Power distribution network concealment fault identification method and system based on multi-scale energy entropy, and storage medium

Abstract

The invention relates to a power distribution network concealment fault identification method and system based on multi-scale energy entropy and a storage medium, and belongs to the technical field of power distribution network fault monitoring and diagnosis. The method comprises the following steps: acquiring a fault recording signal of a feeder line of the power distribution network and preprocessing the fault recording signal to obtain a zero-sequence current recording signal; performing multi-scale decomposition on the zero-sequence current recording signal to obtain a scale component set of the fault recording signal under multiple scales; within the wave recording time range of the zero-sequence current wave recording signal, calculating an energy value of each scale component, and determining an energy ratio according to the energy value; calculating a multi-scale energy entropy of the zero-sequence current recording signal according to the energy ratio; and comparing the multi-scale energy entropy with a preset energy entropy threshold interval to judge the type of the hidden fault. The objective of the invention is to realize unified quantification and identification of three types of hidden faults based on a single zero-sequence current recording signal without depending on complex setting of multiple time windows.

Description

A method, system, and storage medium for identifying hidden faults in distribution networks based on multi-scale energy entropy. Technical Field

[0001] This invention relates to a method, system, and storage medium for identifying hidden faults in distribution networks based on multi-scale energy entropy, belonging to the field of distribution network fault monitoring and diagnosis technology. Background Technology

[0002] Overhead power distribution lines have a wide coverage area and operate in a complex environment. Affected by factors such as tree growth and wind swaying, pollution accumulation and humidity / smog, lightning strikes and component aging, they are prone to hidden faults such as tree contact, insulator flashover, and aging and deterioration of surge arresters. These faults typically exhibit high resistance, intermittency, and weak characteristics. The fault current amplitude is small, the waveform is distorted and unstable, and the duration is short or recurring. They are easily confused with transient disturbances such as load fluctuations, switching operations, and external interference, making it difficult for traditional relay protection and conventional criteria to identify and alarm in a timely and accurate manner. This can lead to fault expansion, further insulation deterioration, and even permanent faults, affecting power supply reliability and operational safety.

[0003] Existing fault identification technologies for distribution networks mainly include threshold-based zero-sequence current / zero-sequence voltage amplitude criteria, waveform distortion and harmonic content analysis, time-frequency energy distribution feature extraction, and multi-feature fusion classification based on machine learning. Among these, amplitude and phase criteria have limited applicability to low-current grounding systems or high-resistance concealed faults, and are prone to missed detection due to weak fault current and strong background noise; harmonic and distortion features are easily affected by load nonlinearity, resonance, and measurement errors, resulting in insufficient robustness; some time-frequency analysis methods require the selection of multiple time windows, frequency bands, or manual experience features, leading to strong feature coupling and parameter dependence, which restricts the universality and transferability of engineering deployment; and multi-feature fusion learning models often rely on large-scale labeled samples and complex training processes, and their performance is easily degraded when operating modes change, line differences occur, or data distribution drifts, and their interpretability is weak, which is not conducive to understanding and tracing the source of faults by on-site operation and maintenance personnel.

[0004] Furthermore, the three types of concealed faults—tree contact, insulator flashover, and arrester aging—differ in their energy distribution and spectral diffusion: tree contact is often accompanied by intermittent discharge and random arcing, causing energy to be more dispersed across multiple scales; insulator flashover typically exhibits a relatively stable discharge channel and phased evolution; arrester aging is mostly characterized by changes in leakage current characteristics and low-intensity disturbances, with its energy distribution being relatively concentrated. Current technology lacks a solution that can utilize these differences and, without relying on complex multi-time-window settings, achieve unified quantification and reliable identification of these three types of concealed faults based on a single zero-sequence current waveform signal.

[0005] Therefore, it is necessary to propose a method, system, and storage medium for identifying hidden faults in distribution networks based on multi-scale energy entropy. By performing multi-scale decomposition on the zero-sequence current waveform signal and calculating the energy entropy over the entire time period, the energy entropy interval mapping can be used to distinguish between tree obstruction, insulator flashover, and surge arrester aging, thereby improving the accuracy, robustness, and engineering deployability of hidden fault identification. Summary of the Invention

[0006] The purpose of this invention is to provide a method for identifying hidden faults in distribution networks based on multi-scale energy entropy. It aims to achieve unified quantification and identification of three types of hidden faults based on a single zero-sequence current waveform signal without relying on complex settings of multiple time windows.

