Power grid cable real-time current detection method and system

Abstract

The invention provides a power grid cable real-time current detection method and system, and the method comprises the steps: carrying out the multi-frequency-band synchronous collection of the current of a cable shielding layer for a high-voltage DC power transmission system with multiple loops laid in parallel, and obtaining the time-space information data; an electromagnetic coupling transfer function is calculated based on the data, and an electromagnetic coupling coefficient matrix is constructed, so that near-end cross induction current is separated to obtain a real current signal; establishing a cable dispersion model by using a real current signal, and performing fault feature recognition by combining a time domain criterion, a frequency domain criterion and time-frequency joint entropy analysis; and finally, calculating a credibility index by combining the temperature gradient and the electromagnetic environment parameters, correcting a preliminary judgment result and generating a protection control signal. According to the method, the near-end cross induction current and the real fault current can be effectively distinguished, the problem that the protection system of the multi-end high-voltage direct current system operates by mistake or does not operate in the power-on transient period is solved, and the reliability of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power grids, and in particular, to a method and system for real-time current detection of power grid cables. Background Art

[0002] In a multi-terminal flexible high-voltage DC power transmission system, cables are usually laid in a multi-loop parallel manner, forming a complex spatial electromagnetic field distribution structure. A cable consists of a conductor, an insulating layer, a shielding layer, and an outer sheath. The shielding layer is usually grounded at specific positions to prevent external electromagnetic interference and protect personal safety. The current real-time current detection methods for power grid cables mainly analyze the current characteristics of a single loop. In a multi-loop parallel laying environment, the electromagnetic fields generated during the energization of adjacent cables will induce current signals in the shielding layer of the measured cable that are highly similar to the fault characteristics. Traditional detection methods are difficult to distinguish this proximal cross-induced current from the real fault current, resulting in possible misoperation or refusal of the protection system. Especially during the power-on transient period of a multi-terminal high-voltage DC system, this interference is the most significant, seriously affecting the system reliability. Summary of the Invention

[0003] The main purpose of the present invention is to solve the technical problem that the existing real-time current detection method for power grid cables is difficult to effectively distinguish proximal cross-induced current from real fault current in a high-voltage DC power transmission system with multi-loop parallel laying; The first aspect of the present invention provides a method for real-time current detection of power grid cables, and the method for real-time current detection of power grid cables includes: Multiband synchronous acquisition of the current in the shielding layer of cables in a multi-loop parallel high-voltage DC power transmission system, and adding time stamps and position identifiers to the acquisition data of each measurement point during the acquisition process to obtain multiband shielding layer current data containing spatio-temporal information; Calculate the electromagnetic coupling transfer function between multi-loop cables according to the multiband shielding layer current data, construct an electromagnetic coupling coefficient matrix, and use the electromagnetic coupling coefficient matrix to separate the proximal cross-induced current in the multiband shielding layer current data to obtain a real current signal; According to the real current signal, establish a cable dispersion model including group velocity estimation parameters, and perform fault current characteristic recognition through a combination of time domain criteria, frequency domain criteria, and time-frequency joint entropy analysis based on the cable dispersion model to obtain a preliminary fault judgment result; According to the preliminary fault judgment result, combine the real-time monitored temperature gradient and electromagnetic environment parameters, calculate a confidence index for fault judgment, and correct the preliminary fault judgment result based on the confidence index to generate a cable fault judgment result and a corresponding protection control signal.

[0004] Optionally, in the first implementation manner of the first aspect of the present invention, the multi-band synchronous acquisition of the cable shielding layer current of the multi-circuit parallel high-voltage direct current transmission system, and adding a time stamp and a position identifier to the acquisition data of each measurement point during the acquisition process to obtain the multi-band shielding layer current data including spatio-temporal information includes: A shielding layer current sensor array is arranged at a preset interval along the length direction of the cables of the multi-circuit parallel high-voltage direct current transmission system; Perform temperature compensation processing on the data collected by the shielding layer current sensor array, and perform segmented sampling on the compensated data to obtain a sampling signal; Add a time stamp and position identifier information to the sampling signal, and transmit the sampling signal to the data processing unit through an optical fiber network to form multi-band shielding layer current data including spatio-temporal information.

[0005] Optionally, in the second implementation manner of the first aspect of the present invention, the calculation of the electromagnetic coupling transfer function between multi-circuit cables according to the multi-band shielding layer current data, constructing an electromagnetic coupling coefficient matrix, and using the electromagnetic coupling coefficient matrix to separate the proximal cross-induced current in the multi-band shielding layer current data to obtain a true current signal includes: Perform adaptive wavelet packet decomposition on the multi-band shielding layer current data to obtain the time-frequency two-dimensional characteristics of each measurement point; Perform spatial correlation analysis on the time-frequency two-dimensional characteristics of multiple measurement points, and combine the position information of the measurement points to construct a frequency-time-space three-dimensional characteristic map; According to the three-dimensional characteristic map, extract the time-domain characteristic parameters and spatial distribution characteristics of signals in different frequency bands, and based on the time-domain characteristic parameters and spatial distribution characteristics, combine the relative position relationship of the cables to construct a matrix equation set, and solve to obtain the electromagnetic coupling coefficient matrix; Substitute the electromagnetic coupling coefficient matrix into a pre-trained filter model offline to obtain corresponding filter coefficients; Use the filter coefficients to perform filtering processing on the multi-band shielding layer current data, separate the proximal cross-induced current component, and obtain a true current signal.

[0006] Optionally, in the third implementation manner of the first aspect of the present invention, the extraction of the time-domain characteristic parameters and spatial distribution characteristics of signals in different frequency bands according to the three-dimensional characteristic map, and based on the time-domain characteristic parameters and spatial distribution characteristics, combining the relative position relationship of the cables to construct a matrix equation set, and solving to obtain the electromagnetic coupling coefficient matrix includes: Perform peak detection on the three-dimensional characteristic map in each frequency band, and extract the time-domain characteristic parameters, including signal rise time, peak time, decay rate, and oscillation frequency, to form a time-domain characteristic vector; Calculate the propagation characteristics of the current wave based on the phase difference and amplitude ratio of the signals at different measurement points, and obtain the spatial distribution characteristics including phase delay, amplitude attenuation, and propagation speed; Based on the time-domain feature vectors and spatial distribution characteristics, and combined with the physical laying position and spacing information of the cables, establish a linear equation system for the electromagnetic coupling relationship between each loop; Apply the singular value decomposition method and the weighted least squares method to solve the linear equation system to obtain the initial electromagnetic coupling coefficient, and perform a physical constraint test on the initial electromagnetic coupling coefficient to obtain the electromagnetic coupling coefficient matrix.

[0007] Optionally, in the fourth implementation manner of the first aspect of the present invention, establish a cable dispersion model including group velocity estimation parameters according to the true current signal, and perform fault current feature recognition through the combination of time-domain criterion, frequency-domain criterion, and time-frequency joint entropy analysis based on the cable dispersion model. The preliminary fault judgment results include: Perform spectral analysis on the true current signal at multiple measurement points, calculate the arrival time difference of different frequency components of the same wavefront, and generate frequency-velocity relationship data; Use the frequency-velocity relationship data for polynomial fitting to establish the functional relationship between the group velocity and frequency, and obtain the group velocity estimation parameters characterizing the cable dispersion characteristics; Construct a cable dispersion model according to the group velocity estimation parameters and cable structure parameters, perform dispersion compensation processing on the true current signal, and obtain the compensated current signal; Calculate the time-domain criterion, frequency-domain criterion, and time-frequency joint entropy for the compensated current signal respectively to form a multi-dimensional feature vector; According to the preset fault type feature library, perform pattern matching on the multi-dimensional feature vector to obtain the fault type judgment result and the estimated fault location value, and form the preliminary fault judgment result.

