Cable grounding system defect live detection method

Abstract

The invention relates to the technical field of cable detection, and discloses a cable grounding system defect live-line detection method, which comprises the following steps: injecting a low-frequency detection current signal into a cable grounding system, and synchronously acquiring a response voltage signal; obtaining an equivalent pure resistance value through frequency domain decomposition and complex number deconstruction, and obtaining a multi-dimensional signal feature by combining voltage signal waveform characterization; based on the typical physical defect type of the cable grounding system, performing mode matching on the multi-dimensional signal features to obtain a feature defect mapping table, and constructing a fault feature database according to the feature defect mapping table; searching and comparing the on-site detected multi-dimensional signal feature vector with the fault feature database, and performing intelligent diagnosis on the comparison result based on the hybrid expert network to obtain the specific defect type and severity of the cable grounding system; carrying out situation study and judgment on the specific defect type and severity to obtain early warning information; according to the invention, the efficiency of cable grounding system defect live-line detection can be improved.

Description

Technical Field

[0001] This invention relates to the field of cable testing technology, and in particular to a method for detecting defects in cable grounding systems while they are energized. Background Technology

[0002] As a crucial guarantee for the safe operation of power systems, the accuracy and timeliness of live-line defect detection in cable grounding systems directly impact the quality of power grid operation and maintenance. Current traditional detection methods lack precise frequency domain analysis and complex deconstruction techniques in the signal processing stage, making it difficult to effectively separate the effective components of the loop impedance. This results in incomplete capture of defect-related electrical characteristics, thus affecting the reliability of defect assessment.

[0003] Existing technologies have significant shortcomings in feature extraction and diagnostic modeling. The integration of multi-dimensional signal features lacks scientific rigor, the construction of fault feature databases does not fully integrate the physical mechanisms of typical defects with actual operating data, and the adaptability of diagnostic models is poor, making it impossible to efficiently process detection information under complex working conditions. This results in long time consumption and insufficient accuracy in defect type identification and severity assessment. Therefore, how to improve the efficiency of live defect detection in cable grounding systems has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a method for detecting live defects in cable grounding systems to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for live detection of defects in a cable grounding system, comprising: S1. Inject the low-frequency detection current signal into the cable grounding system, and simultaneously collect the response voltage signal generated by the low-frequency detection current signal in the cable grounding system; S2. Perform frequency domain decomposition on the response voltage signal and the low-frequency detection current signal to obtain the loop impedance of the cable grounding system, and perform complex decomposition on the loop impedance to obtain the equivalent pure resistance value of the loop impedance. S3. The equivalent pure resistance value and the waveform of the response voltage signal are characterized by state to obtain the multi-dimensional signal characteristics of the cable grounding system; S4. Based on the typical physical defect types of the cable grounding system, perform pattern matching on the multidimensional signal features to obtain the feature defect mapping table of the cable grounding system, and construct the fault feature database of the cable grounding system according to the feature defect mapping table. S5. The multidimensional signal feature vector detected on site is retrieved and compared with the fault feature database, and intelligent diagnosis is performed based on the comparison results using a hybrid expert network to obtain the specific defect type and severity of the cable grounding system. S6. Assess the specific defect type and its severity to obtain early warning information for the cable grounding system.

[0006] In a preferred embodiment, injecting a low-frequency probe current signal into the cable grounding system and simultaneously acquiring the response voltage signal generated by the low-frequency probe current signal in the cable grounding system includes: The low-frequency detection current signal is input to the grounding lead of the cable grounding system; Receive the induced current in the loop of the grounding lead-down conductor; The voltage generated by the induced current on the cable grounding system is collected to obtain the response voltage signal of the cable grounding system.

[0007] In a preferred embodiment, the step of frequency domain decomposition of the response voltage signal and the low-frequency probe current signal to obtain the loop impedance of the cable grounding system, and complex decomposition of the loop impedance to obtain the equivalent pure resistance value of the loop impedance, includes: The response voltage signal and the low-frequency detection current signal are filtered and extracted to obtain the target frequency voltage signal and current signal of the cable grounding system. The target frequency voltage signal and current signal are transformed in the frequency domain to obtain the fundamental voltage phasor and fundamental current phasor of the cable grounding system. The loop impedance of the cable grounding system is obtained by performing a complex quotient operation on the fundamental voltage phasor and the fundamental current phasor. The resistance component of the loop impedance is separated to obtain the equivalent pure resistance value of the loop impedance.

[0008] In a preferred embodiment, the step of performing state characterization on the equivalent pure resistance value and the waveform of the response voltage signal to obtain the multidimensional signal characteristics of the cable grounding system includes: Time-domain statistical analysis was performed on the equivalent pure resistance value to obtain the resistance statistical characteristics of the cable grounding system; The amplitude ratio and distortion degree of the harmonic components in the response voltage signal are extracted to generate the harmonic distortion characteristics of the cable grounding system. The noise power spectrum of the response voltage signal is analyzed to obtain the noise energy characteristics of the cable grounding system. By identifying the chaotic characteristics of the time series of the response voltage signal, the nonlinear dynamic characteristics of the cable grounding system are obtained. The resistance statistical characteristics, harmonic distortion characteristics, noise energy characteristics, and nonlinear dynamic characteristics are integrated into the multidimensional signal characteristics of the cable grounding system.

[0009] In a preferred embodiment, the step of identifying the chaotic characteristics of the time series of the response voltage signal to obtain the nonlinear dynamic characteristics of the cable grounding system includes: The time series of the response voltage signal is reconstructed in state space to obtain the reconstructed phase space trajectory of the cable grounding system; Information entropy is extracted from the reconstructed phase space trajectory to obtain the entropy characteristics of the response voltage signal; Correlation dimension analysis is performed on the reconstructed phase space trajectory to obtain the dimension characteristics of the cable grounding system; By aggregating the entropy feature and the dimension feature into multidimensional features, the nonlinear dynamic features of the cable grounding system are obtained.

[0010] In a preferred embodiment, the step of performing pattern matching on the multidimensional signal features based on the typical physical defect types of the cable grounding system to obtain a feature defect mapping table of the cable grounding system, and constructing a fault feature database of the cable grounding system based on the feature defect mapping table, includes: Obtain cable grounding system samples corresponding to typical physical defect types in the cable grounding system; Supervised feature mapping is performed on the cable grounding system samples to establish the correlation between the multidimensional signal features and the typical physical defect types; Based on the aforementioned correlation, pattern mining is performed on the multidimensional signal features to obtain the core feature representation patterns of the typical physical defect types. By summarizing the pre-stored fault mechanism domain knowledge with the core feature manifestation patterns, a feature defect mapping table of the cable grounding system is obtained. Based on the aforementioned feature defect mapping table, the multidimensional signal features of the cable grounding system samples are vector space projected to obtain the standardized feature vectors of the cable grounding system samples. The standardized feature vector is associated and encapsulated with the defect type label, sample identifier, and acquisition environment parameters corresponding to the cable grounding system sample to obtain the fault feature database of the cable grounding system.

