Dynamic identification method, system and storage medium for dual-frequency coupled partial discharge type in switchgear

Abstract

This invention provides a method, system, and storage medium for dynamic identification of dual-frequency coupled partial discharge (PD) types in switchgear. The method includes acquiring a set of potential PD locations in the target switchgear; pre-setting a dual-frequency monitoring array based on the set of potential PD locations; deploying a synchronous acquisition system; performing dual-frequency coupled collaborative acquisition through the dual-frequency monitoring array and the acquisition system to obtain a dual-frequency signal pair array; dynamically fusing spatiotemporal-frequency domain multidimensional features of the dual-frequency signal pair array; inputting the fused spatiotemporal-frequency domain multidimensional feature array into a pre-set PD feature decision forest for PD type discrimination; and dynamically adjusting the weights of the PD type discrimination results. This invention, by deploying a spatiotemporally correlated dual-frequency monitoring array and acquisition system, constructs a three-level dynamic triggering mechanism, significantly enhancing the collaborative capture capability of transient pulses and intermittent discharges, and effectively solving the problem of missed detection of weak discharge signals by a single sensor.

Description

Technical Field

[0001] This invention relates to the field of partial discharge identification technology, specifically to a method, system, and storage medium for dynamic identification of dual-frequency coupled partial discharge types in switchgear. Background Technology

[0002] As a critical piece of equipment in the power system, the insulation condition of switchgear directly affects the reliability of power supply. Partial discharge (PD) is an early sign of insulation degradation, and accurately identifying the type of discharge (such as surface discharge of insulators, internal air gap discharge, floating potential discharge, etc.) is crucial for fault early warning and precise operation and maintenance.

[0003] Currently, partial discharge monitoring in switchgear mainly relies on two types of sensing technologies: Transient Ground Voltage (TEV) and Ultra-High Frequency (UHF). However, single sensors (TEV or UHF) are susceptible to electromagnetic interference and signal attenuation, and their ability to capture low-energy discharges or transient pulses is insufficient. Traditional fixed threshold triggering mechanisms are difficult to adapt to the randomness and intermittency of discharge signals, leading to a significant increase in the missed detection rate of weak discharge events. Furthermore, existing methods are mostly based on extracting time-domain (e.g., pulse amplitude, phase) or frequency-domain (e.g., spectral centroid) features from a single sensor, making it difficult to comprehensively characterize the physical nature of the discharge. For example, TEV signals are sensitive to internal discharges but have low positioning accuracy, while UHF signals can spatially locate discharges but are easily shielded by the cabinet structure. The lack of cross-modal features leads to insufficient differentiation of similar discharge types, such as a high misjudgment rate for insulator surface discharges and metal tip discharges. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method, system, and storage medium for dynamic identification of dual-frequency coupled partial discharge types in switchgear, thereby resolving the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic identification of dual-frequency coupled partial discharge type in switchgear, comprising the following steps:

[0006] Obtain the set of potential partial discharge points of the target switchgear, and preset a dual-frequency monitoring array based on the set of potential partial discharge points;

[0007] Deploy a synchronous data acquisition system;

[0008] The acquisition system includes a reference clock, an acquisition channel, and an FPGA trigger logic unit.

[0009] The dual-frequency signal pair array is obtained by performing dual-frequency coupled collaborative acquisition through the dual-frequency monitoring array and acquisition system.

[0010] The dual-frequency signal pair array is subjected to dynamic fusion of spatiotemporal and frequency domain multidimensional features;

[0011] The partial discharge feature parameter array, which integrates spatiotemporal and frequency domain multidimensional features, is input into a preset partial discharge feature decision forest to determine the discharge type.

[0012] The results of discharge type discrimination are dynamically weighted to obtain the final discharge type identification result;

[0013] The final result of the output discharge type identification.

[0014] As a further preferred embodiment, the process of obtaining the set of potential partial discharge points of the target switchgear and pre-setting a dual-frequency monitoring array based on the set of potential partial discharge points includes:

[0015] Based on the three-dimensional structural model of the switchgear, historical fault database and electromagnetic field simulation analysis, the set of potential locations for partial discharge inside the target switchgear is interactively determined.

