strategy for the target distribution network according to the weak link data is added. Among them, the specific contents of steps S210 - S220 are substantially the same as those of steps S110 - S120 in the first embodiment, so they will not be elaborated in this embodiment. As Figure 2 shown, the method includes:

[0108] S210. Obtain the operation-related data of the target distribution network.

[0109] Optionally, obtaining the operation-related data of the target distribution network includes: collecting each node of the target distribution network through a distributed sensor network to obtain original operation data, where the original operation data includes voltage data, current data, temperature data, and load rate data; performing wavelet transform denoising processing on the original operation data to generate wavelet transform denoised data; performing time-series frequency change identification on the wavelet transform denoised data to generate a frequency change identification result; when the frequency change identification result is not within the preset frequency change range, generating time-series frequency change data according to the frequency change identification result, and performing adaptive sampling on the time-series frequency change data to obtain adaptive sampling data; when the frequency change identification result is within the preset frequency change range, generating data to be reconstructed according to the frequency change identification result; reconstructing the adaptive sampling data and the data to be reconstructed to obtain reconstructed data; performing standardization processing on the reconstructed data to generate operation-related data.

[0110] S220. Perform a preliminary state assessment on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix.

[0111] Optionally, performing a preliminary state assessment on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix includes: performing spectrum analysis on the operation-related data to generate a frequency domain feature set; performing frequency response analysis on each node of the target distribution network based on the frequency domain feature set to obtain a frequency response matrix; performing singular value decomposition on the frequency response matrix to obtain a reduced-dimensional feature vector set; performing trend analysis on the reduced-dimensional feature vector set to obtain a trend analysis result; using a preset long short-term memory network to perform similarity prediction and matching on the trend analysis result to obtain a similarity matrix; performing state correlation analysis on each node of the target distribution network based on the similarity matrix to obtain a state correlation matrix; using a Bayesian network to perform health state probability estimation on each node of the target distribution network based on the state correlation matrix to obtain a state assessment result corresponding to each node degree; constructing a preliminary state matrix according to the state assessment result corresponding to each node.

[0112] S230. Locate weak links based on the preliminary state matrix to obtain weak link data.

[0113] Optionally, weak link positioning is performed based on the preliminary state matrix to obtain weak link data, including: structuring the preliminary state matrix through a preset topology analysis algorithm to obtain the topology structure matrix of the target distribution network, where the topology structure matrix includes the connection relationships between nodes, the branch impedance parameters corresponding to each node, and the electrical characteristics of each node; performing network modeling on the target distribution network based on the topology structure matrix to obtain network feature vectors; performing weak link analysis on the target distribution network according to the network feature vectors to obtain node vulnerability indicators; comprehensively evaluating each node of the target distribution network based on the node vulnerability indicators to obtain the risk scores of each node; performing spatial clustering analysis on the target distribution network according to the risk scores of each node through a self-organizing mapping neural network algorithm to obtain a risk hot spot map; determining the weak link positions in the risk hot spot map through a preset risk assessment algorithm, and monitoring the weak links at the weak link positions to obtain weak link data.

[0114] Optionally, spatial clustering analysis is performed on the target distribution network according to the risk scores of each node through a self-organizing mapping neural network algorithm to obtain a risk hot spot map, including: performing non-linear dimensionality reduction mapping on the risk scores of the nodes through a self-organizing mapping neural network algorithm to obtain the risk distribution of each node on the mapping grid; using the kernel density estimation method to calculate the relative risk density of each node in the corresponding neighborhood based on the risk distribution of each node to obtain the risk density map of each node; performing local depth analysis on the node density distribution in the risk density map of each node to obtain a preliminary risk area division map, where the preliminary risk area division map is used to distinguish different risk level areas in the target distribution network; optimizing the risk area boundary of the preliminary risk area division map to obtain a risk hot spot map.

[0115] S240. Determine the weak link type corresponding to the weak link data.

