Substation high-voltage chamber power equipment insulation fault early warning and positioning method based on supercomputing
By employing a supercomputing-based approach, real-time acquisition of partial discharge signals, signal classification, and multi-feature fusion judgment, the problem of accurately locating insulation faults in power equipment in high-voltage rooms of substations was solved. This enabled reliable early warning and accurate location of faults, thereby improving the safety and stability of the power system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG POWER GRID CO LTD DONGGUAN POWER SUPPLY BUREAU
- Filing Date
- 2026-03-13
- Publication Date
- 2026-04-28
AI Technical Summary
Existing ultrasonic testing-based technologies for locating insulation faults in substation high-voltage room power equipment suffer from large positioning errors, failing to meet the need for precise location of internal fault points. Furthermore, traditional optimization algorithms are prone to getting trapped in local optima, leading to difficulties in maintenance work.
By employing a supercomputing-based approach, through real-time acquisition of partial discharge signals, signal type classification, narrowband and broadband signal processing, spatial localization of partial discharge sources, and hierarchical early warning, combined with the CTOA algorithm and multi-feature fusion judgment, accurate localization and reliable early warning of partial discharge sources are achieved.
It improves the accuracy of partial discharge source location, reduces the risk of false alarms, enables early and accurate fault identification and real-time location, and ensures the safe and stable operation of the power system.
Smart Images
Figure CN121933892A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of substation early warning and location technology, specifically to a method for early warning and location of insulation faults in high-voltage room power equipment in substations based on supercomputing. Background Technology
[0002] As the core hub of the power system, the insulation condition of key equipment such as transformers, circuit breakers, and switchgear in the high-voltage room of a substation directly determines the safety and stability of the power grid operation. With the development of power systems towards higher voltage, larger capacity, and greater intelligence, the insulation layer of high-voltage equipment is prone to localized aging and damage under long-term electrical stress, thermal stress, and environmental corrosion. This can lead to partial discharge (PD). If the PD source is not detected and located in time, the insulation fault will gradually expand, eventually causing major accidents such as equipment tripping and power grid outages, resulting in huge economic losses. Currently, various PD detection technologies have been developed in the field of power equipment insulation fault detection, mainly including ultrasonic testing, ultra-high frequency testing, and infrared thermal imaging. Among them, ultrasonic testing is widely used in the detection of PD sources in the high-voltage room of substations due to its advantages such as strong anti-electromagnetic interference capability, moderate equipment cost, and convenient operation. However, existing early warning and positioning technologies based on ultrasonic detection still have the following drawbacks: existing positioning technologies mostly construct direction finding lines based on multiple sets of DOA estimation results and solve for the coordinates of partial discharge sources through the geometric intersection method. However, due to the complex spatial structure of the high-voltage chamber (equipment obstruction, signal reflection), the direction finding lines are prone to deviation, and traditional optimization algorithms (such as the simplex method and basic particle swarm optimization algorithm) are prone to getting trapped in local optima, resulting in positioning errors generally exceeding 1m. This cannot meet the requirements for accurate positioning of fault points inside the equipment and brings great difficulties to maintenance work. Summary of the Invention
[0003] To address the shortcomings of existing technologies, this invention provides a method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, thus solving the problems mentioned in the background section.
[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, comprising the following steps: S1. Real-time acquisition of partial discharge signal: Start the partial discharge ultrasonic sensor and environmental sensor, and set the sampling frequency. Sampling period; data is identified by array number, acquisition time, and signal type, and transmitted in real time to the supercomputing high-speed storage module in binary format; S2. Signal type classification: The supercomputer's 32 threads perform frequency analysis on each frame of collected data, using formulas... Calculate the signal bandwidth, where For the maximum frequency, Minimum frequency; Then it is classified and enters the narrowband signal processing flow; Then it is classified and enters the broadband signal processing flow; S3, Narrowband and Wideband Signal Processing: Narrowband signals use the FDOA algorithm, while wideband signals use the RSS algorithm, outputting azimuth angles respectively. With elevation angle ; S4. Spatial location of partial discharge source: Direction finding lines are constructed based on the DOA estimation results of the three arrays, and the optimal partial discharge source coordinates are searched using the CTOA algorithm. S5, Tiered Early Warning Trigger: Extracting the amplitude of the partial discharge signal Frequency ,energy Three core features, based on preset warning thresholds: a single feature exceeding the threshold triggers a Level 1 warning, while multiple features exceeding the threshold or consecutive Level 1 warnings trigger a Level 2 warning.