[0016] Based on the set of potential discharge points, a dual-frequency sensor layout is designed. A transient ground voltage sensor array is deployed at key points on the cabinet grounding wire and metal surface, and an ultra-high frequency sensor array is deployed at electromagnetic wave leakage points such as cabinet gaps and observation windows.

[0017] Establish a sensor spatial location mapping model to ensure that each transient ground voltage monitoring point has a corresponding UHF monitoring point to form a spatial correlation pair;

[0018] When determining the potential locations of partial discharge, high-risk areas such as insulator surfaces, busbar connections, and cable terminals should be highlighted.

[0019] As a further preferred embodiment, the process of obtaining the dual-frequency signal pair array includes:

[0020] The specific process for performing dual-frequency coupled collaborative acquisition to obtain dual-frequency signal pairs from the array includes:

[0021] Dynamic discharge monitoring is implemented on the target switchgear, and a three-level triggering strategy is used to acquire the dual-frequency signal timing array:

[0022] The three-level triggering strategy includes: initiating main triggering when the UHF sensor detects that the pulse leading edge exceeds the threshold;

[0023] The auxiliary trigger is activated when the TEV sensor detects that the energy exceeds the threshold.

[0024] Joint triggering is initiated when the UHF and TEV signals satisfy the logical AND relationship.

[0025] Based on the dual-frequency signal timing array, a time-aligned dual-frequency signal pair array is generated, with each signal pair containing a time-synchronized TEV signal and a UHF signal.

[0026] As a further optimization, a balance model between mold-changing costs and production efficiency is constructed. The specific process for obtaining the optimal production path based on the balance model includes:

[0027] Wavelet packet decomposition is performed on the TEV signal in each signal pair to calculate its energy entropy. A Hankel matrix is ​​constructed for singular value decomposition, and the first 8 singular values ​​are extracted as features. The S-transform is performed on the UHF signal to obtain the time spectrum, and the dominant frequency offset and bandwidth factor are calculated.

[0028] Calculate the cross-correlation function of TEV and UHF signals in the same signal pair, extract the peak position of cross-correlation as the propagation delay difference, calculate the logarithmic ratio of UHF signal energy to TEV signal energy, and use the delay difference information of several sensor pairs to establish a system of equations to find the three-dimensional spatial coordinates of the power source.

[0029] For multiple consecutive discharge pulses, the power frequency phase and propagation delay difference of each pulse are statistically analyzed, a phase-delay joint histogram is constructed, and the phase distribution entropy and delay fluctuation coefficient are extracted from it;

[0030] After weighted fusion of dynamic features, a partial discharge feature parameter array with fused spatiotemporal-frequency domain multidimensional features is obtained.

[0031] As a further preferred option, the specific process of the dynamic feature weighted fusion includes:

[0032] The single-modal features, cross-modal joint features, and dynamic PRPD features are combined into an 8-dimensional feature vector;

[0033] The features in the feature vector are weighted and adjusted according to signal quality (such as the singular value stability of TEV and the signal-to-noise ratio of UHF).

[0034] The adjusted feature vector is processed by a sliding window, scanning the signal stream (feature vector) with a window length of 100ms and a step size of 20ms. Processing is triggered when the number of effective pulses in the window is ≥5.

[0035] PCA is performed on the feature matrix (dimension n×8, where n is the number of pulses) of all pulses within the window. The first 5 dimensions of features retain 95% of the energy information, forming a dimension-reduced feature parameter array, thus obtaining a partial discharge feature parameter array that integrates spatiotemporal and frequency domain multidimensional features.

[0036] As a further preferred embodiment, the root node of the decision tree in the partial discharge feature decision forest is divided into a stable discharge branch and an unstable discharge branch using the time delay difference standard deviation feature.