[0116] Optionally, determining the weak link type corresponding to the weak link data includes: extracting features from the weak link data to obtain a weak link vector, where the weak link vector includes a voltage fluctuation vector, a harmonic content vector, a power factor vector, and a load imbalance degree vector; generating a time-frequency feature vector according to the weak link vector, and performing non-linear dynamic modeling on the time-frequency feature vector to obtain a dynamic modeling vector; performing fault mode recognition on the weak link position based on the dynamic modeling vector through a preset fuzzy inference algorithm to obtain a fault mode recognition result; performing dynamic classification on the fault mode recognition result to obtain the weak link type.

[0117] Among them, the types of weak links include voltage problems (such as overvoltage, undervoltage), current problems (such as overcurrent, short circuit), equipment failures (such as transformer failures, switch failures), line problems (such as broken wires, ground faults), etc. Determining the types of weak links helps to formulate targeted emergency response strategies. Voltage fluctuation refers to the change of the grid voltage within a short period of time. Harmonics refer to the current or voltage components in the grid whose frequencies are integer multiples of the fundamental frequency. Excessive harmonic content will have an adverse impact on grid equipment and power quality. By performing harmonic analysis on the current or voltage signals at the weak link positions, the content of each harmonic can be obtained, thus forming a harmonic content vector. The power factor is the ratio of the active power to the apparent power, which reflects the utilization efficiency of electric energy in the grid. The load unbalance degree refers to the unbalance degree among the three-phase loads. In a three-phase power grid, if the three-phase loads are unbalanced, it will lead to the unbalance of the grid voltage and current, affecting the stability and reliability of the power grid.

[0118] It should be noted that time-frequency analysis is an analysis method that simultaneously considers the characteristics of signals in both the time and frequency domains. For the voltage fluctuation vector, harmonic content vector, power factor vector, and load unbalance degree vector in the weak link vector, time-frequency analysis methods such as short-time Fourier transform, wavelet transform, etc. can be used to convert them into characteristic representations in the time-frequency domain, thereby obtaining time-frequency feature vectors. The time-frequency feature vectors can more comprehensively reflect the dynamic characteristics of the weak links. Specifically, the time-frequency feature vectors can be calculated using the following formula (19):

[0119] (19)

[0120] Where, is the time-frequency feature vector, W is the weak link vector, is the scaling function, representing the decomposition at the s-th scale is the frequency function, representing the decomposition at the frequency, is the number of scales, is the number of frequencies.

[0121] Specifically, nonlinear dynamic modeling is a method used to describe the dynamic behavior of complex systems. For the time-frequency feature vectors, nonlinear dynamic modeling methods such as neural networks, support vector machines, etc. can be used to establish their dynamic models. By training and learning the time-frequency feature vectors, dynamic modeling vectors can be obtained, and these vectors can better reflect the nonlinear dynamic characteristics of the weak links. Specifically, the dynamic modeling vectors can be calculated using the following formula (20):

[0122] (20)

[0123] Where, is the dynamic modeling vector, is the th time-frequency feature vector, is a non-linear dynamic function that acts on the th time-frequency feature vector, is a non-linear dynamic function that acts on the th time-frequency feature vector, is the total number of time-frequency feature vectors.

[0124] Specifically, fuzzy inference is a reasoning method based on fuzzy logic that can handle uncertain and fuzzy information. Through a preset fuzzy inference algorithm, the fault mode recognition of the weak link position is performed based on the dynamic modeling vector, and the fault mode recognition result is obtained using the following formula (21):

[0125] (21)

[0126] where the fault mode recognition result, is the fuzzy inference function, is the th fuzzy membership function that acts on the dynamic modeling vector , is the th fuzzy rule that acts on the weak link position L, is the dynamic modeling vector, L is the weak link position vector, is the number of fuzzy rules.

[0127] Based on the fault mode recognition result, a dynamic classification method can be used to more accurately classify the type of weak link, as shown in the following formula (22):

[0128] (22)

[0129] where is the type of weak link, is the weight coefficient of the th class, is the dynamic classification function that acts on the th fault mode recognition result and the classification criterion , is the th fault mode recognition result, is the th classification criterion, K is the number of classification categories.