[0007] To achieve the above objectives, the technical solution of this invention is: a method for identifying hidden faults in distribution networks based on multi-scale energy entropy. This method relies solely on zero-sequence current recording signals, constructs multi-scale energy distributions and calculates energy entropy throughout the entire recording period, and uses a "high entropy-medium entropy-low entropy" interval mapping to distinguish the types of tree obstructions, insulator flashover, and arrester aging, thereby reducing dependence on multiple time windows and multi-source signals and improving the robustness and deployability of identification. The method includes the following steps: S1: Acquire fault recording signals from distribution network feeders and preprocess them to obtain zero-sequence current recording signals; S2: Perform multi-scale decomposition on the zero-sequence current recording signals to obtain the set of scale components of the fault recording signals at multiple scales; S3: In the zero-sequence current recording signals... Within the recording time range of the current recording signal, the energy value of each scale component is calculated, and the energy ratio is determined based on the energy value; S4: The multi-scale energy entropy of the zero-sequence current recording signal is calculated based on the energy ratio; S5: The multi-scale energy entropy is compared with a preset energy entropy threshold range to determine the type of concealed fault; wherein, the preset energy entropy threshold range includes a first threshold and a second threshold, and the first threshold is greater than the second threshold. When the multi-scale energy entropy is not less than the first threshold, it is determined to be a tree-damaged concealed fault; when the multi-scale energy entropy is less than the first threshold and greater than the second threshold, it is determined to be an insulator flashover concealed fault; when the multi-scale energy entropy is not greater than the second threshold, it is determined to be an arrester aging concealed fault.

[0008] Optionally, the multi-scale decomposition of the zero-sequence current waveform signal is performed using variational mode decomposition.

[0009] Optionally, calculating the energy value of each scale component within the recording time range of the zero-sequence current recording signal, and determining the energy proportion based on the energy value, includes: calculating the energy of the k-th scale component within the recording time range, expressed as:

[0010] in, This represents the energy of the k-th scale component during the recording period. The discrete amplitude of the k-th scale component at the n-th sampling point is represented by N, where N represents the total number of sampling points during the recording period; the energy proportion of the k-th scale component is calculated. The expression is:

[0011] Where K is the number of scale components. Let be the energy value of the j-th scale component during the recording period.

[0012] Optionally, the expression for calculating the multi-scale energy entropy of the zero-sequence current recording signal based on the energy ratio is:

[0013] Where H is the multi-scale energy entropy. The energy percentage of the k-th scale component.

[0014] Optionally, after determining the type of concealed fault, the method further includes: outputting the determined concealed fault type, the corresponding multi-scale energy entropy, and alarm information; wherein the alarm information includes the fault occurrence feeder identifier, waveform recording timestamp, and distance margin from the threshold interval.

[0015] Furthermore, to achieve the above objectives, this application also provides a distribution network hidden fault identification system based on multi-scale energy entropy. The system includes: a signal acquisition module for acquiring and preprocessing fault recording signals from distribution network feeders to obtain a zero-sequence current recording signal; a decomposition module for performing multi-scale decomposition on the zero-sequence current recording signal to obtain a set of scale components of the fault recording signal at multiple scales; an energy calculation module for calculating the energy value of each scale component within the recording time range of the zero-sequence current recording signal and determining the energy proportion based on the energy value; and an energy entropy calculation module for calculating the energy proportion based on the energy proportion. The system calculates the multi-scale energy entropy of the zero-sequence current recording signal; a type determination module compares the multi-scale energy entropy with a preset energy entropy threshold range to determine the type of concealed fault; specifically, the type determination module determines the following: when the multi-scale energy entropy is not less than a first threshold, it is determined to be a tree-damaged concealed fault; when the multi-scale energy entropy is less than the first threshold but greater than a second threshold, it is determined to be an insulator flashover concealed fault; when the multi-scale energy entropy is not greater than the second threshold, it is determined to be a surge arrester aging concealed fault; an output module outputs the determined concealed fault type, the corresponding multi-scale energy entropy, and alarm information.

[0016] In addition, to achieve the above objectives, this application also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a distribution network hidden fault identification and prediction program based on multi-scale energy entropy, wherein when the distribution network hidden fault identification program based on multi-scale energy entropy is executed by a processor, it implements the steps of the distribution network hidden fault identification method based on multi-scale energy entropy as described in any of the preceding claims.

