Method, equipment, storage medium and product for locating leakage fault of branch line before meter

Abstract

The present invention discloses a method, device, storage medium, and product for locating leakage faults in a pre-meter branch line. The method comprises obtaining the total meter voltage, user meter voltage, user meter active power, and user meter reactive power of a substation within a measurement period; calculating the user virtual impedance; extracting global correlation features and neighboring correlation features from the user virtual impedance; calculating the symmetric relative entropy of the global correlation features and neighboring correlation features based on the extracted global correlation features and neighboring correlation features; calculating a comprehensive anomaly score of the user virtual impedance based on the user virtual impedance and the symmetric relative entropy; and locating the leakage fault based on the comprehensive anomaly score of the user virtual impedance. The present invention improves the efficiency of leakage fault locating and effectively enhances the efficiency of low-voltage distribution network operation and maintenance.

Description

Technical Field

[0001] The present invention belongs to the technical field of abnormal positioning of low-voltage distribution networks, and in particular relates to a method, device, storage medium and product for positioning leakage faults in branch lines before user meters in a substation. Background Art

[0002] Due to factors such as insulation moisture or aging, some insulation of overhead lines or underground pipelines in the power supply lines of the substation (referring to the power supply range or power supply area of ​​a transformer) is damaged, resulting in leakage, and the leakage location is highly hidden.

[0003] Currently, leakage problems caused by insulation damage mainly rely on manual experience and traditional instrument measurement methods. That is, operation and maintenance personnel usually use tools such as multimeters and clamp ammeters to climb the pole from the center point of the main line, and narrow the scope of leakage fault investigation by measuring the residual current section by section. This method has limitations in real-time, accuracy and intermittent leakage detection and positioning capabilities. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, device, storage medium and product for locating leakage faults in a branch line before a meter, so as to solve the problems of poor real-time performance, poor accuracy and low efficiency of traditional leakage locating methods.

[0005] The present invention solves the above technical problems through the following technical solutions: a method for locating leakage faults in branch lines before a meter, comprising:

[0006] Obtain the total meter voltage, user meter voltage, user meter active power, and user meter reactive power within the measurement period;

[0007] Calculate the user virtual impedance based on the total meter voltage, user meter voltage, user meter active power and user meter reactive power;

[0008] Extracting global correlation features and proximity correlation features from the user virtual impedance;

[0009] Calculate the symmetric relative entropy of the global correlation feature and the neighboring correlation feature based on the extracted global correlation feature and the neighboring correlation feature;

[0010] Calculating a comprehensive anomaly score of the user virtual impedance according to the user virtual impedance and the symmetric relative entropy;

[0011] The leakage fault is located according to the comprehensive abnormality score of the user virtual impedance.

[0012] Furthermore, based on the total meter voltage, the user meter voltage, the user meter active power, and the user meter reactive power, the user virtual impedance is solved, including:

[0013] Based on the user meter voltage, user meter active power and user meter reactive power, the total meter voltage estimation function is constructed. The specific expression is:

[0014] ;

[0015] in, It represents the estimated value of the total meter voltage corresponding to the nth user at the tth moment, It represents the active power of the nth user's meter at the tth moment, It represents the reactive power of the nth user's meter at the tth moment, Indicates the nth user virtual resistance at the tth moment, represents the virtual reactance of the nth user at the tth moment, Indicates the voltage of the nth user's meter at the tth moment;

[0016] The user virtual impedance is solved according to the total meter voltage estimation function and the total meter voltage; wherein the solution function is:

[0017] ;

[0018] ;

[0019] Where T represents the number of sampling moments in the measurement period, N represents the number of user meters in the area, represents the total meter voltage at the tth moment, Indicates the virtual impedance of the nth user at the tth moment.