[0037] As a further optimization, the specific process for dynamically adjusting the weights of the discharge type discrimination results includes:

[0038] Calculate the weighting coefficients for UHF and TEV based on the signal quality of the dual-frequency monitoring array;

[0039] The weighting coefficients for UHF and TEV are calculated using the following formulas:

[0040] ;

[0041] ;

[0042] in, Indicates UHF weights, Indicates the UHF signal-to-noise ratio. This represents the normalized UHF noise energy. Indicates the stability of the TEV channel. This indicates the power frequency harmonic distortion rate of the TEV channel. Indicates the TEV weight;

[0043] The weighting coefficients of UHF and TEV are probabilistically fused with the results of discharge feature decision forest discharge type discrimination to obtain the final identification result of discharge type;

[0044] The probability fusion formula is:

[0045] ;

[0046] ;

[0047] in, This indicates the fusion probability result. This represents the type probability output by the decision forest based on UHF features. This represents the type probability output by the decision forest based on TEV features. This indicates the final fusion result.

[0048] As a further preferred embodiment, the final discharge type identification result includes discharge type, confidence level, weight allocation, key feature values, and signal quality indicators.

[0049] As a further preferred embodiment, a dynamic identification system for dual-frequency coupled partial discharge type of switchgear is provided to implement the above-mentioned dynamic identification method for dual-frequency coupled partial discharge type of switchgear, comprising:

[0050] The three-dimensional discharge point localization and array deployment module is used to acquire the set of potential partial discharge points of the target switchgear, and to preset a dual-frequency monitoring array and deploy a synchronous acquisition system based on the set of potential partial discharge points;

[0051] The dual-frequency collaborative signal processing module is used to perform dual-frequency coupled collaborative acquisition through the dual-frequency monitoring array and the acquisition system to obtain a dual-frequency signal pair array, and to perform spatiotemporal-frequency domain multidimensional feature dynamic fusion on the dual-frequency signal pair array;

[0052] The hierarchical discharge type decision module is used to input the partial discharge feature parameter array, which integrates spatiotemporal and frequency domain multidimensional features, into a preset partial discharge feature decision forest to determine the discharge type.

[0053] The confidence fusion and result output module is used to dynamically adjust the weights of the discharge type discrimination results, obtain the final discharge type identification result, and output it.

[0054] As a further preferred embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described dynamic identification method for dual-frequency coupled partial discharge type of switchgear.

[0055] This invention provides a method, system, and storage medium for dynamic identification of dual-frequency coupled partial discharge types in switchgear, which has the following advantages: By deploying a spatiotemporally correlated dual-frequency monitoring array and acquisition system, a three-level dynamic triggering mechanism is constructed, significantly enhancing the collaborative capture capability of transient pulses and intermittent discharges, effectively solving the problem of missed detection of weak discharge signals by a single sensor; based on a spatiotemporal-frequency domain multi-dimensional feature dynamic fusion mechanism, cross-modal features such as signal energy entropy, main frequency offset, and propagation delay difference are extracted in parallel, and an adaptive fusion strategy with signal quality weighting is introduced to overcome the limitations of traditional single feature representation and achieve a holographic representation of discharge characteristics. This method employs a hierarchical decision forest model, incorporating threshold logic based on key features such as propagation delay difference stability, UHF frequency offset range, and energy ratio, to establish a layered discrimination path for discharge types. This significantly improves the accuracy of distinguishing typical defects such as surface discharge and internal air gap discharge in insulators. An innovative dual-channel probabilistic dynamic fusion mechanism is designed, adjusting weight allocation in real time based on UHF signal-to-noise ratio and TEV channel stability, ensuring robustness even under strong electromagnetic interference or signal attenuation scenarios. Combining PCA dimensionality reduction of pulse groups within a sliding time window with dynamic PRPD phase analysis, online tracking and early warning of discharge mode evolution are achieved. This method overcomes the reliance of traditional partial discharge monitoring on continuous and stable discharge, increasing the detection rate of intermittent discharge by over 40%. Furthermore, through multi-dimensional feature fusion and a dynamic decision mechanism, it achieves minute-level real-time discrimination of discharge types under complex operating conditions, providing technical support for proactive defense and precise operation and maintenance of switchgear insulation status. Attached Figure Description

[0056] Figure 1 This is a flowchart of the dynamic identification method for dual-frequency coupled partial discharge type in switchgear according to the present invention;

[0057] Figure 2 This is a block diagram of the dynamic identification system for dual-frequency coupled partial discharge type of switchgear according to the present invention. Detailed Implementation

[0058] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0059] The following disclosure provides many different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. In addition, examples of various specific processes and materials are provided in this invention, but those skilled in the art will recognize the application of other processes and / or the use of other materials.