[0017] The beneficial effects of this invention are: 1. This invention uses only the zero-sequence current waveform signal of the feeder as input to complete the identification of concealed faults, without the need to simultaneously acquire multi-source measurements such as three-phase voltage / current or zero-sequence voltage, which reduces the complexity of on-site measurement point configuration, communication transmission and data synchronization, and facilitates rapid deployment in edge terminals such as FTU / DTU.

[0018] 2. This invention constructs a multi-scale energy distribution and calculates energy entropy by using the entire recording period as the analysis interval. It does not require setting multiple time windows, window lengths, or window shifts, thus avoiding feature drift and misjudgment caused by improper selection of time windows and improving the consistency and reproducibility of engineering applications.

[0019] 3. This invention first performs DC removal and normalization on the zero-sequence current, and uses energy proportion to construct energy distribution, so that the energy entropy index is not sensitive to the overall amplitude scaling. It can effectively suppress the influence of different line impedances, different load levels, transformer ratio differences and noise disturbances on the identification results, and improve the applicability across lines and scenarios. Attached Figure Description

[0020] Figure 1 is a flowchart illustrating a method for identifying hidden faults in a distribution network based on multi-scale energy entropy, provided in an embodiment of this application; Figure 2 is a structural diagram illustrating a system for identifying hidden faults in a distribution network based on multi-scale energy entropy, provided in an embodiment of this application; Figure 3 is a schematic diagram illustrating the entropy interval discrimination of three types of faults in a method for identifying hidden faults in a distribution network based on multi-scale energy entropy, provided in an embodiment of this application; Figure 4 is a deployment diagram of an architecture for identifying hidden faults in a distribution network based on multi-scale energy entropy, provided in an embodiment of this application. Detailed Implementation

[0021] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0022] To make the above-mentioned objectives, technical solutions and beneficial effects of the present invention clearer and easier to understand, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] Example 1: As shown in Figure 1, the multi-scale energy entropy-based method for identifying hidden faults in distribution networks proposed in this invention is applicable to online identification scenarios of hidden faults in overhead lines of distribution networks. These hidden faults include tree-related hidden faults, insulator flashover hidden faults, and surge arrester aging hidden faults. The method uses the zero-sequence current waveform signal collected by the feeder terminal device as the sole input. It preprocesses the zero-sequence current waveform signal, performs variational mode decomposition, multi-scale energy proportion calculation, and energy entropy calculation, and outputs the fault type based on the comparison between energy entropy and a threshold interval. The method includes the following steps: S1: Acquire the fault waveform signal of the distribution network feeder and preprocess it to obtain the zero-sequence current waveform signal; specifically, acquire the zero-sequence current waveform signal of the distribution network feeder within a single waveform recording period to obtain a discrete sequence. ,in N is the number of sampling points during this recording period; the This signal serves as the input signal for subsequent processing links in this invention, and together with metadata such as the feeder identifier, sampling rate, and start and end times of the recording, it constitutes the data object for an identification task.

[0024] It should be noted that zero-sequence current is a continuous-time signal in a physical sense. The waveform recording device uses a sampling period Sampling was performed Therefore, the subsequent processing in this embodiment is performed in the form of a discrete sequence.

[0025] Since hidden faults typically exhibit intermittent, random discharge and local energy diffusion characteristics, this invention does not manually segment or divide the recording period into multiple time windows. Instead, it uses the entire recording period as the analysis interval to ensure a complete characterization of the energy distribution state throughout the fault process.

[0026] To avoid incomparable energy indicators due to data acquisition link bias, transformer errors, and differences in operating conditions, the original zero-sequence current sequence is preprocessed. This preprocessing includes removing the DC component and amplitude normalization. Specifically, the zero-sequence current sequence... After removing the DC component, the DC-removed sequence is obtained:

[0027] in, This is the de-DC zero-sequence current sequence obtained after performing DC removal processing on the original zero-sequence current sequence. The value of the original zero-sequence current discrete sequence at the nth sampling point, representing all terms to the right of the minus sign. The original zero-sequence current sequence contains the DC component throughout the entire recording period; after removing the DC component, amplitude normalization is performed to obtain the preprocessed sequence. :

[0028] The result obtained by removing the DC component and normalizing the process This will serve as the input for subsequent multi-scale decomposition and energy calculation; the denominator term... The root mean square value of the DC zero-sequence current sequence is used to normalize the signal amplitude.

