Distribution line concealment fault identification method and system

Abstract

The embodiment of the invention discloses a distribution line concealment fault identification method and system, and the method comprises the steps: carrying out the training according to a fault data set, and obtaining a first fault identification model and a second fault identification model; acquiring to-be-detected power distribution circuit data; respectively inputting the to-be-detected power distribution circuit data into the first fault recognition model and the second fault recognition model to obtain a first fault type output by the first fault recognition model and a second fault type output by the second fault recognition model; and determining a target fault type corresponding to the to-be-detected power distribution circuit data according to the first fault type and the second fault type. According to the method, the two independent fault identification models are trained and the prediction results are comprehensively utilized, so that the misjudgment risk possibly existing in a single model is effectively reduced, and the identification accuracy and reliability of the hidden fault of the distribution line are improved.

Description

Technical Field

[0001] The present invention relates to fault identification technology, and in particular to a method and system for identifying hidden faults of distribution lines. Background Art

[0002] In the field of research and management of hidden faults in distribution lines, although some technologies and methods have been applied to monitoring, prediction and experimental verification, there are still many problems and challenges.

[0003] First, in terms of data collection and processing, it is difficult to accumulate enough historical data to train the model due to the low frequency of some hidden faults. In addition, the data formats from different sources are not unified, and there is a lack of effective data fusion mechanisms, which makes it difficult to comprehensively consider multiple factors. Secondly, the existing feature extraction and selection methods mainly focus on electrical parameters, ignoring the impact of environmental factors (such as temperature, humidity), geographical location, and socioeconomic activities on faults. For specific hidden faults, such as small animal bridging, existing methods are difficult to accurately capture key features, limiting the predictive ability of the model. In the process of model construction, physical models are often based on idealized assumptions and fail to fully consider complex factors in the actual environment, such as material aging and changes in environmental conditions. Traditional statistical models perform poorly when dealing with nonlinear and high-dimensional data, and it is difficult to capture the complex dynamic characteristics of hidden faults. Although deep learning models can handle such problems, they require a large amount of high-quality data for training, while the amount of data for hidden faults is usually small and difficult to obtain. In addition, existing research methods are insufficient in dealing with the spatiotemporal dynamic evolution characteristics of faults, making it difficult to fully reveal the evolution laws of faults in time and space. The development of hidden faults is affected by multiple factors. Existing models can only handle a single factor and lack research on the coupling of multiple factors. In terms of simulation tests, existing platforms can usually only simulate a single type of fault and lack the ability to comprehensively test multiple hidden faults. The construction and operation costs of simulation test platforms are high, which limits their large-scale application and continuous improvement capabilities. It is difficult to fully reproduce all variables in the actual environment under laboratory conditions, especially fault simulation under extreme weather conditions.

[0004] Finally, in the integrated management system, the existing system performs poorly in the early identification of hidden faults and the prediction of dynamic evolution characteristics, and lacks forward-looking operation and maintenance recommendations. Summary of the invention

[0005] The main purpose of the present invention is to provide a method and system for identifying hidden faults in distribution lines, so as to solve the technical problem that the detection accuracy of hidden faults in distribution lines is not high.

[0006] To achieve the above-mentioned purpose, the first aspect of the present application provides a method for identifying hidden faults of distribution lines, which is applied to a hidden fault identification system of distribution lines. The method comprises:

[0007] According to the fault data set, a first fault identification model and a second fault identification model are obtained by training;

[0008] Acquire data of the power distribution circuit to be detected;

[0009] Inputting the to-be-detected power distribution circuit data into the first fault identification model and the second fault identification model respectively, obtaining a first fault type output by the first fault identification model and a second fault type output by the second fault identification model;

[0010] According to the first fault type and the second fault type, a target fault type corresponding to the power distribution circuit data to be detected is determined.

[0011] On the other hand, the present application also provides a distribution line hidden fault identification device, comprising:

[0012] A training module, used for training a first fault recognition model and a second fault recognition model according to a fault data set;

[0013] An acquisition module, used for acquiring data of a power distribution circuit to be detected;

[0014] A determination module, used for inputting the to-be-detected power distribution circuit data into the first fault identification model and the second fault identification model respectively, to obtain a first fault type output by the first fault identification model and a second fault type output by the second fault identification model;

[0015] The determination module is further used to determine a target fault type corresponding to the power distribution circuit data to be detected according to the first fault type and the second fault type.

[0016] On the other hand, the present application also provides a distribution line hidden fault identification system, including a memory and a processor, the memory storing a computer program, and when the computer program is executed by the processor, the processor executes the steps of the first aspect and any possible implementation method thereof.

[0017] On the other hand, the present application further provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes each step of the method described in the first aspect.