[0060] like Figure 1 As shown, this embodiment of the invention provides a method for dynamic identification of dual-frequency coupled partial discharge type in switchgear, including the following steps:

[0061] S1: Obtain the set of potential partial discharge points of the target switchgear, and preset a dual-frequency monitoring array based on the set of potential partial discharge points;

[0062] Specifically, the process of step S1 includes:

[0063] Based on the three-dimensional structural model of the switchgear, historical fault database and electromagnetic field simulation analysis, the set of potential locations for partial discharge inside the target switchgear is interactively determined.

[0064] Based on the set of potential discharge points, a dual-frequency sensor layout is designed: Transient ground voltage (TEV) sensor arrays are deployed at key points on the cabinet grounding wire and metal surface; Ultra-high frequency (UHF) sensor arrays are deployed at electromagnetic wave leakage points such as cabinet gaps and observation windows.

[0065] Establish a sensor spatial location mapping model to ensure that each transient ground voltage monitoring point has a corresponding UHF monitoring point to form a spatial correlation pair.

[0066] When determining the potential locations of partial discharge, high-risk areas such as insulator surfaces, busbar connections, and cable terminals should be highlighted.

[0067] S2: Deploy a synchronous data acquisition system;

[0068] The acquisition system includes a reference clock, acquisition channels, and FPGA trigger logic units.

[0069] Specifically, a high-precision atomic clock disciplined clock source is configured as the system's reference clock, and a dual-channel synchronous acquisition hardware architecture is constructed. A GPS disciplined clock is used as the reference clock source, which is transmitted to each acquisition channel through a clock distribution unit. An FPGA trigger logic unit realizes cross-channel joint trigger control, ensuring that the time synchronization accuracy of all sensor data is within 1 nanosecond.

[0070] It should be noted that the hardware configuration can be selected according to actual needs. The FPGA trigger logic unit to realize cross-channel joint trigger control is an existing technology and will not be described in detail here.

[0071] S3: Perform dual-frequency coupled collaborative acquisition through the dual-frequency monitoring array and acquisition system to obtain a dual-frequency signal pair array;

[0072] Specifically, the process of performing dual-frequency coupled collaborative acquisition to obtain dual-frequency signal pairs from the array includes:

[0073] Dynamic discharge monitoring is implemented on the target switchgear, and a three-level triggering strategy is used to acquire the dual-frequency signal timing array:

[0074] The three-level triggering strategy includes: initiating main triggering when the UHF sensor detects that the pulse leading edge exceeds the threshold;

[0075] The auxiliary trigger is activated when the TEV sensor detects that the energy exceeds the threshold.

[0076] Joint triggering is initiated when the UHF and TEV signals satisfy the logical AND relationship.

[0077] Based on the dual-frequency signal timing array, a time-aligned dual-frequency signal pair array is generated, with each signal pair containing a time-synchronized TEV signal and a UHF signal.

[0078] S4: Perform dynamic fusion of spatiotemporal-frequency domain multidimensional features on the dual-frequency signal pair array;

[0079] Specifically, the spatiotemporal-frequency domain multidimensional feature dynamic fusion extraction is a hierarchical feature fusion process, the specific steps of which include:

[0080] Parallel extraction of time-frequency features: Wavelet packet decomposition is performed on the TEV signal in each signal pair to calculate its energy entropy, and a Hankel matrix is ​​constructed for singular value decomposition to extract the first 8 singular values ​​as features; S-transform is performed on the UHF signal to obtain the time spectrum, and the dominant frequency offset and bandwidth factor are calculated.

