A 10kv power distribution network line fault type identification method

Abstract

This invention proposes a method for identifying fault types in 10kV distribution network lines. First, the amplitude and phase angle features of the initial fault stage are extracted from voltage and current signals to form a preliminary trajectory. Then, wavelet transform decomposition is used to obtain millisecond-scale energy distribution to capture transient characteristic differences, and key components are selected for preliminary classification. Next, fine-grained features from multiple time windows are integrated, and support vector machines are used to calculate the similarity with sample trajectories to obtain probability estimates. Finally, transient trends and steady-state features are combined for final determination, thereby achieving accurate fault type identification. Finally, by clustering and summarizing stable feature patterns and feeding back to optimize the wavelet scale and model, a continuously adaptive identification framework is constructed, significantly improving the accuracy and reliability of fault classification and identification.

Description

Technical Field

[0001] This invention relates to a method for identifying fault types in 10kV distribution network lines. Background Technology

[0002] Distribution network fault types include single-phase grounding, phase-to-phase short circuits, transient faults, and permanent faults. Different types of faults exhibit significant differences in their occurrence and development. Timely differentiation of these types can guide targeted handling measures and avoid unnecessary power outages or equipment damage. Current fault identification methods mostly rely on the overall characteristics of steady-state electrical quantities or transient components after a fault occurs. However, these methods reveal significant shortcomings when facing the actual operating environment of distribution networks: due to the numerous branches, complex load types, and dispersed fault locations in distribution networks, faults often undergo drastic changes from initial germination to full formation within a very short period. Existing identification methods struggle to capture the subtle but crucial differences in electrical parameters during these rapid evolutions, leading to confusion in the judgment of certain complex fault scenarios. In particular, the similarity between transient and permanent faults in their early stages results in a high misjudgment rate.

[0003] The evolution of faults in distribution networks exhibits fine-grained characteristics on a millisecond-level time scale. Electrical parameters display unique morphological change trajectories at different micro-stages. However, existing methods generally neglect this fine-grained division of the time dimension, resulting in insufficient extraction of the dynamic characteristics of the entire fault process. Because of the lack of segmented characterization of each stage of a fault from its precursory appearance and rapid development to its stable continuation, it is impossible to effectively accumulate and compare the unique fingerprints of different fault types along their evolution paths, making it difficult to form a reliable basis for type differentiation in the early stages of a fault. When a single-phase ground fault occurs in a real line, small intermittent discharge characteristics may appear in the early stages. If only the overall waveform after several weeks is observed, it is easy to confuse it with a permanent short circuit. Once the subtle differences in the early evolution trajectory are missed, subsequent processing strategies may be biased. Therefore, how to accurately extract and analyze the minute changes in electrical parameters at each stage of the fault evolution process within a millisecond-level time window, thereby revealing the essential differences between transient and permanent faults on the complete evolution trajectory, has become a key issue for the refined identification of fault types in distribution network lines. Summary of the Invention

[0004] This invention proposes a fault type identification method for 10kV distribution network lines. By integrating waveform feature extraction, multi-scale energy analysis and dynamic trajectory matching, it can effectively distinguish the trajectory evolution differences between transient and permanent faults in millisecond-level transient data, thereby realizing the identification and classification of fault types.

[0005] The technical solution of this invention is implemented as follows: A method for identifying fault types in a 10kV distribution network line includes: Step S1: Collect voltage and current signals of the distribution network lines using a high sampling rate sensor to obtain millisecond-level time series data. Extract waveform amplitude and phase angle change characteristics of the initial stage of single-phase grounding fault or phase-to-phase short-circuit fault from the data to obtain a preliminary trajectory representation characterizing the fault evolution trend. Step S2: The initial trajectory representation signal is decomposed into multi-scale components using wavelet transform. Based on the difference in transient characteristics between transient and permanent faults, the energy distribution of the components on the millisecond scale is calculated to obtain fine-grained feature indices characterizing the details of the fault. Step S3: Select key components related to the fault type from the fine-grained feature indicators. If the energy distribution of the key component exceeds the preset judgment threshold, it is judged as a potential trajectory of a permanent fault; otherwise, it is classified as a dynamic feature of a transient fault, and an intermediate result of fault classification is obtained. Step S4: By integrating intermediate results from multiple time windows with corresponding fine-grained feature indicators, the support vector machine algorithm is used to train the pre-labeled fault samples, calculate the similarity between the trajectory and the sample trajectory, and obtain the probability estimate of the fault category. Step S5: Combine the transient component change trends of adjacent time periods to analyze steady-state electrical characteristics. If the similarity is higher than the preset confirmation threshold, it is determined to be a permanent fault; otherwise, the identification result is adjusted to a transient fault to obtain the final fault type label.

