Distribution line fault identification method and handheld intelligent terminal
Through the fault identification method based on the semi-supervised learning data model, the fault categories of distribution lines are identified, and the problems of insufficient accuracy and high cost in the prior art are solved, thereby achieving more efficient and reliable fault identification.
Patent Information
- Application Number
- CN202510176853.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-18
- Publication Date
- 2025-06-13
AI Technical Summary
The prior art cannot effectively identify potential safety hazards of distribution lines, resulting in insufficient identification accuracy and high cost.
The fault identification method based on the semi-supervised learning data model is adopted to pre-process the relevant operating data of the distribution line by obtaining the relevant operation data of the power distribution line, and the fault category is identified in the trained fault identification model.
It improves the accuracy of distribution line fault identification, reduces costs, and enhances the generalization ability and reliability of the system.
Smart Images
Figure CN120145138A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power systems, and in particular, to a method for identifying distribution line faults and a handheld intelligent terminal. Background Art
[0002] Distribution lines are characterized by a large number of equipment points and a wide area, as well as a complex operating environment, making them prone to faults during operation. Currently, common distribution line faults mainly include phase-to-phase short-circuit faults, overload faults, single-phase grounding faults, and open-phase faults, etc. Among them, short-circuit faults are relatively common faults during the operation of distribution lines in power systems. Once they occur, they will have a huge impact on the entire power system, thus triggering other faults; overload faults mainly refer to the situation where the current that the distribution line can withstand is less than the actual current of the line operation, causing the line to be severely overloaded during operation, resulting in damage to line equipment or even fire; single-phase grounding faults are also common fault types during the operation of distribution lines. The reasons for such faults are the breakdown of the single-phase insulation of distribution line cables, the breakdown of the single-phase insulation of overhead lines, etc. If they cannot be isolated or eliminated in time, there will be greater operation risks; open-phase faults often occur in distribution lines. If the distribution line operates with an open phase, it will have a serious impact on the equipment on the user side. Thus, it is necessary to effectively identify distribution line faults to improve the operation safety of the power grid.
[0003] In the prior art, when an actual distribution line fails, the collected current and voltage values are instantaneous and local, so it is impossible to comprehensively and comprehensively judge the faults of the distribution lines in the entire monitored area (such as open-phase faults, power theft, grounding, etc.). Traditional methods for identifying distribution line faults often rely on experienced operation and maintenance personnel to judge based on experience or install fixed on-line monitoring devices at a certain point on the distribution line. This not only has a high cost and is easy to miss, but also makes data collection and analysis inaccurate, resulting in insufficient predictability of potential safety hazards of distribution lines. Therefore, it cannot meet the requirements of modern power systems for safe, reliable, and efficient operation and maintenance. Therefore, a method that can process complex data situations to identify distribution line faults and has a low cost is needed to enhance the generalization ability and reliability of the system. Summary of the Invention
[0004] The technical problem to be solved by the embodiments of the present invention is to provide a method for identifying distribution line faults and a handheld intelligent terminal, which can solve the problems such as insufficient predictability of potential safety hazards of distribution lines by traditional distribution line fault identification methods, not only improving the identification accuracy but also reducing the cost.
[0005] To solve the above technical problem, the embodiments of the present invention provide a method for identifying distribution line faults, and the method includes the following steps:
[0006] Obtain the relevant operation data of the distribution line collected in real time by a preset acquisition device;
[0007] Preprocess the relevant operation data of the obtained distribution line, and import the preprocessed data into a trained fault identification model to identify the fault category of the distribution line; wherein, the fault identification model is constructed based on a semi-supervised learning data model and is iteratively trained through a model training stage, a marking stage, and a backtest optimization stage.
[0008] Among them, the relevant operation data of the distribution line includes the current change value, voltage change value, waveform signal, line phase sequence, leakage current, and measured power when the distribution line fails.
[0009] Among them, the preprocessing of the relevant operation data of the distribution line includes data normalization processing and data noise processing; wherein, the data noise processing includes Gaussian smoothing, median filtering, and bilateral filtering.