[0081] Cross-modal spatiotemporal feature fusion: Calculate the cross-correlation function of TEV and UHF signals in the same signal pair, extract the peak position of cross-correlation as the propagation delay difference; calculate the logarithmic ratio of UHF signal energy to TEV signal energy; use the delay difference information of several sensor pairs to establish a system of equations to find the three-dimensional spatial coordinates of the power supply.

[0082] Dual-frequency phase correlation analysis: For multiple consecutive discharge pulses, the power frequency phase and propagation delay difference of each pulse are statistically analyzed, a phase-delay joint histogram is constructed, and the phase distribution entropy and delay fluctuation coefficient are extracted from it;

[0083] Dynamic feature weighted fusion: combines single-modal features, cross-modal joint features, and dynamic PRPD features into an 8-dimensional feature vector;

[0084] The features in the feature vector are weighted and adjusted according to signal quality (such as the singular value stability of TEV and the signal-to-noise ratio of UHF).

[0085] The adjusted feature vector is processed by a sliding window, scanning the signal stream (feature vector) with a window length of 100ms and a step size of 20ms. Processing is triggered when the number of effective pulses in the window is ≥5.

[0086] PCA is performed on the feature matrix (dimension n×8, where n is the number of pulses) of all pulses within the window. The first 5 dimensions of features retain 95% of the energy information, forming a dimension-reduced feature parameter array, thus obtaining a partial discharge feature parameter array that integrates spatiotemporal and frequency domain multidimensional features.

[0087] Specifically, it utilizes time domain (e.g., energy entropy), frequency domain (e.g., dominant frequency offset, bandwidth factor), spatiotemporal domain (propagation delay difference, spatial positioning), and phase statistical features simultaneously to overcome the limitations of single feature representation and comprehensively describe discharge characteristics. Features are dynamically weighted using signal quality assessment factors, with weights reduced when UHF signals are interfered with and when TEV signals are weak, improving feature robustness. Different discharge types (e.g., internal discharge, surface discharge, corona discharge) exhibit significant differences in propagation delay difference, energy ratio, and phase distribution; fused features can effectively distinguish them. Spatial positioning information not only provides the location of the discharge source, but its positioning residual can also serve as an indicator of the severity of insulation defects. Sliding time window processing, combined with in-window pulse statistical features, can capture the dynamic evolution of discharge modes.

[0088] S5: Input the partial discharge feature parameter array that integrates spatiotemporal and frequency domain multidimensional features into the preset partial discharge feature decision forest to determine the discharge type;

[0089] The root node of the decision tree in the partial discharge characteristic decision forest is divided into a stable discharge branch and an unstable discharge branch using the time delay difference standard deviation feature;

[0090] For example, the root node is partitioned using the time delay difference feature 'a':

[0091] When a < 2ns, it enters a stable discharge branch;

[0092] When a ≥ 2ns, it enters the unstable discharge branch;

[0093] Stable discharge branch (a < 2ns):

[0094] The secondary node adopts the TEV oscillation period characteristic (b):

[0095] If 1ns < b < 5ns: Enter the third-level PRPD phase characteristic node (c);

[0096] c > 0.8: Insulator surface discharge is detected;

[0097] c ≤ 0.8: Entering the fourth-order energy ratio node (d);

[0098] d > -10dB: Determined as floating discharge;

[0099] d ≤ -10dB: Determined as internal air gap discharge;

[0100] If b ≤ 1ns or ≥ 5ns: mark as abnormal discharge;

[0101] Unstable discharge branch (a ≥ 2ns):

[0102] The secondary node adopts the UHF frequency offset range characteristic (e):

[0103] 300MHz < e < 600MHz: determined to be surface discharge;

[0104] e > 800MHz: Determined as a tip discharge.

[0105] The hierarchical feature decision tree structure includes branches for typical discharge types such as insulator surface discharge, internal air gap discharge, and suspension discharge.