[0005] Furthermore, in step S3, the narrowband signal processing flow is as follows: The 8-element array is divided into 5 overlapping subarrays, with 6 elements in each subarray. A supercomputer with 32 threads simultaneously calculates the covariance matrix of each subarray. ,in, For the first Data is collected from individual subarrays. This is the conjugate transpose. For the expected outcome; take each The first column is used as a subspace And default local source number The formula for global signal subspace extraction is as follows: in, Number of subarrays; Parallel traversal of azimuth angles Angle of elevation The power spectrum is calculated using the following formula: in, Given the identity matrix, this formula outputs the estimated value corresponding to the peak value. .
[0006] Furthermore, in step S3, the broadband signal processing flow is as follows: Through formula ( , (Number of frequency points), the broadband signal frequency range is discretized into 20 frequency points, reference frequency. The supercomputing thread performs a subspace extraction step (i.e., the global signal subspace extraction formula) on each of the 20 frequency points to obtain the results for each frequency point. ; Applying the focusing matrix to Rotational transformation: get , among which, among which Reference frequency The corresponding wavelength, i.e. , The speed of light in a vacuum; This is an initial rough estimate of the angle; Representation and frequency and initial rough estimate angle The relevant diagonal matrix; This constructs a diagonal matrix; the elements within the square brackets form the main diagonal elements of the diagonal matrix. Imaginary unit, satisfying ; Represents the spacing between elements in the sensor array; For the first One frequency component; Indicates the number of elements in the sensor array; For the process The signal matrix after rotation transformation; The specific formula for fusing subspaces of multiple data sets is as follows: The formula is used to express that... Subspace matrix corresponding to the group of data Perform an arithmetic average to obtain the merged subspace matrix. The merged subspace matrix Substituting into the FDOA power spectrum formula: in, The array manifold vector is used; the peak position of the power spectrum is calculated using the FDOA power spectrum formula, thereby outputting the estimated azimuth angle of the broadband signal. With elevation angle .
[0007] Furthermore, in step S4, the specific process of searching for the optimal partial discharge source coordinates using the CTOA algorithm is as follows: Based on 3 sets of array coordinates Cosine of DOA direction The following formula is used: Construct 3 direction finding lines ; Supercomputing generates 200 spatial individuals in parallel. Follows a normal cloud distribution , And cover the high-pressure chamber space; The objective function value, i.e., the sum of distances, for each individual is calculated using the following formula: Select the top 10% of individuals as elites And calculate the mean: To preserve the best-performing solution in the current iteration; through parallel execution. Guided by elite solutions, the population iteratively evolves towards better solutions, among which... Indicates the updated number One data value; Represents the original first One data value; The step size parameter is used to control the magnitude of each update. The average expected value of the data serves as a reference benchmark for updates; It is a noise intensity control parameter used to adjust the degree of influence of random noise; It represents a random number that follows a standard normal distribution with a mean of 0 and a variance of 1. It is used to introduce random disturbances to enhance the randomness of the data or to simulate uncertainties in reality. When the number of iterations reaches 100 or When the error threshold is reached, the output is... Used as coordinates for partial discharge source.
[0008] Furthermore, in step S5, the calculation formulas for the three core features are as follows: Amplitude ; Frequency ,in The pulse threshold; And energy: in .
[0009] Furthermore, in step S5, the first-level warning is a suspected fault; if... or or At this time, record the warning time and DOA information; a level 2 warning confirms the fault. and and Or, a Level 1 warning is triggered after three consecutive sampling cycles, and the positioning error... Output the fault location, warning level, and fault severity. The larger the size, the more serious the malfunction.