[0011] In a preferred embodiment, the step of performing rule-based induction of pre-stored fault mechanism domain knowledge and the core feature manifestation patterns to obtain the feature defect mapping table of the cable grounding system includes: Based on the pre-stored fault mechanism domain knowledge, rules are extracted from the typical physical defect types to obtain the rule knowledge base of the cable grounding system; The consistency between the rule knowledge base and the core feature representation pattern is verified to obtain the verification result of the core feature representation pattern. Based on the verification results, the core feature representation pattern is optimized using knowledge guidance to obtain a refined feature pattern of the core feature representation pattern. By binding the refined feature patterns with the typical physical defect types, a feature defect mapping table for the cable grounding system is obtained.

[0012] In a preferred embodiment, the step of comparing the multi-dimensional signal feature vector detected on-site with the fault feature database, and performing intelligent diagnosis based on a hybrid expert network to obtain the specific defect type and severity of the cable grounding system includes: Based on the standardized parameters of the fault feature database, the multidimensional signal feature vector detected on site is reduced to obtain the standardized on-site feature vector of the multidimensional signal feature vector. The standardized field feature vector is input into a hybrid expert network, and the probability distribution of the standardized field feature vector being assigned to the expert network for processing is calculated through the gating network of the hybrid expert network. Based on the probability distribution, the standardized field feature vectors are collaboratively inferred to obtain the diagnostic output of the cable grounding system; Based on the probability distribution, the diagnostic output is aggregated to obtain the specific defect type and severity of the cable grounding system.

[0013] In a preferred embodiment, the probability distribution is calculated using the following formula: ; In the formula, The standardized field feature vector Assigned to the The probability processed by an expert network For the first The weight vector associated with each expert network, For the first Bias terms associated with an expert network, These are preset hyperparameters. For the penalty term function, This is for performing normalization exponent operations on the vector within the parentheses.

[0014] In a preferred embodiment, the step of assessing the specific defect type and its severity to obtain early warning information for the cable grounding system includes: The warning level of the cable grounding system is determined based on the specific defect type and its severity. By performing maintenance knowledge mining on the warning level and the maintenance history of similar defects recorded in the fault feature database, a warning description and handling suggestions for the cable grounding system can be obtained. The warning level, the warning description, and the handling suggestions are encapsulated into the warning information for the cable grounding system.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention accurately injects low-frequency detection current signals into the cable grounding system and simultaneously collects response voltage signals. After frequency domain decomposition and complex deconstruction to separate the equivalent pure resistance value of the loop impedance, it integrates resistance statistics, harmonic distortion, noise energy and nonlinear dynamic characteristics to form multi-dimensional signal features, comprehensively capturing the electrical state information of the cable grounding system. This significantly improves the completeness and accuracy of feature extraction and provides reliable data support for defect diagnosis.

[0016] 2. This invention establishes a mapping relationship between multidimensional signal features and typical defects, constructs a standardized fault feature database, and combines intelligent collaborative inference with a hybrid expert network to achieve efficient identification of defect types and severity. Then, through situation analysis, it generates early warning information including warning levels and handling suggestions, which greatly improves the intelligence and efficiency of detection, provides scientific guidance for the timely operation and maintenance of cable grounding systems, and ensures the safe and stable operation of power systems. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a method for detecting live defects in a cable grounding system according to an embodiment of the present invention. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a method for detecting live defects in a cable grounding system. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for detecting live defects in a cable grounding system can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a method for detecting live defects in a cable grounding system according to an embodiment of the present invention. In this embodiment, the method for detecting live defects in a cable grounding system includes: S1. Inject the low-frequency detection current signal into the cable grounding system, and simultaneously collect the response voltage signal generated by the low-frequency detection current signal in the cable grounding system; In this embodiment of the invention, injecting a low-frequency detection current signal into the cable grounding system and simultaneously acquiring the response voltage signal generated by the low-frequency detection current signal in the cable grounding system includes: The low-frequency detection current signal is input to the grounding lead of the cable grounding system; Receive the induced current in the loop of the grounding lead-down conductor; The voltage generated by the induced current on the cable grounding system is collected to obtain the response voltage signal of the cable grounding system.

[0021] Select a low-frequency detection current signal generator with a frequency range of 30Hz-300Hz. The output of the signal generator is firmly connected to the designated connection point of the grounding down conductor of the cable grounding system through a shielded wire. The connection point needs to be mechanically polished to remove the surface oxide layer and dirt, ensuring that the contact resistance of the connection part does not exceed 0.1Ω. Then start the signal generator and output a low-frequency detection current signal according to the preset constant amplitude, so that the signal is continuously and stably injected into the grounding down conductor.

[0022] A high-precision current sensor with an accuracy class of 0.05 is used. The sensor's detection coil is tightly wound around the loop path of the grounding lead. The number of turns of the coil is fixed at 5, and the inner wall of the coil is in close contact with the surface of the lead without gaps. The signal output terminal of the sensor is connected to the input terminal of the data acquisition module through a dedicated signal line. Through the real-time sampling function of the data acquisition module, the induced current generated in the grounding lead loop due to the injection of low-frequency detection current signal is captured, ensuring that the waveform, amplitude and trend of the induced current are completely captured without signal distortion.

[0023] At the connection point between the grounding electrode and the grounding down conductor, and at three evenly distributed key nodes where the grounding electrode contacts the earth in the cable grounding system, voltage acquisition probes with an accuracy class of 0.02 are installed. The positive and negative electrodes of the probes are reliably in contact with the metal surface of the detection node by bolt fixing. Special conductive paste is applied to the electrode surface to control the contact impedance below 0.05Ω. The voltage acquisition probes are connected to a data recording device with real-time recording function through a double-shielded cable. The device collects the potential difference between each detection node in real time at a sampling frequency of 1kHz, which is generated by the induced current flowing through the cable grounding system. These potential difference data collected continuously in time sequence are integrated into a continuous electrical signal sequence, which finally forms the response voltage signal of the cable grounding system.

[0024] The beneficial effects are that by clarifying signal parameters, standardizing connection methods and testing equipment configuration, stable injection of low-frequency detection current signals, complete capture of induced current, and accurate acquisition of response voltage signals can be achieved. This ensures that the acquired raw signals have high fidelity and integrity, providing reliable data support for subsequent steps such as calculating the loop impedance of the cable grounding system and extracting multi-dimensional signal features, thus guaranteeing the stability of the entire defect detection process and the accuracy of the detection results.

[0025] S2. Perform frequency domain decomposition on the response voltage signal and the low-frequency detection current signal to obtain the loop impedance of the cable grounding system, and perform complex decomposition on the loop impedance to obtain the equivalent pure resistance value of the loop impedance. In this embodiment of the invention, the step of performing frequency domain decomposition on the response voltage signal and the low-frequency detection current signal to obtain the loop impedance of the cable grounding system, and performing complex decomposition on the loop impedance to obtain the equivalent pure resistance value of the loop impedance, includes: The response voltage signal and the low-frequency detection current signal are filtered and extracted to obtain the target frequency voltage signal and current signal of the cable grounding system. The target frequency voltage signal and current signal are transformed in the frequency domain to obtain the fundamental voltage phasor and fundamental current phasor of the cable grounding system. The loop impedance of the cable grounding system is obtained by performing a complex quotient operation on the fundamental voltage phasor and the fundamental current phasor. The resistance component of the loop impedance is separated to obtain the equivalent pure resistance value of the loop impedance.