[0130] Specific application scenarios: First, edge computing technology is applied to the processing of weak link data to achieve real-time feature extraction. By extracting features from weak link data, the system can generate an extracted weak link vector, which contains information in multiple dimensions such as voltage fluctuation vector, harmonic content vector, power factor vector, and load imbalance vector. These feature vectors can fully reflect the operating status of the distribution network and provide key data support for subsequent analysis. For example, in an electric power system, voltage fluctuations may lead to unstable operation of equipment. Therefore, by monitoring this vector, potential problems can be discovered in a timely manner. After obtaining the extracted weak link vector, the system will perform multi-scale decomposition on it to obtain a time-frequency feature vector. Through this process, the system can capture the changing characteristics of the signal at different time scales and frequency ranges, so that the dynamic behavior of the weak link can be more deeply understood. At this time, the time-frequency feature vector can not only reveal the transient characteristics of the equipment operation, but also analyze the possible long-term trends, providing a multi-level perspective for subsequent fault analysis. Next, the system performs nonlinear dynamic modeling on the time-frequency feature vector to obtain a dynamic modeling vector. The modeling process aims to capture the nonlinear relationship within the system so that the model can more accurately reflect the complex dynamic behavior of the distribution network. Through such dynamic modeling, the system can identify specific patterns and trends, providing a more accurate data basis for fault pattern identification. Subsequently, through the preset fuzzy reasoning algorithm, based on the dynamic modeling vector, the system performs fault pattern identification on the weak link location to obtain the fault pattern identification result. The fuzzy reasoning algorithm plays a key role in this process because it can handle uncertainty and ambiguity, enabling the system to more accurately identify potential fault modes in a complex operating environment. For example, when a weak link fails due to excessive load, the fuzzy reasoning algorithm can combine historical data and real-time monitoring information to comprehensively determine the type of fault and its possible impact. Finally, the system will dynamically classify the fault pattern identification results to obtain the type of weak link. Through this classification process, operation and maintenance personnel can quickly identify the types corresponding to different weak links, providing guidance for subsequent maintenance and management. For example, after a certain area is identified as a load imbalance problem, relevant personnel can carry out targeted load allocation to avoid equipment damage or system crash, thereby improving the reliability and stability of the entire distribution network. Therefore, the implementation of this series of steps not only improves the accuracy of identifying weak links, but also provides important support for the optimized management of distribution networks.

[0131] S250, perform semantic analysis on the weak link type to obtain a semantic feature vector.

[0132] Among them, semantic parsing is to convert the weak link types described in natural language into semantic feature representations that can be understood by computers. Natural language processing technologies such as lexical analysis, syntactic analysis, and semantic role labeling can be adopted. For example, for the weak link type of "transformer overheating fault", key semantic elements such as "transformer", "overheating", and "fault" can be extracted through semantic parsing. Then, the extracted semantic elements are converted into numerical vector representations, that is, semantic feature vectors. Word vector technologies such as Word2Vec and GloVe can be used to represent each semantic element as a low-dimensional vector. Finally, the vectors of all semantic elements are combined to form a semantic feature vector.

[0133] S260. Screen the historical emergency strategies in the preset database based on the semantic feature vector to obtain a candidate strategy set.

[0134] Among them, the preset database is a database containing historical emergency strategies, and the strategies therein are formulated for different types of weak links and fault situations. The database can include information such as the description of the strategy, applicable conditions, and implementation steps.

[0135] S270. Generate an emergency response strategy according to the candidate strategy set.

[0136] Optionally, generating an emergency response strategy according to the candidate strategy set includes: performing strategy evaluation on the candidate strategy set through a preset Pareto front search algorithm to obtain a strategy evaluation result; performing weight assignment on the strategy evaluation result to obtain a strategy assignment weight; performing strategy ranking on the candidate strategy set based on the strategy assignment weight to obtain an optimal emergency response strategy; performing intelligent strategy matching on the target distribution network based on the optimal emergency response strategy to obtain an emergency response strategy.

[0137] Among them, the Pareto front search algorithm is a multi-objective optimization algorithm used to find the set of optimal solutions among multiple objectives. In the evaluation of emergency response strategies, usually multiple objectives need to be considered, such as minimizing power outage time, minimizing economic losses, and maximizing system reliability. Weight assignment is to determine the importance of each objective in the comprehensive evaluation. Different objectives may have different importance for the selection of emergency response strategies. For example, in some cases, minimizing power outage time may be more important than minimizing economic losses. Through weight assignment, the relative importance of each objective can be adjusted according to the actual situation.