[0006] In another aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the method described thereon.

[0007] In another aspect, the present invention also provides a computer program product, comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method described herein.

[0008] The beneficial effects of this invention are as follows: First, the amplitude and phase angle features of the initial stage of the fault are extracted from the voltage and current signals to form a preliminary trajectory. Then, wavelet transform decomposition is used to obtain the millisecond-scale energy distribution to capture transient characteristic differences, and key components are selected for preliminary classification. Next, fine-grained features from multiple time windows are integrated, and support vector machines are used to calculate the similarity with the sample trajectory to obtain probability estimates. Finally, transient trends and steady-state features are combined for final judgment, thereby achieving accurate identification of fault types. Finally, by clustering and summarizing stable feature patterns and feeding back to optimize the wavelet scale and model, a continuously adaptive identification framework is constructed, which significantly improves the accuracy and reliability of fault classification and identification. Detailed Implementation

[0009] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0010] In one aspect, this invention provides a method for identifying fault types in 10kV distribution network lines, comprising: Step S1: The voltage and current signals of the distribution network lines are collected by a high sampling rate sensor to obtain millisecond-level time series data. The waveform amplitude and phase angle change characteristics of the initial stage of single-phase grounding fault or phase-to-phase short-circuit fault are extracted from the data to obtain a preliminary trajectory representation of the fault evolution trend.

[0011] A millisecond-level sequence is constructed by reading transient voltage and current values ​​from real-time electrical signal conversion. The initial fault data window is determined and extracted from the abnormal disturbance starting point where the instantaneous rate of change exceeds the mutation threshold. The instantaneous magnitude and phase offset of the decomposed signal are calculated for the initial fault data window to obtain the decomposed waveform amplitude features and phase angle change features. The waveform amplitude features and phase angle change features are mapped to dynamic coordinate points and connected to obtain a preliminary trajectory representation characterizing the fault evolution trend.

[0012] In one specific embodiment, a sensor with a sampling rate of 10kHz is used to collect 10,000 data points per second. For a 10kV distribution line, three-phase voltage and current signals are collected to obtain waveform data of 100 sampling points per millisecond. The data is stored in the form of timestamp and corresponding amplitude pairs. For example, when the timestamp is 0.001s, the voltage amplitude of phase A is 8.16kV and the current is 200A. Subsequently, high-frequency noise is removed by data preprocessing algorithms, such as wavelet denoising, to ensure signal quality and provide a reliable basis for subsequent analysis. Next, for single-phase grounding faults or phase-to-phase short-circuit faults, the waveform amplitude and phase angle change characteristics in the initial stage of the fault were extracted from the data. For example, in the first 5 ms after the fault occurred, the voltage amplitude of phase A was detected to drop sharply from 8.16kV to 2.5kV, and the current amplitude was detected to surge from 200A to 1500A. At the same time, the phase angle change was calculated by Fourier transform algorithm, and it was found that the phase angle difference between phase A voltage and current changed abruptly from 30 degrees to 85 degrees, indicating that there was a significant phase angle shift in the initial stage of the fault. These characteristic data were further quantified by time series analysis to form two indicators: amplitude decrease rate and phase angle change rate. The amplitude decrease rate was calculated as (8.16-2.5) / 5=1.132kV / ms, and the phase angle change rate was calculated as (85-30) / 5=11 degrees / ms. Finally, based on the extracted features, a preliminary trajectory representation of the fault evolution trend is constructed. For example, the amplitude decrease rate and phase angle change rate are used as two-dimensional coordinates to draw the fault evolution trajectory. The initial point is (0,0), and the trajectory point is (1.132,11) after 5ms. The trajectory curve is fitted by a linear interpolation algorithm to predict the change trend in the next 10ms. At the same time, the pattern matching of similar trajectories in the historical fault database is combined to determine that the current fault may evolve into a persistent ground fault, with a matching degree of 85%.