[0026] A passive RC bandpass filter is selected. This filter consists of a fixed resistor and a capacitor connected in series in a specific manner, and then connected in parallel with another set of fixed resistors. The lower cutoff frequency is precisely set to 45Hz and the upper cutoff frequency is precisely set to 55Hz, which perfectly matches the standard frequency of the injected low-frequency detection current signal of 50Hz. A double-shielded conductor with a diameter of 1.5mm² is selected, and its outer shield is grounded to isolate electromagnetic interference. The collected response voltage signal is connected to the voltage input terminal of the filter through this conductor. When wiring, ensure that the conductor core is completely flush with the input terminal and tighten it with screws. The low-frequency detection current signal is simultaneously connected to the current input terminal of the filter through the same double-shielded conductor. When the signal flows through the filter, the attenuation of noise signals with frequencies below 45Hz is not less than 40dB, and the attenuation of noise signals with frequencies above 55Hz is also not less than 40dB, while the pass rate of the target frequency signal of 50Hz is not less than 98%. Finally, the target frequency voltage signal and the target frequency current signal of the cable grounding system are stably obtained from the output terminal of the filter.

[0027] Frequency domain transformation is performed using an analog Fourier transform circuit. Internally, this circuit consists of a series-connected frequency selection network composed of capacitors and inductors. Each frequency selection branch corresponds to a fixed frequency signal selection. The obtained target frequency voltage signal is connected to the circuit's voltage signal input port via a BNC interface, and the target frequency current signal is connected to the current signal input port via the same BNC interface. When connecting the interfaces, ensure the latches are fully locked to prevent signal loosening. After the time-domain signal is input, it is filtered by the frequency selection network. Signals other than 50Hz are blocked by the branches, allowing only the 50Hz fundamental wave signal to pass. The detection unit in the circuit accurately acquires the amplitude of the fundamental wave signal. The phase detection unit obtains phase data by comparing the time difference between the signal and the reference oscillation signal. The amplitude and phase data of the fundamental voltage are then integrated in the form of "amplitude-phase" pairs to form the fundamental voltage phasor. Similarly, the amplitude and phase data of the fundamental current are integrated to form the fundamental current phasor.

[0028] A dedicated complex number operation circuit is constructed to perform complex number quotient operations. This circuit includes an amplitude division operation unit and a phase subtraction operation unit, both composed of high-precision analog operational amplifiers. The amplitude and phase data of the fundamental voltage phasor are input to the circuit via the data bus in a timing sequence of "amplitude first, phase second". Subsequently, the amplitude and phase data of the fundamental current phasor are input. After receiving the two sets of amplitude data, the amplitude division operation unit divides the fundamental voltage amplitude by the fundamental current amplitude according to a fixed operation logic to obtain the amplitude ratio. After receiving the two sets of phase data, the phase subtraction operation unit subtracts the fundamental voltage phase from the fundamental current phase to obtain the phase difference. The amplitude ratio is used as the modulus of the complex number, and the phase difference is used as the argument of the complex number. The two are correlated to form the loop impedance of the cable grounding system in complex form. The impedance data is stored in the circuit's built-in buffer in the format of "modulus-argument".

[0029] The loop impedance is processed using an impedance component separation module. This module is equipped with a standard BNC interface, through which the complex loop impedance signal is input. The phase discrimination unit inside the module distinguishes between the real and imaginary parts by analyzing the phase characteristics of the signal. When the signal phase is consistent with the reference phase, it is determined to be the real part; when the signal phase deviates from the reference phase by 90°, it is determined to be the imaginary part. The module's built-in high-precision acquisition unit acquires the amplitude of the real part signal with a resolution of 0.001Ω. Then, the acquired voltage signal is linearly converted into a resistance value through a preset calibration curve. The calibration curve is plotted based on the measured data of a standard resistance sample. There is no additional signal distortion during the conversion process, and the final output value is the equivalent pure resistance value of the loop impedance.

[0030] The beneficial effects are as follows: by clarifying the circuit structure, shielding wire specifications, and clutter attenuation standards of the passive RC bandpass filter, the high purity and high throughput of the target frequency signal are ensured; the frequency selection network structure, data acquisition unit, and phasor integration form of the analog Fourier transform circuit are refined to ensure the integrity and accuracy of the fundamental voltage phasor and fundamental current phasor; the internal unit structure, data input timing, and impedance storage format of the complex number operation circuit are clarified to achieve accurate calculation of loop impedance; and the interface type, phase discrimination logic, and linear conversion calibration method of the impedance component separation module are standardized to ensure the reliability of the equivalent pure resistance value. These refined designs together provide high-precision and high-stability basic electrical parameters for subsequent multi-dimensional signal feature extraction, ensuring the continuity of the cable grounding system defect detection process and the accuracy of the detection results from the source.

[0031] S3. The equivalent pure resistance value and the waveform of the response voltage signal are characterized by state to obtain the multi-dimensional signal characteristics of the cable grounding system; In this embodiment of the invention, the step of performing state characterization on the equivalent pure resistance value and the waveform of the response voltage signal to obtain the multidimensional signal characteristics of the cable grounding system includes: Time-domain statistical analysis was performed on the equivalent pure resistance value to obtain the resistance statistical characteristics of the cable grounding system; The amplitude ratio and distortion degree of the harmonic components in the response voltage signal are extracted to generate the harmonic distortion characteristics of the cable grounding system. The noise power spectrum of the response voltage signal is analyzed to obtain the noise energy characteristics of the cable grounding system. By identifying the chaotic characteristics of the time series of the response voltage signal, the nonlinear dynamic characteristics of the cable grounding system are obtained. The resistance statistical characteristics, harmonic distortion characteristics, noise energy characteristics, and nonlinear dynamic characteristics are integrated into the multidimensional signal characteristics of the cable grounding system.

[0032] The chaotic characteristic identification of the time series of the response voltage signal to obtain the nonlinear dynamic characteristics of the cable grounding system includes: The time series of the response voltage signal is reconstructed in state space to obtain the reconstructed phase space trajectory of the cable grounding system; Information entropy is extracted from the reconstructed phase space trajectory to obtain the entropy characteristics of the response voltage signal; Correlation dimension analysis is performed on the reconstructed phase space trajectory to obtain the dimension characteristics of the cable grounding system; By aggregating the entropy feature and the dimension feature into multidimensional features, the nonlinear dynamic features of the cable grounding system are obtained.