[0008] Optionally, in the fifth implementation manner of the first aspect of the present invention, the performing pattern matching on the multi-dimensional feature vector according to the preset fault type feature library to obtain the fault type judgment result and the estimated fault location value, and forming the preliminary fault judgment result includes: Perform normalization processing on the multi-dimensional feature vector to map the feature values of each dimension to a unified interval, and calculate the Mahalanobis distance between the normalized feature vector and each fault template in the fault type feature library to obtain a distance matrix; Apply a weighted voting mechanism to the distance matrix, assign different weights to the candidate fault types according to the distance, calculate the cumulative score, determine the corresponding fault type, and obtain the fault type judgment result; Using the group velocity parameter in the cable dispersion model, combining the arrival time difference of current waves at different measurement points, constructing a fault location equation, solving to obtain an estimated fault location value, and combining the fault type judgment result and the estimated fault location value to form a preliminary fault judgment result.

[0009] Optionally, in the sixth implementation manner of the first aspect of the present invention, the calculating a confidence index for fault judgment based on the preliminary fault judgment result, combining the temperature gradient and electromagnetic environment parameters monitored in real time, and correcting the preliminary fault judgment result based on the confidence index to generate a cable fault judgment result and a corresponding protection control signal includes: Obtain the temperature distribution data and electromagnetic environment noise level along the cable, and construct an environmental impact assessment model as input parameters; Associate the environmental impact assessment model with the characteristic parameters in the preliminary fault judgment result, and calculate the theoretical deviation degree of the characteristic parameters under the current environmental conditions; Taking the preliminary fault judgment result as a prior probability according to the theoretical deviation degree, calculate the posterior probability of the fault occurrence, and generate a confidence index for fault judgment; Set a confidence threshold according to the operating state of the cable, compare the confidence index with the threshold, correct the preliminary fault judgment result to obtain a cable fault judgment result, and generate a protection control signal corresponding to the cable fault judgment result according to a preset protection strategy.

[0010] The second aspect of the present invention provides a power grid cable real-time current detection system, and the power grid cable real-time current detection system includes: A data acquisition module, configured to perform multi-band synchronous acquisition on the cable shielding layer current of a high-voltage direct current transmission system with multiple parallel circuits, and add a time stamp and a position identifier to the acquired data at each measurement point during the acquisition process to obtain multi-band shielding layer current data containing spatio-temporal information; An induction separation module, configured to calculate the electromagnetic coupling transfer function between multiple circuits of cables according to the multi-band shielding layer current data, construct an electromagnetic coupling coefficient matrix, and separate the proximal cross-induction current in the multi-band shielding layer current data by using the electromagnetic coupling coefficient matrix to obtain a true current signal; A feature recognition module, configured to establish a cable dispersion model including group velocity estimation parameters according to the true current signal, and perform fault current feature recognition through a time domain criterion, a frequency domain criterion, and combined time-frequency joint entropy analysis based on the cable dispersion model to obtain a preliminary fault judgment result; A judgment correction module is used to calculate a confidence index for fault judgment based on the preliminary fault judgment result, in combination with the temperature gradient and electromagnetic environment parameters monitored in real time, and correct the preliminary fault judgment result based on the confidence index to generate a cable fault judgment result and a corresponding protection control signal.

[0011] The above-mentioned real-time current detection method and system for power grid cables are aimed at a high-voltage DC transmission system with multiple circuits laid in parallel. By synchronously collecting spatio-temporal information data for the current of the cable shielding layer in multiple frequency bands; calculating the electromagnetic coupling transfer function based on this data and constructing an electromagnetic coupling coefficient matrix, so as to separate the proximal cross-induced current to obtain a true current signal; using the true current signal to establish a cable dispersion model, and combining time-domain criteria, frequency-domain criteria and time-frequency joint entropy analysis for fault feature recognition; finally, calculating a confidence index in combination with the temperature gradient and electromagnetic environment parameters, correcting the preliminary judgment result and generating a protection control signal. The present invention can effectively distinguish the proximal cross-induced current from the true fault current, solves the problem of misoperation or refusal to operate of the protection system during the power-on transient period of a multi-terminal high-voltage DC system, and improves the system reliability.

[0012] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention are achieved and obtained by the structures specifically pointed out in the specification, claims and drawings.

[0013] In order to make the above-mentioned objectives, features and advantages of the present invention more obvious and understandable, the following specifically gives preferred embodiments and, in conjunction with the accompanying drawings, makes a detailed description as follows. Description of the Drawings

[0014] Figure 1 It is a schematic diagram of the first embodiment of the real-time current detection method for power grid cables in the embodiments of the present invention; Figure 2 It is a schematic diagram of an embodiment of the real-time current detection system for power grid cables in the embodiments of the present invention. Detailed Embodiments

[0015] In order to make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0016] As used in the embodiments of the present invention, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally further include other unlisted steps or units, or may optionally further include other steps or units inherent to these processes, methods, products or devices.

[0017] For ease of understanding of this embodiment, first, a method for real-time current detection of power grid cables disclosed in the embodiments of the present invention will be introduced in detail. As Figure 1 shown, this method includes the following steps: 101. Synchronously collect the cable shielding layer currents of a high-voltage direct current transmission system with multiple parallel circuits in multiple frequency bands, and add time stamps and position identifiers to the collected data at each measurement point during the collection process to obtain multi-band shielding layer current data containing spatio-temporal information; In an embodiment of the present invention, the step of synchronously collecting the cable shielding layer currents of a high-voltage direct current transmission system with multiple parallel circuits in multiple frequency bands, and adding time stamps and position identifiers to the collected data at each measurement point during the collection process to obtain multi-band shielding layer current data containing spatio-temporal information includes: arranging a shielding layer current sensor array at preset intervals along the length direction of the cables of the high-voltage direct current transmission system with multiple parallel circuits; performing temperature compensation processing on the data collected by the shielding layer current sensor array, and performing segmented sampling on the compensated data to obtain a sampling signal; adding time stamp and position identifier information to the sampling signal, and transmitting the sampling signal to a data processing unit through an optical fiber network to form multi-band shielding layer current data containing spatio-temporal information.

[0018] Specifically, in a high-voltage direct current (HVDC) power transmission system with multi-circuit parallel laying, in order to achieve multi-band synchronous acquisition of cable shielding layer current, a shielding layer current sensor array is arranged along the cable length direction at a preset interval. Usually, in an HVDC (multi-terminal high-voltage direct current) cable system, the interval of the sensor array is set between 100 meters and 500 meters, and a combined sensor is installed at each measurement point, including a Hall current sensor, a Rogowski coil, and a high-frequency current transformer. The Hall current sensor is responsible for collecting low-frequency signals from 0 to 1 kHz. Based on the closed-loop Hall principle, it converts the magnetic field intensity into a voltage signal. The Rogowski coil is used for mid-frequency band acquisition from 1 to 10 kHz and detects the current change rate using the electromagnetic induction principle. The high-frequency current transformer is responsible for collecting high-frequency signals from 10 to 100 kHz and realizes it using the magnetic flux coupling principle. This multi-sensor configuration can comprehensively capture the full-spectrum current signal of the cable shielding layer to completely record the fault waveform and cross-induction characteristics. The sensor installation adopts a special double-layer structure. The outer layer is composed of silicon steel sheets or nanocrystalline alloys to form a magnetic shielding layer, and the inner layer uses a low magnetic permeability amorphous material to form a gradient shielding, thereby suppressing the electromagnetic interference of adjacent cables and improving the single-loop signal acquisition accuracy.