[0010] Among them, the fault category of the distribution line is one of phase-to-phase short circuit fault, overload fault, single-phase ground fault, and open-phase fault.
[0011] The embodiment of the present invention also provides a handheld intelligent terminal, including:
[0012] An operation data real-time acquisition unit for obtaining the relevant operation data of the distribution line collected in real time by a preset acquisition device;
[0013] A distribution line fault identification unit for preprocessing the relevant operation data of the obtained distribution line, and importing the preprocessed data into a trained fault identification model to identify the fault category of the distribution line; wherein, the fault identification model is constructed based on a semi-supervised learning data model and is iteratively trained through a model training stage, a marking stage, and a backtest optimization stage.
[0014] Among them, the relevant operation data of the distribution line includes the current change value, voltage change value, waveform signal, line phase sequence, leakage current, and measured power when the distribution line fails.
[0015] Among them, the preprocessing of the relevant operation data of the distribution line includes data normalization processing and data noise processing; wherein, the data noise processing includes Gaussian smoothing, median filtering, and bilateral filtering.
[0016] Among them, the fault category of the distribution line is one of phase-to-phase short circuit fault, overload fault, single-phase ground fault, and open-phase fault.
[0017] Implementing the embodiment of the present invention has the following beneficial effects:
[0018] The present invention obtains the relevant operation data of the distribution line and imports it into a fault identification model constructed based on a semi-supervised learning data model to identify the fault category, thereby effectively reducing the computational amount of fault defect identification, solving the problems such as insufficient predictability of potential safety hazards of traditional distribution line fault identification methods, not only improving the identification accuracy, but also reducing the cost. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, obtaining other drawings based on these drawings without creative efforts still belongs to the scope of the present invention.
[0020] Figure 1 It is a flowchart of a distribution line fault identification method provided by an embodiment of the present invention;
[0021] Figure 2 It is a logical working diagram of a semi-supervised learning data model in a distribution line fault identification method provided by an embodiment of the present invention;
[0022] Figure 3 It is a schematic structural diagram of a handheld intelligent terminal provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0023] To make the purpose, technical solutions and advantages of the present invention clearer, the following will further describe the present invention in detail with reference to the drawings.
[0024] As Figure 1 shown, in an embodiment of the present invention, a distribution line fault identification method is proposed, which is implemented on a handheld intelligent terminal. The method includes the following steps:
[0025] Step S1: Obtain the relevant operation data of the distribution line collected in real time by a preset collection device;
[0026] Step S2: Preprocess the obtained relevant operation data of the distribution line, and import the preprocessed data into a trained fault identification model to identify the fault category of the distribution line; wherein, the fault identification model is constructed based on a semi-supervised learning data model and is iteratively trained through a model training stage, a marking stage and a backtest optimization stage.
[0027] Specifically, in step S1, one or more acquisition devices (including various types of sensors) are pre-arranged at key monitoring points of the distribution line to collect relevant operation data of the distribution line in real time; among them, the relevant operation data of the distribution line includes the current change value, voltage change value, waveform signal, line phase sequence, leakage current, measured power, etc. when the distribution line fails.
[0028] Secondly, through wireless communication, the handheld intelligent terminal establishes communication with each acquisition device and obtains the relevant operation data of the distribution line in real time.
[0029] In step S2, first, a fault identification model is constructed based on a semi-supervised learning data model. This semi-supervised learning data model combines a semi-supervised learning algorithm and a technology for extracting current and voltage characteristics of the distribution line. By using labeled empirical data to improve the model performance and applying unlabeled data to extract more fault characteristics and data of the distribution line operation.
[0030] At this time, the training of the semi-supervised learning data model includes three stages, specifically: the model training stage, the labeling stage, and the backtest optimization stage, as Figure 2 shown.
[0031] 1) Model training stage
[0032] (1) Preparation of labeled data: From the long-term distribution line maintenance and fault identification database, extract the current and voltage change values, waveform signals, line phase sequences, etc. when the distribution line fails, as well as the corresponding fault judgment results. First, preprocess the data, including data normalization and data noise processing; among them, data noise processing includes Gaussian smoothing, median filtering, and bilateral filtering; then, through experienced engineers, perform data annotation and then label it to form preliminary labeled data.