[0033] A continuous data segment of equivalent pure resistance values ​​was selected, with the acquisition time strictly set to 10 seconds and the sampling interval consistent with the response voltage signal at 1 millisecond. Therefore, this data segment contains a total of 10,000 equivalent pure resistance value data points. A high-precision data logger with a sampling accuracy of 0.001Ω was used to extract all equivalent pure resistance value data within this time segment. The data is stored in binary format, with each data point occupying 32 bits of storage space. The recorder's built-in hardware processing unit calculates the arithmetic mean, maximum value, minimum value, and variance of these data. The arithmetic mean is calculated by summing all 10,000 data points one by one and then dividing the sum by the total number of data points, 10,000. The maximum value is selected by comparing all data points sequentially using a hardware comparator. The minimum value is selected by comparing all data points in reverse using the same hardware comparator. The variance is calculated by first calculating the difference between each data point and the arithmetic mean, squaring each difference, summing all the squared results, and finally dividing the sum by the total number of data points, 10,000. These four statistical results are combined in the order of "arithmetic mean - maximum value - minimum value - variance" to form the resistance statistical characteristics of the cable grounding system.

[0034] A multi-channel passive filter circuit is used to separate harmonics in the response voltage signal. This circuit consists of four independent filter branches: a fundamental channel and dedicated channels for the 3rd, 5th, and 7th harmonics. The center frequency of the fundamental channel is set to 50Hz, the 3rd harmonic channel to 150Hz, the 5th harmonic channel to 250Hz, and the 7th harmonic channel to 350Hz. Each channel achieves frequency filtering through precisely matched resistor and capacitor parameters. The response voltage signal is connected to the common input terminal of the filter circuit via a shielded wire. When the signal flows through the circuit, only the harmonic components corresponding to the center frequency are allowed to pass through each channel, with other frequency components attenuated by at least 40dB. A 0.01mV-level high-precision voltage acquisition module is connected to the output terminals of each channel. Before acquisition, the module undergoes three-point calibration using a standard voltage source to ensure that the amplitude measurement error does not exceed ±0.1%. This module is used to measure the amplitude of the fundamental channel output signal and the amplitudes of the 3rd, 5th, and 7th harmonic channel output signals. Each harmonic amplitude is input into an analog divider and divided by the fundamental amplitude to obtain the ratio of each harmonic amplitude to the fundamental amplitude. At the same time, the sum of squares of all harmonic amplitudes is calculated by an adder, and then the sum of squares is input into another analog divider and divided by the square of the fundamental amplitude. The ratio is defined as the total harmonic distortion rate. The ratios of all amplitudes and the total harmonic distortion rate together constitute the harmonic distortion characteristics of the cable grounding system.

[0035] The response voltage signal waveform is analyzed using an analog spectrum analysis module. This module internally consists of a resonant network composed of multiple inductors and capacitors with different parameters. Each resonant unit corresponds to a specific frequency for signal resonance detection, with the noise identification frequency range set to below 20Hz and above 200Hz. The response voltage signal is connected to the module input via a BNC interface. During connection, the latches are ensured to be fully locked to prevent signal attenuation. After signal input, the resonant network decomposes the signal. Noise signals below 20Hz and above 200Hz are captured by the resonant units. The module's built-in power detection unit measures the voltage across a 1kΩ high-precision standard resistor and calculates the power value of each noise frequency component using the power calculation formula (power = voltage squared / resistance). An integrator circuit then accumulates and integrates the power values ​​of all noise components within this frequency range. The integration time is strictly consistent with the signal acquisition duration at 10 seconds, and the time constant of the integrator circuit is set to 1 second to ensure a stable and distortion-free integration process. The final sum of integrated power values ​​represents the noise energy characteristics of the cable grounding system.

[0036] The time delay τ = 5 sampling points are determined based on the autocorrelation function method. Specifically, the autocorrelation function of the response voltage signal time series is calculated. When the autocorrelation function value drops to 1 / e of the initial value, the corresponding number of sampling points is 5. The spurious nearest neighbor method is used to determine the embedding dimension m = 3. By gradually increasing the embedding dimension, the proportion of spurious nearest neighbors under each dimension is calculated. When the proportion is less than 5%, the embedding dimension is determined to be 3. The response voltage signal time series is reconstructed in the format "x(t), x(t+τ), x(t+2τ)", where t is the starting sampling point of the time series, ranging from 1 to N-2τ (N is the total number of sampling points of the response voltage signal). All sampling points meeting the conditions are sequentially reconstructed to form a set of three-dimensional data points. The coordinate format of each three-dimensional data point is (x(t), x(t+5), x(t+10)). Using a data visualization module, these three-dimensional data points are sequentially connected in a three-dimensional coordinate system in ascending order of t value, forming a continuous spatial curve. This curve is the reconstructed phase space trajectory of the cable grounding system.

[0037] The three-dimensional space containing the reconstructed phase space is uniformly divided. First, the maximum and minimum values ​​of the x, y, and z dimensions are calculated to obtain the span of each dimension. Each span is then divided into 10 equal parts, forming 10 × 10 × 10 = 1000 small cubes of equal volume. The side length of each small cube is one-tenth of the corresponding dimensional span. A hardware traversal unit scans all points on the reconstructed phase space trajectory along a preset path. Each point passed triggers a counter in the corresponding small cube, incrementing by 1, and the number of trajectory points contained in each small cube is counted. The number of trajectory points in each small cube is input into a division unit and divided by the total number of trajectory points to obtain the ratio of the number of points in each small cube to the total number of trajectory points. This ratio is stored as a probability value. A logarithmic unit takes the natural logarithm of each probability value, and the logarithmic result is multiplied by the corresponding probability value. All multiplication results are input into an accumulator for summation. Finally, an inverter takes the negative value of the summation, and the resulting value is the entropy characteristic of the response voltage signal.

[0038] 1000 uniformly distributed sampling points are selected on the reconstructed phase space trajectory. The sampling point selection method is to calculate the interval K = round(M / 1000) based on the total number of points M on the trajectory. Starting from the first point, a sampling point is selected every K points to ensure that the sampling points cover the entire trajectory. The Euclidean distance between any two sampling points is calculated through the three-dimensional distance calculation module. In the calculation, the differences between the x, y, and z coordinates of the two points are first calculated separately. Each difference is squared, and the sum of the three squares is taken as the square root to obtain the Euclidean distance value. The distance scale range is set to 0.01Ω to 10Ω, and 1000 distance scales are selected with a step size of 0.01Ω. For each distance scale, the counting module counts the number of point pairs whose distance is less than that scale. This number is input into the division unit and divided by the total number of point pairs to obtain the ratio. Using the logarithm of the distance scale as the x-axis and the logarithm of the ratio as the y-axis, a curve is plotted using the coordinate plotting module. The correlation coefficient calculation module calculates the correlation coefficient of each line segment on the curve. Line segments with a correlation coefficient greater than 0.95 are selected, and the coordinates of these line segments are input into the least squares fitting module. The module calculates the slope value that minimizes the sum of squared deviations between the fitted line and the points on the line segment. This slope value is the correlation dimension, which serves as the dimensional characteristic of the cable grounding system.

[0039] The specific values ​​of the entropy and dimension features are combined in the order of "entropy-dimension" to form a two-dimensional feature vector. The first element of the vector is the entropy feature value, and the second element is the dimension feature value. This two-dimensional vector is input into a data integration module composed of a digital signal processor. The module performs format standardization processing on the values, first converting them to 32-bit floating-point format, and then rounding them to four decimal places. During the processing, a verification unit ensures that there is no loss of precision. The standardized two-dimensional vector is the nonlinear dynamic characteristic of the cable grounding system.