[0019] To address the problem of sensor temperature characteristic drift, temperature compensation is performed on the acquired data to eliminate measurement errors caused by load changes, ambient temperature fluctuations, and cable self-heating effects. In an HVDC cable system, the operating temperature range of the sensor is from -40°C to 85°C, which may cause sensitivity and zero-point drift. The compensation process first uses a PT100 platinum resistance temperature sensor integrated inside the sensor to monitor the temperature in real time and calculate the sensitivity change at the current temperature, where is the sensitivity at the standard temperature and and They are the first-order and second-order temperature coefficients respectively. The compensation algorithm corrects the original acquired signal through an inverse function relationship. The compensated data is sampled in segments by a 16-bit high-precision ADC, and the sampling rate is dynamically adjusted according to different frequency bands: the sampling rate in the low-frequency band is 10 kHz, in the medium-frequency band is 100 kHz, and in the high-frequency band is 1 MHz. The time-division multiplexing strategy is adopted to merge the sampling data of each frequency band to form a complete sampling signal. To achieve spatio-temporal correlation analysis, time stamps and position identification information need to be added to the sampling signal. Time synchronization uses the IEEE1588 Precision Time Protocol (PTP), which provides sub-microsecond accuracy and makes the local clocks of all measurement points consistent with the master station atomic clock. Specifically, each sampling point is appended with a 64-bit time stamp, including 32-bit second count and 32-bit nanosecond count, to ensure that the sampling time accuracy reaches the order of 10 nanoseconds. The position identification uses a 36-bit code, covering 8-bit cable number, 8-bit measurement point serial number, and 20-bit GPS coordinate information. The sampling signal with time stamp and position identification is transmitted to the data processing unit through a single-mode fiber optic network at a transmission rate of 1 Gbps, and the industrial Ethernet protocol is adopted, which conforms to the data frame format of IEC61850 standard, including preamble, destination address, source address, data length, data content, and check code. Through the above method, the system finally obtains the multi-band shield current data with complete spatio-temporal information correlation.

[0020] 102. Calculate the electromagnetic coupling transfer function between multi-loop cables based on the multi-band shield current data, construct an electromagnetic coupling coefficient matrix, and use the electromagnetic coupling coefficient matrix to separate the proximal cross-induced current in the multi-band shield current data to obtain the true current signal; In an embodiment of the present invention, the calculating the electromagnetic coupling transfer function between multi-loop cables based on the multi-band shield current data, constructing an electromagnetic coupling coefficient matrix, and using the electromagnetic coupling coefficient matrix to separate the proximal cross-induced current in the multi-band shield current data to obtain the true current signal includes: performing adaptive wavelet packet decomposition on the multi-band shield current data to obtain the time-frequency two-dimensional characteristics of each measurement point; performing spatio-temporal correlation analysis on the time-frequency two-dimensional characteristics of multiple measurement points, combining the position information of the measurement points, and constructing a frequency-time-space three-dimensional characteristic map; according to the three-dimensional characteristic map, extracting the time-domain characteristic parameters and spatial distribution characteristics of signals in different frequency bands, and based on the time-domain characteristic parameters and spatial distribution characteristics, combining the relative position relationship of the cables, constructing a matrix equation set, and solving to obtain the electromagnetic coupling coefficient matrix; substituting the electromagnetic coupling coefficient matrix into a pre-offline trained filter model to obtain the corresponding filter coefficients; using the filter coefficients to perform filtering processing on the multi-band shield current data to separate the proximal cross-induced current component and obtain the true current signal.

[0021] Specifically, during the process of adaptively decomposing the multi-band shield current data, in the HVDC cable system, the shield current signal usually contains various components such as fault current, cross-induced current, and background noise, and these components have different characteristics in the time-frequency domain. The adaptive wavelet packet decomposition uses the Daubechies wavelet basis function family, and particularly selects the DB4 wavelet, which has good time-frequency localization characteristics and is suitable for analyzing transient signals in the cable. The decomposition process first performs a 5-layer wavelet packet decomposition on the original signal to form a complete binary tree structure, and each node represents the signal component of a specific frequency band. Then, the Shannon entropy criterion is used to evaluate the information content of each node, and adaptive pruning is performed according to a preset threshold to retain the nodes with rich information content. In addition, considering the characteristics of the HVDC system, the algorithm uses signal extension technology to handle the boundary effect during the decomposition process and uses the soft threshold method to suppress high-frequency noise. Through this processing, the current signal at each measurement point is converted into a time-frequency two-dimensional feature representation, which clearly shows the energy distribution of the signal at different times and frequencies, laying a foundation for subsequent analysis. These two-dimensional features are essentially composed of a coefficient matrix, where each row represents a specific frequency band, each column represents a specific time window, and the coefficient value represents the energy magnitude at the corresponding time-frequency point.

[0022] During the process of performing spatial correlation analysis on the time-frequency two-dimensional features of multiple measurement points, since there are multiple measurement points distributed along the cable, the same current wave will pass through these measurement points in sequence, showing specific spatial correlation in terms of time and amplitude. The spatial correlation analysis first calculates the correlation between measurement points. For the two-dimensional feature matrices Mi and Mj of the i-th and j-th measurement points, calculate their cross-correlation function to determine the signal propagation delay time. Then, arrange the two-dimensional features of all measurement points according to the actual physical position to construct an initial three-dimensional data cube. Considering the signal propagation delay, perform time alignment processing on the time axis of each measurement point to compensate for the propagation delay and align the features of the same current wave at different measurement points. Then, introduce a spatial interpolation algorithm to fill the data between measurement points, using the cubic spline interpolation method to ensure a smooth transition in the spatial dimension. In this way, the finally constructed frequency-time-space three-dimensional feature map is a three-dimensional matrix, where represents the frequency dimension, the time dimension, and the spatial dimension. This three-dimensional feature map intuitively shows the propagation process of the current signal in space and the differences in propagation characteristics at different frequencies. For example, there are significant differences in the propagation speed between high-frequency components and low-frequency components, and this difference is of great significance for distinguishing proximal cross-induced current and true fault current.

[0023] In the process of extracting the time-domain characteristic parameters and spatial distribution characteristics of signals in different frequency bands according to the three-dimensional characteristic spectrum, in the time-domain characteristic extraction, first, the main waveform events are identified on each frequency-band slice of the three-dimensional spectrum, including the rising edge, peak point, and decay segment. The key-point positions are determined by calculating the first and second derivatives, and then the following time-domain characteristic parameters are extracted: rise time, peak time, decay rate, number of polarity conversions, and oscillation amplitude. For the spatial distribution characteristics, the phase difference, amplitude attenuation ratio, and propagation speed of the current wave propagating along the spatial dimension are calculated. The phase difference is calculated from the phase spectrum after Fourier transform, the amplitude attenuation ratio is determined by the peak ratio of adjacent measurement points, and the propagation speed is obtained by dividing the known distance between measurement points by the time difference of wavefront arrival. Based on the extracted characteristic parameters, combined with the actual physical laying position and relative distance relationship of the cables, a matrix equation set describing the electromagnetic coupling relationship is constructed. This equation set is based on the principle of electromagnetic induction, and the induced current in the measured cable is expressed as a linear combination of the original currents of adjacent cables. The solution process uses the weighted least squares method and combines the singular value decomposition technique to handle the possible ill-conditioned problems of the equation set, and finally an accurate electromagnetic coupling coefficient matrix is obtained.