[0010] Furthermore, , , These are preset thresholds for amplitude, frequency, and energy, respectively. Adaptive dynamic calibration is achieved through the following steps: S6. Synchronous acquisition of environmental parameters: Temperature and humidity sensors and electromagnetic interference sensors are deployed in the high-voltage chamber. Both are synchronized with the partial discharge signal acquisition in step S1, collecting environmental data at the same sampling period and marking them as... , For temperature; Humidity; Electromagnetic interference level; S7. Parallel extraction of interference features: After receiving environmental data, the supercomputer uses wavelet packet decomposition algorithm to jointly analyze the environmental data and partial discharge signal, extracting interference correlation features. Specifically: The coupling coefficient between temperature and signal amplitude is ;in, This indicates the degree of coupling between temperature and signal amplitude; The amplitude of the partial discharge signal With ambient temperature The covariance is used to measure the correlation trend between the two during the process of change; The amplitude of the partial discharge signal The standard deviation reflects the degree of dispersion of the signal amplitude; For ambient temperature The standard deviation reflects the fluctuation of temperature data; Humidity and signal energy attenuation coefficient are ;in, Used to characterize the degree to which humidity affects signal energy; It is the partial discharge signal energy With ambient humidity The correlation coefficient measures the degree of linear correlation between the two. Electromagnetic interference and signal frequency interference coefficients are ;in, This is the reference threshold for electromagnetic interference. This represents a counting function used to count the number of samples that meet certain conditions. The sampled signal; This is the signal amplitude threshold. For real-time monitoring of electromagnetic interference intensity; This represents the total number of samples taken.
[0011] Furthermore, adaptive dynamic calibration also includes the following steps: S8. Threshold dynamic calibration model construction: Based on historical environmental and fault data, a multivariate nonlinear regression model was trained using a supercomputing machine. in, These are the trained and optimized weight coefficients, with values ranging from 0.01. 0.1, thereby enabling environmental adaptive adjustment of the early warning threshold.
[0012] Furthermore, after each threshold adjustment, the partial discharge signal and fault verification results from the subsequent three sampling cycles are combined; Calibration error is calculated in parallel using a supercomputing system. ,like Then the weight coefficients are updated dynamically. To continuously optimize the calibration model until .
[0013] Furthermore, the electromagnetic interference reference threshold The calculation was based on electromagnetic interference data from one month of normal operation of the high-voltage chamber. The calculation formula is as follows: in, This represents the average electromagnetic interference level during normal operation. The standard deviation is used, and the data is re-statistically updated quarterly via supercomputing. ; The trigger condition for threshold dynamic calibration is: when the change in environmental parameters meets the following conditions. , or When the threshold calibration process is initiated, the supercomputer immediately starts the calibration process; if there are no significant environmental changes, calibration is performed every 10 sampling cycles to ensure threshold stability.
[0014] This invention provides a method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, which has the following beneficial effects: 1. This supercomputing-based method for early warning and location of insulation faults in high-voltage substation power equipment employs adaptive algorithms for both wide and narrow band partial discharge signals. Through subarray optimization and focusing matrix improvement, the accuracy of direction-of-arrival estimation is enhanced, effectively adapting to complex signal environments. The location algorithm integrates population initialization to avoid local optima, achieving precise spatial location of the partial discharge source. Combining multi-feature fusion judgment and location results reduces the false alarm risk of traditional early warning technologies and improves early warning reliability. The overall solution achieves early and accurate fault identification, real-time location, and reliable early warning, providing scientific support for equipment operation and maintenance, effectively preventing major power grid accidents, and ensuring the safe and stable operation of the power system.
[0015] 2. This supercomputing-based method for early warning and location of insulation faults in high-voltage substation equipment constructs a dynamic calibration model through joint analysis of environmental parameters and partial discharge signals. This effectively solves the signal distortion problem caused by factors such as temperature, humidity, and electromagnetic interference in complex environments, overcoming the core pain point of false alarms or missed alarms caused by fixed thresholds. At the same time, by adopting a multivariate nonlinear regression and real-time verification iteration mechanism, combined with the parallel computing capabilities of supercomputing, the speed and accuracy of threshold adjustment are greatly improved. In addition, environmental factors such as electromagnetic interference, temperature, and humidity are quantified as interference coefficients and incorporated into threshold calibration, which improves the early warning accuracy in complex environments. Moreover, it can automatically adapt to the environmental differences of high-voltage substations without frequent manual adjustments, significantly improving the universality and engineering applicability of the method. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the partial discharge source localization and graded early warning process of an early warning and localization method for insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, according to the present invention. Figure 2 This is a schematic diagram of the threshold adaptive dynamic calibration process for an early warning and location method for insulation faults in high-voltage substation power equipment based on supercomputing, according to the present invention. Detailed Implementation
[0017] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.