[0040] The arithmetic mean, maximum, minimum, and variance of the resistance statistical characteristics, the ratio of the amplitudes of the 3rd, 5th, and 7th harmonics to the fundamental amplitude and the total harmonic distortion rate of the harmonic distortion characteristics, the total noise power of the noise energy characteristics, and the two-dimensional feature vector of the nonlinear dynamic characteristics are arranged in the above order. Each feature parameter is stored in 32-bit floating-point format to form a comprehensive feature vector containing 11 consecutively stored feature parameters. This feature vector is the multidimensional signal feature of the cable grounding system.

[0041] The beneficial effects are that by clarifying the equipment specifications, calibration methods, calculation logic, and data processing details in each feature extraction process, the entire process from signal acquisition to feature integration is refined, ensuring that the extraction of resistance statistical features, harmonic distortion features, noise energy features, and nonlinear dynamic features has extremely high accuracy and consistency. Each feature comprehensively describes the actual state of the cable grounding system from different dimensions such as electrical parameter stability, signal distortion degree, interference energy magnitude, and nonlinear operating state. The integrated multidimensional signal features not only ensure the integrity of the information but also have clear physical meaning and distinguishability, providing highly reliable and highly available feature support for subsequent pattern matching and intelligent diagnosis, and ensuring the accuracy and efficiency of cable grounding system defect detection from the core link.

[0042] S4. Based on the typical physical defect types of the cable grounding system, perform pattern matching on the multidimensional signal features to obtain the feature defect mapping table of the cable grounding system, and construct the fault feature database of the cable grounding system according to the feature defect mapping table. In this embodiment of the invention, the step of performing pattern matching on the multidimensional signal features based on the typical physical defect types of the cable grounding system to obtain a feature defect mapping table of the cable grounding system, and constructing a fault feature database of the cable grounding system based on the feature defect mapping table, includes: Obtain cable grounding system samples corresponding to typical physical defect types in the cable grounding system; Supervised feature mapping is performed on the cable grounding system samples to establish the correlation between the multidimensional signal features and the typical physical defect types; Based on the aforementioned correlation, pattern mining is performed on the multidimensional signal features to obtain the core feature representation patterns of the typical physical defect types. By summarizing the pre-stored fault mechanism domain knowledge with the core feature manifestation patterns, a feature defect mapping table of the cable grounding system is obtained. Based on the aforementioned feature defect mapping table, the multidimensional signal features of the cable grounding system samples are vector space projected to obtain the standardized feature vectors of the cable grounding system samples. The standardized feature vector is associated and encapsulated with the defect type label, sample identifier, and acquisition environment parameters corresponding to the cable grounding system sample to obtain the fault feature database of the cable grounding system.

[0043] The step of summarizing the pre-stored fault mechanism domain knowledge with the core feature manifestation patterns to obtain the feature defect mapping table of the cable grounding system includes: Based on the pre-stored fault mechanism domain knowledge, rules are extracted from the typical physical defect types to obtain the rule knowledge base of the cable grounding system; The consistency between the rule knowledge base and the core feature representation pattern is verified to obtain the verification result of the core feature representation pattern. Based on the verification results, the core feature representation pattern is optimized using knowledge guidance to obtain a refined feature pattern of the core feature representation pattern. By binding the refined feature patterns with the typical physical defect types, a feature defect mapping table for the cable grounding system is obtained.

[0044] Common physical defect types in cable grounding systems were selected, including four categories: broken grounding down conductor, corrosion of grounding body, excessive grounding resistance, and loose connection. For each defect type, cable grounding systems with an operating age of 1-10 years and installed in outdoor open-air, underground tunnel, and humid environments were selected as samples. At least 50 valid samples were collected for each defect type. For each sample, the specific location of the defect, the cause of the defect, and the operating condition information were clearly recorded. At the same time, the multi-dimensional signal characteristics of each sample were fully collected according to the aforementioned detection method to ensure that the defect status of the sample corresponds one-to-one with the signal characteristics.

[0045] Each cable grounding system sample is labeled with a unique defect type tag in the format of "defect type-serial number". A hardware-based feature association unit is used to bind and store the multidimensional signal features of each sample with the corresponding defect type tag. The value of each feature parameter is associated with the tag in a fixed way, forming an association pair of "multidimensional signal feature set-defect type tag" to clarify the correspondence between the two.

[0046] For all sample pairs corresponding to each type of typical physical defect, the multidimensional signal feature parameters under the same defect type are analyzed one by one through the feature statistics unit. The numerical distribution range, frequency of occurrence and central tendency of each feature parameter are statistically analyzed. Abnormal values ​​with a dispersion of more than 5% in each feature parameter are removed. The feature parameters and numerical ranges common to all samples under the defect type are selected. These common feature parameters are integrated in their original order to form the core feature performance pattern that corresponds only to the defect type. Each pattern clearly reflects the common features of the defect type.

[0047] The pre-stored fault mechanism knowledge includes the physical action mechanism and electrical response law of various defects, such as "a broken grounding lead will lead to incomplete circuit conduction, a sudden change in the equivalent pure resistance value, and an increase in harmonic distortion rate" and "corrosion of the grounding body will reduce the conductive cross-sectional area, increase the variance of the resistance statistical characteristics, and increase the noise energy." Based on this knowledge, clear judgment rules are extracted for each typical physical defect type, such as "equivalent pure resistance value exceeds 10Ω and total harmonic distortion rate is greater than 5% → corresponding grounding lead breakage" and "annual increase in equivalent pure resistance value exceeds 2Ω and noise energy is greater than 0.5W → corresponding grounding body corrosion." All the extracted judgment rules are classified and organized according to defect type to form a structured cable grounding system rule knowledge base.

[0048] The judgment rules for each defect type in the rule knowledge base are compared parameter by parameter with the corresponding core feature performance pattern. It is determined whether the numerical range of each feature parameter in the core feature performance pattern falls completely within the rule threshold. If more than 90% of the feature parameters in a core feature performance pattern meet the threshold requirements of the corresponding rule, the verification is deemed to have passed. If the compliance rate is less than 70%, the verification is deemed to have failed. The non-compliant feature parameters and the range of difference are recorded in detail, and finally, a clear verification result is formed for each core feature performance pattern.

[0049] For core feature patterns that pass verification, all feature parameters and ranges are directly retained. For core feature patterns that fail verification, the feature parameter ranges are adjusted based on the differences in the verification results and knowledge of the fault mechanism domain. For example, if the noise energy range in the core feature is inconsistent with the rules, the mechanism of "grounding corrosion leading to increased noise energy" is referenced, and the noise energy data of the defect type sample is re-statistically analyzed. Outliers caused by interference from the acquisition environment are removed. The adjusted feature parameter ranges must 100% conform to the corresponding judgment rules in the rule knowledge base to form a refined feature pattern.