[0029] It is understandable that this embodiment, through DC component removal and amplitude normalization, can eliminate amplitude differences caused by acquisition bias, residual magnetization of the current transformer, and different line operating conditions, resulting in... The mean is zero and the energy scale is unified, making the multi-scale energy proportions comparable under different feeders, load levels and recording devices, thus ensuring the stability of subsequent energy entropy threshold discrimination.

[0030] S2: Perform multi-scale decomposition on the zero-sequence current waveform signal to obtain the set of scale components of the fault waveform signal at multiple scales; optionally, perform preprocessing on the sequence. Perform variational mode decomposition, with a preset number of modes. The modal component set is obtained. and its corresponding center frequency And satisfy the refactoring constraints:

[0031] It is important to understand that variational mode decomposition is achieved by solving a variational optimization problem with reconstruction constraints. The objective is to minimize the sum of the bandwidths of each mode component, thereby expanding the non-stationary zero-sequence current into multiple narrow-band components in the scale domain. This allows the degree of energy diffusion caused by different faults to be clearly characterized in the scale domain, expressed in continuous form as follows:

[0032] in, For impulse functions, For convolution operations, For the derivative with respect to time, It is a norm 2. The imaginary unit, For the first The center frequency of each modal component The original zero-sequence current waveform signal to be decomposed is shown. For the first One modal component.

[0033] Furthermore, for ease of engineering implementation, this embodiment transforms the above-mentioned constrained optimization problem into an augmented Lagrangian form and employs the alternating direction multiplier method for iterative solution. The augmented Lagrangian function is: in, To constrain the augmented Lagrangian function corresponding to the optimization problem, As a penalty factor, It is a Lagrange multiplier.

[0034] In the frequency domain iterative implementation, the first... The updates for each modal component and the center frequency are as follows:

[0035] Lagrange multipliers are updated to:

[0036] in, Indicates the first In the nth iteration Representation of each modal component in the frequency domain Indicates the corresponding center frequency. The Fourier transform of the original zero-sequence current waveform signal is represented. Indicates the first The frequency domain representation of the Lagrange multipliers in the next iteration. For step size parameters, This represents the Fourier transform.

[0037] The iteration termination condition is set as follows:

[0038] in, The convergence threshold is used to output the value after the termination condition is met. Modal components In the discrete implementation, this embodiment uses an alternating iterative approach to update. and Continue until the convergence condition is met.

[0039] Understandably, this embodiment introduces variational mode decomposition for multi-scale energy characterization of zero-sequence current recording signals, and constructs the scale energy proportion of the entire recording period based on the variational mode decomposition results. Through variational mode decomposition, intermittent discharges and random arcs caused by tree-type faults will lead to energy diffusion across more scale components. Insulator flashover often exhibits the formation and development of staged discharge channels, with a moderate degree of energy dispersion. Changes in leakage current characteristics caused by arrester aging usually show that energy is more concentrated in a few scale components, thus providing a scale domain basis for the subsequent "high entropy-medium entropy-low entropy" discrimination logic.

[0040] S3: Within the recording time range of the zero-sequence current recording signal, calculate the energy value of each scale component, and determine the energy proportion based on the energy value; optionally, calculating the energy value of each scale component within the recording time range of the zero-sequence current recording signal and determining the energy proportion based on the energy value includes: calculating the energy of the k-th scale component within the recording time range, expressed as:

[0041] in, This represents the energy of the k-th scale component during the recording period. The discrete amplitude of the k-th scale component at the n-th sampling point is represented by N, where N represents the total number of sampling points during the recording period; the energy proportion of the k-th scale component is calculated. The expression is:

[0042] Where K is the number of scale components. Let be the energy value of the j-th scale component during the recording period.

[0043] It is understood that this embodiment is provided by The multi-scale energy distribution is constructed. Since it uses energy percentage rather than absolute energy, the multi-scale energy distribution is not sensitive to the overall scaling of the zero-sequence current amplitude. Combined with preprocessing normalization, the impact of differences in the operating modes of different lines on the discrimination results can be further reduced.

[0044] S4: Calculate the multi-scale energy entropy of the zero-sequence current waveform signal based on the energy ratio; optionally, the expression for calculating the multi-scale energy entropy of the zero-sequence current waveform signal based on the energy ratio is:

[0045] Where H is the multi-scale energy entropy. The energy percentage of the k-th scale component.