[0020] Furthermore, a pre-trained neural network model is used to extract global correlation features and proximity correlation features of the user virtual impedance;

[0021] The neural network model includes a feature extraction module, which is used to extract global correlation features and proximity correlation features from the user virtual impedance. The specific process includes:

[0022] According to the user virtual impedance, the query vector, key vector, value vector and Gaussian kernel scale parameter are calculated. The specific calculation formula is:

[0023] , , , ;

[0024] in, 、 、 and They represent the query vector, key vector, value vector and Gaussian kernel scale parameter of the lth layer in the feature extraction module respectively; Represents the input of the lth layer in the feature extraction module. When l=1, ; 、 、 and Respectively represent the weight parameter matrix of the lth layer in the feature extraction module;

[0025] According to the query vector , key vector Sum vector Calculate the global correlation features. The specific calculation formula is:

[0026] ;

[0027] in, represents the global correlation feature of the lth layer in the feature extraction module, represents the normalization function, the superscript T represents the matrix transpose, Table 1 shows the input dimension of the lth layer in the feature extraction module;

[0028] According to the Gaussian kernel scale parameter Calculate the adjacent correlation features. The specific calculation formula is:

[0029] ;

[0030] in, represents the neighboring correlation features of the lth layer in the feature extraction module, represents the normalization function, T represents the number of sampling moments in the measurement period, 、 They respectively represent the sequence numbers of the time points in the measurement period.

[0031] Furthermore, the neural network model further includes a first calculation module, which calculates the symmetric relative entropy of the global correlation feature and the neighboring correlation feature based on the extracted global correlation feature and the neighboring correlation feature using the first calculation module. The specific calculation formula is:

[0032] ;

[0033] in, represents the symmetric relative entropy between global correlation features and neighboring correlation features, L represents the number of layers of the feature extraction module, represents relative entropy.

[0034] Furthermore, calculating a comprehensive anomaly score of the user virtual impedance according to the user virtual impedance and the symmetric relative entropy specifically includes:

[0035] The reconstruction error is calculated based on the user virtual impedance and the symmetric relative entropy. The specific calculation formula is:

[0036] ;

[0037] in, represents the reconstruction error; represents the virtual impedance of the nth user at the tth moment; Indicates the virtual impedance of the nth user at time t The predicted value of represents the loss coefficient; represents the F norm; represents the 1 norm;

[0038] The abnormal score of the user virtual impedance is calculated according to the symmetric relative entropy and the reconstruction error. The specific calculation formula is:

[0039] ;

[0040] in, represents the abnormal score of the nth user's virtual impedance at the tth moment, represents element-wise product, represents the normalization function;

[0041] The anomaly scores of the user virtual impedance at all times are summed to obtain the comprehensive anomaly score of the user virtual impedance.

[0042] Furthermore, the leakage fault is located according to the comprehensive abnormal score of the user virtual impedance, specifically including:

[0043] The leakage fault location is determined based on the number of anomalies in a single leakage fault location and the comprehensive anomaly score of all user virtual impedances.

[0044] Furthermore, the method further comprises:

[0045] After eliminating the leakage fault according to the leakage fault location result, obtain the residual current of the substation, and determine whether the leakage fault still exists according to the residual current of the substation; if so, repeat the leakage fault location step.

[0046] Based on the same concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program / instruction stored in the memory, wherein the processor executes the computer program / instruction to implement the above-mentioned method for locating leakage faults in branch lines before a meter.

[0047] Based on the same concept, the present invention further provides a computer-readable storage medium having a computer program / instruction stored thereon. When the computer program / instruction is executed by a processor, the method for locating leakage fault of a branch line before a meter as described above is implemented.

[0048] Based on the same concept, the present invention further provides a computer program product, including a computer program / instruction, which implements the above-mentioned method for locating leakage faults in branch lines before a meter when the computer program / instruction is executed by a processor.