[0106] It should be noted that each branch sets characteristic discrimination nodes: TEV oscillation period threshold (1-5ns), UHF frequency offset range (300-1500MHz), energy ratio range (-20~10dB), time delay standard deviation limit (<2ns), PRPD phase distribution characteristics (surface discharge exhibits bimodal characteristics), and other discrimination conditions.

[0107] S6: Dynamically adjust the weights of the discharge type discrimination results to obtain the final discharge type identification result;

[0108] The specific process for dynamically adjusting the weights of the discharge type determination results includes:

[0109] Calculate the weighting coefficients for UHF and TEV based on the signal quality of the dual-frequency monitoring array;

[0110] The weighting coefficients for UHF and TEV are calculated using the following formulas:

[0111] ;

[0112] ;

[0113] in, Indicates UHF weights, Indicates the UHF signal-to-noise ratio. This represents the normalized UHF noise energy. Indicates the stability of the TEV channel. This indicates the power frequency harmonic distortion rate of the TEV channel. Indicates the TEV weight;

[0114] The weighting coefficients of UHF and TEV are probabilistically fused with the results of discharge feature decision forest discharge type discrimination to obtain the final identification result of discharge type;

[0115] The probability fusion formula is:

[0116] ;

[0117] ;

[0118] in, This indicates the fusion probability result. This represents the type probability output by the decision forest based on UHF features. This represents the type probability output by the decision forest based on TEV features. This indicates the final fusion result.

[0119] It should be noted that the formula for calculating the UHF signal-to-noise ratio is as follows: The formula for calculating TEV channel stability is: ;

[0120] in, Indicates the peak amplitude of the partial discharge pulse. The standard deviation of the background noise signal. This represents the standard deviation of the TEV baseline voltage. This indicates the rated voltage of the equipment.

[0121] S7: Output discharge type final identification result.

[0122] The final discharge type identification result includes: discharge type, confidence level, weight allocation, key feature values, and signal quality indicators.

[0123] This application presents a dynamic identification method for dual-frequency coupled partial discharge types in switchgear. By deploying a spatiotemporally correlated dual-frequency monitoring array and acquisition system, a three-level dynamic triggering mechanism is constructed, significantly enhancing the collaborative capture capability of transient pulses and intermittent discharges, effectively solving the problem of missed detection of weak discharge signals by a single sensor. Based on a spatiotemporal-frequency domain multi-dimensional feature dynamic fusion mechanism, cross-modal features such as signal energy entropy, dominant frequency offset, and propagation delay difference are extracted in parallel, and an adaptive fusion strategy with signal quality weighting is introduced to overcome the limitations of traditional single feature representation and achieve a holographic characterization of discharge characteristics. By constructing a hierarchical... The decision forest model, combined with threshold logic based on key features such as propagation delay difference stability, UHF frequency offset range, and energy ratio, establishes a hierarchical discrimination path for discharge types, significantly improving the accuracy of distinguishing typical defects such as surface discharge and internal air gap discharge of insulators. An innovative dual-channel probabilistic dynamic fusion mechanism is designed, adjusting weight allocation in real time based on UHF signal-to-noise ratio and TEV channel stability, ensuring robustness even under strong electromagnetic interference or signal attenuation scenarios. Combining PCA dimensionality reduction of pulse groups within a sliding time window with dynamic PRPD phase analysis, online tracking and early warning of discharge mode evolution are achieved. This method breaks through the dependence of traditional partial discharge monitoring on continuous stable discharge, increasing the detection rate of intermittent discharge by more than 40%. Simultaneously, through multi-dimensional feature fusion and dynamic decision-making mechanisms, it achieves minute-level real-time discrimination of discharge types under complex operating conditions, providing core technical support for proactive defense and precise operation and maintenance of switchgear insulation status.