[0046] It is understandable that this embodiment introduces multi-scale energy entropy. Used to characterize the degree of energy dispersion across multiple scale modes: when energy is concentrated in a few modes, the energy distribution is more "sharp". The energy distribution is smaller; as energy diffuses to more modes, the energy distribution becomes more "flat". The magnitude is relatively large. Based on multi-scale energy entropy, this embodiment maps the differences in the degree of energy diffusion at different scales of concealed faults into a unified scalar index that can be directly compared. This allows for the unified quantification of the differences between tree obstruction, flashover, and surge arrester aging into a "high entropy-medium entropy-low entropy" interval discrimination criterion.

[0047] S5: Compare the multi-scale energy entropy with a preset energy entropy threshold range to determine the type of concealed fault; wherein, the preset energy entropy threshold range includes a first threshold and a second threshold, and the first threshold is greater than the second threshold. When the multi-scale energy entropy is not less than the first threshold, it is determined to be a tree-damaged concealed fault; when the multi-scale energy entropy is less than the first threshold and greater than the second threshold, it is determined to be an insulator flashover concealed fault; when the multi-scale energy entropy is not greater than the second threshold, it is determined to be an arrester aging concealed fault.

[0048] Optionally, to form stable and deployable engineering criteria, the first and second thresholds are set offline. Statistical analysis is performed on the multi-scale energy entropy distribution of three types of faults—tree obstruction, insulator flashover, and surge arrester aging—based on historical labeled samples. The first and second thresholds are determined according to a preset misclassification rate or maximum classification interval criterion. Specifically, a labeled sample set is constructed, including tree obstruction sample set, insulator flashover sample set, and surge arrester aging sample set. The corresponding energy entropy is calculated for each sample, and the average energy entropy values ​​for the three types of samples—tree obstruction sample set, insulator flashover sample set, and surge arrester aging sample set—are obtained as follows: , , and satisfy .

[0049] Furthermore, the first threshold is determined based on the mean. Second threshold The expression is:

[0050] and ,Will and The threshold value is solidified into an online identification threshold. During the online identification phase, the energy entropy corresponding to the waveform to be measured is determined. The fault type is compared with a threshold and output as shown in Figure 3: When When, it is determined to be a tree obstacle concealment fault (high entropy); when When the fault is identified as a concealed insulator flashover fault (medium entropy); when When this occurs, it is determined to be a hidden fault due to aging of the surge arrester (low entropy).

[0051] Understandably, the aforementioned online identification stage relies solely on scalars. The comparison relationship between the two thresholds avoids the tuning complexity of multi-feature coupled criteria, while ensuring the consistency between the discrimination logic and the physical mechanism.

[0052] Optionally, after determining the type of concealed fault, the method further includes: outputting the determined concealed fault type, the corresponding multi-scale energy entropy, and alarm information; wherein the alarm information includes the fault occurrence feeder identifier, waveform recording timestamp, and distance margin from the threshold interval.

[0053] Specifically, when acquiring fault recording signals of distribution network feeders, the feeder identifier and recording start and end timestamps corresponding to the fault recording signals are acquired and stored simultaneously as associated metadata for this hidden fault identification task.

[0054] Specifically, the distance margin of the threshold interval is used to reflect the degree of confidence in the judgment: when it is determined to be a tree obstacle, the output margin is... When the fault is determined to be insulator flashover, the output margin is... When the surge arrester is determined to be aging, the output margin is... .

[0055] Understandably, margin The larger the value, the higher the degree to which the energy entropy falls within the corresponding range, making it easier for operations and maintenance personnel to classify and review alarms.

[0056] It is understood that the technical solution provided in this embodiment is applicable to the application scenario of identifying hidden faults in overhead lines of distribution networks, especially to the application scenario of online identification of hidden fault types such as tree obstruction, insulator flashover, and surge arrester aging based on zero-sequence current recording signals collected by feeder terminal devices. In practical applications, the distribution network hidden fault identification method proposed in this invention preprocesses and decomposes the zero-sequence current recording signal into multiple scales, calculates the energy proportion of each scale component within the recording period, and further calculates the multi-scale energy entropy. Based on the distribution relationship of energy entropy within a preset threshold range, the fault type is determined. When the energy entropy is in the high entropy range, it is determined to be a tree obstruction hidden fault; when the energy entropy is in the medium entropy range, it is determined to be an insulator flashover hidden fault; and when the energy entropy is in the low entropy range, it is determined to be a surge arrester aging hidden fault. This achieves accurate identification and alarm output of hidden fault types in the distribution network.