[0138] Specifically, when performing policy sorting, the controller multiplies the performance metrics of each policy on each target by the corresponding weights, then sums them to obtain a comprehensive score, and then sorts the candidate policy set according to the comprehensive score. Intelligent policy matching is to match and adjust the optimal emergency response policy with the specific situation of the target distribution network to ensure the effectiveness and feasibility of the policy. The target distribution network may have different topological structures, equipment characteristics, and operating states, so it is necessary to appropriately adjust and optimize the optimal emergency response policy.

[0139] Specific application scenario: First, the system performs semantic parsing on the type of weak link, aiming to extract the key information of the weak link, thereby generating a semantic feature vector. This feature vector can comprehensively reflect the specific characteristics, potential risks, and related operating status of the weak link, providing the necessary basic data for subsequent policy matching. For example, if the weak link type indicates a voltage fluctuation problem, the semantic feature vector will contain relevant voltage characteristic information to help the system identify emergency strategies applicable to this problem. Next, based on the semantic feature vector, the system will search and filter the historical emergency strategies in the preset database to form a candidate strategy set. In this process, the system uses an intelligent matching algorithm to screen out the strategies most relevant to the current situation by analyzing the similarity between the historical emergency strategies and the current weak link characteristics. This step ensures that the selected strategies are not only targeted but also can provide effective solutions for the current weak link based on historical data and previous response experiences. For example, if a similar voltage fluctuation event was successfully handled in a historical strategy, then that strategy will be included in the candidate strategy set. Subsequently, the system will perform policy evaluation on the candidate strategy set through a preset Pareto front search algorithm to obtain the policy evaluation results. The Pareto front search algorithm can effectively balance multiple objectives, enabling the system to consider multiple factors simultaneously when evaluating strategies, such as response time, resource consumption, and potential risks. Through this evaluation, the system can identify the advantages and disadvantages of each candidate strategy, ensuring that the finally selected strategy is not only effective but also feasible. After obtaining the policy evaluation results, the system will perform weight assignment to assign different importance to different strategies. This process is based on the specific performance of the policy evaluation. For example, if a certain strategy has a high historical success rate in dealing with voltage fluctuations, it will be given a higher weight during weight assignment. In this way, the system can ensure that more attention is paid to those strategies that perform well in actual applications during decision-making. Finally, based on the weights assigned to the policies, the system will rank the candidate strategy set to obtain the optimal emergency response strategy. This ranking process not only considers the historical performance of the strategies but also combines the specific situation of the current weak link, ensuring that the selected strategy can minimize risks in the shortest time and at the lowest cost. For example, in the case of large voltage fluctuations, the system preferentially selects those proven fast-response strategies to immediately stabilize the operation of the power grid. Through the above steps, the system will achieve intelligent policy matching for the target distribution network, ensuring that when facing weak links, it can quickly formulate effective emergency response strategies to maintain the safe and stable operation of the distribution network. The intelligence and efficiency of this process greatly improve the response ability and flexibility of the distribution network in the face of emergencies, ultimately ensuring the reliability and continuity of power supply.

[0140] The technical solution of the embodiment of the present invention can locate weak links through the preliminary state matrix to obtain weak link data, quickly and accurately find the problematic parts in the distribution network, and improve the efficiency of problem discovery. Generate an emergency response strategy according to the weak link data, so that effective countermeasures can be taken quickly when a fault or abnormality occurs in the distribution network, improve the overall operation reliability of the distribution network, and ensure the stability of power supply. It can dynamically track the changes of the distribution network and adapt to the changing power grid operation environment and requirements.

[0141] Embodiment III

[0142] Figure 3 It is a schematic structural diagram of a monitoring system for weak links in a distribution network provided by Embodiment III of the present invention. As Figure 3 shown, the system includes: an operation-related data acquisition module 310, configured to acquire operation-related data of a target distribution network;

[0143] a preliminary state matrix generation module 320, configured to perform a preliminary state assessment on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix;

[0144] an emergency response strategy generation module 330, configured to locate weak links based on the preliminary state matrix to obtain weak link data, and generate an emergency response strategy for the target distribution network according to the weak link data.