[0124] like Figure 2 As shown, this embodiment also provides a dynamic identification system for dual-frequency coupled partial discharge type of switchgear, used to implement the above-mentioned dynamic identification method for dual-frequency coupled partial discharge type of switchgear, including:

[0125] The three-dimensional discharge point localization and array deployment module is used to acquire the set of potential partial discharge points of the target switchgear, and to preset a dual-frequency monitoring array and deploy a synchronous acquisition system based on the set of potential partial discharge points;

[0126] The dual-frequency collaborative signal processing module is used to perform dual-frequency coupled collaborative acquisition through the dual-frequency monitoring array and the acquisition system to obtain a dual-frequency signal pair array, and to perform spatiotemporal-frequency domain multidimensional feature dynamic fusion on the dual-frequency signal pair array;

[0127] The hierarchical discharge type decision module is used to input the partial discharge feature parameter array, which integrates spatiotemporal and frequency domain multidimensional features, into a preset partial discharge feature decision forest to determine the discharge type.

[0128] The confidence fusion and result output module is used to dynamically adjust the weights of the discharge type discrimination results, obtain the final discharge type identification result, and output it.

[0129] This embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described dynamic identification method for dual-frequency coupled partial discharge types in switchgear.

[0130] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A switch cabinet dual-frequency coupling partial discharge type dynamic identification method, characterized in that, The method comprises the following steps: acquire a set of potential partial discharge points of a target switch cabinet, and preset a dual-frequency monitoring array according to the set of potential partial discharge points; The dual-frequency monitoring array is specifically a transient earth voltage (TEV) sensor array and an ultra-high frequency (UHF) sensor array; deploy a synchronous acquisition system; The acquisition system comprises a reference clock, an acquisition channel, and an FPGA trigger logic unit; Perform dual-frequency coupled cooperative acquisition through the dual-frequency monitoring array and the acquisition system to obtain a dual-frequency signal pair array; Each dual-frequency signal pair comprises a time-synchronized TEV signal and a UHF signal; Perform dynamic fusion of time-space-frequency multidimensional features on the dual-frequency signal pair array; Input the partial discharge feature parameter array fused with the time-space-frequency multidimensional features into a preset partial discharge feature decision forest to determine the discharge type; Perform dynamic weight adjustment on the result of the discharge type determination to obtain a final discharge type recognition result; Output the final discharge type recognition result; The dynamic fusion of time-space-frequency multidimensional features is specifically hierarchical feature fusion, and the specific process comprises the following steps: Parallel extraction of time-frequency features: perform wavelet packet decomposition on the TEV signal in each signal pair, calculate the energy entropy, construct a Hankel matrix for singular value decomposition, and extract the first 8 singular values as features; perform S transform on the UHF signal to obtain a time-frequency spectrum, and calculate the main frequency offset and bandwidth factor; Cross-modal time-space feature fusion: calculate the cross-correlation function of the TEV and UHF signals in the same signal pair, extract the cross-correlation peak position as the propagation time delay difference, calculate the logarithmic ratio of the UHF signal energy and the TEV signal energy, and use the propagation time delay differences of a plurality of sensor pairs to establish an equation set to solve the three-dimensional spatial coordinates of the discharge source; Dual-frequency phase correlation analysis: for a plurality of continuous discharge pulses, count the power frequency phase and the propagation time delay difference of each pulse, construct a phase-time delay joint histogram, and extract the phase distribution entropy and the time delay fluctuation coefficient from the histogram; Dynamic feature weighted fusion: combine the single-modal features, the cross-modal joint features, and the dynamic PRPD features into an 8-dimensional feature vector; According to the signal quality, weight and adjust each feature in the feature vector; Perform sliding window processing on the adjusted feature vector to scan the signal stream with a 100ms window length and a 20ms step length, and trigger processing when the number of effective pulses in the window is greater than or equal to 5; Perform PCA on the feature matrix of all pulses in the window, retain the first 5 features with 95% energy information to form a reduced feature parameter array, and obtain the partial discharge feature parameter array fused with the time-space-frequency multidimensional features.