[0049] Compared with the prior art, the advantages of the present invention are:

[0050] The leakage fault locating method of the present invention does not require human intervention. The leakage fault of the user branch line can be located in real time only through the total meter voltage, user meter voltage, user meter active power and user meter reactive power, thereby improving the efficiency of leakage fault locating and effectively improving the operation and maintenance efficiency of the low-voltage distribution network. The present invention integrates user virtual impedance modeling and global-adjacent two-dimensional correlation feature analysis to accurately capture the user virtual impedance mutation characteristics caused by leakage, significantly improving the leakage detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only one embodiment of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 This is a flow chart of a method for locating leakage faults in branch lines before user meters in a substation area according to an embodiment of the present invention;

[0053] Figure 2 : is a heat map of the proximity correlation feature distribution of the virtual impedance of normal users in an embodiment of the present invention;

[0054] Figure 3 : is a global correlation feature distribution heat map of normal user virtual impedance in an embodiment of the present invention;

[0055] Figure 4 : is a heat map of the proximity correlation feature distribution of the virtual impedance of the leakage user in an embodiment of the present invention;

[0056] Figure 5 : is a global correlation feature distribution heat map of the virtual impedance of the leakage user in an embodiment of the present invention;

[0057] Figure 6 This is a visualization diagram of the leakage fault location result in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following is a clear and complete description of the technical solutions of the present invention in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts are within the scope of protection of the present invention.

[0059] The following specific embodiments are used to describe the technical solution of the present application in detail. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0060] Example 1

[0061] In order to solve the problem that traditional leakage fault location mainly relies on manual experience and traditional instrument measurement, resulting in poor real-time performance and accuracy, and poor intermittent leakage detection and positioning capabilities, the present invention provides a leakage fault location method for branch lines before user meters in substations, which is applied to leakage identification and positioning of branch lines before user meters in low-voltage substations. Figure 1 As shown, the leakage fault location method includes the following steps:

[0062] Step 1: When leakage fault location is required, obtain the substation total meter voltage, customer meter voltage, customer meter active power, and customer meter reactive power within the measurement period.

[0063] Real-time monitoring of the residual current in the substation area. When the residual current in the substation area is greater than the current threshold, it indicates that there is leakage in the substation area and leakage fault location is required. The current threshold is the tripping threshold of the leakage protection element. In this embodiment, the current threshold is 300mA.

[0064] During the measurement period, according to the sampling interval, the user information collection system is used to collect the total meter voltage of the substation, the user meter voltage, the user meter active power and the user meter reactive power, so as to obtain the total meter voltage at each sampling time t within the measurement period T. , the voltage of each user's meter n at each sampling time t (i.e., user meter voltage), active power of each user meter n at each sampling time t and reactive power (i.e., the active power of the user's meter and the reactive power of the user's meter).

[0065] Step 2: Based on the total meter voltage , User meter voltage , user meter active power and user meter reactive power , solve the user virtual impedance .

[0066] There is a shortest power supply path between each user's electricity meter and the transformer. The loop impedance on this shortest power supply path is defined as the user's virtual impedance. The specific expression is:

[0067] (1)

[0068] in, represents the virtual impedance of the nth user at the tth moment; Indicates the nth user virtual resistance at the tth moment; represents the virtual reactance of the nth user at the tth moment, and j represents the imaginary unit. The voltage drop on the shortest power supply path The longitudinal component Add lateral component The form is expressed as:

[0069] (2)

[0070] (3)

[0071] (4)

[0072] in, Indicates the voltage drop of the nth user at the tth moment; It represents the longitudinal component of the voltage drop of the nth user at the tth moment; It represents the horizontal component of the voltage drop of the nth user at the tth moment.

[0073] According to formula (1) to formula (4), the total meter voltage estimation function can be constructed with the user virtual impedance as the independent variable and the user meter voltage, user meter active power and user meter reactive power as the dependent variables:

[0074] (5)

[0075] in, It represents the total voltage estimate corresponding to the nth user at the tth time.

[0076] The problem of solving the user virtual impedance is transformed into an objective function optimization problem of minimizing the difference between the total meter voltage in the substation area and the estimated total meter voltage. The specific expression is:

[0077] (6)

[0078] Where T represents the number of sampling moments in the measurement period, N represents the number of user meters in the area, Represents the objective function. Optimize and solve formula (6) to obtain 、 , and then according to formula (1) the user virtual impedance can be solved From this, we can get the virtual impedance of each user at each sampling moment during the measurement period:

[0079] (7)

[0080] Formula (7) is flattened into a one-dimensional form to facilitate subsequent feature extraction operations.