[0050] Each typical physical defect type is fixedly bound to its name, unique identifier, and corresponding refined feature pattern. The binding process is implemented through a hardware storage unit. Each binding relationship forms an independent entry, which contains three core fields: "defect type name - defect identifier - refined feature pattern". All entries are sorted and organized according to defect type to form a structured and searchable characteristic defect mapping table for cable grounding systems.

[0051] An 11-dimensional standard feature space corresponding to the multi-dimensional signal features is constructed, with each dimension corresponding to a feature parameter. Based on the parameter range of each refined feature mode in the feature defect mapping table, the multi-dimensional signal feature parameters of each cable grounding system sample are linearly projected and transformed. During the transformation, the minimum value of the refined feature mode parameter is taken as the starting point of the dimension and the maximum value is taken as the ending point. The sample feature parameter values ​​are mapped to the standard interval [0,1]. The 11 standard values ​​after transformation are arranged in the original feature order to form the standardized feature vector of the sample.

[0052] The standardized feature vector of each sample is associated and integrated with the corresponding defect type label, unique sample identifier, and collection environment parameters through the data encapsulation unit. Each integrated data group is stored in a fixed format. All data groups are classified and stored in the distributed storage module according to defect type. The storage module supports retrieval by feature parameters, defect type, and environmental parameters, forming a complete fault feature database of the cable grounding system.

[0053] The beneficial effects are as follows: by clarifying the specific standards for sample collection, the binding method of feature mapping, the statistical logic of pattern mining, the basis for rule extraction, and the verification and optimization process, the accuracy and relevance of the feature defect mapping table are ensured. The generation of standardized feature vectors enables the data to have unified comparability. The associated encapsulation of the fault feature database realizes the complete storage of feature, defect, and environmental information, providing a basic data support with standardized structure, reliable data, and strong reproducibility for subsequent on-site detection feature retrieval and comparison, and ensuring the accuracy and efficiency of subsequent intelligent diagnosis.

[0054] S5. The multidimensional signal feature vector detected on site is retrieved and compared with the fault feature database, and intelligent diagnosis is performed based on the comparison results using a hybrid expert network to obtain the specific defect type and severity of the cable grounding system. In this embodiment of the invention, the step of comparing the multi-dimensional signal feature vector detected on-site with the fault feature database, and performing intelligent diagnosis based on a hybrid expert network to obtain the specific defect type and severity of the cable grounding system includes: Based on the standardized parameters of the fault feature database, the multidimensional signal feature vector detected on site is reduced to obtain the standardized on-site feature vector of the multidimensional signal feature vector. The standardized field feature vector is input into a hybrid expert network, and the probability distribution of the standardized field feature vector being assigned to the expert network for processing is calculated through the gating network of the hybrid expert network. Based on the probability distribution, the standardized field feature vectors are collaboratively inferred to obtain the diagnostic output of the cable grounding system; Based on the probability distribution, the diagnostic output is aggregated to obtain the specific defect type and severity of the cable grounding system.

[0055] The formula for calculating the probability distribution is as follows: ; In the formula, The standardized field feature vector Assigned to the The probability processed by an expert network For the first The weight vector associated with each expert network, For the first Bias terms associated with an expert network, These are preset hyperparameters. For the penalty term function, This is for performing normalization exponent operations on the vector within the parentheses.

[0056] The standardized range parameters corresponding to each feature parameter are extracted from the fault feature database, namely the minimum and maximum values ​​of each of the 11 feature parameters. These parameters are stored in a dedicated data storage unit, and each parameter corresponds one-to-one with its position in the feature vector. A multidimensional signal feature vector from the field detection is obtained, containing 4 parameters for resistance statistics, 4 parameters for harmonic distortion, 1 parameter for noise energy, and 2 parameters for nonlinear dynamics, for a total of 11 parameters. For each field feature parameter, a reduction calculation is performed using a hardware computing module. The calculation method is to subtract the minimum value of the corresponding parameter in the database from the field parameter value, and then divide by the difference between the maximum and minimum values ​​of the corresponding parameter in the database. If the calculation result is less than 0, it is taken as 0; if it is greater than 1, it is taken as 1, ensuring that each reduced parameter value falls within the [0,1] interval. These 11 reduced parameters are arranged in their original order to form a standardized field feature vector of the multidimensional signal feature vector.

[0057] The hybrid expert network comprises four expert networks, each corresponding to one of the four typical physical defects in a cable grounding system. The gated network consists of hardware-based weighting units, bias superposition units, and normalization units. The 11 parameters of the standardized field feature vector are input into the gated network via a data bus. In the pre-stored weight matrix accessed by the weighting unit, each feature parameter corresponds to a unique weight for one of the four expert networks; this weight is the weight of the first expert network. The weight vector associated with each expert network It is determined through supervised training based on standardized feature vectors and corresponding defect type labels stored in the fault feature database. During the training process, the standardized feature vectors are input into the corresponding expert network, and the consistency between the network output and the defect type label is compared. The value of each element in the vector is continuously adjusted until the feature recognition accuracy of the expert network for the same type of defect sample reaches a preset standard of more than 95%. Finally, it is fixed in the weight matrix.

[0058] The bias stacking unit adds a fixed bias term to each preliminary calculation result, i.e., the first... Bias terms associated with each expert network Its weight vector During supervised training, the specific details are determined simultaneously for the first... Each expert network corresponds to all samples of a defect type. It calculates the overall offset between the standardized feature vector and the ideal feature vector of the defect. The offset is preset to offset the influence of feature offset and ensure the accuracy of feature matching.

[0059] Preset hyperparameters Based on the feature distribution density of samples in the fault feature database, the proportion of various defect samples, and the training convergence speed of the expert network, the value that maximizes the overall diagnostic accuracy is determined by testing the impact of different fixed values ​​on the accuracy of the intelligent diagnosis results. Once determined, it remains constant throughout the entire testing process.

[0060] Penalty term function Based on the Constructing refined feature patterns for each defect type within an expert network, and calculating standardized field feature vectors. The deviation of each feature parameter from the corresponding parameter in the refined feature pattern is calculated by summing the differences between each parameter and the range of parameters in the refined feature pattern. The larger the sum of the differences, the stronger the penalty term function. The larger the output value, the smaller it is, which is used to suppress defects that differ significantly from the defect characteristics corresponding to the expert network. Assign it to the network.

[0061] The weighted computation unit multiplies each feature parameter by the weight of the corresponding expert network and then sums the results to obtain four preliminary computation results. The bias stacking unit adds a fixed bias term of the corresponding expert network to each preliminary computation result to obtain four biased computation results. This process corresponds to the formula " The operation logic is then followed by preset hyperparameters. With penalty term function The product of , that is, completing all the operations within the parentheses in the formula, yields the product of , that is, the product of , i.e., completing all the operations within the parentheses in the formula, and obtaining the product of . The original assigned scores of the expert network.

[0062] The normalization unit performs a normalized exponent calculation on the raw allocation scores of all expert networks. This involves calculating the natural exponent value for each raw score separately, and then dividing that natural exponent value by the sum of the natural exponent values ​​of all raw scores from all expert networks. This process corresponds to the formula... The four values ​​obtained from the calculation are the standardized field feature vector. The probability assigned to the corresponding expert network for processing The sum of the four probabilities equals 1, forming a probability distribution. This formula integrates factors such as feature matching degree, offset correction, and deviation penalty to achieve [the desired outcome]. Reasonable allocation ensures that the most suitable expert network participates in subsequent diagnosis.