[0018] This method can significantly improve the recognition accuracy and reliability of hidden faults in distribution lines by training two independent fault identification models and comprehensively utilizing their prediction results. That is, this method effectively reduces the risk of misjudgment that may exist in a single model through a multi-model fusion strategy, ensuring the comprehensiveness and accuracy of fault diagnosis. At the same time, this integrated learning method can enhance the system's ability to recognize complex fault modes, thereby accelerating fault response speed, reducing maintenance costs, and ultimately improving the operational safety and stability of distribution lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] in:

[0021] Figure 1 A schematic diagram of a flow chart of a method for identifying hidden faults in a power distribution line provided in an embodiment of the present application;

[0022] Figure 2 A schematic diagram of the structure of a first initial model provided in an embodiment of the present application;

[0023] Figure 3 A schematic diagram of the structure of a hidden fault identification device for a power distribution line provided in an embodiment of the present application;

[0024] Figure 4 A schematic diagram of the structure of a distribution line hidden fault identification system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0026] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0027] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0028] The technical solution of the present application can be applied to various distribution line hidden fault identification scenarios.

[0029] In addition to the existing problems raised by the above background technology, the following problems also exist: For specific fault types, the existing technology also has the following specific problems: tree barrier monitoring means are difficult to detect changes in the distance between trees and conductors in real time; changes in the spacing between buildings and conductors in line room faults are not easy to monitor, and there is a lack of effective early warning mechanisms; the aging monitoring method for insulator breakdown is not sensitive enough and difficult to detect in time; the performance degradation process of lightning arresters is slow, and existing monitoring means are difficult to capture subtle changes; the status monitoring of drop-out fuses is difficult, especially under severe weather conditions; the bridging activities of small animals are difficult to predict, and the existing protective measures are limited in effect; changes in the electrical properties of cement poles after being damp are difficult to monitor in real time.

[0030] This method can significantly improve the recognition accuracy and reliability of hidden faults in distribution lines by training two independent fault identification models and comprehensively utilizing their prediction results. That is, this method effectively reduces the risk of misjudgment that may exist in a single model through a multi-model fusion strategy, ensuring the comprehensiveness and accuracy of fault diagnosis. At the same time, this integrated learning method can enhance the system's ability to recognize complex fault modes, thereby accelerating fault response speed, reducing maintenance costs, and ultimately improving the operational safety and stability of distribution lines.

[0031] The embodiments of the present application are described below in conjunction with the drawings in the embodiments of the present application.

[0032] See also Figure 1 , is a flow chart of a method for identifying hidden faults in a power distribution line provided in an embodiment of the present application, such as Figure 1As shown, the method includes:

[0033] 101. According to the fault data set, a first fault identification model and a second fault identification model are trained.

[0034] The above fault data set contains data on historical faults, including normal operation data and data under various fault conditions. The fault data is used to train the model so that it can learn the characteristics and patterns of the faults.

[0035] Specifically, the fault types in the fault data set may be hidden faults such as tree barriers, line room faults, insulator breakdown, lightning arrester breakdown, drop-out fuse abnormality, small animal bridging, and wet cement poles, and there is no unique limitation here.

[0036] Among them, the above-mentioned first fault identification model can be a one-dimensional convolutional neural network (1D CNN) model, and the first fault identification model is used for fault type classification (FTC) and line fault detection (LFI) of distribution lines. The first fault identification model directly processes the original fault data in the fault data set for model training without any preprocessing. By utilizing the 1D convolution layer, the first fault identification model can effectively extract features from the three-phase voltage and current measurement values ​​and achieve high-precision fault identification. The first fault identification model focuses on the processing of numerical data sets. The first fault identification model performs functions such as feature extraction and reducing network size through convolution layers, pooling layers, and fully connected layers. This structure enables the model to directly process raw data without preprocessing. The proposed 1D CNN model consists of ten levels.

[0037] Among them, the above-mentioned second fault identification model can be an artificial neural network (ANN) model, including a feedforward neural network (FNN) and a recurrent neural network (RNN). Among them, long short-term memory (LSTM) networks are particularly suitable for time series classification because they can solve problems related to gradient explosion and disappearance. The second fault identification model has undergone two training methods: Normal Classification (NC) and Cross-Fold Validation (CV). NC involves dividing the data set into a training set and a test set, while CV involves dividing the data set into 10 levels, each level has different training and test sets.

[0038] Among them, although the above-mentioned first fault identification model and the second fault identification model are different in structure and the way of processing data, they are both aimed at achieving the same goal: identifying and classifying different fault types from the measurement data of the power system.

[0039] It can be seen that the trained first and second fault identification models in this embodiment provide accurate identification for fault diagnosis of distribution lines, which helps to timely identify and respond to potential faults, thereby reducing power outages and maintenance costs.

[0040] 102. Obtain data of the power distribution circuit to be tested.

[0041] The above-mentioned power distribution circuit data to be detected may include but is not limited to real-time data and historical data.

[0042] Specifically, real-time data includes real-time measurement values ​​of parameters such as voltage, current, frequency, temperature, and power. The real-time data reflects the current working status of the distribution circuit; historical data records the operating data of the distribution circuit over a period of time in the past, including operation records, maintenance records, fault records, etc.