[0145] Optionally, the operation-related data acquisition module 310 is specifically configured to: collect each node of the target distribution network through a distributed sensor network to obtain original operation data, where the original operation data includes voltage data, current data, temperature data, and load rate data; perform wavelet transform denoising processing on the original operation data to generate wavelet transform denoised data; perform time-series frequency change recognition on the wavelet transform denoised data to generate a frequency change recognition result; when the frequency change recognition result is not within a preset frequency change range, generate time-series frequency change data according to the frequency change recognition result, and perform adaptive sampling on the time-series frequency change data to obtain adaptive sampling data; when the frequency change recognition result is within the preset frequency change range, generate data to be reconstructed according to the frequency change recognition result; reconstruct the adaptive sampling data and the data to be reconstructed to obtain reconstructed data; perform normalization processing on the reconstructed data to generate operation-related data.

[0146] Optionally, the preliminary state matrix generation module 320 is specifically configured to: perform spectrum analysis on operation-related data to generate a frequency-domain feature set; perform frequency response analysis on each node of the target distribution network based on the frequency-domain feature set to obtain a frequency response matrix; perform singular value decomposition on the frequency response matrix to obtain a dimensionality-reduced eigenvector set; perform trend analysis on the dimensionality-reduced eigenvector set to obtain a trend analysis result; use a preset long short-term memory network to perform similarity prediction and matching on the trend analysis result to obtain a similarity matrix; perform state correlation analysis on each node of the target distribution network based on the similarity matrix to obtain a state correlation matrix; use a Bayesian network to estimate the health state probability of each node of the target distribution network based on the state correlation matrix to obtain a state evaluation result corresponding to each node degree; and construct a preliminary state matrix according to the state evaluation result corresponding to each node.

[0147] Optionally, the emergency response strategy generation module 330 specifically includes: a weak-link data generation unit, configured to: perform structural processing on the preliminary state matrix through a preset topology analysis algorithm to obtain a topology structure matrix of the target distribution network, where the topology structure matrix includes connection relationships between nodes, branch impedance parameters corresponding to each node, and electrical characteristics of each node; perform network modeling on the target distribution network based on the topology structure matrix to obtain network feature vectors; perform weak-link analysis on the target distribution network according to the network feature vectors to obtain node vulnerability indicators; perform comprehensive evaluation on each node of the target distribution network based on the node vulnerability indicators to obtain a risk score for each node; perform spatial clustering analysis on the target distribution network according to the risk scores of each node through a self-organizing mapping neural network algorithm to obtain a risk hot spot map; determine the weak-link positions in the risk hot spot map through a preset risk assessment algorithm, and perform weak-link monitoring at the weak-link positions to obtain weak-link data.

[0148] Optionally, the weak-link data generation unit specifically includes: a risk hot spot map generation subunit, configured to: perform non-linear dimensionality reduction mapping on the node risk scores through a self-organizing mapping neural network algorithm to obtain the risk distribution of each node on the mapping grid; use a kernel density estimation method to calculate the relative risk density of each node in the corresponding neighborhood based on the risk distribution of each node to obtain a risk density map of each node; perform local depth analysis on the node density distribution in the risk density map of each node to obtain a preliminary risk area division map, where the preliminary risk area division map is used to distinguish different risk level areas in the target distribution network; and optimize the risk area boundary of the preliminary risk area division map to obtain a risk hot spot map.

[0149] Optionally, the emergency response strategy generation module 330 specifically includes: an emergency response strategy generation unit, configured to: determine the weak link type corresponding to the weak link data; perform semantic analysis on the weak link type to obtain a semantic feature vector; screen the historical emergency strategies in the preset database based on the semantic feature vector to obtain a candidate strategy set; and generate an emergency response strategy according to the candidate strategy set.

[0150] Optionally, the emergency response strategy generation unit specifically includes: a weak link type determination subunit, configured to: extract features from the weak link data to obtain a weak link vector, where the weak link vector includes a voltage fluctuation vector, a harmonic content vector, a power factor vector, and a load unbalance degree vector; generate a time-frequency feature vector according to the weak link vector, and perform non-linear dynamic modeling on the time-frequency feature vector to obtain a dynamic modeling vector; perform fault mode recognition on the weak link position based on the dynamic modeling vector through a preset fuzzy inference algorithm to obtain a fault mode recognition result; and perform dynamic classification on the fault mode recognition result to obtain the weak link type.