[0013] Step S2: The initial trajectory representation signal is decomposed into multi-scale components using wavelet transform. Based on the difference in transient characteristics between transient and permanent faults, the energy distribution of the components on the millisecond scale is calculated to obtain fine-grained feature indices that characterize the details of the fault.

[0014] The preliminary trajectory representation is decomposed into multiple layers using wavelet basis functions. Square integrals are then performed on the high-frequency detail components and low-frequency approximate components to obtain instantaneous frequency band energy values. Multi-scale permutation entropy is calculated based on the instantaneous frequency band energy values ​​to extract energy singularities. An energy evolution feature vector describing the time-varying characteristics of fault energy is constructed and mapped to a high-dimensional feature space to generate fine-grained feature indices characterizing fault details.

[0015] In one specific embodiment, the system first performs multi-scale decomposition on the acquired voltage and current signals using wavelet transform, specifically employing the Daubechies wavelet basis function (Db4) to decompose the signal into five scale levels, obtaining multiple components from low frequency to high frequency. In the first high-frequency component, transient disturbance signals with a frequency range of 2.5kHz to 5kHz are extracted, with a peak amplitude of 0.8kV. In the third low-frequency component, continuous fluctuation signals with a frequency range of 0.6kHz to 1.2kHz are extracted, with an amplitude stable at 0.3kV, and so on. These component data are stored in the form of a time-frequency matrix using an automated algorithm. Next, considering the differences in transient characteristics between transient and permanent faults, the system calculates the energy distribution of each scale component: the square value of each component is integrated over time. For example, in the first 10 ms after the fault occurs, the energy value of the first-level high-frequency component is 12.5 J, and the energy value of the third-level low-frequency component is 3.8 J. By comparison, it is found that the high-frequency energy accounts for 76.5%, indicating that the fault has obvious transient impact characteristics. Subsequently, the energy distribution data is divided into 10 time windows according to the millisecond scale, and the energy change rate is calculated in each window. For example, the high-frequency energy change rate is 2.1 J / ms in the first 5 ms, and then decreases to 0.4 J / ms in the next 5 ms, reflecting the concentrated energy release characteristics in the early stage of the fault. Finally, based on energy distribution and rate of change, the system generates fine-grained feature indicators. For example, the high-frequency energy proportion indicator is defined as 76.5%, and the energy attenuation rate indicator is (2.1-0.4) / 5=0.34J / ms². These indicators are automatically classified into current fault types by comparing them with the energy feature thresholds of transient and permanent faults in the historical database. If the high-frequency energy proportion exceeds 70% and the attenuation rate is greater than 0.3J / ms², it is judged as a transient fault with a classification confidence level of 82.3%, thus providing data basis for subsequent protection strategy optimization.

[0016] Step S3: Select key components related to the fault type from the fine-grained feature indicators. If the energy distribution of the key component exceeds the preset judgment threshold, it is judged as a potential trajectory of a permanent fault; otherwise, it is classified as a dynamic feature of a transient fault, and an intermediate result of fault classification is obtained. The mutual information algorithm is used to filter out key components from fine-grained feature indicators, and a sliding time window integration operation is performed to obtain the energy distribution value. If the energy distribution value exceeds the preset judgment threshold, it is determined to be a potential trajectory of a permanent fault; otherwise, it is determined to be a dynamic feature of a transient fault. The potential trajectory of the permanent fault and the dynamic feature of the transient fault are integrated to obtain the intermediate result of fault classification.

[0017] In one specific embodiment, after obtaining fine-grained feature indicators, the system uses a feature importance ranking algorithm based on random forest, according to a historical fault mode library, to screen out key components strongly correlated with the fault type from all components. For fault data of 10kV lines, the Gini impurity reduction of the second-scale component (corresponding to a frequency range of 1.25kHz to 2.5kHz) and the fourth-scale component (corresponding to a frequency range of 0.3125kHz to 0.625kHz) is evaluated to be 0.15 and 0.09 respectively, and is therefore identified as a key component. Subsequently, the system calls a preset judgment logic to perform threshold judgment on the energy distribution of these two key components. The average energy of the fourth-scale component is calculated within a time window from 20ms to 100ms after the fault occurs and compared with a preset threshold of 0.5J. If the average value exceeds the threshold, for example, a measured value of 0.68J, it indicates that low-frequency energy persists, which is consistent with the characteristics of permanent fault arc reignition or continuous insulation degradation, and the system marks it as a potential trajectory of a permanent fault. Conversely, if the mean is only 0.25J, below the threshold, and the energy of the second-scale component rapidly decays to the background noise level within the first cycle after the fault, the system classifies it as a dynamic characteristic of a transient fault, attributing it to lightning strikes or transient flashovers. The intermediate results generated in this classification process, along with the energy time-series curves of the key components, will be encapsulated into a structure and directly passed to the subsequent fault location and protection action decision-making module, forming a continuous analysis chain.