[0043] Optionally, the frequency of data collection of the distribution circuit to be inspected depends on the monitoring requirements. For real-time monitoring, data collection may be continuous or at a high frequency (such as every second or every minute); for non-critical monitoring, data collection may be performed at a lower frequency (such as every hour or every day).

[0044] It can be seen that in this embodiment, by acquiring the data of the power distribution circuit to be detected, a more in-depth analysis is further performed to identify potential problems, predict future failures, and optimize the operating efficiency of the power distribution system.

[0045] 103. Input the to-be-detected power distribution circuit data into the first fault identification model and the second fault identification model respectively to obtain a first fault type output by the first fault identification model and a second fault type output by the second fault identification model.

[0046] The power distribution circuit data to be detected are input into two fault identification models respectively, which means that the same set of data will be analyzed independently by the two models respectively.

[0047] The first fault type and the second fault type may be specific fault category labels, such as short circuit, overload, insulation aging, etc., or probability distribution of faults, indicating the possibility that the data belongs to each fault category.

[0048] Furthermore, the first fault type and the second fault type are prediction results of the two models respectively. The two results may be the same or different, depending on the training data, algorithm and structure of the model.

[0049] It can be seen that in this embodiment, the first fault identification model and the second fault identification model are used to quickly identify and classify potential faults in the power distribution circuit, thereby improving the reliability and operation and maintenance efficiency of the power distribution system.

[0050] 104. Determine a target fault type corresponding to the power distribution circuit data to be detected according to the first fault type and the second fault type.

[0051] Among them, in order to determine the final target fault type, the first fault type and the second fault type need to be analyzed and fused. The specific implementation process may include but is not limited to the following processes: voting mechanism, if both models predict the same fault type, then this type is considered to be the target fault type; probability fusion, if the model outputs a probability distribution, the target fault type can be determined by calculating the average or weighted average of the output probabilities of the two models; decision rules, some decision rules can be set, for example, if the prediction confidence of one model is higher than that of another, then the prediction result of the model with higher confidence is given priority.

[0052] It can be seen that in this embodiment, multiple prediction results are obtained through the prediction of multiple models, and the target fault type is further determined based on the multiple prediction results to improve the accuracy and reliability of fault detection, thereby providing strong support for the maintenance and management of the power distribution system.

[0053] This method can significantly improve the recognition accuracy and reliability of hidden faults in distribution lines by training two independent fault identification models and comprehensively utilizing their prediction results. That is, this method effectively reduces the risk of misjudgment that may exist in a single model through a multi-model fusion strategy, ensuring the comprehensiveness and accuracy of fault diagnosis. At the same time, this integrated learning method can enhance the system's ability to recognize complex fault modes, thereby accelerating fault response speed, reducing maintenance costs, and ultimately improving the operational safety and stability of distribution lines.

[0054] In one embodiment, the training of the first fault identification model and the second fault identification model based on the fault data set includes: performing fault simulation on a simulation platform in the distribution line hidden fault identification system to obtain a fault data set; training the first initial model and the second initial model respectively according to the fault data set to obtain the first fault identification model and the second fault identification model, wherein the first fault identification model is a model obtained after the training of the first initial model is completed, and the second fault identification model is a model obtained after the training of the second initial model is completed.

[0055] Among them, fault simulation is to simulate various hidden faults on the simulation platform of the distribution line hidden fault identification system, such as tree barriers, line room faults, insulator breakdown, etc. The purpose of these simulated faults is to generate fault data that is difficult to obtain in the real world for model training.

[0056] Among them, the fault data set includes electrical parameters such as current, voltage, power, fault characteristics and corresponding fault labels, etc. The fault data set also includes normal operation data so that the model can learn the difference between normal state and fault state.

[0057] The first initial model and the second initial model may have different architectures or parameter settings to capture different types of fault features or provide complementary information. Optionally, the first initial model may be a one-dimensional convolutional neural network, and the second initial model may be an artificial neural network (ANN) model.

[0058] It can be seen that in this example, the data set generated by the simulation platform is used to train a model that can accurately identify distribution line faults, thereby improving the efficiency and accuracy of fault detection.

[0059] In one embodiment, after the first initial model and the second initial model are trained respectively according to the fault data set to obtain the first fault identification model and the second fault identification model, the method further includes: selecting verification fault data in the fault data set; inputting the verification fault data into the simulation platform, the first fault identification model and the second fault identification model respectively to obtain the third fault type output by the simulation platform, the fourth fault type output by the first fault identification model and the fifth fault type output by the second fault identification model; comparing the fourth fault type and the fifth fault type respectively according to the third fault type to obtain a first comparison result corresponding to the fourth fault type and a second comparison result corresponding to the fifth fault type; if the comparison results of the first comparison result and / or the second comparison result are inconsistent, adjusting the first fault identification model and / or the second fault identification model.

[0060] The verification fault data set does not participate in the training of the model, but is used for subsequent model verification, that is, to evaluate the performance of the model and ensure that the model can accurately identify the fault type.