[0024] In the process of substituting the electromagnetic coupling coefficient matrix into the filter model pre-trained offline, the offline-trained filter model is constructed using a deep learning-based method. This model uses a long short-term memory network structure and can capture the temporal characteristics of the current signal. The input of the model is the eigenvector of the electromagnetic coupling coefficient matrix, and the output is the optimal filter coefficient. In the offline training stage, a large amount of historical data is used, including current records under different operating states, different fault types, and different environmental conditions. The network parameters are optimized through the backpropagation algorithm. In real-time applications, the currently calculated electromagnetic coupling coefficient matrix is fed into the trained model, and the filter coefficients under the current working conditions are quickly obtained through forward calculation. The entire calculation process is completed within milliseconds, meeting the requirements of real-time processing. These filter coefficients essentially constitute the coefficient vector of a finite impulse response filter. The filter order is set between 32 and 64 according to system requirements, which is sufficient to capture complex induction characteristics while maintaining computational efficiency. The filter design adopts a multi-channel structure, and sub-filters are designed separately for different frequency bands. Finally, the full-band filtering result is obtained through weighted synthesis. This method makes full use of the frequency selectivity characteristics of the cross-induction phenomenon in the HVDC system and improves the filtering accuracy.

[0025] In the process of filtering the multi - band shield current data using the filter coefficients to separate the proximal cross - induction current component, the filtering process is implemented by time - domain convolution operation. The filter coefficient vector is convolved with the multi - band shield current data to obtain the estimated proximal cross - induction current. In the convolution operation process, the fast convolution algorithm is adopted. The time - domain convolution is converted into a frequency - domain product using the fast Fourier transform and then returned to the time - domain through the inverse transform, significantly improving the calculation efficiency. For boundary processing, the mirror extension method is used to avoid edge distortion. After obtaining the estimated proximal cross - induction current, this component is subtracted from the original shield current to obtain the true current signal. To ensure the accuracy of the filtering result, the system also sets up a self - verification mechanism. By calculating the power - spectrum characteristics of the residual signal, the separation effect is evaluated. When there is still an obvious correlation structure in the residual, the fine - tuning process of the filter parameters is triggered. The finally obtained true current signal no longer contains the interference components generated by adjacent cables and can truly reflect the current state of the cable under test, especially at the moment of power - on transient and fault occurrence, and the signal characteristics are more pure and clear.

[0026] Further, extracting the time - domain characteristic parameters and spatial distribution characteristics of signals in different frequency bands according to the three - dimensional characteristic spectrum, and based on the time - domain characteristic parameters and spatial distribution characteristics, combined with the relative position relationship of the cables, constructing a matrix equation set and solving to obtain the electromagnetic coupling coefficient matrix includes: performing peak detection on the three - dimensional characteristic spectrum within each frequency band to extract the time - domain characteristic parameters, including signal rise time, peak time, decay rate, and oscillation frequency, forming a time - domain characteristic vector; calculating the propagation characteristics of the current wave according to the phase difference and amplitude ratio of the signals at different measurement points to obtain the spatial distribution characteristics including phase delay, amplitude attenuation, and propagation speed; establishing a linear equation set of the electromagnetic coupling relationship between each loop based on the time - domain characteristic vector and spatial distribution characteristics, combined with the physical laying position and spacing information of the cables; applying the singular - value decomposition method and weighted least - squares method to solve the linear equation set to obtain the initial electromagnetic coupling coefficient, and performing physical constraint verification on the initial electromagnetic coupling coefficient to obtain the electromagnetic coupling coefficient matrix.

[0027] Specifically, during the process of performing peak detection on the three-dimensional feature spectrum, the three-dimensional feature spectrum contains the complete information of the current signal in the three dimensions of frequency, time, and space. By analyzing on different frequency band slices, more detailed time-domain features can be obtained. The peak detection process uses an extreme point search algorithm. First, two-dimensional slicing is performed on the time-space plane of each frequency band, and then the sliding window method is applied to the time series at a fixed spatial position to identify local maximum points. To improve the detection accuracy, the algorithm adopts an adaptive threshold mechanism, and the threshold value is dynamically set according to 3 times the standard deviation of the signal average energy, so that peak points can be accurately identified at different frequency bands and different energy levels. For each identified peak point, four key time-domain feature parameters are extracted: signal rise time (the time interval required from 10% peak to 90% peak), peak time (the absolute time when the signal reaches the maximum value), decay rate (the exponential decay coefficient after the signal peak, obtained by exponential fitting of the decay curve), and oscillation frequency (the oscillation period frequency during the decay process, obtained by statistical analysis of the zero-crossing intervals). The same feature extraction process is performed for each frequency band of each measurement point, and finally, the feature parameters of each frequency band are organized into a unified time-domain feature vector. This feature vector is a multi-dimensional array that contains the complete feature description of the signal in the time dimension, especially reflecting the dynamic change process of the current waveform. These features have significant differences in distinguishing normal energization and fault states in the HVDC system.

[0028] During the process of calculating the propagation characteristics of the current wave based on the phase difference and amplitude ratio of the signals at different measurement points, in a multi-circuit parallel HVDC system, the current wave exhibits an obvious spatial distribution law when propagating in the cable. For phase difference calculation, first, the Fourier transform is performed on the signals at each measurement point to obtain the phase information of each frequency component, and then the phase difference between adjacent measurement points is calculated to obtain the phase delay at different frequencies. In actual processing, the short-time Fourier transform is used to segment the signal, the window length is set to 1024 points, and the overlap rate is 50% to balance the time-frequency resolution. The amplitude ratio calculation is obtained by comparing the ratio of the peak amplitudes of adjacent measurement points, and this ratio reflects the attenuation of the signal during propagation. The propagation speed calculation uses the known physical distance between measurement points divided by the time difference of the wavefront arrival to obtain the propagation speed of different frequency components. For the case involving multiple measurement points, the least squares method is used to fit the relationship curve between propagation time and distance to improve the accuracy of speed estimation. These calculation results comprehensively form the spatial distribution characteristics, which include three key indicators: phase delay, amplitude attenuation, and propagation speed. The spatial distribution characteristics intuitively reflect the propagation behavior of the current wave in the spatial dimension, especially the different propagation characteristics exhibited by different frequency components. These characteristics are important bases for distinguishing proximal cross-inductive current and true fault current because there are essential differences in the spatial propagation characteristics between cross-inductive current and fault current.

[0029] In the process of establishing a system of linear equations for the electromagnetic coupling relationship between circuits based on the time-domain eigenvector and spatial distribution characteristics, combined with the physical laying position and spacing information of the cables, the electromagnetic coupling phenomenon in the HVDC cable system is essentially a multi-input multi-output system, in which the currents of each circuit affect each other. The establishment of the system of linear equations is based on the principle of electromagnetic induction, and the induced current in the shield of the measured cable is expressed as a linear combination of the original currents of all adjacent cables. When constructing the equations, the time-domain eigenvector and spatial distribution characteristics are used as system parameters, and the initial estimate of the coupling strength is determined according to the relative position relationship of the cables. For a typical n-circuit system, an n×n coupling equation system is constructed, where the diagonal elements represent the contribution of the self-current, and the non-diagonal elements represent the cross-induction contribution. The construction of the equations also takes into account the geometric structure of the cable laying, including factors such as the parallel laying distance, cross angle, and depth difference. In practical applications, the cable laying information is obtained from the geographic information system database, including accurate three-dimensional coordinates and relative position relationships. In addition, the physical parameters of the cable itself, such as the resistivity of the shield and the dielectric constant, are incorporated into the equations, and these parameters have a direct impact on the electromagnetic coupling strength. By comprehensively considering these factors, the finally established system of linear equations can accurately describe the complex electromagnetic coupling relationship in the multi-circuit HVDC cable system, providing a mathematical basis for the subsequent solution of the electromagnetic coupling coefficient matrix.