[0018] Step S4: By integrating intermediate results from multiple time windows with corresponding fine-grained feature indicators, the support vector machine algorithm is used to train the pre-labeled fault samples, calculate the similarity between the trajectory and the sample trajectory, and obtain the probability estimate of the fault category.

[0019] The intermediate results from multiple time windows are concatenated with the fine-grained features to construct a multidimensional temporal feature vector sequence. A radial basis function is used to map the multidimensional temporal feature vector sequence to obtain a high-dimensional feature mapping set. The high-dimensional feature mapping set is then input into a fault discrimination model constructed based on pre-labeled fault samples. The normal vector projection distance of the high-dimensional feature mapping set in the fault discrimination model is calculated to quantify the trajectory similarity value. The trajectory similarity value is normalized using a logistic regression function to obtain a probability estimate of the fault category.

[0020] In one specific embodiment, the system first constructs a comprehensive feature matrix by integrating intermediate results from multiple time windows with corresponding fine-grained feature data. The system extracts feature data from two time windows after the fault occurs: 0ms to 50ms and 50ms to 150ms. This data is normalized to form feature vectors. Next, the system trains a pre-labeled fault sample library using a support vector machine (SVM) algorithm. This library contains 500 known fault cases, including 300 persistent faults caused by equipment aging and 200 transient faults caused by external interference. During training, a radial basis function kernel is used, with a penalty parameter C set to 1.0 and a kernel parameter γ set to 0.5. After training, the model achieves an accuracy of 92.3%. Subsequently, the system inputs the feature vector of the current fault into the trained SVM model and calculates its similarity to the sample trajectory. Specifically, it calculates the similarity using Euclidean distance in the feature space, finding that the average distance between the current trajectory and persistent fault samples is 0.21, and the average distance with transient fault samples is 0.45. Ultimately, based on the similarity results and the probability estimates of the fault categories output by the softmax function, the system calculates that the probability of the current fault being a persistent fault is 0.78, and the probability of it being a transient fault is 0.22. These probability results are recorded as the basis for subsequent analysis. To form a complete logical chain, the system performs correlation analysis between the probability estimates and historical fault trend data, automatically generating classification labels and storing them in the database for use by the subsequent fault warning module, ensuring the continuity of the analysis process and the business loop.

[0021] Step S5: Combine the transient component change trends of adjacent time periods to analyze steady-state electrical characteristics. If the similarity is higher than the preset confirmation threshold, it is determined to be a permanent fault; otherwise, the identification result is adjusted to a transient fault to obtain the final fault type label.

[0022] The original electrical data corresponding to the probability estimation is obtained, and transient component sequences and steady-state electrical feature data are extracted from them. The gradient change rate is calculated based on the transient component sequences to obtain the transient component change trend. The transient component change trend is concatenated with the steady-state electrical feature data. The concatenated feature data is input into a preset fault feature library for matching, and a similarity value is calculated. If the similarity value is higher than a preset confirmation threshold, it is determined to be a permanent fault; otherwise, a transient fault identification result is generated to obtain the final fault type label.

[0023] In one specific embodiment, the system analyzes the fault data of a 10kV line based on the calculated probability estimate and the transient component change trend in adjacent time periods. It extracts transient component data within 100ms before and after the fault, finding that the transient voltage fluctuation amplitude decreases from the initial 0.15pu to 0.05pu, showing a clear attenuation trend. The system performs spectral analysis on the transient signal using a Fourier transform algorithm, obtaining a dominant frequency component of 50Hz, accompanied by a second harmonic component accounting for 0.03. Subsequently, combined with the probability estimate of 0.85 (indicating a tendency towards a permanent fault), the system further extracts steady-state electrical characteristics and analyzes the data within 200ms after the fault. The steady-state current data within s showed that the current value was stable at 1.2 times the rated value, deviating from the normal range of 0.9 to 1.1 times. The system used a time series-based anomaly detection algorithm to calculate that the deviation between the steady-state characteristics and historical normal data reached 0.25, exceeding the preset deviation threshold of 0.1. At the same time, the transient decay trend and steady-state deviation characteristics were input into the pre-constructed decision tree model, and the comprehensive similarity score was calculated to be 0.88, which is higher than the preset confirmation threshold of 0.75. Based on this, the system determined that the current fault was a permanent fault. If the similarity score was lower than the threshold, the system would automatically adjust to the identification result of a transient fault and generate the final fault type label as "permanent fault".