[0061] Among them, the third fault type is the prediction or simulation result of the fault type by the simulation platform based on the verification data set.

[0062] Specifically, the output of the simulation platform (the third fault type) is compared with the outputs of the two models (the fourth fault type and the fifth fault type) to evaluate the accuracy of the model or to evaluate the performance of the model, such as accuracy, recall, F1 score, etc.

[0063] Specifically, inconsistent comparison results may mean that the model has problems such as overfitting, underfitting or data imbalance; the consistency of the comparison results can serve as an important basis for model adjustment.

[0064] Optionally, the process of adjustment may include, but is not limited to, retraining the model, changing the model structure, adjusting training parameters, increasing the diversity of the data set, etc. After adjusting the model, the validation process should be re-performed to ensure that the performance of the model is improved.

[0065] It can be seen that in this embodiment, the accuracy and reliability of the fault identification model are further ensured through the verification and adjustment process.

[0066] In one embodiment, the fault simulation is performed on a simulation platform in the distribution line hidden fault identification system to obtain a fault data set, including: obtaining at least one hidden fault; simulating at least one hidden fault type according to a three-phase fault block in the simulation platform; and collecting simulation data for the at least one hidden fault type to obtain a fault data set.

[0067] Among them, at least one hidden fault may include but is not limited to tree obstacles, line room faults, insulator breakdown, lightning arrester breakdown, drop-out fuse abnormality, small animal bridging, wet cement poles and other hidden faults.

[0068] Among them, the three-phase fault block refers to the fault conditions simulated in the three-phase power system, which may include single-phase ground fault, two-phase short circuit, two-phase ground fault or three-phase short circuit.

[0069] Among them, in the simulation platform, it is necessary to set fault parameters, such as the time, location, severity, etc. of the fault. The simulation may need to consider different operating conditions, such as load changes, ambient temperature, humidity and other factors.

[0070] The collected data may include time series data of parameters such as voltage, current, power, temperature, and frequency. Data collection should cover the entire time period before, during, and after the fault occurs, so that the model can learn the complete characteristics of the fault.

[0071] In the specific implementation, use MATLAB's Simulink or Power System Blockset to build a single-line diagram of the IEEE 6-bus system. The IEEE 6-bus system consists of three traditional voltage sources. Open MATLAB and start Simulink, create a new blank model in Simulink, and add components: Voltage source: add three voltage sources (Generators), connected to buses 1, 2, and 3 respectively; Transformer: add necessary transformers (Transformer) to connect different buses; Transmission line: add transmission line (Line) to connect each bus; Load: add load (Load) to simulate the electrical equipment in the system; Set parameters: set the parameters of each component according to the standard parameters of the IEEE 6-bus system. For example, the rated voltage of the voltage source, the transformation ratio of the transformer, the resistance and inductance of the transmission line, etc. Further, add a three-phase fault block, find the "Three-Phase Fault" block in the Simulink library browser, and add it to the model, connect the three-phase fault block to the transmission line that needs to simulate the fault; set the fault type of the three-phase fault block, you can choose single-phase ground fault (AG, BG, CG), two-phase short circuit fault (AB, BC, CA), three-phase short circuit fault (ABC), etc.; set the specific fault type by selecting different phases (A, B, C) and ground G; divide the transmission line into two parts and insert a breakpoint in the middle; change the resistance and inductance values ​​of each part to simulate faults in different locations. For example, the resistance and inductance of the first half can be set to lower values, and the second half can be set to higher values ​​to simulate a fault close to one side. To further set the load curve, add a "From Workspace" block in Simulink to input different expected load curves; import the load curve data into the MATLAB workspace and connect it to the corresponding load block; add measurement modules, add current measurement modules (Current Measurement) and voltage measurement modules (Voltage Measurement), and connect them to both sides of the transmission line respectively. These modules will record the current and voltage data during the fault; set the simulation time to ensure that the time period when the fault occurs and lasts is covered; run the simulation, collect current and voltage data, and finally obtain the fault data set.

[0072] It can be seen that in this example, the fault scenario is simulated through the simulation platform to obtain the fault data set. The data set is the basis for training the fault identification model. By learning these data, the model can improve its ability to identify hidden faults in actual distribution lines.

[0073] In one embodiment, the first initial model is trained according to the fault data set to obtain a first fault identification model, including: inputting the fault data set into the first convolutional layer group in the first initial model to obtain a first feature map; inputting the first feature map into the first pooling layer in the first initial model to obtain a second feature map processed by the first pooling layer; inputting the second feature map into the second convolutional layer group in the first initial model to obtain a third feature map; inputting the third feature map into the second pooling layer in the first initial model to obtain a fourth feature map processed by the second pooling layer; inputting the fourth feature map into the flattening layer in the first initial model to obtain a one-dimensional feature vector; inputting the one-dimensional feature vector into the first fully connected layer in the first initial model to obtain a second feature vector processed by the fully connected layer; inputting the two-dimensional feature vector into the output layer in the first initial model to obtain a first training fault type; training the first initial model according to the fault data set in sequence until a first preset stop condition is met, then stopping the training to obtain the first fault identification model.