[0018] like Figures 1-2 As shown, the present invention provides a technical solution: an early warning and location method for insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, comprising the following steps: S1. Real-time acquisition of partial discharge signal: Start the partial discharge ultrasonic sensor and environmental sensor, and set the sampling parameters: sampling frequency. Sampling period; The sampled data is identified by array number, acquisition time, and signal type, and transmitted in real time to the supercomputing high-speed storage module in binary format; S2. Signal type classification: The supercomputer's 32 threads perform frequency analysis on each frame of collected data, using formulas... Calculate the signal bandwidth, where For the maximum frequency, Minimum frequency; Then it is classified and enters the narrowband signal processing flow; Then it is classified and enters the broadband signal processing flow; S3, Narrowband and Wideband Signal Processing: The narrowband signal processing flow is as follows: The 8-element array is divided into 5 overlapping subarrays, and each subarray has 6 elements. The supercomputer with 32 threads simultaneously calculates the covariance matrix of each subarray. ,in, For the first Data is collected from individual subarrays. This is the conjugate transpose. For the expected outcome; take each The first column is used as a subspace And default local source number The formula for global signal subspace extraction is as follows: in, Number of subarrays; Parallel traversal of azimuth angles Angle of elevation The power spectrum is calculated using the following formula: in, Given the identity matrix, this formula outputs the estimated value corresponding to the peak value. ; The broadband signal processing flow is as follows: (using the formula...) ( , (Number of frequency points), the broadband signal frequency range is discretized into 20 frequency points, reference frequency. The supercomputing thread performs a subspace extraction step (i.e., the global signal subspace extraction formula) on each of the 20 frequency points to obtain the results for each frequency point. ; Applying the focusing matrix to Rotational transformation: get , among which, among which Reference frequency The corresponding wavelength, i.e. , The speed of light in a vacuum; This is an initial rough estimate of the angle; Representation and frequency and initial rough estimate angle The relevant diagonal matrix; This constructs a diagonal matrix; the elements within the square brackets form the main diagonal elements of the diagonal matrix. Imaginary unit, satisfying ; Represents the spacing between elements in the sensor array; For the first One frequency component; Indicates the number of elements in the sensor array; For the process The signal matrix after rotation transformation; The specific formula for fusing subspaces of multiple data sets is as follows: The formula is used to express that... Subspace matrix corresponding to the group of data Perform an arithmetic average to obtain the merged subspace matrix. The merged subspace matrix Substituting into the FDOA power spectrum formula: in, The array manifold vector is used; the peak position of the power spectrum is calculated using the FDOA power spectrum formula, thereby outputting the estimated azimuth angle of the broadband signal. With elevation angle ; S4. Spatial location of partial discharge source: Based on 3 sets of array coordinates Cosine of DOA direction The following formula is used: Construct 3 direction finding lines ; Supercomputing generates 200 spatial individuals in parallel. Follows a normal cloud distribution , And cover the high-pressure chamber space; The objective function value, i.e., the sum of distances, for each individual is calculated using the following formula: Select the top 10% of individuals as elites And calculate the mean: To preserve the best-performing solution in the current iteration; through parallel execution. Guided by elite solutions, the population iteratively evolves towards better solutions, among which... Indicates the updated number One data value; Represents the original first One data value; The step size parameter is used to control the magnitude of each update. The average expected value of the data serves as a reference benchmark for updates; It is a noise intensity control parameter used to adjust the degree of influence of random noise; It represents a random number that follows a standard normal distribution with a mean of 0 and a variance of 1. It is used to introduce random disturbances to enhance the randomness of the data or to simulate uncertainties in reality. When the number of iterations reaches 100 or When the error threshold is reached, the output is... As the coordinates of the partial discharge source; S5, Tiered Early Warning Trigger: Three core characteristics of synchronous computation in supercomputing threads: Amplitude Frequency ,in For the pulse threshold; and the energy: in ; like or or If a fault is detected, it is recorded as a suspected fault. In this case, the warning time and DOA information are recorded. The DOA information includes the signal arrival angle, time difference, and signal strength difference. , , These are the preset thresholds for amplitude, frequency, and energy, respectively. like and and Or, a Level 1 warning is triggered after three consecutive sampling cycles, and the positioning error... To confirm the fault, the system will output the fault location, warning level, and fault severity. The larger the size, the more severe the malfunction; Based on the above description, this invention