[0024] Step S6: Extract dynamic parameters that match the distribution network operation scenario from the final fault type label, use cluster analysis to summarize similar fault trajectories, determine stable feature patterns that characterize fault types, and obtain an optimized fault feature set.

[0025] The dynamic electrical parameters associated with the final fault type label are mapped to a multi-dimensional state vector to construct fault trajectory data. A similarity matrix is ​​calculated based on the fault trajectory data, and trajectories with similar evolution paths are grouped into a trajectory cluster set. The dispersion of each moment in the trajectory cluster set is calculated, and the center point values ​​below the convergence threshold are locked to determine the stable feature pattern. The stable feature pattern is mapped to the original feature space and the noise dimension is removed to obtain the optimized fault feature set.

[0026] In one specific embodiment, the system automatically filters parameters such as voltage fluctuation rate, current surge value, and power factor change during the fault occurrence period. Specifically, the data includes a voltage fluctuation rate of 0.12, a current surge value of 1.5 times the rated value, and a power factor decrease to 0.85. The system then compares these parameters with historical fault records, using a K-means clustering algorithm to group and analyze the parameters. With 3 cluster centers and 10 iterations, the Euclidean distance between the current fault parameters and a certain type of historical fault trajectory is calculated to be 0.08, lower than the preset distance threshold of 0.1, indicating a high degree of similarity. The system then summarizes the common characteristics of similar fault trajectories, such as current surge values ​​fluctuating between 1.4 and 1.6 times the rated value. Based on this result, the system further analyzes... By analyzing stable feature patterns in the trajectory data and extracting key variables using principal component analysis, it was found that voltage fluctuation rate and current mutation value contribute 0.45 and 0.38 to the characterization of fault type, respectively. Based on this, the system uses these two variables as primary features, supplemented by power factor change as a secondary feature, to form an optimized fault feature set. The specific feature values ​​are voltage fluctuation rate 0.12, current mutation value 1.5, and power factor 0.85. Finally, the system stores this set in the feature database and maps it to the load characteristics of the distribution network operation scenario. For example, when the load rate increases from 0.7 to 0.9, the feature value change trend is consistent, thus forming a complete logical chain from parameter extraction to feature optimization, ensuring that the subsequent fault prediction module can call these features for accurate analysis.

[0027] Step S7: Based on the optimized fault feature set, the scale parameters of the wavelet transform are adjusted. For subsequent signal acquisition and processing loops, it is determined whether the recognition performance has degraded to decide whether to update the support vector machine model, thus constructing a continuously optimized fault recognition framework.

[0028] The frequency band energy distribution density is calculated based on the optimized fault feature set to determine the feedback-adjusted wavelet transform scale parameters. The scale parameters are obtained by mapping the high-density frequency band to the center frequency of the wavelet basis function. The time-frequency wavelet coefficients are obtained using the feedback-adjusted wavelet transform scale parameters and input into the current support vector machine model to quantify the recognition performance index. If the recognition performance index is lower than a preset threshold, the weight vector and bias parameters are reconstructed based on the selected boundary feature vectors to generate a corrected model adapted to the new feature distribution. The corrected model is deployed to the online monitoring process to build a continuously optimized fault recognition framework.