[0063] Each expert network has a built-in refined feature pattern for the corresponding defect type, containing a standard range of 11 feature parameters. When the standardized field feature vector is input into the expert network, the feature comparison unit of the expert network matches the field feature parameters with the built-in standard range one by one. If the field parameter falls within the standard range, it is considered a match and recorded as a valid match; otherwise, it is an invalid match. The number of valid matches for each expert network is counted, and the number of valid matches is divided by the total number of feature parameters (11) to obtain the feature matching degree of that expert network. The feature matching degree is multiplied by the corresponding probability assigned by the gating network to obtain the diagnostic confidence of that expert network for the corresponding defect type. Each expert network outputs a single diagnostic result containing the corresponding defect type name and diagnostic confidence. The single diagnostic results of all expert networks are summarized to form the diagnostic output of the cable grounding system.

[0064] The diagnostic confidence scores of all individual diagnostic results are extracted from the diagnostic output and sorted in descending order. The defect type corresponding to the diagnostic result with the highest confidence score is selected as the candidate defect type. The diagnostic confidence score corresponding to this candidate defect type is extracted, and three confidence score thresholds are set: 0.8 and above is the high confidence interval, 0.5 to 0.79 is the medium confidence interval, and below 0.5 is the low confidence interval. If the diagnostic confidence score of a candidate defect type is in the high confidence interval, the severity is determined to be "severe"; in the medium confidence interval, it is determined to be "moderate"; and in the low confidence interval, it is determined to be "minor". At the same time, the number of supporting diagnostic outputs for this candidate defect type is counted, that is, the number of other expert network outputs where the confidence score of this defect type is not lower than 0.3. If the number of supporting outputs is not less than 2, the candidate defect type is confirmed to be valid. Finally, the valid candidate defect types and their corresponding severity scores are integrated to obtain the specific defect type and severity of the cable grounding system.

[0065] The beneficial effects are as follows: by clarifying the computational logic and boundary processing rules of data specification, the standardized field feature vectors are ensured to have uniform comparability; the hardware design of the gated network, combined with the synergistic application of weight vectors, preset bias terms, hyperparameters, and penalty term functions trained and solidified based on the fault feature database, combined with the rigorous computational logic of the formula, enables accurate calculation and reasonable allocation of probability distribution; the combination of feature matching and confidence calculation of the expert network, and collaborative inference to form a comprehensive diagnostic output; and the data aggregation method based on confidence ranking, threshold determination, and support for quantitative verification, ensures the accuracy and reliability of the determination of specific defect types and severity, providing a solid basis for the precise handling of defects in cable grounding systems.

[0066] S6. Assess the specific defect type and its severity to obtain early warning information for the cable grounding system.

[0067] In this embodiment of the invention, the step of assessing the specific defect type and its severity to obtain early warning information for the cable grounding system includes: The warning level of the cable grounding system is determined based on the specific defect type and its severity. By performing maintenance knowledge mining on the warning level and the maintenance history of similar defects recorded in the fault feature database, a warning description and handling suggestions for the cable grounding system can be obtained. The warning level, the warning description, and the handling suggestions are encapsulated into the warning information for the cable grounding system.

[0068] Based on the specific defect types of the cable grounding system—grounding lead breakage, grounding body corrosion, excessive grounding resistance, and loose connection—and their severity levels (severe, moderate, and minor), a three-level warning system is pre-defined. A "severe" severity level is directly classified as a Level 1 warning, requiring the initiation of a response process within 24 hours; a "moderate" severity level is classified as a Level 2 warning, requiring the initiation of a response process within 72 hours; and a "minor" severity level is classified as a Level 3 warning, requiring the initiation of a response process within 168 hours. The warning level determination is implemented through a hardware logic judgment unit. This unit stores the aforementioned judgment criteria, and upon inputting the specific defect type and severity level, automatically matches the corresponding warning level and outputs it.

[0069] The fault characteristic database stores maintenance history records for various defects. Each record includes the defect type, warning level, maintenance implementation time, maintenance operation procedure, tools and materials used, post-treatment operation monitoring data (continuous monitoring for 3 months), and a stable operating time of ≥3 months as an evaluation of the treatment effect for effective treatment. Using a hardware search module, based on the specific defect type and the determined warning level, all maintenance records of similar defects from the past 5 years are filtered from the database. Effective treatment identifiers for each record are statistically analyzed, and maintenance operation procedures with an effective treatment rate of over 90% are extracted. The core steps of these procedures are summarized; for example, the first-level warning maintenance procedure for a broken grounding lead includes: power off and voltage testing, removal of the broken lead, replacement with a new lead of the same specification, bolt tightening, and continuity testing as treatment suggestions. The warning description combines the defect type, severity, and typical consequences of not promptly addressing similar defects in historical records, such as a broken grounding lead leading to grounding circuit interruption and the risk of equipment insulation breakdown. It is generated in a fixed format: "Defect Type: XX, Severity: XX, Potential Risk: XX, Treatment Time Limit Requirement: XX hours," ensuring the description information is complete and unambiguous.

[0070] The data encapsulation module integrates the warning level, warning description, and handling suggestions. The encapsulation adopts a structured data format, and the data contains three fixed fields: the "Warning Level" field has a value of Level 1 / Level 2 / Level 3 warning; the "Warning Description" field has a value of the specific description text generated above; and the "Handling Suggestions" field has a value of the core steps arranged in the order of operation, with steps separated by semicolons. During the encapsulation process, the integrity of each field content is verified by a verification unit to ensure that there are no missing or incorrect data. The structured data is then stored in a dedicated storage chip and synchronized to the operation and maintenance management terminal via an RS485 communication interface. After receiving the data, the terminal can directly display the complete warning information for operation and maintenance personnel to view and execute.

[0071] The beneficial effects include ensuring timely early warning response by clearly defining the criteria for judging early warning levels and the time limits for handling them; knowledge mining based on maintenance history verified in practice in the fault feature database makes the early warning descriptions accurately reflect the risks and the handling suggestions highly practical, avoiding blind maintenance; and the standardized packaging format ensures the transmission and display of early warning information, providing maintenance personnel with clear and specific action guidelines, effectively shortening the handling cycle, improving maintenance quality, and reducing the power safety risks caused by defects in the cable grounding system.

[0072] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0073] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for live detection of defects in a cable grounding system, characterized in that, The method includes: S1. Inject the low-frequency detection current signal into the cable grounding system, and simultaneously collect the response voltage signal generated by the low-frequency detection current signal in the cable grounding system; S2. Perform frequency domain decomposition on the response voltage signal and the low-frequency detection current signal to obtain the loop impedance of the cable grounding system, and perform complex decomposition on the loop impedance to obtain the equivalent pure resistance value of the loop impedance. S3. The equivalent pure resistance value and the waveform of the response voltage signal are characterized by state to obtain the multi-dimensional signal characteristics of the cable grounding system; S4. Based on the typical physical defect types of the cable grounding system, perform pattern matching on the multidimensional signal features to obtain the feature defect mapping table of the cable grounding system, and construct the fault feature database of the cable grounding system according to the feature defect mapping table. S5. The multidimensional signal feature vector detected on site is retrieved and compared with the fault feature database, and intelligent diagnosis is performed based on the comparison results using a hybrid expert network to obtain the specific defect type and severity of the cable grounding system. S6. Assess the specific defect type and its severity to obtain early warning information for the cable grounding system.