employs adaptation algorithms for both wide-band and narrow-band partial discharge (PD) signal characteristics. Through subarray optimization and focusing matrix improvement, the accuracy of direction-of-arrival (DOA) estimation is enhanced, effectively adapting to complex signal environments. The localization algorithm integrates population initialization to avoid local optima, achieving precise spatial localization of the PD source. Combining multi-feature fusion judgment and localization results reduces the false alarm risk of traditional early warning technologies and improves early warning reliability. The overall solution achieves early and accurate fault identification, real-time localization, and reliable early warning, providing scientific support for equipment operation and maintenance, effectively preventing major power grid accidents, and ensuring the safe and stable operation of the power system. , , These are preset thresholds for amplitude, frequency, and energy, respectively. Adaptive dynamic calibration is achieved through the following steps: S6. Synchronous acquisition of environmental parameters: Temperature and humidity sensors and electromagnetic interference sensors are deployed in the high-voltage chamber. Both are synchronized with the partial discharge signal acquisition in step S1, collecting environmental data at the same sampling period and marking them as... , For temperature; Humidity; Electromagnetic interference level; S7. Parallel extraction of interference features: After receiving environmental data, the supercomputer uses wavelet packet decomposition algorithm to jointly analyze the environmental data and partial discharge signal, extracting interference correlation features. Specifically: The coupling coefficient between temperature and signal amplitude is ;in, This indicates the degree of coupling between temperature and signal amplitude; The amplitude of the partial discharge signal With ambient temperature The covariance is used to measure the correlation trend between the two during the process of change; The amplitude of the partial discharge signal The standard deviation reflects the degree of dispersion of the signal amplitude; For ambient temperature The standard deviation reflects the fluctuation of temperature data; Humidity and signal energy attenuation coefficient are ;in, Used to characterize the degree to which humidity affects signal energy; It is the partial discharge signal energy With ambient humidity The correlation coefficient measures the degree of linear correlation between the two. Electromagnetic interference and signal frequency interference coefficients are ;in, This is the reference threshold for electromagnetic interference. This represents a counting function used to count the number of samples that meet certain conditions. The sampled signal; This is the signal amplitude threshold. For real-time monitoring of electromagnetic interference intensity; This represents the total number of samples. Electromagnetic Interference Reference Threshold The calculation was based on electromagnetic interference data from one month of normal operation of the high-voltage chamber. The calculation formula is as follows: in, This represents the average electromagnetic interference level during normal operation. The standard deviation is used, and the data is re-statistically updated quarterly via supercomputing. ; S8. Threshold dynamic calibration model construction: Based on historical environmental and fault data, a multivariate nonlinear regression model was trained using a supercomputing machine. in, These are the trained and optimized weight coefficients, with values ranging from 0.01. 0.1, thereby achieving environmental adaptive adjustment of the early warning threshold. After each threshold adjustment, the partial discharge signal and fault verification results of the subsequent 3 sampling cycles are combined. The calibration error is calculated in parallel using a supercomputing system. ,like Then the weight coefficients are updated dynamically. To continuously optimize the calibration model until ; The trigger condition for threshold dynamic calibration is: when the change in environmental parameters meets the following conditions. , or When the threshold calibration process is initiated, the supercomputer immediately starts the calibration procedure. If there are no significant environmental changes, calibration is performed every 10 sampling cycles to ensure threshold stability. Based on the above description, this invention constructs a dynamic calibration model through joint analysis of environmental parameters and partial discharge signals. This effectively solves the signal distortion problem caused by factors such as temperature, humidity, and electromagnetic interference in complex environments, and overcomes the core pain point of false alarms or missed alarms caused by fixed thresholds. At the same time, by adopting a multivariate nonlinear regression and real-time verification iteration mechanism, combined with supercomputing parallel computing capabilities, the speed and accuracy of threshold adjustment are greatly improved. In addition, environmental factors such as electromagnetic interference, temperature, and humidity are quantified as interference coefficients and incorporated into threshold calibration, which improves the early warning accuracy in complex environments. Furthermore, it can automatically adapt to the environmental differences of high-voltage rooms in different substations without the need for frequent manual adjustments, significantly improving the universality and engineering applicability of the method.