[0074] The specific structure of the first initial model can be as follows: Figure 2 As shown. Figure 2 In the structure from left to right, 2x(Conv1D+tanh)(6x 16) represents the first convolutional layer group, 2x(Conv1D+tanh)(30x 32) represents the second convolutional layer group, Maxpooling(30x 16) represents the first pooling layer, Maxpooling(15x 32) represents the second pooling layer, FlattenLayer(480x 1) represents the flattening layer, FCT represents the fully connected layer, and FLE and FLU constitute the output layer.

[0075] Among them, the first convolutional layer group can include two convolutional layers 2x(Conv1D+tanh)(6x 16), the convolution kernel size of the first convolution layer is 6, and the number of output channels is 16; the convolution kernel size of the second convolution layer is 6, and the number of output channels is 16.

[0076] Specifically, the first convolutional layer group extracts local features in the input data through a convolution operation to generate a first feature map.

[0077] Among them, the first pooling layer includes a pooling window size of 2 and a step size of 2. The first pooling layer reduces the size of the feature map through a downsampling operation while retaining important feature information to obtain a second feature map.

[0078] Among them, each convolution layer in the second convolution layer group has 32 filters and the convolution kernel size is 6, so the output feature map size is 22x32.

[0079] Among them, the second pooling layer includes a pooling window size of 2 and a step size of 2, so the output feature map size is 10x32.

[0080] Among them, the flattening layer expands all feature maps into a one-dimensional vector to prepare for the fully connected layer.

[0081] Among them, the one-dimensional feature vector in the first fully connected layer is input into the first fully connected layer, and the fully connected layer performs weighted summation on the features to obtain the second feature vector.

[0082] Among them, the output layer outputs the probability distribution after being processed by the softmax activation function, which indicates the probability that the input data belongs to each category, and the first training fault type is obtained.

[0083] Specifically, the model will be iteratively trained multiple times according to the fault data set until a first preset stop condition is met, such as reaching a certain accuracy rate or number of iterations, at which time the training is stopped to obtain the final first fault recognition model.

[0084] The model was trained using a 10-fold cross validation method and a custom train / test split. For cross validation, the model was trained for 50 epochs; for the custom split, the model was trained for 100 epochs with a batch size of 20.

[0085] The first preset stop condition may be reaching a certain accuracy rate or number of iterations, which is not limited here.

[0086] Specifically, the first initial model directly processes the original data samples in the fault data set for model training without any preprocessing. The first initial model performs functions such as feature extraction and network size reduction through convolutional layers, pooling layers, and fully connected layers.

[0087] It can be seen that in this example, through the 1D convolution layer in the first initial model, the model can effectively extract features from the three-phase voltage and current measurement values ​​and achieve high-precision fault identification. Therefore, the trained first fault identification model can learn the characteristics of hidden faults from the distribution line data and use them for actual fault diagnosis.

[0088] In one embodiment, the second initial model is trained according to the fault data set to obtain the second fault identification model, including: performing data set division on the fault data set to obtain a fault training set and a fault test set; performing data preprocessing on the fault training set to obtain a processed fault training set; inputting the processed fault training set into the first hidden layer in the second initial model to obtain a first activation value output by the first hidden layer; inputting the first activation value into the second hidden layer in the second initial model to obtain a second activation value output by the second hidden layer; inputting the second activation value into the third hidden layer in the second initial model to obtain a third activation value output by the third hidden layer; inputting the third activation value into the second fully connected layer in the second initial model to obtain a second training fault type; and training the second initial model according to the fault data set in sequence until a second preset stop condition is met, and then stopping the training to obtain the second fault identification model.

[0089] The fault data set is divided into two parts, a fault training set and a fault test set. The fault training set is used to train the model, while the fault test set is used to evaluate the performance of the model. Specifically, the data set division ratio can be based on empirical values ​​or manually set, and is not limited here.

[0090] Among them, data preprocessing includes normalization, standardization, removal of missing values, feature engineering, etc., which are not limited here.

[0091] Specifically, the processed fault training set is input into the first hidden layer of the second initial model, the neurons in the hidden layer perform weighted summation on the input data, and generate a first activation value through an activation function; the first activation value is passed as input to the second hidden layer, and is again processed through the weighted summation of the neurons and the activation function to obtain a second activation value; similarly, the second activation value is input into the third hidden layer, and a third activation value is generated; the third activation value is then input into the second fully connected layer, and the fully connected layer performs weighted summation on the activation values ​​to obtain a prediction of the second training fault type; the fault training set is further used for multiple iterative training, and the weights of the network are updated in each iteration until a second preset stop condition is met.

[0092] Among them, the second preset stopping condition can be reaching a specific performance indicator or number of iterations. The second stopping condition should be based on actual application requirements and model performance. Reasonable stopping conditions help prevent overfitting.

[0093] It can be seen that in this example, the second fault recognition model obtained by model training with preprocessed data can learn features from the fault data of the distribution line and be used to identify and classify different fault types.