[0151] The technical solution of the embodiment of the present invention locates the weak link through the preliminary state matrix to obtain the weak link data, which can quickly and accurately find out the problematic part in the distribution network and improve the efficiency of problem discovery. Generating an emergency response strategy according to the weak link data enables effective countermeasures to be taken quickly when a fault or abnormality occurs in the distribution network, improves the overall operation reliability of the distribution network, and ensures the stability of power supply. It can dynamically track the changes of the distribution network and adapt to the changing grid operation environment and requirements.

[0152] A monitoring system for weak links in a distribution network provided by an embodiment of the present invention can execute a method for monitoring weak links in a distribution network provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the method.

[0153] Embodiment 4

[0154] Figure 4 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital assistant, a cellular phone, a smart phone, a wearable device (such as a helmet, glasses, a watch, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described herein and / or claimed.

[0155] As Figure 4As shown, the electronic device 10 includes at least one processor 11 and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0156] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0157] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a monitoring method for weak links in a distribution network.

[0158] It should be understood that various forms of the processes shown above can be used, reordering, adding, or deleting steps. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0159] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for monitoring weak links in a distribution network, characterized in that: include: Obtaining operation-related data of the target distribution network; Performing a preliminary state assessment on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix; Based on the preliminary state matrix, weak link location is performed to obtain weak link data, and an emergency response strategy of the target distribution network is generated according to the weak link data; The step of performing a preliminary state evaluation on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix includes: performing spectrum analysis on the operation-related data to generate a frequency domain feature set; Performing frequency response analysis on each node of the target distribution network based on the frequency domain feature set to obtain a frequency response matrix; Performing singular value decomposition on the frequency response matrix to obtain a reduced-dimensional feature vector set; Performing trend analysis on the dimension-reduced feature vector set to obtain a trend analysis result; Using a preset long short-term memory network to perform similarity prediction and matching on the trend analysis results to obtain a similarity matrix; Based on the similarity matrix, a state correlation analysis is performed on each node of the target distribution network to obtain a state correlation matrix; Using a Bayesian network, based on the state association matrix, the health state probability of each node of the target distribution network is estimated to obtain a state assessment result corresponding to each node; Constructing the preliminary state matrix according to the state evaluation results corresponding to each node; Wherein, the locating of weak links based on the preliminary state matrix to obtain weak link data includes: The preliminary state matrix is ​​structured by a preset topology analysis algorithm to obtain a topology matrix of the target distribution network, wherein the topology matrix includes connection relationships between nodes, branch impedance parameters corresponding to each node, and electrical characteristics of each node; Performing network modeling on the target distribution network based on the topological structure matrix to obtain a network feature vector; Performing a weak link analysis on the target distribution network according to the network feature vector to obtain a node vulnerability index; Comprehensively evaluating each node of the target distribution network based on the node vulnerability index to obtain a risk score for each node; Performing spatial clustering analysis on the target distribution network according to the risk score of each node by using a self-organizing map neural network algorithm to obtain a risk hotspot map; The weak link location is determined in the risk hot spot map by a preset risk assessment algorithm, and weak link monitoring is performed at the weak link location to obtain weak link data.

2. The method according to claim 1, characterized in that The obtaining of operation-related data of the target distribution network includes: Collecting data from each node of the target distribution network through a distributed sensor network to obtain original operation data, wherein the original operation data includes voltage data, current data, temperature data and load rate data; Performing wavelet transform denoising processing on the original operation data to generate wavelet transform denoised data; Performing time series frequency change recognition on the wavelet transform denoised data to generate a frequency change recognition result; When the frequency change recognition result is not within the preset frequency change range, generating time series frequency change data according to the frequency change recognition result, and adaptively sampling the time series frequency change data to obtain adaptive sampling data; When the frequency change identification result is within a preset frequency change range, generating data to be reconstructed according to the frequency change identification result; Reconstructing the adaptive sampling data and the data to be reconstructed to obtain reconstructed data; The reconstructed data is normalized to generate operation-related data.