[0019] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.
Claims
1. A method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, characterized in that: Includes the following steps: S1. Real-time acquisition of partial discharge signal: Start the partial discharge ultrasonic sensor and environmental sensor, and set the sampling frequency. Sampling period; data is identified by array number, acquisition time, and signal type, and transmitted in real time to the supercomputing high-speed storage module in binary format; S2. Signal type classification: The supercomputer's 32 threads perform frequency analysis on each frame of collected data, using formulas... Calculate the signal bandwidth, where For the maximum frequency, Minimum frequency; Then it is classified and enters the narrowband signal processing flow; Then it is classified and enters the broadband signal processing flow; S3, Narrowband and Wideband Signal Processing: Narrowband signals use the FDOA algorithm, while wideband signals use the RSS algorithm, outputting azimuth angles respectively. With elevation angle ; S4. Spatial location of partial discharge source: Direction finding lines are constructed based on the DOA estimation results of the three arrays, and the optimal partial discharge source coordinates are searched using the CTOA algorithm. S5, Tiered Early Warning Trigger: Extracting the amplitude of the partial discharge signal Frequency ,energy Three core features, based on preset warning thresholds: a single feature exceeding the threshold triggers a Level 1 warning, while multiple features exceeding the threshold or consecutive Level 1 warnings trigger a Level 2 warning.
2. The method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, as described in claim 1, is characterized in that: In step S3, the narrowband signal processing flow is as follows: The 8-element array is divided into 5 overlapping subarrays, with 6 elements in each subarray. A supercomputer with 32 threads simultaneously calculates the covariance matrix of each subarray. ,in, For the first Data is collected from individual subarrays. This is the conjugate transpose. For the expected outcome; take each The first column is used as a subspace And default local source number The formula for global signal subspace extraction is as follows: in, Number of subarrays; Parallel traversal of azimuth angles Angle of elevation The power spectrum is calculated using the following formula: in, Given the identity matrix, this formula outputs the estimated value corresponding to the peak value. .
3. The method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, as described in claim 1, is characterized in that: In step S3, the broadband signal processing flow is as follows: Through formula ( , (Number of frequency points), the broadband signal frequency range is discretized into 20 frequency points, reference frequency. The supercomputing thread performs a subspace extraction step (i.e., the global signal subspace extraction formula) on each of the 20 frequency points to obtain the results for each frequency point. ; Applying the focusing matrix to Rotational transformation: get , among which, among which Reference frequency The corresponding wavelength, i.e. , The speed of light in a vacuum; This is an initial rough estimate of the angle; Representation and frequency and initial rough estimate angle The relevant diagonal matrix; This constructs a diagonal matrix; the elements within the square brackets form the main diagonal elements of the diagonal matrix. Imaginary unit, satisfying ; Represents the spacing between elements in the sensor array; For the first One frequency component; Indicates the number of elements in the sensor array; For the process The signal matrix after rotation transformation; The specific formula for fusing subspaces of multiple data sets is as follows: The formula is used to express that... Subspace matrix corresponding to the group of data Perform an arithmetic average to obtain the merged subspace matrix. The merged subspace matrix Substituting into the FDOA power spectrum formula: in, The array manifold vector is used; the peak position of the power spectrum is calculated using the FDOA power spectrum formula, thereby outputting the estimated azimuth angle of the broadband signal. With elevation angle .
4. The method for early warning and location of insulation faults in high-voltage substation power equipment based on supercomputing, as described in claim 1, is characterized in that: In step S4, the specific process of searching for the optimal partial discharge source coordinates using the CTOA algorithm is as follows: Based on 3 sets of array coordinates Cosine of DOA direction The following formula is used: Construct 3 direction finding lines ; Supercomputing generates 200 spatial individuals in parallel. Follows a normal cloud distribution , And cover the high-pressure chamber space; The objective function value, i.e., the sum of distances, for each individual is calculated using the following formula: Select the top 10% of individuals as elites And calculate the mean: To preserve the best-performing solution in the current iteration; through parallel execution. Guided by elite solutions, the population iteratively evolves towards better solutions, among which... Indicates the updated number One data value; Represents the original first One data value; The step size parameter is used to control the magnitude of each update. The average expected value of the data serves as a reference benchmark for updates; It is a noise intensity control parameter used to adjust the degree of influence of random noise; It represents a random number that follows a standard normal distribution with a mean of 0 and a variance of 1. It is used to introduce random disturbances to enhance the randomness of the data or to simulate uncertainties in reality. When the number of iterations reaches 100 or When the error threshold is reached, the output is... Used as coordinates for partial discharge source.