[0057] Furthermore, as shown in Figure 2, this embodiment also provides a distribution network hidden fault identification system based on multi-scale energy entropy. The system includes a zero-sequence current acquisition device 10, a data processing device 20, and a terminal 30. The zero-sequence current acquisition device 10 is used to acquire the zero-sequence current waveform signal of the distribution network feeder and send it to the data processing device 20; the data processing device 20 is used to process and analyze the zero-sequence current waveform signal to determine the fault type, and send the fault type and alarm information to the terminal 30; after receiving the fault type and alarm information, the terminal 30 issues a prompt message to the operation and maintenance personnel. The terminal 30 is a mobile terminal or a computer terminal.

[0058] The zero-sequence current acquisition device 10 and data processing device 20 can be deployed in a feeder terminal unit (FTU / DTU) or a station-end waveform recording device. The data processing device 20 and the terminal 30 are connected via wired or wireless means to transmit and display fault types and alarm information. Specifically, the data processing device 20 is implemented by a computer device, which includes a processor, a memory, and a network interface. The processor, memory, and network interface are connected via a system bus. The memory stores a computer program. When the processor executes the computer program, it executes all the processing flows corresponding to steps S1 to S5 and outputs the judgment results of tree obstruction, insulator flashover, or surge arrester aging.

[0059] Further, as shown in Figure 2, the data processing device 20 includes: a preprocessing module 2010, a multi-scale decomposition module 2020, an energy calculation module 2030, an energy entropy calculation module 2040, a threshold discrimination module 2050, and an output module 2060.

[0060] The preprocessing module 2010 receives the zero-sequence current waveform signal output by the zero-sequence current acquisition device 10 and performs DC removal and normalization; the multi-scale decomposition module 2020 performs variational mode decomposition on the preprocessed signal and outputs modal components; the energy calculation module 2030 calculates... and Energy entropy calculation module 2040 calculation Threshold discrimination module 2050 based on , , Output fault type; the output module 2060 sends the fault type, energy entropy value and alarm information to the terminal 30.

[0061] Furthermore, as shown in Figure 4, this embodiment also provides a distribution network hidden fault identification architecture based on multi-scale energy entropy. The architecture adopts an edge-cloud collaborative architecture and includes an edge-side identification unit and a central-side threshold management unit. The edge-side identification unit completes zero-sequence current acquisition, preprocessing, multi-scale decomposition, energy entropy calculation, and threshold comparison and discrimination, and sends the results up. The central-side threshold management unit completes sample management and offline threshold tuning and sends threshold parameters down to the edge side to ensure the consistency of discrimination between different lines.

[0062] Furthermore, this application also provides a computer-readable storage medium storing a distribution network hidden fault identification and prediction program based on multi-scale energy entropy, wherein when the distribution network hidden fault identification program based on multi-scale energy entropy is executed by a processor, it implements the steps of the distribution network hidden fault identification method based on multi-scale energy entropy as described in any of the preceding claims.

[0063] In summary, this invention collects fault recording data from distribution network feeders; performs preprocessing and standardization of the recorded signals using the fault trigger time as a reference; performs variational mode multi-scale decomposition on the preprocessed recorded signals to obtain several narrowband mode components and their corresponding scale features; calculates the energy and energy proportion of each mode component within the recording period, and calculates multi-scale energy entropy to construct an energy entropy feature set; performs feature correlation constraint and redundancy suppression processing on the energy entropy feature set to achieve feature decoupling and quantization, obtaining a quantitative index vector for fault identification; inputs the quantitative index into the fault identification model, outputs the hidden fault type and its development stage, and generates alarm information. The system provided by this invention includes a data acquisition module, a preprocessing and variational mode decomposition module, an energy entropy calculation module, a threshold discrimination module, and a result output module; a computer-readable storage medium stores instructions, and a processor executes the instructions to implement the above method. Compared with existing technologies, this invention utilizes variational mode decomposition to perform multi-scale decomposition on non-stationary waveform signals, which can more fully characterize the multi-scale energy distribution and evolution features of hidden faults, reduce the risk of misjudgment caused by feature coupling and redundancy, and improve the robustness and engineering applicability of fault identification.