[0081] Step 3: Extract global correlation features and proximity correlation features of user virtual impedance.

[0082] In a specific embodiment of the present invention, a pre-trained neural network model is used to perform feature extraction, symmetric relative entropy calculation, and anomaly score calculation. The neural network model of this embodiment is based on the original Transformer model and adds a feature extraction module, a first calculation module, and a second calculation module. The newly added feature extraction module is used to extract global correlation features and neighboring correlation features of the user virtual impedance. The newly added first calculation module is used to calculate the symmetric relative entropy of the global correlation features and the neighboring correlation features based on the extracted global correlation features and neighboring correlation features. The newly added second calculation module is used to calculate the comprehensive anomaly score of the user virtual impedance based on the user virtual impedance and the symmetric relative entropy. The original Transformer model is used to predict the user virtual impedance to obtain the user virtual impedance prediction value. .

[0083] In a specific embodiment of the present invention, a new feature extraction module is used to extract global correlation features and neighboring correlation features of the user virtual impedance, specifically including:

[0084] Step 3.1: Calculate the query vector, key vector, value vector, and Gaussian kernel scale parameter based on the user virtual impedance. The specific calculation formula is:

[0085] (8)

[0086] (9)

[0087] (10)

[0088] (11)

[0089] in, 、 、 and They represent the query vector, key vector, value vector and Gaussian kernel scale parameter of the lth layer in the feature extraction module respectively; Represents the input of the lth layer in the feature extraction module. When l=1, ; 、 、 and They represent the weight parameter matrix of the lth layer in the feature extraction module respectively.

[0090] Step 3.2: Based on the query vector , key vector Sum vector Calculate the global correlation features. The specific calculation formula is:

[0091] (12)

[0092] in, represents the global correlation feature of the lth layer in the feature extraction module, represents the normalization function, the superscript T represents the matrix transpose, Represents the input dimension of layer l in the feature extraction module. The global correlation feature uses the multi-head attention mechanism in the original Transformer model to quantify the inherent electrical interconnectivity between different users.

[0093] Step 3.3: According to the Gaussian kernel scale parameter Calculate the adjacent correlation features. The specific calculation formula is:

[0094] (13)

[0095] in, represents the neighboring correlation features of the lth layer in the feature extraction module, represents the normalization function, T represents the number of sampling moments in the measurement period, The neighboring correlation feature uses an adaptive Gaussian kernel to capture the neighboring leakage users that are most affected by the leakage point.

[0096] Figure 2 and Figure 3 The neighboring correlation feature distribution and global correlation feature distribution of normal user virtual impedance are shown respectively. Figure 4 and Figure 5 The neighboring correlation feature distribution and global correlation feature distribution of the virtual impedance of the leakage user are shown respectively, where the numbers represent the correlation feature values, and the closer to 1, the greater the correlation.

[0097] The pre-training process of the neural network model is similar to the leakage fault location process. The weight parameter matrix of the feature extraction module is determined through pre-training. 、 、 and To obtain training samples, a simulation was performed in a real-world distribution network laboratory. The simulated substation contained 117 simulated users (N = 117). The measurement period was one month, with a sampling interval of 15 minutes per time (T = 2880). The total meter voltage, simulated user meter voltage, simulated user meter active power, and simulated user meter reactive power were obtained. The data were then classified according to the scenarios in Table 1, with each scenario divided into a training set and a test set in a ratio of 7:3, resulting in a total of 24,750 samples.

[0098] Table 1 All leakage scenarios included in the sample

[0099]

[0100] The pre-training of the neural network model includes the training of the original Transformer model and the training of the feature extraction module. The original Transformer model is trained first, and then the feature extraction module is trained. The training process of the original Transformer model is as follows:

[0101] Calculate the simulated user virtual impedance based on the total meter voltage, simulated user meter voltage, simulated user meter active power and simulated user meter reactive power (such as formula (5) and formula (6));

[0102] The original Transformer model is trained and tested with the total meter voltage, simulated user meter voltage, simulated user meter active power and simulated user meter reactive power as input quantities and the calculated simulated user virtual impedance as the true label.