[0033] Collection of unlabeled data: Build a simulated test experimental environment in the laboratory, simulate various distribution line faults, install mobile acquisition sensor devices at key monitoring points of the distribution line, collect data such as the operation current and voltage data values, waveform signals, line phase sequences of the distribution line, and perform preliminary validity and correctness checks on the collected data to filter out invalid data. First, use the fast independent component algorithm and wavelet packet algorithm to extract its features, and then preprocess the data, including data normalization and data noise processing; among them, data noise processing includes Gaussian smoothing, median filtering, and bilateral filtering, and then initially assume these data as normal operation values and initially define them as unlabeled data.
[0034] (2) Dataset definition: Divide the dataset into a labeled dataset (D l ) and an unlabeled dataset (D u ).
[0035] Suppose there is a dataset containing n samples, among which m samples have labels and n - m samples have no labels. Denote the labeled dataset as Denote the unlabeled dataset as Denote the model as M(x; w), where w represents the parameters of the model.
[0036] Construction of the initial model: In the actual model construction process, based on the labeled dataset, by training the unlabeled dataset, from the following loss function relationship L(W)1, we can know how to ensure the minimum loss of the model. We only hope to find an optimal model parameter w, so as to achieve the gradient descent of data loss. Its loss function is expressed as:
[0037]
[0038] where l(y i , M(x i ; w)) represents the loss of the labeled data on the initial model, R(w) represents the regularization term of the model, and λ is the regularization parameter.
[0039] From the above theoretical analysis, using the labeled dataset D l (the current and voltage change values, waveform signals, line phase sequence, etc. when the distribution line fails) as the definition standard, adopting a semi-supervised method to label the unlabeled dataset (D u ), and training the model, the initial model M 0 can be obtained.
[0040] 2) Labeling stage
[0041] By deploying the application scenarios in different environments on-site, installing the mobile acquisition sensor device at different monitoring points of the distribution line, and collecting the current and voltage data, waveform signals, line phase sequence, etc. of the distribution line irregularly and continuously as the unlabeled dataset, and using the fast independent component algorithm and wavelet packet algorithm to extract its current characteristics. A large amount of data is input into the model for training, and the initial model generated in the first stage is called as the criterion to judge and analyze the unlabeled dataset. The self-learning algorithm program makes a clustering hypothesis for the assumed collected current and voltage data. Assuming that there is a similar cluster structure in the data model, this data is labeled with the same category label. If the data is assumed to be distributed on another manifold structure and neighboring samples have similar output values, then the data is labeled with an unrestricted classification label through the manifold hypothesis, and then returned for model backtesting and optimization.
[0042] 3) Backtesting and optimization stage
[0043] Backtesting and optimization: Redefine the above returned unlabeled data as the unlabeled dataset D u , and for the initial model M0 Optimize to obtain the final model M f . On the unlabeled dataset, we can optimize the model parameters w by leveraging the unlabeled dataset, thereby reducing the loss on the labeled dataset. This can be achieved by adding an additional loss term, such as defining that neighboring samples have similar output values multiple times and through multiple comparisons to reach the final identification. For example: In multiple monitoring areas, when similar faults occur in the distribution lines, the measured current values are all similar and close. We define a popular structure and use it as an additional loss, input it into the semi-supervised algorithm generator for optimization, and its loss function is expressed as:
[0044]
[0045] where P(w) represents the loss of the unlabeled data on the model, and μ is the regularization parameter. By this optimized loss function, a more powerful model can be iteratively constructed on the limited labeled data.
[0046] High-precision data model: Based on the labeled data generated in the second and third stages, repeat the model training, continuously iterate and optimize the model to form the final high-precision data model as the finally trained fault identification model.
[0047] Secondly, import the above preprocessed data into the trained fault identification model to identify one of the fault categories of the distribution line as phase-to-phase short circuit fault, overload fault, single-phase ground fault, and open-phase fault.