5. The method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, as described in claim 1, is characterized in that: In step S5, the calculation formulas for the three core features are as follows: Amplitude ; Frequency ,in The pulse threshold; And energy: in .
6. The method for early warning and location of insulation faults in high-voltage substation power equipment based on supercomputing, as described in claim 1, is characterized in that: In step S5, the first-level warning is a suspected fault. or or At this time, record the warning time and DOA information; a level 2 warning confirms the fault. and and Or, a Level 1 warning is triggered after three consecutive sampling cycles, and the positioning error... Output the fault location, warning level, and fault severity. The larger the size, the more serious the malfunction.
7. The method for early warning and location of insulation faults in high-voltage substation power equipment based on supercomputing, as described in claim 6, is characterized in that: , , These are preset thresholds for amplitude, frequency, and energy, respectively. Adaptive dynamic calibration is achieved through the following steps: S6. Synchronous acquisition of environmental parameters: Temperature and humidity sensors and electromagnetic interference sensors are deployed in the high-voltage chamber. Both are synchronized with the partial discharge signal acquisition in step S1, collecting environmental data at the same sampling period and marking them as... , For temperature; Humidity; Electromagnetic interference level; S7. Parallel extraction of interference features: After receiving environmental data, the supercomputer uses wavelet packet decomposition algorithm to jointly analyze the environmental data and partial discharge signal, extracting interference correlation features. Specifically: The coupling coefficient between temperature and signal amplitude is ;in, This indicates the degree of coupling between temperature and signal amplitude; The amplitude of the partial discharge signal With ambient temperature The covariance is used to measure the correlation trend between the two during the process of change; The amplitude of the partial discharge signal The standard deviation reflects the degree of dispersion of the signal amplitude; For ambient temperature The standard deviation reflects the fluctuation of temperature data; Humidity and signal energy attenuation coefficient are ;in, Used to characterize the degree to which humidity affects signal energy; It is the partial discharge signal energy With ambient humidity The correlation coefficient measures the degree of linear correlation between the two. Electromagnetic interference and signal frequency interference coefficients are ;in, This is the reference threshold for electromagnetic interference. This represents a counting function used to count the number of samples that meet certain conditions. The sampled signal; This is the signal amplitude threshold. For real-time monitoring of electromagnetic interference intensity; This represents the total number of samples taken.
8. The method for early warning and location of insulation faults in power equipment in high-voltage rooms of substations based on supercomputing, as described in claim 7, is characterized in that: Adaptive dynamic calibration also includes the following steps: S8. Threshold dynamic calibration model construction: Based on historical environmental and fault data, a multivariate nonlinear regression model was trained using a supercomputing machine. in, These are the trained and optimized weight coefficients, with values ranging from 0.
01. 0.1, thereby enabling environmental adaptive adjustment of the early warning threshold.
9. A method for early warning and location of insulation faults in high-voltage substation power equipment based on supercomputing, as described in claim 8, characterized in that: After each threshold adjustment, the partial discharge signal and fault verification results of the subsequent three sampling cycles are combined; Calibration error is calculated in parallel using a supercomputing system. ,like Then the weight coefficients are updated dynamically. To continuously optimize the calibration model until .
10. A method for early warning and location of insulation faults in high-voltage substation power equipment based on supercomputing, as described in claim 7, characterized in that: Electromagnetic interference reference threshold The calculation was based on electromagnetic interference data from one month of normal operation of the high-voltage chamber. The calculation formula is as follows: in, This represents the average electromagnetic interference level during normal operation. The standard deviation is used, and the data is re-statistically updated quarterly via supercomputing. ; The trigger condition for threshold dynamic calibration is: when the change in environmental parameters meets the following conditions. , or When the threshold calibration process is initiated, the supercomputer immediately starts the calibration process; if there are no significant environmental changes, calibration is performed every 10 sampling cycles to ensure threshold stability.