[0030] In the process of applying the singular value decomposition method and the weighted least squares method to solve the system of linear equations, the singular value decomposition is a powerful matrix analysis tool that can decompose the original equations into the product form of three matrices: A = UΣV^T, where U and V are orthogonal matrices, and Σ is a diagonal matrix, and the elements on the diagonal are the singular values. This method first decomposes the coefficient matrix of the system of linear equations, identifies the singular value distribution of the system, and judges the ill-condition degree of the equations by analyzing the magnitudes of the singular values. In the HVDC cable system, due to reasons such as measurement noise and weak coupling between some circuits, the equations usually have a certain degree of ill-condition characteristics. To solve this problem, the algorithm adopts the truncated singular value decomposition technique, retains the singular values greater than the preset threshold, and filters out the too-small singular values. In practical applications, the threshold is usually set to 1% of the largest singular value. Then, the weighted least squares method is applied to solve the processed equations, and this method assigns different weights to the data at different measurement points, and the weight values are positively correlated with the signal-to-noise ratio of the signals. This weighting strategy makes the measurement data with good signal quality contribute more in the solution process, improving the reliability of the solution. The finally obtained initial electromagnetic coupling coefficient is a matrix, and each element in it represents the coupling strength of a specific circuit to the target circuit.

[0031] During the process of physically constraining and verifying the initial electromagnetic coupling coefficient, the initially solved coupling coefficient may have physically unreasonable values due to factors such as numerical calculation errors, measurement noise, or incomplete linearity of the equations. The physical constraint verification first conducts boundary condition verification based on electromagnetic field theory. For example, the coupling coefficient should not exceed the theoretical maximum value of 1, and the coefficient value should be inversely proportional to the distance between the cables. The verification process uses an electromagnetic coupling attenuation model, which takes into account factors such as the distance between the cables, laying depth, and shield layer structure, to calculate the theoretical range of the coupling coefficient. When the initial solution result exceeds the theoretical range, a constraint optimization algorithm based on a physical model is used for correction. This algorithm approaches the predicted value of the theoretical model while maintaining the data fitting accuracy. In addition, the system also performs an energy conservation check to ensure that the sum of the squares of all coupling coefficients does not exceed a specific threshold, usually set to 1.5 times the number of loops. After completing all physical constraint verifications, the coupling coefficients are normalized so that the main diagonal elements are 1, indicating the complete contribution of the self-current. The finally generated electromagnetic coupling coefficient matrix is an n×n matrix that conforms to physical laws and has numerical stability.

[0032] 103. Based on the true current signal, establish a cable dispersion model including group velocity estimation parameters, and based on the cable dispersion model, perform fault current feature recognition through the combination of time-domain criteria, frequency-domain criteria, and time-frequency joint entropy analysis to obtain a preliminary fault judgment result. In an embodiment of the present invention, the step of establishing a cable dispersion model including group velocity estimation parameters based on the true current signal and performing fault current feature recognition through the combination of time-domain criteria, frequency-domain criteria, and time-frequency joint entropy analysis based on the cable dispersion model to obtain a preliminary fault judgment result includes: performing spectral analysis on the true current signal at multiple measurement points, calculating the arrival time difference of different frequency components of the same wavefront, and generating frequency-velocity relationship data; using the frequency-velocity relationship data for polynomial fitting to establish the functional relationship between the group velocity and frequency, and obtaining the group velocity estimation parameters characterizing the cable dispersion characteristics; constructing a cable dispersion model according to the group velocity estimation parameters and cable structure parameters, performing dispersion compensation processing on the true current signal to obtain the compensated current signal; calculating the time-domain criterion, frequency-domain criterion, and time-frequency joint entropy for the compensated current signal respectively to form a multi-dimensional feature vector; according to a preset fault type feature library, performing pattern matching on the multi-dimensional feature vector to obtain a fault type judgment result and a fault location estimation value, and forming a preliminary fault judgment result.

[0033] Specifically, in an HVDC cable system, current waves exhibit obvious dispersion phenomena during propagation, that is, the propagation speeds of different frequency components are different. Spectrum analysis first applies the short-time Fourier transform to the real current signals at each measurement point, using a Hanning window function with a window length of 2048 points and an overlap rate of 75% to obtain a high-precision time-frequency representation. For each identified current wave event, mark the time when its wavefront arrives at each measurement point, especially paying attention to obvious moments such as the mutation edge and peak points. By comparing the arrival times of the same current wave at different measurement points and combining the known physical distances between the measurement points, calculate the propagation speed of the wavefront. Since the current wave contains multiple frequency components, the signal is decomposed into multiple frequency bands during the analysis process, usually divided into 10 frequency bands, with a range covering from 500 Hz to 50 kHz. For the main frequency components in each frequency band, calculate their propagation time delays between the measurement points respectively, and then obtain the propagation speed at that frequency. For the complex waveforms commonly found in HVDC systems, the wavelet ridge tracking algorithm is used to enhance the ability to identify the propagation characteristics of different frequency components. This analysis process finally generates frequency-velocity relationship data, that is, the propagation speed values corresponding to each frequency point. These data points intuitively reflect the dispersion characteristics of electromagnetic waves in the cable and lay the foundation for the calculation of group velocity estimation parameters.

[0034] During the process of polynomial fitting using the frequency-velocity relationship data, the group velocity is a physical quantity that describes the propagation speed of wave packet energy. In a cable system, accurate group velocity estimation is crucial for fault location. Polynomial fitting uses the least squares method to construct the best fitting curve based on the frequency-velocity scatter plot. For the characteristics of HVDC cables, a polynomial model of order 3 to 5 is usually selected. A weighting mechanism is introduced during the fitting process to assign higher weights to the data points in the key frequency bands. These key frequency bands are usually the main frequency distribution intervals of HVDC system fault signals, such as 2 kHz to 20 kHz. To improve the fitting accuracy, the algorithm uses the cross-validation method to determine the optimal polynomial order. By dividing the data set into a training set and a validation set, select the model with the smallest validation error. In some cases, a single polynomial is difficult to describe the behavior of the entire frequency range. At this time, piecewise polynomial fitting is used, and different polynomial models are used in different frequency bands. After fitting, extract the polynomial coefficients as group velocity estimation parameters, and these parameters fully characterize the dispersion characteristics of the cable. In addition, the frequency derivative of the group velocity, that is, the dispersion coefficient, is also calculated. This coefficient describes the broadening degree of the wave packet during propagation and is another important parameter of the cable dispersion model.