[0103] The original Transformer model after training and testing can be used to predict the user's virtual impedance and obtain the user's virtual impedance prediction value. After completing the training and testing of the original Transformer model, the feature extraction module is trained. The specific training process of the feature extraction module is as follows:

[0104] The calculated simulated user virtual impedance is subjected to global correlation features and neighboring correlation features extraction. The symmetric relative entropy of the global correlation features and neighboring correlation features is calculated based on the extracted global correlation features and neighboring correlation features. The comprehensive anomaly score of the simulated user virtual impedance is calculated based on the simulated user virtual impedance and the symmetric relative entropy. The predicted leakage simulated user is determined based on the comprehensive anomaly score of the simulated user virtual impedance. The parameters in the feature extraction module are reversely adjusted based on the predicted leakage simulated users and the actual leakage simulated users to achieve pre-training of the feature extraction module. If the number of predicted leakage simulated users is equal to the number of anomalies located in a single leakage fault, and the predicted leakage simulated users are the same as the actual leakage simulated users, the prediction is correct. Otherwise, the parameters in the feature extraction module are adjusted. Table 2 shows the parameter settings for model training.

[0105] Table 2 Model training parameter settings

[0106]

[0107] After the training is completed, the test set is used to evaluate the neural network model. The evaluation index results are shown in Table 3.

[0108] Table 3 Neural network model evaluation index results

[0109]

[0110] Step 4: Calculate the symmetric relative entropy of the global correlation features and the neighboring correlation features based on the extracted global correlation features and neighboring correlation features.

[0111] Symmetric relative entropy is used to quantify the feature differences between normal users and leakage users. The first calculation module calculates the symmetric relative entropy of global correlation features and neighboring correlation features based on the extracted global correlation features and neighboring correlation features. The specific calculation formula is:

[0112] (14)

[0113] in, represents the symmetric relative entropy between global correlation features and neighboring correlation features, L represents the number of layers of the feature extraction module, represents relative entropy.

[0114] Step 5: Calculate the comprehensive anomaly score of the user virtual impedance based on the user virtual impedance and the symmetric relative entropy.

[0115] In a specific embodiment of the present invention, the second calculation module is used to calculate the comprehensive abnormality score of the user virtual impedance according to the user virtual impedance and the symmetric relative entropy, specifically including:

[0116] Step 5.1: Calculate the reconstruction error based on the user virtual impedance and symmetric relative entropy. The specific calculation formula is:

[0117] (15)

[0118] in, represents the reconstruction error; Indicates the virtual impedance of the nth user at time t The predicted value of Predicted by the original Transformer model; represents the loss coefficient; represents the F norm, Represents the 1-norm.

[0119] The reconstruction error calculation enhances the model's ability to extract neighboring correlation features and global correlation features.

[0120] Step 5.2: Calculate the abnormal score of the user virtual impedance based on the symmetric relative entropy and reconstruction error. The specific calculation formula is:

[0121] (16)

[0122] in, represents the abnormal score of the nth user's virtual impedance at the tth moment, represents element-wise product, The anomaly score of the virtual impedance of each user at each moment combines the symmetric relative entropy and the reconstruction error, which improves the scoring accuracy.

[0123] Step 5.3: Sum the anomaly scores of the user virtual impedance at all times to obtain the comprehensive anomaly score of the user virtual impedance.

[0124] For each user, the anomaly scores of the virtual impedance at all times are summed up to obtain the total anomaly score of the user's virtual impedance, which transforms the leakage location problem of the branch line before the meter into a time series anomaly detection problem.

[0125] Step 6: Locate the leakage fault based on the comprehensive abnormality score of the user's virtual impedance.

[0126] In a specific embodiment of the present invention, the leakage fault location is determined according to the number of anomalies of a single leakage fault location and the comprehensive anomaly score of all user virtual impedances.

[0127] The number of abnormalities in a single leakage fault location refers to the number of users with leakage faults identified during a single leakage fault location. When leakage fault location is required, there must be at least one user with leakage faults, so the number of abnormalities in a single leakage fault location must be greater than or equal to 1.