[0048] As Figure 3 shown, in an embodiment of the present invention, a handheld intelligent terminal is provided, including:
[0049] An operation data real-time acquisition unit 110 for acquiring the relevant operation data of the distribution line collected in real time by a preset acquisition device;
[0050] A distribution line fault identification unit 120 for preprocessing the acquired relevant operation data of the distribution line and importing the preprocessed data into the trained fault identification model to identify the fault category of the distribution line; wherein, the fault identification model is constructed based on a semi-supervised learning data model and is iteratively trained through a model training stage, a labeling stage, and a backtest optimization stage.
[0051] Among them, the relevant operation data of the distribution line includes the current change value, voltage change value, waveform signal, line phase sequence, leakage current, and measured power when the distribution line fails.
[0052] Among them, the preprocessing of the relevant operation data of the distribution line includes data normalization processing and data noise processing; among them, the data noise processing includes Gaussian smoothing, median filtering, and bilateral filtering.
[0053] Among them, the fault category of the distribution line is one of the phase-to-phase short-circuit fault, overload fault, single-phase grounding fault, and open-phase fault.
[0054] Implementing the embodiments of the present invention has the following beneficial effects:
[0055] The present invention can effectively reduce the computational complexity of fault defect identification by obtaining the relevant operation data of the distribution line and importing it into a fault identification model constructed based on a semi-supervised learning data model, and solves the problems such as insufficient predictability of potential safety hazards of traditional distribution line fault identification methods. It not only improves the identification accuracy but also reduces the cost.
[0056] It should be noted that in the above device embodiments, the various device modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional modules are only for easy distinction from each other and do not limit the protection scope of the present invention.
[0057] Those of ordinary skill in the art can understand that all or part of the steps in implementing the above method embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium, such as ROM / RAM, disk, optical disc, etc.
[0058] The above-disclosed are only the preferred embodiments of the present invention, and of course, the scope of the rights of the present invention cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.
Claims
1. A method for identifying a distribution line fault, characterized in that: The method comprises the following steps: Obtain relevant operation data of the distribution line collected in real time by a preset collection device; The acquired relevant operation data of the distribution line is preprocessed, and the preprocessed data is imported into the trained fault identification model to identify the fault type of the distribution line; wherein the fault identification model is constructed based on the semi-supervised learning data model, and is iteratively trained through the model training stage, the labeling stage and the backtest optimization stage.
2. The method for identifying a distribution line fault according to claim 1, characterized in that: The relevant operating data of the distribution line includes the current change value, voltage change value, waveform signal, line phase sequence, leakage current and measured power when a fault occurs in the distribution line.
3. The method for identifying a distribution line fault according to claim 2, characterized in that: The preprocessing of the relevant operating data of the distribution line includes data normalization processing and data noise processing; wherein the data noise processing includes Gaussian smoothing, median filtering, and bilateral filtering.
4. The method for identifying a distribution line fault according to claim 1, wherein: The fault type of the distribution line is one of a phase-to-phase short circuit fault, an overload fault, a single-phase grounding fault and a phase loss fault.
5. A handheld intelligent terminal, characterized in that: include: An operation data real-time acquisition unit, used to acquire the relevant operation data of the distribution line collected in real time by a preset collection device; The distribution line fault identification unit is used to pre-process the relevant operating data of the distribution line obtained, and import the pre-processed data into the trained fault identification model to identify the fault type of the distribution line; wherein the fault identification model is constructed based on the semi-supervised learning data model, and is iteratively trained through the model training stage, the labeling stage and the backtesting optimization stage.
6. The handheld intelligent terminal according to claim 5, characterized in that: The relevant operating data of the distribution line includes the current change value, voltage change value, waveform signal, line phase sequence, leakage current and measured power when a fault occurs in the distribution line.
7. The handheld intelligent terminal according to claim 6, characterized in that: The preprocessing of the relevant operating data of the distribution line includes data normalization processing and data noise processing; wherein the data noise processing includes Gaussian smoothing, median filtering, and bilateral filtering.
8. The handheld intelligent terminal according to claim 5, characterized in that: The fault type of the distribution line is one of a phase-to-phase short circuit fault, an overload fault, a single-phase grounding fault and a phase loss fault.