[0035] Based on the group velocity estimation parameters and cable structure parameters, a cable dispersion model is constructed. During the process of performing dispersion compensation on the real current signal, the cable dispersion model is represented by a distributed parameter equivalent circuit based on the transmission line theory. Each small section of the cable is described by four basic parameters: resistance, inductance, capacitance, and conductance. Different from the traditional fixed-parameter model, this model introduces frequency-dependent parameters, especially the characteristics of the inductance L(f) and capacitance C(f) varying with frequency, which directly determine the propagation speed of electromagnetic waves. In the process of model construction, theoretical calculations and measured data are combined. The theoretical part is based on the physical structure parameters of the cable, such as conductor radius, insulation layer thickness, dielectric constant, etc.; the measured part uses the aforementioned group velocity estimation parameters to adjust the model parameters to make the propagation characteristics predicted by the model consistent with the actual observation results. The constructed dispersion model can accurately simulate the propagation behavior of current waves in the cable, including phenomena such as wavefront distortion, energy attenuation, and phase change. The dispersion compensation process is based on the inverse filtering principle. First, the real current signal is transformed into the frequency domain, and then a phase compensation function is applied, which is calculated by the dispersion model and can compensate for the phase delay differences of different frequency components. The compensated signal is restored to the time domain through the inverse Fourier transform to obtain the compensated current signal. These compensated signals eliminate the waveform distortion caused by the dispersion effect, making the fault waveform characteristics more obvious and facilitating subsequent feature recognition.

[0036] During the process of calculating the time-domain criterion, frequency-domain criterion, and time-frequency joint entropy for the compensated current signal respectively, the calculation of the time-domain criterion mainly focuses on the morphological characteristics of the current waveform, including parameters such as maximum amplitude, rise rate, integral value, duration, number of polarity reversals, and waveform asymmetry. Among them, the number of polarity reversals is particularly suitable for distinguishing energization transients and fault transients because fault currents usually accompany multiple polarity reversals. The calculation of the frequency-domain criterion is based on the power spectral density analysis of the signal, and features such as the main frequency component, frequency band energy ratio, harmonic content, and spectral centroid are extracted. These features reflect the frequency characteristic differences of different types of faults. For example, there are significant differences in the proportion of high-frequency components between single-pole grounding faults and inter-pole short-circuit faults. The time-frequency joint entropy is a new index for measuring the complexity of the signal, based on information theory, combining the short-time Fourier transform and Shannon entropy calculation. In the specific implementation, first, the time-frequency distribution matrix of the signal is calculated, then the matrix is normalized to a probability distribution, and finally the entropy value of this distribution is calculated. The time-frequency joint entropy can effectively distinguish random noise and structured fault signals because fault signals usually show strong structure in the time-frequency plane and have a lower entropy value. After all the calculated characteristic parameters are standardized, they are combined to form a multi-dimensional feature vector, usually with a dimension between 15 and 25, and each dimension represents a characteristic parameter, jointly constituting a comprehensive description of the fault state.

[0037] In the process of performing pattern matching on the multi-dimensional feature vector according to a preset fault type feature library to obtain a fault type judgment result and a fault location estimation value, the fault type feature library is a knowledge base constructed through a large amount of historical data and simulation analysis, and contains typical feature templates for various faults. Common fault types in the HVDC system are predefined in the library, such as single-pole ground fault, double-pole ground fault, pole-to-pole short circuit fault, and non-fault states such as normal power-on transient. Each fault type corresponds to a set of feature templates, which describe the typical feature vector distribution of that type of fault. The pattern matching process uses the Mahalanobis distance calculation method, which takes into account the correlation between the dimensions of the features and provides a more accurate similarity measure than the Euclidean distance. Calculate the distances between the current feature vector and each template in the library to obtain a distance matrix. Based on the distance matrix, apply the k-nearest neighbor algorithm to determine the closest fault type and calculate the matching confidence. For fault location estimation, use the group velocity parameter in the cable dispersion model, combined with the wavefront arrival time differences at multiple measurement points, to construct an overdetermined system of equations. Solve this system of equations by the weighted least squares method to obtain the distances from the fault point to each measurement point, and then determine the fault location. In practical applications, the location estimation also considers the influence of special points such as cable joints and bends, which will cause changes in wave impedance and wave reflection and need special treatment. Finally, combine the fault type judgment result and the location estimation value to form a preliminary fault judgment result including the fault type, location, and confidence level.

[0038] Further, the process of performing pattern matching on the multi-dimensional feature vector according to a preset fault type feature library to obtain a fault type judgment result and a fault location estimation value, and forming a preliminary fault judgment result includes: normalizing the multi-dimensional feature vector to map the feature values of each dimension to a unified interval, and calculating the Mahalanobis distances between the normalized feature vector and each fault template in the fault type feature library to obtain a distance matrix; applying a weighted voting mechanism to the distance matrix, assigning different weights to candidate fault types according to the distance, calculating the cumulative score, determining the corresponding fault type, and obtaining the fault type judgment result; using the group velocity parameter in the cable dispersion model, combined with the current wave arrival time differences at different measurement points, to construct a fault location equation, solving it to obtain the fault location estimation value, and combining the fault type judgment result and the fault location estimation value to form a preliminary fault judgment result.

[0039] Specifically, in HVDC cable fault detection, multi-dimensional feature vectors usually contain parameters with different physical meanings and orders of magnitude, such as voltage amplitudes in millivolts, time parameters in microseconds, and frequency characteristics in kilohertz, etc. These differences will cause features with large dimensions to dominate in distance calculation, affecting the judgment accuracy. The normalization process adopts the maximum-minimum normalization method to map each feature value to the interval from 0 to 1. For features sensitive to outliers, Z-score normalization is used, which normalizes through the mean and standard deviation. In the actual system, the normalization parameters are obtained through statistics of a large amount of historical data and are updated regularly to adapt to system changes. After normalization, the Mahalanobis distance between the feature vector and each template in the fault type feature library is calculated. This distance metric takes into account the correlation between features and can more accurately reflect the similarity in high-dimensional space. The fault type feature library in the HVDC system usually contains multiple preset templates such as single-pole ground fault, double-pole ground fault, pole-to-pole short circuit fault, shield layer breakage fault, etc. The template for each fault type is further divided into multiple subclasses according to factors such as the fault occurrence location and severity. During the calculation process, a parallel processing strategy is adopted to calculate the distances to multiple templates simultaneously, improving the real-time performance. The final obtained distance matrix is an N×M matrix, where N is the number of feature vectors currently processed, and M is the number of templates in the feature library. Each element in the matrix represents the Mahalanobis distance value between a specific feature vector and a specific template. The smaller the value, the higher the matching degree.

[0040] In the HVDC system, a single criterion often has difficulty dealing with complex fault scenarios. The weighted voting mechanism improves the judgment reliability by integrating multiple evidences. This mechanism first converts the distance matrix into a weight matrix and calculates the weights using an exponential decay function, so that templates with small distances obtain higher weights, and the weights of templates with large distances decay rapidly, reflecting the principle of "the closer ones have greater weights". After the weight calculation, the weights of all subclass templates of each fault type are accumulated to obtain the total score of this type. During the accumulation process, a template reliability coefficient is introduced, and a larger coefficient is given to templates with a high historical judgment accuracy. The coefficient value is determined through statistical analysis of long-term operation data. After the score calculation is completed, the system uses the threshold discrimination method to determine the final fault type. If the highest score exceeds the preset threshold (usually set to 1.5 times the second-highest score), it is confirmed as the corresponding fault type; if the score distribution is relatively average, it is marked as an "uncertain type" and further analysis is required. In addition, the system also sets a boundary threshold between faults and non-faults. When the scores of all types are lower than this threshold, it is determined as a non-fault state. This weighted voting mechanism is particularly suitable for dealing with fuzzy boundary cases in the HVDC system, such as the distinction between minor shield layer breakage and normal noise fluctuations. The final obtained fault type judgment result contains two parts of information: the fault type code and the confidence level, providing a basis for subsequent protection decisions.