[0064] The above description is merely illustrative of embodiments of the present invention and is not intended to limit the scope of protection of the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the above embodiments without departing from the concept of the present invention should fall within the scope of protection of the present invention.

Claims

1. A method for identifying hidden faults in distribution networks based on multi-scale energy entropy, characterized in that, The method includes the following steps: S1: Acquire fault recording signals of distribution network feeders and preprocess them to obtain zero-sequence current recording signals; S2: Perform multi-scale decomposition on the zero-sequence current recording signals to obtain a set of scale components of the fault recording signals at multiple scales; S3: Calculate the energy value of each scale component within the recording time range of the zero-sequence current recording signals, and determine the energy ratio based on the energy value; S4: Calculate the multi-scale energy entropy of the zero-sequence current recording signals based on the energy ratio; S5: Compare the multi-scale energy entropy with a preset energy entropy threshold range to determine the type of concealed fault; wherein, the preset energy entropy threshold range includes a first threshold and a second threshold, and the first threshold is greater than the second threshold. When the multi-scale energy entropy is not less than the first threshold, it is determined to be a tree-damaged concealed fault; when the multi-scale energy entropy is less than the first threshold and greater than the second threshold, it is determined to be an insulator flashover concealed fault; when the multi-scale energy entropy is not greater than the second threshold, it is determined to be an arrester aging concealed fault.

2. The method for identifying hidden faults in a distribution network based on multi-scale energy entropy according to claim 1, characterized in that, The multi-scale decomposition of the zero-sequence current recording signal employs variational mode decomposition.

3. The method for identifying hidden faults in a distribution network based on multi-scale energy entropy according to claim 1, characterized in that, The step of calculating the energy value of each scale component within the recording time range of the zero-sequence current recording signal, and determining the energy proportion based on the energy value, includes: calculating the energy of the k-th scale component within the recording time range, expressed as: ;in, This represents the energy of the k-th scale component during the recording period. The discrete amplitude of the k-th scale component at the n-th sampling point is represented by N, where N represents the total number of sampling points during the recording period; the energy proportion of the k-th scale component is calculated. The expression is: Where K is the number of scale components. Let be the energy value of the j-th scale component during the recording period.

4. The method for identifying hidden faults in a distribution network based on multi-scale energy entropy according to claim 1, characterized in that, The expression for calculating the multi-scale energy entropy of the zero-sequence current recording signal based on the energy ratio is as follows: Where H is the multi-scale energy entropy, The energy percentage of the k-th scale component.

5. The method for identifying hidden faults in a distribution network based on multi-scale energy entropy according to claim 1, characterized in that, After determining the type of concealed fault, the method further includes: outputting the determined concealed fault type, the corresponding multi-scale energy entropy, and alarm information; wherein, the alarm information includes the fault occurrence feeder identifier, waveform recording timestamp, and distance margin from the threshold interval.

6. A system for identifying hidden faults in a distribution network based on multi-scale energy entropy, characterized in that, The system includes: a signal acquisition module for acquiring fault recording signals from distribution network feeders and preprocessing them to obtain zero-sequence current recording signals; a decomposition module for performing multi-scale decomposition on the zero-sequence current recording signals to obtain a set of scale components of the fault recording signals at multiple scales; an energy calculation module for calculating the energy value of each scale component within the recording time range of the zero-sequence current recording signals and determining the energy proportion based on the energy value; and an energy entropy calculation module for calculating the multi-scale energy entropy of the zero-sequence current recording signals based on the energy proportion. The type determination module is used to compare the multi-scale energy entropy with a preset energy entropy threshold range to determine the type of hidden fault. Specifically, the type determination module determines the fault as follows: when the multi-scale energy entropy is not less than a first threshold, it is determined to be a tree-damaged hidden fault; when the multi-scale energy entropy is less than the first threshold but greater than a second threshold, it is determined to be an insulator flashover hidden fault; when the multi-scale energy entropy is not greater than the second threshold, it is determined to be a surge arrester aging hidden fault. The output module is used to output the determined hidden fault type, the corresponding multi-scale energy entropy, and alarm information.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a distribution network hidden fault identification and prediction program based on multi-scale energy entropy. When the distribution network hidden fault identification program based on multi-scale energy entropy is executed by a processor, it implements the steps of the distribution network hidden fault identification method based on multi-scale energy entropy as described in any one of claims 1 to 5.

Images