[0029] In one specific embodiment, the system, based on an optimized fault feature set (e.g., feature vectors containing a high-frequency energy proportion of 0.35% for zero-sequence current and a transient voltage distortion rate of 0.18), automatically adjusts the scale parameters of the wavelet transform to enhance the targeting of feature extraction. Specifically, the number of wavelet packet decomposition layers is adjusted from 5 to 6, and the frequency range corresponding to the third scale is refined from 1.2kHz to 2.4kHz to 1.0kHz to 2.0kHz, to more accurately capture the specific frequency band fault transient signal indicated by the feature set. In subsequent real-time signal acquisition and processing loops, the system applies the adjusted wavelet transform to extract features from the current signals of 10 consecutively acquired time windows and calculates their cosine similarity with the optimized feature set. When the average similarity continuously decreases from 0.92 to 0.78, and falls below the performance threshold of 0.85 three times consecutively, the system automatically determines that the current feature recognition performance has degraded. To address this situation, the system triggers a model update process, retrieving the most recent 1000 fault samples containing new features from the historical database. A grid search algorithm is used to optimize the penalty factor C and kernel function parameter γ of the support vector machine model, with C ranging from [0.1, 10] and γ from [0.01, 1] expanded logarithmically. Five-fold cross-validation determines the optimal parameter combination as C=2.5 and γ=0.1. Subsequently, the system retrains the support vector machine model using these parameters and deploys the updated model to the online recognition module, establishing a new performance baseline. This constructs a closed-loop optimization framework from feature-guided signal processing to adaptive model updates.

[0030] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0031] In another aspect, the present invention also proposes a computer program product, comprising a computer program, characterized in that the computer program implements the above-described method when executed by a processor.

[0032] In particular, according to some embodiments of this disclosure, the processes described above can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods described above. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, it performs the functions defined in the methods of some embodiments of this disclosure.

[0033] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a task data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated task data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0034] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol, such as HTTP (Hypertext Transfer Protocol), and can interconnect with digital task data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0035] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: determine the network connection status of the switch production line management application in response to detecting a query operation on a production collaboration document in the switch production line management application; replace the webpage entry information corresponding to the production collaboration document with target entry file information and load target webpage resource information in response to determining that the network connection status of the switch production line management application indicates an offline state, so as to display the webpage of the production collaboration document offline in the switch production line management application, wherein the target entry file information is the file information of the entry file corresponding to the webpage of the production collaboration document downloaded in advance, and the target webpage resource information is the resource information corresponding to the webpage stored locally; in response to determining that the network connection status of the switch production line management application indicates an online state and that the webpage resource information corresponding to the production collaboration document is not stored locally, download the webpage resource information of the webpage from the production line document server, wherein the webpage resource information includes an entry file and resource information; display the webpage of the production collaboration document in the switch production line management application according to the webpage resource information, and store the webpage resource information in a local database.

[0036] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including product-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0037] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying fault types in a 10kV distribution network line, characterized in that, include: Step S1: Collect voltage and current signals of the distribution network lines using a high sampling rate sensor to obtain millisecond-level time series data. Extract waveform amplitude and phase angle change characteristics of the initial stage of single-phase grounding fault or phase-to-phase short-circuit fault from the data to obtain a preliminary trajectory representation characterizing the fault evolution trend. Step S2: The initial trajectory representation signal is decomposed into multi-scale components using wavelet transform. Based on the difference in transient characteristics between transient and permanent faults, the energy distribution of the components on the millisecond scale is calculated to obtain fine-grained feature indices characterizing the details of the fault. Step S3: Select key components related to the fault type from the fine-grained feature indicators. If the energy distribution of the key component exceeds the preset judgment threshold, it is judged as a potential trajectory of a permanent fault; otherwise, it is classified as a dynamic feature of a transient fault, and an intermediate result of fault classification is obtained. Step S4: By integrating intermediate results from multiple time windows with corresponding fine-grained feature indicators, the support vector machine algorithm is used to train the pre-labeled fault samples, calculate the similarity between the trajectory and the sample trajectory, and obtain the probability estimate of the fault category. Step S5: Combine the transient component change trends of adjacent time periods to analyze steady-state electrical characteristics. If the similarity is higher than the preset confirmation threshold, it is determined to be a permanent fault; otherwise, the identification result is adjusted to a transient fault to obtain the final fault type label.

2. The method for identifying fault types in a 10kV distribution network line as described in claim 1, characterized in that, In step S1, The transient voltage and transient current values ​​of real-time electrical signal conversion are read to construct a millisecond-level sequence. The data window at the initial stage of the fault is determined and extracted from the abnormal disturbance starting point where the instantaneous rate of change exceeds the mutation threshold. The instantaneous magnitude and phase offset of the decomposed signal are calculated for the initial fault data window to obtain the amplitude characteristics and phase angle change characteristics of the decomposed extracted waveform. The waveform amplitude characteristics and phase angle change characteristics are mapped to dynamic coordinate points and connected to obtain a preliminary trajectory representation characterizing the fault evolution trend.