[0094] In one embodiment, after the second initial model is trained in sequence according to the fault data set until a second preset stop condition is met, the training is stopped, and the second fault identification model is obtained, the method further includes: inputting the fault test set into the second fault identification model to obtain a test fault type; obtaining a data label in the fault test set; comparing the data label with the test fault type to obtain a performance indicator of the second fault identification model; and determining, based on the performance indicator, whether the second fault identification model is used for hidden fault identification of distribution lines.

[0095] The fault test set is used to evaluate the performance of the second fault identification model, and based on the evaluation result, it is decided whether to apply the model to the identification of hidden faults of distribution lines.

[0096] Among them, the data label is the actual fault type of each sample in the fault test set. These labels are known and are used to compare with the prediction results of the model.

[0097] Among them, the test fault type output by the model is compared with the actual data label of the fault test set to evaluate the prediction accuracy of the model.

[0098] Among them, performance indicators are the key to evaluating the quality of the model. They provide quantitative information about how the model may perform in actual applications. By comparison, a series of performance indicators can be obtained, such as accuracy, recall, F1 score, confusion matrix, etc., which quantify the performance of the model.

[0099] Specifically, based on the results of the performance indicators, it is decided whether to deploy the second fault identification model into actual applications. If the performance indicators meet the predetermined threshold, the model can be used to identify hidden faults; if the performance does not meet the standard, it may be necessary to further adjust the model or collect more data for training. Furthermore, if the performance indicators show that the model is insufficient, it may be necessary to tune the model, including adjusting the model structure, changing the training parameters, increasing the size or diversity of the data set, etc.

[0100] It can be seen that in this example, the second fault identification model is tested by the fault test set to ensure that the second fault identification model has met the standard before being applied to the hidden fault identification of the distribution line.

[0101] In one embodiment, determining the target fault type corresponding to the distribution circuit data to be detected based on the first fault type and the second fault type includes: when the first fault type and the second fault type are the same fault type, determining that the target fault type corresponding to the distribution circuit data to be detected is the first fault type and the second fault type; or, when the first fault type and the second fault type are not the same fault type, performing data analysis on the first fault type and the second fault type to obtain an analysis result; and based on the analysis result, determining that the first fault type or the second fault type is the target fault type corresponding to the distribution circuit data to be detected.

[0102] Among them, the same fault type is the same fault type output by the first fault identification model and the second fault identification model, that is, they both identify the same type of fault, then it can be directly determined that the target fault type of the distribution circuit data to be detected is the fault type jointly identified by these two models.

[0103] Among them, the fault types output by the first fault identification model and the second fault identification model are not the same fault type; then it is necessary to further analyze the two fault types, which may include comparing the confidence of the two models, considering the historical performance of the models, analyzing the severity and urgency of the fault, etc.

[0104] Specifically, confidence analysis is to compare the prediction confidence of the two models for their respective fault types, and select the fault type with higher confidence; historical performance records are the performance of the reference model on historical data, and the fault type identified by the model with more stable historical performance is selected.

[0105] Furthermore, based on the result of the data analysis, it is determined which fault type is more credible or more consistent with the actual situation, thereby determining the target fault type of the power distribution circuit data to be detected.

[0106] It can be seen that in this example, even when the fault types are inconsistent, the target fault type of the power distribution circuit data to be detected can be determined as accurately as possible, thereby improving the efficiency and accuracy of fault processing.

[0107] In one embodiment, the data analysis of the first fault type and the second fault type is performed to obtain an analysis result, including: weighting the first fault type and the second fault type respectively to obtain a first weighted value corresponding to the first fault type and a second weighted value corresponding to the second fault type; among the first weighted value and the second weighted value, selecting the largest weighted value as the target weighted value, and the fault type corresponding to the target weighted value is the analysis result.

[0108] The weighted processing process is to assign certain weights to the fault types (the first fault type and the second fault type) output by the two models respectively. The weight reflects the recognition credibility of the model for a specific fault type or the importance of the fault type in practical applications.

[0109] Among them, through weighted processing, weighted values ​​of each of the two fault types are obtained. These weighted values ​​may be calculated based on factors such as model performance, historical data, and the probability of fault occurrence, and are not limited here.

[0110] Optionally, weighted processing can be based on multiple logics. If a model performs better on historical data, its output may be given a higher weight; if the consequences of a certain fault type are more serious, then this fault type may be given a higher weight. Weighting can be achieved in the following ways: fixed weights, that is, fixed weights are set in advance based on expert knowledge or experience; dynamic weights, dynamically adjusting weights based on the real-time performance of the model or the real-time data of the fault; learning weights, learning the best weight distribution through machine learning algorithms. In some cases, a decision threshold needs to be set. Only when the weighted value exceeds this threshold, the fault type is considered credible.

[0111] Specifically, the fault type corresponding to the target weighted value is identified as the analysis result, that is, the fault type finally determined by the system.