[0041] In an HVDC cable system, the current wave generated by a fault propagates in both directions and passes through each measuring point in chronological order. The fault location equation is constructed based on the relationship between the wave propagation time difference and distance, taking into account the dispersion characteristics of the cable. In the specific implementation, first, the arrival time of the wavefront is accurately extracted from the compensated current signal by using a method combining wavelet transform and threshold detection to locate the mutation point of the waveform. Since the fault waveform in the HVDC system usually contains multiple frequency components, the most significant characteristic frequency bands (usually in the range of 5 kHz to 15 kHz) are selected for analysis during the location process, where the waveform clarity is the highest. For each pair of adjacent measuring points, based on the wavefront arrival time difference and the distance between the measuring points, the group velocity parameter in the dispersion model is used to establish the fault location equation. When there are n measuring points in the system, multiple equations can be constructed to form an overdetermined system of equations. Due to factors such as measurement errors and imperfect dispersion models, these equations usually do not have an exact solution, so the weighted least squares method is used to solve them, and the weights are proportional to the signal-to-noise ratio and distance of the measuring point signals. During the solution process, cable geometric constraints are also introduced, such as the fault point must be within the cable length range, and topological constraints in a branched cable system. The finally obtained fault location estimate is a specific distance value, representing the distance of the fault point from the reference end (usually the starting end of the cable).

[0042] Combining the fault type judgment result and the fault location estimate value, first, a consistency check is performed on the fault type and location because specific types of faults usually occur at specific locations. For example, ground faults often occur at cable joints or terminals. The check is implemented through a cross-validation mechanism. The system maintains a historical fault database that records the statistical distribution of various faults occurring at different locations. When the current judgment result deviates significantly from the historical statistical rules, the system will re-evaluate the judgment reliability and adjust the location estimate or type judgment if necessary. After the consistency check passes, the system generates a preliminary fault judgment result in a standard format, which includes five key fields: fault type code, fault type description text, fault location estimate value (accurate to meters), location estimate error range, and comprehensive confidence index. Among them, the comprehensive confidence index is the weighted average of the type judgment confidence and the location estimate confidence, reflecting the reliability of the overall judgment. In the HVDC system, the preliminary fault judgment result also includes an assessment of the fault severity, which is classified into three levels: minor, medium, and severe according to the amplitude, duration, and spectral characteristics of the fault current waveform. This classification is crucial for the selection of subsequent protection strategies. For example, minor faults may only require recording and monitoring, while severe faults require immediate tripping and isolation.

[0043] 104. According to the preliminary fault judgment result, combined with the real-time monitored temperature gradient and electromagnetic environment parameters, calculate the trust index of the fault judgment, and based on the trust index, correct the preliminary fault judgment result to generate a cable fault judgment result and the corresponding protection control signal.

[0044] In one embodiment of the present invention, based on the preliminary fault judgment result, combining the temperature gradient and electromagnetic environment parameters monitored in real time, calculating the confidence index of the fault judgment, and correcting the preliminary fault judgment result based on the confidence index to generate a cable fault judgment result and a corresponding protection control signal, which includes: obtaining the temperature distribution data along the cable and the electromagnetic environment noise level, and constructing an environmental impact assessment model with them as input parameters; associating the environmental impact assessment model with the characteristic parameters in the preliminary fault judgment result, and calculating the theoretical deviation degree of the characteristic parameters under the current environmental conditions; taking the preliminary fault judgment result as a prior probability according to the theoretical deviation degree, calculating the posterior probability of the fault occurrence, and generating the confidence index of the fault judgment; setting a confidence threshold according to the operating state of the cable, comparing the confidence index with the threshold, correcting the preliminary fault judgment result to obtain a cable fault judgment result, and generating a protection control signal corresponding to the cable fault judgment result according to a preset protection strategy.

[0045] Specifically, in the HVDC cable system, temperature and electromagnetic environment have a significant impact on the propagation characteristics of current signals, thereby affecting the accuracy of fault judgment. The temperature distribution data is obtained through a temperature sensor network arranged at key points of the cable. The sensor type mainly adopts a distributed optical fiber temperature measurement system, which uses the principle of Raman scattering of optical fibers and can provide a temperature measurement point every 1 meter along the entire cable. After the collected original temperature data is filtered by median filtering to remove outliers, a cubic spline interpolation algorithm is used to generate a continuous temperature distribution curve. The electromagnetic environment noise level is measured by electromagnetic field sensors set at key nodes of the cable system. These sensors include broadband magnetic field detection coils and electric field probes, and the monitoring frequency range is from 50 Hz to 100 kHz. After the noise level data is collected, spectrum analysis is carried out to extract the noise intensity indexes in different frequency bands, and special attention is paid to the background noise in the fault characteristic frequency band of the HVDC system. All monitoring data are added with time stamps and location identifiers and are transmitted to the data processing center in real time through a dedicated environmental monitoring network. These temperature distribution data and electromagnetic environment noise levels together serve as input parameters for the environmental impact assessment model, which reflect the real-time operating environment state of the cable system and provide necessary basic data for accurately evaluating the impact of environmental factors on fault judgment.

[0046] In the process of associating the environmental impact assessment model with the characteristic parameters in the preliminary fault judgment result and calculating the theoretical deviation degree of the characteristic parameters under the current environmental conditions, the environmental impact assessment model is constructed by a hybrid method combining data-driven and physical models. The model structure includes two parts: a temperature impact sub-model and an electromagnetic noise impact sub-model. The temperature impact sub-model is established based on the thermodynamic characteristics of the cable and the law of electrical parameters changing with temperature, mainly considering the impact of temperature on parameters such as the resistivity and dielectric constant of the cable shielding layer, and then analyzing the impact of these changes on the signal propagation characteristics of the current. This sub-model uses the finite element method to calculate the cable parameter distribution at different temperatures, and then derives the change in signal propagation characteristics through transmission line theory. The electromagnetic noise impact sub-model analyzes the interference degree of noise in different frequency bands on feature extraction, and establishes a mapping relationship between the noise intensity and the deviation of characteristic parameters through signal-to-noise ratio analysis and statistical methods. The outputs of the two sub-models are combined to form a complete environmental impact assessment result. In the association process, the system associates the multi-dimensional feature vector used in the preliminary fault judgment result with the environmental impact assessment model, and calculates the theoretical deviation degree of each dimension of characteristic parameters under the current environmental conditions. The specific calculation uses the environmental sensitivity matrix method, which represents the sensitivity of characteristic parameters to changes in environmental factors and is obtained through statistical analysis of a large amount of historical data. The finally obtained theoretical deviation degree is a vector with the same dimension as the feature vector, and each element represents the expected deviation amount of the corresponding characteristic parameter under the current environment.

[0047] Based on the theoretical deviation degree, the preliminary fault judgment result is used as the prior probability, and the posterior probability of the fault occurrence is calculated. The generation of the confidence index is realized based on the Bayesian probability theory. The preliminary fault judgment is regarded as a prior judgment based on imperfect observations and is corrected through environmental impact analysis. The specific implementation uses a Bayesian network model, and the network structure includes three layers: the environmental parameter layer, the feature parameter layer, and the fault type layer. The nodes in the environmental parameter layer include environmental factors such as temperature gradient and electromagnetic noise; the nodes in the feature parameter layer are the various parameters of the multi-dimensional feature vector; the nodes in the fault type layer correspond to various preset fault types. The conditional probability table of the network is obtained through training with historical data, which reflects the statistical relationship between the feature parameters and the fault types under different environmental conditions. During the calculation process, first, the environmental parameters and the theoretical deviation degree are substituted into the Bayesian network to update the probability distribution of the feature parameter layer, and then, based on the posterior distribution of the feature parameters, the posterior probabilities of each fault type are calculated. In the HVDC system, the Bayesian network usually considers five key fault types: monopole ground fault, bipolar ground fault, inter-pole short circuit fault, shield layer breakage, and non-fault state. The calculated posterior probability is directly converted into a confidence index for fault judgment. This index is a value between 0 and 1, indicating the confidence level in the fault judgment result under the current environmental conditions. A high confidence level indicates that the environmental factors have little impact on the judgment result and the judgment result is reliable; a low confidence level indicates that the environmental factors have significantly affected the feature parameters and the judgment result needs to be treated with caution.