2. The switch cabinet double-frequency coupling partial discharge type dynamic identification method according to claim 1, characterized in that, The process of acquiring the set of potential partial discharge points of the target switch cabinet and presetting the dual-frequency monitoring array based on the set of potential partial discharge points comprises the following steps: Based on the three-dimensional structure model of the switch cabinet, the historical fault database, and the electromagnetic field simulation analysis, interactively determine the set of potential partial discharge points inside the target switch cabinet; Based on the set of potential partial discharge points, design the layout of the dual-frequency sensors, arrange the transient earth voltage sensor array at the key points of the cabinet ground wire and the metal surface, and arrange the ultra-high frequency sensor array at the electromagnetic wave leakage points such as the cabinet gaps and the observation windows. A sensor spatial position mapping model is established to ensure that each transient voltage monitoring point has a corresponding UHF monitoring point to form a spatial correlation pair. In the determination of the set of potential partial discharge points, the high-risk areas of the insulator surface, the bus connection, and the cable terminal are highlighted.

3. The switch cabinet double-frequency coupling partial discharge type dynamic identification method according to claim 2, characterized in that, The process of obtaining the double-frequency signal pair array includes: Dynamic discharge monitoring is performed on the target switch cabinet, and a three-level trigger strategy is used to obtain a double-frequency signal timing array: The three-level trigger strategy includes: starting the main trigger when the UHF sensor detects a pulse front exceeding a threshold value; starting the auxiliary trigger when the TEV sensor detects energy exceeding a threshold value; starting the joint trigger when the UHF and TEV signals satisfy a logical AND relationship; Based on the double-frequency signal timing array, a time-stamped double-frequency signal pair array is generated.

4. The switch cabinet double-frequency coupling partial discharge type dynamic identification method according to claim 1, characterized in that, The decision tree root node of the partial discharge feature decision forest is divided into stable discharge branches and non-stable discharge branches using the time delay difference standard deviation feature.

5. The switch cabinet double-frequency coupling partial discharge type dynamic identification method according to claim 1, characterized in that, The specific process of dynamically adjusting the results of discharge type discrimination includes: Based on the signal quality of the double-frequency monitoring array, the weight coefficients of UHF and TEV are calculated. The weight coefficient calculation formula of UHF and TEV is: ; ; wherein, represents the UHF weight, represents the UHF signal-to-noise ratio, represents the normalized UHF noise energy, represents the TEV channel stability, represents the TEV channel power frequency harmonic distortion rate, represents the TEV weight; The weight coefficients of UHF and TEV are combined with the results of discharge type discrimination of the discharge feature decision forest to obtain the final recognition result of the discharge type. The probability fusion formula is: ； ； wherein, represents the fusion probability result, represents the type probability output by the decision forest based on the UHF features, represents the type probability output by the decision forest based on the TEV features, represents the final fusion result.

6. The switch cabinet double-frequency coupling partial discharge type dynamic identification method according to claim 1, characterized in that, The final recognition result of the discharge type includes the discharge type, confidence, weight distribution, key feature value, and signal quality index.

7. A switch cabinet dual-frequency coupled partial discharge type dynamic identification system for implementing the switch cabinet dual-frequency coupled partial discharge type dynamic identification method of any one of claims 1-6, characterized in that, It includes: A three-dimensional discharge point positioning and array deployment module is used to obtain a set of potential partial discharge points of the target switch cabinet, and a double-frequency monitoring array is preset based on the set of potential partial discharge points and a synchronous acquisition system is deployed; A double-frequency cooperative signal processing module is used to perform double-frequency coupling cooperative acquisition through the double-frequency monitoring array and the acquisition system, obtain a double-frequency signal pair array, and perform dynamic fusion of time-space-frequency domain multi-dimensional features on the double-frequency signal pair array; A hierarchical discharge type decision module is used to input the local discharge feature parameter array fused with time-space-frequency domain multi-dimensional features into a pre-set local discharge feature decision forest to discriminate the discharge type; A confidence fusion and result output module is used to dynamically adjust the results of discharge type discrimination, obtain the final recognition result of the discharge type, and output it.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to realize the steps of the switch cabinet double-frequency coupling partial discharge type dynamic identification method of any one of claims 1-6.

Images