[0128] For example, the number of users in the substation area is 117, and the number of anomalies in a single leakage fault location is 1, which means that when the leakage fault location method of the present invention is used to locate a leakage fault, one leakage fault user needs to be located. Therefore, the maximum comprehensive anomaly score is selected from the comprehensive anomaly scores of all user virtual impedances, and the branch line before the user meter corresponding to the maximum comprehensive anomaly score is the leakage fault location, such as Figure 6 shown. Figure 6 In the example, when the number of anomalies in a single leakage fault location is 1, the method of the present invention can calculate the comprehensive anomaly scores of the virtual impedances of 117 users, and locate the branch line before the meter of the 40th user as the leakage fault location; Figure 6 The threshold δ in [1] is determined based on the number of anomalies detected in a single leakage fault location and the combined anomaly score of all user virtual impedances. In this embodiment, the threshold δ is 0.21. That is, when the combined anomaly score of a user's virtual impedance exceeds the threshold δ, the corresponding user's pre-meter branch line is located at a leakage fault.

[0129] Step 7: After eliminating the leakage fault according to the leakage fault location result of step 6, obtain the residual current of the substation and determine whether the leakage fault still exists based on the residual current of the substation. If so, repeat steps 1 to 7.

[0130] Example 2

[0131] An embodiment of the present invention also provides an electronic device, which includes: a memory, a processor and a computer program / instructions stored in the memory, and the processor executes the computer program / instructions to implement the method for locating leakage faults in branch lines before user meters in a substation in an embodiment of the present application.

[0132] Although not shown, the electronic device includes a processor that can perform various appropriate operations and processes based on programs and / or data stored in a read-only memory (ROM) or programs and / or data loaded from a storage portion into a random access memory (RAM). The processor can be a multi-core processor or can include multiple processors. In some embodiments, the processor can include a general-purpose main processor and one or more special coprocessors, such as a central processing unit, a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), etc. Various programs and data required for device operation are also stored in the RAM. The processor, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0133] The processor and memory are used together to execute the program / instructions stored in the memory. When the program / instructions are executed by the computer, the methods, steps or functions described in the above embodiments can be implemented.

[0134] Although not shown, an embodiment of the present invention further provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the method for locating leakage faults in branch lines before user meters in a substation in an embodiment of the present application.

[0135] Computer-readable storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0136] Although not shown, an embodiment of the present invention further provides a computer program product, including: a computer program / instruction, which, when executed by a processor, implements the method for locating leakage faults in branch lines before user meters in a substation in an embodiment of the present application.

[0137] The above disclosure is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or modifications within the technical scope disclosed in the present invention, and they should all be covered by the scope of protection of the present invention.

Claims

1. A method for locating leakage faults in branch lines before a meter, characterized in that: The fault locating method includes: Obtain the total meter voltage, user meter voltage, user meter active power, and user meter reactive power within the measurement period; Calculate the user virtual impedance based on the total meter voltage, user meter voltage, user meter active power and user meter reactive power; Extracting global correlation features and neighboring correlation features from the user virtual impedance; wherein the global correlation features are calculated based on the query vector, key vector, and value vector calculated by the neural network model, and the neighboring correlation features are calculated based on the Gaussian kernel scale parameter calculated by the neural network model; Calculate the symmetric relative entropy of the global correlation feature and the neighboring correlation feature based on the extracted global correlation feature and the neighboring correlation feature; Calculating a reconstruction error according to the user virtual impedance and the symmetric relative entropy, and calculating a comprehensive anomaly score of the user virtual impedance according to the symmetric relative entropy and the reconstruction error; The leakage fault is located according to the comprehensive abnormality score of the user virtual impedance.