[0112] It can be seen that in this example, through weighted processing, the information of multiple models can be more effectively utilized, the accuracy of fault diagnosis can be improved, and more reliable guidance can be provided for maintenance and repair work.

[0113] As another aspect of the embodiment of the present application, the embodiment of the present application provides a distribution line hidden fault identification device. The distribution line hidden fault identification device can be a software module, and the software module includes a number of instructions, which are stored in a memory, and the processor can access the memory and call the instructions for execution to complete the distribution line hidden fault identification method described in the above-mentioned various embodiments.

[0114] See also Figure 3 , Figure 3 Schematic diagram of a hidden fault identification device for a power distribution line provided in an embodiment of the present application. Figure 3 As shown, the hidden fault identification device of the power distribution line includes:

[0115] A training module 301 is used to train a first fault recognition model and a second fault recognition model according to a fault data set;

[0116] An acquisition module 302 is used to acquire the power distribution circuit data to be detected;

[0117] A determination module 303, configured to input the to-be-detected power distribution circuit data into the first fault identification model and the second fault identification model respectively, to obtain a first fault type output by the first fault identification model and a second fault type output by the second fault identification model;

[0118] The determination module 303 is further configured to determine a target fault type corresponding to the power distribution circuit data to be detected according to the first fault type and the second fault type.

[0119] This method can significantly improve the recognition accuracy and reliability of hidden faults in distribution lines by training two independent fault identification models and comprehensively utilizing their prediction results. That is, this method effectively reduces the risk of misjudgment that may exist in a single model through a multi-model fusion strategy, ensuring the comprehensiveness and accuracy of fault diagnosis. At the same time, this integrated learning method can enhance the system's ability to recognize complex fault modes, thereby accelerating fault response speed, reducing maintenance costs, and ultimately improving the operational safety and stability of distribution lines.

[0120] It can be understood that the relevant contents of each module in the above device have been described in detail in the aforementioned method embodiment, and the details can be referred to the contents in the method embodiment; that is, the hidden fault identification device for distribution line provided in the present application can be executed as follows: Figure 1 Any steps in the illustrated embodiment will not be described in detail here.

[0121] See also Figure 4 , Figure 4 1 is a schematic diagram of a system for identifying hidden faults in distribution lines provided in an embodiment of the present application. The system for identifying hidden faults in distribution lines includes one or more processors 41 and a memory 42. The memory 42 is connected to the one or more processors 41, for example, connected to the processors 41 via a bus.

[0122] The processor 41 is configured to support the distribution line hidden fault identification system to perform the corresponding functions in the method in the above method embodiment. The processor 41 can be a central processing unit (CPU), a network processor (NP), a hardware chip or any combination thereof. The above hardware chip can be an application specific integrated circuit (ASIC), a programmable logic device (PLD) or a combination thereof. The above PLD can be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL) or any combination thereof.

[0123] The memory 42 is used to store program codes, etc. The memory 42 may include a volatile memory (VM), such as a random access memory (RAM); the memory 42 may also include a non-volatile memory (NVM), such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD); the memory 42 may also include a combination of the above-mentioned types of memories.

[0124] The memory 42 may be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the method for identifying hidden faults of distribution lines in the embodiment of the present application. The processor 41 executes various functional applications and data processing of the method for identifying hidden faults of distribution lines and the device for identifying hidden faults of distribution lines by running the non-volatile software programs, instructions and modules stored in the memory 42, that is, realizes the functions of the various modules or units of the method for identifying hidden faults of distribution lines and the device for identifying hidden faults of distribution lines provided in the above method embodiment.

[0125] The memory 42 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function. The data storage area may store data created according to the use of the distribution line hidden fault identification device, etc. In some embodiments, the memory 42 may optionally include a memory 42 remotely arranged relative to the processor 41, and these remote memories 42 may be connected to the distribution line hidden fault identification device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0126] The one or more modules are stored in the memory 42, and when executed by the one or more processors 41, execute the distribution line hidden fault identification method in any of the above method embodiments, for example, execute the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.

[0127] In one embodiment, a computer-readable storage medium is further provided. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the processor executes any step in the above method embodiment.

[0128] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0129] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0130] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying hidden faults in distribution lines, characterized in that: Applied to a hidden fault identification system for a power distribution line, the method comprises: According to the fault data set, a first fault identification model and a second fault identification model are obtained by training; Acquire data of the power distribution circuit to be detected; Inputting the to-be-detected power distribution circuit data into the first fault identification model and the second fault identification model respectively, obtaining a first fault type output by the first fault identification model and a second fault type output by the second fault identification model; According to the first fault type and the second fault type, a target fault type corresponding to the power distribution circuit data to be detected is determined.

2. The method for identifying hidden faults in distribution lines according to claim 1, characterized in that: The training of the first fault identification model and the second fault identification model according to the fault data set includes: Performing fault simulation on a simulation platform in the distribution line hidden fault identification system to obtain a fault data set; The first initial model and the second initial model are trained respectively according to the fault data set to obtain a first fault identification model and a second fault identification model, wherein the first fault identification model is a model obtained after the first initial model is trained, and the second fault identification model is a model obtained after the second initial model is trained.