2. The method for detecting live defects in a cable grounding system as described in claim 1, characterized in that, The step of injecting a low-frequency detection current signal into the cable grounding system and simultaneously acquiring the response voltage signal generated by the low-frequency detection current signal in the cable grounding system includes: The low-frequency detection current signal is input to the grounding lead of the cable grounding system; Receive the induced current in the loop of the grounding lead-down conductor; The voltage generated by the induced current on the cable grounding system is collected to obtain the response voltage signal of the cable grounding system.

3. The method for detecting live defects in a cable grounding system as described in claim 1, characterized in that, The step of performing frequency domain decomposition on the response voltage signal and the low-frequency detection current signal to obtain the loop impedance of the cable grounding system, and performing complex decomposition on the loop impedance to obtain the equivalent pure resistance value of the loop impedance, includes: The response voltage signal and the low-frequency detection current signal are filtered and extracted to obtain the target frequency voltage signal and current signal of the cable grounding system. The target frequency voltage signal and current signal are transformed in the frequency domain to obtain the fundamental voltage phasor and fundamental current phasor of the cable grounding system. The loop impedance of the cable grounding system is obtained by performing a complex quotient operation on the fundamental voltage phasor and the fundamental current phasor. The resistance component of the loop impedance is separated to obtain the equivalent pure resistance value of the loop impedance.

4. The method for detecting live defects in a cable grounding system as described in claim 1, characterized in that, The step of performing state characterization on the equivalent pure resistance value and the waveform of the response voltage signal to obtain the multidimensional signal characteristics of the cable grounding system includes: Time-domain statistical analysis was performed on the equivalent pure resistance value to obtain the resistance statistical characteristics of the cable grounding system; The amplitude ratio and distortion degree of the harmonic components in the response voltage signal are extracted to generate the harmonic distortion characteristics of the cable grounding system. The noise power spectrum of the response voltage signal is analyzed to obtain the noise energy characteristics of the cable grounding system. By identifying the chaotic characteristics of the time series of the response voltage signal, the nonlinear dynamic characteristics of the cable grounding system are obtained. The resistance statistical characteristics, harmonic distortion characteristics, noise energy characteristics, and nonlinear dynamic characteristics are integrated into the multidimensional signal characteristics of the cable grounding system.

5. The method for detecting live defects in a cable grounding system as described in claim 4, characterized in that, The chaotic characteristic identification of the time series of the response voltage signal to obtain the nonlinear dynamic characteristics of the cable grounding system includes: The time series of the response voltage signal is reconstructed in state space to obtain the reconstructed phase space trajectory of the cable grounding system; Information entropy is extracted from the reconstructed phase space trajectory to obtain the entropy characteristics of the response voltage signal; Correlation dimension analysis is performed on the reconstructed phase space trajectory to obtain the dimension characteristics of the cable grounding system; By aggregating the entropy feature and the dimension feature into multidimensional features, the nonlinear dynamic features of the cable grounding system are obtained.

6. The method for live-line detection of defects in a cable grounding system as described in claim 1, characterized in that, Based on the typical physical defect types of the cable grounding system, pattern matching is performed on the multidimensional signal features to obtain a feature defect mapping table for the cable grounding system. Then, based on the feature defect mapping table, a fault feature database for the cable grounding system is constructed, including: Obtain cable grounding system samples corresponding to typical physical defect types in the cable grounding system; Supervised feature mapping is performed on the cable grounding system samples to establish the correlation between the multidimensional signal features and the typical physical defect types; Based on the aforementioned correlation, pattern mining is performed on the multidimensional signal features to obtain the core feature representation patterns of the typical physical defect types. By summarizing the pre-stored fault mechanism domain knowledge with the core feature manifestation patterns, a feature defect mapping table of the cable grounding system is obtained. Based on the aforementioned feature defect mapping table, the multidimensional signal features of the cable grounding system samples are vector space projected to obtain the standardized feature vectors of the cable grounding system samples. The standardized feature vector is associated and encapsulated with the defect type label, sample identifier, and acquisition environment parameters corresponding to the cable grounding system sample to obtain the fault feature database of the cable grounding system.

7. The method for live-line detection of defects in a cable grounding system as described in claim 6, characterized in that, The step of summarizing the pre-stored fault mechanism domain knowledge with the core feature manifestation patterns to obtain the feature defect mapping table of the cable grounding system includes: Based on the pre-stored fault mechanism domain knowledge, rules are extracted from the typical physical defect types to obtain the rule knowledge base of the cable grounding system; The consistency between the rule knowledge base and the core feature representation pattern is verified to obtain the verification result of the core feature representation pattern. Based on the verification results, the core feature representation pattern is optimized using knowledge guidance to obtain a refined feature pattern of the core feature representation pattern. By binding the refined feature patterns with the typical physical defect types, a feature defect mapping table for the cable grounding system is obtained.

8. The method for detecting live defects in a cable grounding system as described in claim 1, characterized in that, The process involves comparing the multi-dimensional signal feature vectors detected on-site with the fault feature database, and then using a hybrid expert network to perform intelligent diagnosis based on the comparison results. This yields the specific defect type and severity of the cable grounding system, including: Based on the standardized parameters of the fault feature database, the multidimensional signal feature vector detected on site is reduced to obtain the standardized on-site feature vector of the multidimensional signal feature vector. The standardized field feature vector is input into a hybrid expert network, and the probability distribution of the standardized field feature vector being assigned to the expert network for processing is calculated through the gating network of the hybrid expert network. Based on the probability distribution, the standardized field feature vectors are collaboratively inferred to obtain the diagnostic output of the cable grounding system; Based on the probability distribution, the diagnostic output is aggregated to obtain the specific defect type and severity of the cable grounding system.

9. The method for detecting live defects in a cable grounding system as described in claim 8, characterized in that, The formula for calculating the probability distribution is as follows: ； In the formula, The standardized field feature vector Assigned to the The probability processed by an expert network For the first The weight vector associated with each expert network, For the first Bias terms associated with an expert network, These are preset hyperparameters. For the penalty term function, This is for performing normalization exponent operations on the vector within the parentheses.

10. The method for detecting live defects in a cable grounding system as described in claim 1, characterized in that, The process of assessing the specific defect type and its severity to obtain early warning information for the cable grounding system includes: The warning level of the cable grounding system is determined based on the specific defect type and its severity. By performing maintenance knowledge mining on the warning level and the maintenance history of similar defects recorded in the fault feature database, a warning description and handling suggestions for the cable grounding system can be obtained. The warning level, the warning description, and the handling suggestions are encapsulated into the warning information for the cable grounding system.

Images