2. The method for locating leakage fault of a branch line before a meter according to claim 1, characterized in that: Based on the total meter voltage, user meter voltage, user meter active power, and user meter reactive power, the user virtual impedance is calculated, including: Based on the user meter voltage, user meter active power and user meter reactive power, the total meter voltage estimation function is constructed. The specific expression is: ； in, It represents the estimated value of the total meter voltage corresponding to the nth user at the tth moment, It represents the active power of the nth user's meter at the tth moment, It represents the reactive power of the nth user's meter at the tth moment, Indicates the nth user virtual resistance at the tth moment, represents the virtual reactance of the nth user at the tth moment, Indicates the voltage of the nth user's meter at the tth moment; The user virtual impedance is solved according to the total meter voltage estimation function and the total meter voltage; wherein the solution function is: ； ； Where T represents the number of sampling moments in the measurement period, N represents the number of user meters in the area, represents the total meter voltage at the tth moment, Indicates the virtual impedance of the nth user at the tth moment.

3. The method for locating leakage fault of a branch line before a meter according to claim 1, characterized in that: The neural network model includes a feature extraction module, which is used to extract global correlation features and proximity correlation features from the user virtual impedance. The specific process includes: According to the user virtual impedance, the query vector, key vector, value vector and Gaussian kernel scale parameter are calculated. The specific calculation formula is: ， ， ， ； in, 、 、 and They represent the query vector, key vector, value vector and Gaussian kernel scale parameter of the lth layer in the feature extraction module respectively; Represents the input of the lth layer in the feature extraction module. When l=1, ; 、 、 and Respectively represent the weight parameter matrix of the lth layer in the feature extraction module; According to the query vector , key vector Sum vector Calculate the global correlation features. The specific calculation formula is: ； in, represents the global correlation feature of the lth layer in the feature extraction module, represents the normalization function, the superscript T represents the matrix transpose, Represents the input dimension of the lth layer in the feature extraction module; According to the Gaussian kernel scale parameter Calculate the adjacent correlation features. The specific calculation formula is: ； in, represents the neighboring correlation features of the lth layer in the feature extraction module, represents the normalization function, T represents the number of sampling moments in the measurement period, 、 They respectively represent the sequence numbers of the time points in the measurement period.

4. The method for locating leakage fault of a branch line before a meter according to claim 3, characterized in that: The neural network model further includes a first calculation module, which calculates the symmetric relative entropy of the global correlation feature and the neighboring correlation feature based on the extracted global correlation feature and the neighboring correlation feature using the first calculation module. The specific calculation formula is: ； in, represents the symmetric relative entropy between global correlation features and neighboring correlation features, L represents the number of layers of the feature extraction module, represents relative entropy.

5. The method for locating leakage fault of a branch line before a meter according to claim 1, characterized in that: The reconstruction error is calculated based on the user virtual impedance and the symmetric relative entropy. The specific calculation formula is: ； in, represents the reconstruction error; represents the virtual impedance of the nth user at the tth moment; Indicates the virtual impedance of the nth user at time t The predicted value of represents the loss coefficient; represents the F norm; represents the symmetric relative entropy between global correlation features and neighboring correlation features, represents the 1 norm; The abnormal score of the user virtual impedance is calculated according to the symmetric relative entropy and the reconstruction error. The specific calculation formula is: ； in, represents the abnormal score of the nth user's virtual impedance at the tth moment, represents element-wise product, represents the normalization function; The anomaly scores of the user virtual impedance at all times are summed to obtain the comprehensive anomaly score of the user virtual impedance.

6. The method for locating leakage fault of a branch line before a meter according to claim 1, characterized in that: The leakage fault is located according to the comprehensive abnormal score of the user virtual impedance, specifically including: The leakage fault location is determined based on the number of anomalies in a single leakage fault location and the comprehensive anomaly score of all user virtual impedances.

7. The method for locating leakage faults in branch lines before a meter according to any one of claims 1 to 6, characterized in that: The fault location method further includes: After eliminating the leakage fault according to the leakage fault location result, obtain the residual current of the substation, and determine whether the leakage fault still exists according to the residual current of the substation; if so, repeat the leakage fault location step.

8. An electronic device comprising a memory, a processor, and a computer program / instruction stored in the memory, characterized in that: The processor executes the computer program / instruction to implement the method for locating a leakage fault in a branch line before a meter according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the method for locating leakage faults in a branch line before a meter as claimed in any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the method for locating leakage faults in a branch line before a meter as claimed in any one of claims 1 to 7 is implemented.

Images