3. The method for identifying fault types in a 10kV distribution network line as described in claim 2, characterized in that, In step S2, The preliminary trajectory representation is decomposed into multiple layers using wavelet basis functions. The high-frequency detail components and low-frequency approximate components are then subjected to square integral operations to obtain the instantaneous frequency band energy value. Multi-scale permutation entropy is calculated based on instantaneous frequency band energy values ​​to extract energy singularities. An energy evolution feature vector describing the time-varying characteristics of fault energy is constructed and mapped to a high-dimensional feature space to generate fine-grained feature indicators characterizing fault details.

4. The method for identifying fault types in a 10kV distribution network line as described in claim 3, characterized in that, In step S3, The mutual information algorithm is used to filter out key components from fine-grained feature indicators, and the sliding time window integration operation is performed to obtain the energy distribution value. If the energy distribution value exceeds the preset judgment threshold, it is determined to be a potential trajectory of a permanent fault; otherwise, it is determined to be a dynamic characteristic of a transient fault. By integrating the potential trajectory of the permanent fault with the dynamic characteristics of the transient fault, an intermediate result for fault classification is obtained.

5. The method for identifying fault types in a 10kV distribution network line as described in claim 4, characterized in that, In step S4, The intermediate results from multiple time windows are concatenated with the fine-grained features to construct a multidimensional time-series feature vector sequence. The multidimensional time-series feature vector sequence is mapped using a radial basis kernel function to obtain a high-dimensional feature mapping set. The high-dimensional feature mapping set is then input into a fault discrimination model constructed based on pre-labeled fault samples. The normal vector projection distance of the high-dimensional feature mapping set in the fault discrimination model is calculated to obtain the trajectory similarity value. The trajectory similarity value is normalized using a logistic regression function to obtain the probability estimate of the fault category.

6. The method for identifying fault types in a 10kV distribution network line as described in claim 1, characterized in that, In step S5, Obtain the original electrical data corresponding to the probability estimation, extract transient component sequences and steady-state electrical feature data from them, calculate the gradient change rate based on the transient component sequences to obtain the transient component change trend, and concatenate the transient component change trend with the steady-state electrical feature data. The spliced ​​feature data is input into a preset fault feature library for matching, and a similarity value is calculated. If the similarity value is higher than the preset confirmation threshold, it is determined to be a permanent fault; otherwise, a transient fault identification result is generated, and the final fault type label is obtained.

7. The method for identifying fault types in a 10kV distribution network line as described in claim 1, characterized in that, Also includes: Step S6: Extract dynamic parameters that match the distribution network operation scenario from the final fault type label, use cluster analysis to summarize similar fault trajectories, determine stable feature patterns that characterize fault types, and obtain an optimized fault feature set. Step S7: Based on the optimized fault feature set, the scale parameters of the wavelet transform are adjusted. For subsequent signal acquisition and processing loops, it is determined whether the recognition performance has degraded to decide whether to update the support vector machine model, thus constructing a continuously optimized fault recognition framework.

8. The method for identifying fault types in a 10kV distribution network line as described in claim 7, characterized in that, In step S6, The dynamic electrical parameters associated with the final fault type label are mapped as multi-dimensional state vectors to construct fault trajectory data. A similarity matrix is ​​calculated based on the fault trajectory data, and trajectories with similar evolution paths are grouped into a trajectory cluster set. The dispersion of each moment in the trajectory cluster set is calculated, the center point value below the convergence threshold is locked to determine the stable feature pattern, the stable feature pattern is mapped to the original feature space and the noise dimension is removed to obtain the optimized fault feature set.

9. The method for identifying fault types in a 10kV distribution network line as described in claim 8, characterized in that, In step S7, The frequency band energy distribution density is calculated based on the optimized fault feature set to determine the feedback-adjusted wavelet transform scaling parameters, which are obtained by mapping the high-density frequency band to the center frequency of the wavelet basis function. The time-frequency wavelet coefficients are obtained by using the wavelet transform scaling parameters after feedback adjustment. The time-frequency wavelet coefficients are then input into the current support vector machine model to quantify the recognition performance index. If the recognition performance index is lower than a preset threshold, the weight vector and bias parameters are reconstructed based on the selected boundary feature vectors to generate a corrected model that adapts to the new feature distribution. Deploy the corrective model into the online monitoring process to build a continuously optimized fault identification framework.