3. The method for identifying hidden faults in distribution lines according to claim 2, characterized in that: After respectively training the first initial model and the second initial model according to the fault data set to obtain the first fault identification model and the second fault identification model, the method further includes: Selecting verification fault data from the fault data set; Inputting the verification fault data into the simulation platform, the first fault identification model and the second fault identification model respectively, to obtain a third fault type output by the simulation platform, a fourth fault type output by the first fault identification model and a fifth fault type output by the second fault identification model; Comparing the fourth fault type and the fifth fault type respectively according to the third fault type, to obtain a first comparison result corresponding to the fourth fault type and a second comparison result corresponding to the fifth fault type; If the comparison results of the first comparison result and / or the second comparison result are inconsistent, the first fault identification model and / or the second fault identification model are adjusted.

4. The method for identifying hidden faults in distribution lines according to claim 2, characterized in that: The fault simulation is performed on the simulation platform in the distribution line hidden fault identification system to obtain a fault data set, including: Obtain at least one hidden fault; In the simulation platform, simulating the at least one hidden fault type according to the three-phase fault block; Simulation data for the at least one hidden fault type is collected to obtain a fault data set.

5. The method for identifying hidden faults in distribution lines according to claim 2, characterized in that: The step of training the first initial model according to the fault data set to obtain a first fault identification model comprises: Inputting the fault data set into the first convolutional layer group in the first initial model to obtain a first feature map; Inputting the first feature map into the first pooling layer in the first initial model to obtain a second feature map processed by the first pooling layer; Inputting the second feature map into the second convolutional layer group in the first initial model to obtain a third feature map; Inputting the third feature map into the second pooling layer in the first initial model to obtain a fourth feature map processed by the second pooling layer; Inputting the fourth feature map into the flattening layer in the first initial model to obtain a one-dimensional feature vector; Inputting the one-dimensional feature vector into a first fully connected layer in the first initial model to obtain a second feature vector processed by the fully connected layer; Inputting the two-dimensional feature vector into the output layer of the first initial model to obtain a first training fault type; The first initial model is trained in sequence according to the fault data set until a first preset stop condition is met, and then the training is stopped to obtain the first fault identification model.

6. The method for identifying hidden faults in distribution lines according to claim 2, characterized in that: The step of training the second initial model according to the fault data set to obtain a second fault identification model includes: Dividing the fault data set into a fault training set and a fault test set; Performing data preprocessing on the fault training set to obtain a processed fault training set; Inputting the processed fault training set into the first hidden layer in the second initial model to obtain a first activation value output by the first hidden layer; Inputting the first activation value into a second hidden layer in the second initial model to obtain a second activation value output by the second hidden layer; Inputting the second activation value into a third hidden layer in the second initial model to obtain a third activation value output by the third hidden layer; Inputting the third activation value into a second fully connected layer in the second initial model to obtain a second training fault type; The second initial model is trained in sequence according to the fault data set until a second preset stop condition is met, and the training is stopped to obtain the second fault identification model.

7. The method for identifying hidden faults in distribution lines according to claim 1, characterized in that: After the second initial model is trained sequentially according to the fault data set until a second preset stop condition is met, the training is stopped, and the second fault identification model is obtained, the method further includes: Inputting the fault test set into the second fault identification model to obtain a test fault type; Obtaining data labels in the fault test set; Comparing the data label with the test fault type to obtain a performance indicator of the second fault identification model; According to the performance indicator, it is determined whether the second fault identification model is used for hidden fault identification of distribution lines.

8. The method for identifying hidden faults in distribution lines according to claim 1, characterized in that: The determining, according to the first fault type and the second fault type, a target fault type corresponding to the to-be-detected power distribution circuit data includes: When the first fault type and the second fault type are the same fault type, determining that the target fault type corresponding to the to-be-detected power distribution circuit data is the first fault type and the second fault type; or, When the first fault type and the second fault type are not the same fault type, performing data analysis on the first fault type and the second fault type to obtain an analysis result; According to the analysis result, it is determined that the first fault type or the second fault type is the target fault type corresponding to the power distribution circuit data to be detected.

9. The method for identifying hidden faults in distribution lines according to claim 8, characterized in that: The performing data analysis on the first fault type and the second fault type to obtain analysis results includes: Performing weighted processing on the first fault type and the second fault type respectively to obtain a first weighted value corresponding to the first fault type and a second weighted value corresponding to the second fault type; Among the first weighted value and the second weighted value, the largest weighted value is screened out as the target weighted value, and the fault type corresponding to the target weighted value is the analysis result.

10. A distribution line hidden fault identification system, characterized in that: It includes a memory and a processor, the memory is connected to the processor, the processor is used to execute one or more computer programs stored in the memory, and when the processor executes the one or more computer programs, the distribution line hidden fault identification system implements the method described in any one of claims 1-9.

Images