3. The method according to claim 1, characterized in that The method of performing spatial clustering analysis on the target distribution network according to the risk score of each node by using a self-organizing map neural network algorithm to obtain a risk hotspot map includes: Performing nonlinear dimensionality reduction mapping on the node risk score by using a self-organizing map neural network algorithm to obtain the risk distribution of each node on the mapping grid; Using a kernel density estimation method, based on the risk distribution of each node, the relative risk density of each node in the corresponding neighborhood is calculated to obtain a risk density map of each node; Performing local depth analysis on the node density distribution in the risk density map of each node to obtain a preliminary risk area division map, wherein the preliminary risk area division map is used to distinguish different risk level areas in the target distribution network; The risk area boundaries of the preliminary risk area division map are optimized to obtain a risk hot spot map.

4. The method according to claim 1, characterized in that: The generating of the emergency response strategy of the target distribution network according to the weak link data includes: Determine the weak link type corresponding to the weak link data; Performing semantic analysis on the weak link type to obtain a semantic feature vector; Based on the semantic feature vector, the historical emergency strategies in the preset database are screened to obtain a candidate strategy set; An emergency response strategy is generated according to the candidate strategy set.

5. The method according to claim 4, characterized in that The determining the weak link type corresponding to the weak link data includes: Extracting features from the weak link data to obtain a weak link vector, wherein the weak link vector includes a voltage fluctuation vector, a harmonic content vector, a power factor vector, and a load imbalance vector; Generate a time-frequency feature vector according to the weak link vector, and perform nonlinear dynamic modeling on the time-frequency feature vector to obtain a dynamic modeling vector; By using a preset fuzzy inference algorithm, fault mode identification is performed on the weak link location based on the dynamic modeling vector to obtain a fault mode identification result; The fault mode recognition results are dynamically classified to obtain weak link types.

6. A monitoring system for weak links in a distribution network, characterized in that: include: An operation-related data acquisition module is used to acquire operation-related data of the target distribution network; A preliminary state matrix generating module, used for performing preliminary state evaluation on each node of the target distribution network based on the operation-related data to generate a preliminary state matrix; An emergency response strategy generation module, used to locate weak links based on the preliminary state matrix to obtain weak link data, and generate an emergency response strategy for the target distribution network according to the weak link data; Among them, the preliminary state matrix generation module is specifically used to: perform spectrum analysis on the operation-related data to generate a frequency domain feature set; perform frequency response analysis on each node of the target distribution network based on the frequency domain feature set to obtain a frequency response matrix; perform singular value decomposition on the frequency response matrix to obtain a reduced dimension feature vector set; perform trend analysis on the reduced dimension feature vector set to obtain a trend analysis result; use a preset long short-term memory network to perform similarity prediction and matching on the trend analysis results to obtain a similarity matrix; perform state association analysis on each node of the target distribution network based on the similarity matrix to obtain a state association matrix; use a Bayesian network to estimate the health state probability of each node of the target distribution network based on the state association matrix to obtain a state evaluation result corresponding to each node; construct the preliminary state matrix according to the state evaluation result corresponding to each node; The emergency response strategy generation module specifically includes: a weak link data generation unit, which is used to: perform structured processing on the preliminary state matrix through a preset topology analysis algorithm to obtain a topology matrix of the target distribution network, wherein the topology matrix includes the connection relationship between each node, the branch impedance parameters corresponding to each node, and the electrical characteristics of each node; Performing network modeling on the target distribution network based on the topological structure matrix to obtain a network feature vector; Performing a weak link analysis on the target distribution network according to the network feature vector to obtain a node vulnerability index; Comprehensively evaluating each node of the target distribution network based on the node vulnerability index to obtain a risk score for each node; Performing spatial clustering analysis on the target distribution network according to the risk score of each node by using a self-organizing map neural network algorithm to obtain a risk hotspot map; The weak link location is determined in the risk hot spot map by a preset risk assessment algorithm, and weak link monitoring is performed at the weak link location to obtain weak link data.

7. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method according to any one of claims 1 to 5.

8. A computer storage medium, characterized in that: The computer storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method according to any one of claims 1 to 5 when executed.

Images