[0048] The trust threshold setting is based on the operating status of the cable system, including factors such as load level, voltage level, converter operating mode, etc. The system establishes an operating status - threshold mapping relationship library based on historical operating data, and different operating statuses correspond to different threshold settings. For example, in the high - load operating state, the fault risk increases, and the system tends to set a lower trust threshold to improve protection sensitivity; while in the low - load or maintenance state, a higher threshold is adopted to reduce the probability of misoperation. In specific implementation, the system uses a fuzzy logic controller to dynamically adjust the threshold and determines the currently most suitable threshold value according to the combination of multiple operating parameters. After comparing the trust index with the threshold, the system corrects the preliminary fault judgment according to the comparison result: when the trust is higher than the threshold, the preliminary judgment result is directly adopted; when the trust is lower than the threshold but higher than the second - lowest threshold, the system will reduce the fault level assessment, such as correcting "confirmed fault" to "suspected fault"; when the trust is lower than the second - lowest threshold, the system will mark this judgment as "unreliable judgment" and trigger the manual review mechanism. The corrected result forms the final cable fault judgment result, which includes complete information such as fault type, location, level, and confidence. Based on the final judgment result, the system generates corresponding protection control signals according to the preset protection strategy. Different protection actions corresponding to different fault types, levels, and locations are defined in the protection strategy library, such as immediate tripping, delayed tripping, alarm monitoring, etc. The control signals are transmitted to relevant protection devices through a secure communication network to perform corresponding protection operations.

[0049] In this embodiment, by performing regional semantic perception association on the sensor combinations in different regions and performing three - dimensional association processing on physical data through an adaptive transmission protocol, a structured quality data stream is obtained; according to the structured quality data stream and quality inspection parameters, differential property quality inspection load balancing distribution and multi - stop point collaborative reasoning are performed on the logistics stop point network to obtain semi - structured quality inspection result data; according to the quality inspection result data, topological path quality decreasing analysis and quality anomaly spatio - temporal clustering processing are performed on the goods location and product quality status to obtain a quality knowledge graph; according to the quality knowledge graph, multi - dimensional quality control limit self - learning adjustment and quality early warning grading response processing are performed on the quality inspection parameters of different goods locations and paths. This method realizes the accurate traceability and rapid improvement of product quality problems, and improves the efficiency and accuracy of quality inspection.

[0050] The above described the real - time current detection method for power grid cables in the embodiments of the present invention. Next, the real - time current detection system for power grid cables in the embodiments of the present invention will be described. Please refer to Figure 2 One embodiment of the real - time current detection system for power grid cables in the embodiments of the present invention includes: The data acquisition module 201 is configured to perform multi - band synchronous acquisition on the cable shielding layer current of a high - voltage direct - current power transmission system with multi - loop parallel laying, and add time stamps and location identifiers to the acquisition data of each measurement point during the acquisition process to obtain multi - band shielding layer current data containing spatio - temporal information; The induction separation module 202 is configured to calculate the electromagnetic coupling transfer function between multi - loop cables according to the multi - band shielding layer current data, construct an electromagnetic coupling coefficient matrix, and use the electromagnetic coupling coefficient matrix to separate the proximal cross - induction current in the multi - band shielding layer current data to obtain a true current signal; The feature recognition module 203 is configured to establish a cable dispersion model including group velocity estimation parameters according to the true current signal, and perform fault current feature recognition through a combination of time - domain criteria, frequency - domain criteria, and time - frequency joint entropy analysis based on the cable dispersion model to obtain a preliminary fault judgment result; The judgment correction module 204 is configured to calculate a confidence index for fault judgment according to the preliminary fault judgment result, combined with the temperature gradient and electromagnetic environment parameters monitored in real time, and correct the preliminary fault judgment result based on the confidence index to generate a cable fault judgment result and a corresponding protection control signal.

[0051] In the embodiments of the present invention, the real - time current detection system for grid cables operates the above - mentioned real - time current detection method for grid cables. The real - time current detection system for grid cables is aimed at a high - voltage direct - current power transmission system with multi - loop parallel laying, and obtains spatio - temporal information data by performing multi - band synchronous acquisition on the cable shielding layer current; calculates the electromagnetic coupling transfer function based on this data and constructs an electromagnetic coupling coefficient matrix, so as to separate the proximal cross - induction current to obtain a true current signal; establishes a cable dispersion model using the true current signal, and performs fault feature recognition by combining time - domain criteria, frequency - domain criteria, and time - frequency joint entropy analysis; finally, calculates the confidence index by combining the temperature gradient and electromagnetic environment parameters, corrects the preliminary judgment result, and generates a protection control signal. The present invention can effectively distinguish between proximal cross - induction current and true fault current, solves the problem of misoperation or refusal to operate of the protection system during the power - on transient period of a multi - terminal high - voltage direct - current system, and improves the system reliability.

[0052] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above - described system or device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described in detail here.

[0053] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0054] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. A real-time current detection method for power grid cables, characterized in that: The grid cable real-time current detection method comprises: The cable shield layer current of the high-voltage direct current transmission system with multiple circuits laid in parallel is collected synchronously in multiple frequency bands, and the collected data of each measuring point is timestamped and marked with a location during the collection process to obtain multi-frequency shield layer current data containing time and space information; Calculating the electromagnetic coupling transfer function between the multi-loop cables according to the multi-band shielding layer current data, constructing an electromagnetic coupling coefficient matrix, and using the electromagnetic coupling coefficient matrix to separate the proximal cross-induced current in the multi-band shielding layer current data to obtain a real current signal; According to the real current signal, a cable dispersion model including group velocity estimation parameters is established, and based on the cable dispersion model, fault current characteristics are identified by combining time domain criterion and frequency domain criterion with time-frequency joint entropy analysis to obtain a preliminary fault judgment result; According to the preliminary fault judgment result, combined with the real-time monitored temperature gradient and electromagnetic environment parameters, the confidence index of the fault judgment is calculated, and the preliminary fault judgment result is corrected based on the confidence index to generate a cable fault judgment result and a corresponding protection control signal.

2. The real-time current detection method of power grid cable according to claim 1 is characterized in that: The method of synchronously collecting the cable shielding layer current of a multi-circuit high-voltage direct current transmission system in parallel and adding a timestamp and a position identifier to the collected data of each measuring point during the collection process to obtain multi-band shielding layer current data containing time and space information includes: An array of shield current sensors is arranged at preset intervals along the length direction of a high voltage direct current transmission system cable laid in parallel with multiple circuits; Performing temperature compensation processing on the data collected by the shielding layer current sensor array, and performing segmented sampling on the compensated data to obtain a sampling signal; A timestamp and location identification information are added to the sampling signal, and the sampling signal is transmitted to a data processing unit through an optical fiber network to form multi-band shielding layer current data containing time and space information.

3. The real-time current detection method of power grid cable according to claim 1 is characterized in that: The electromagnetic coupling transfer function between the multi-loop cables is calculated according to the multi-band shielding layer current data, an electromagnetic coupling coefficient matrix is ​​constructed, and the near-end cross-induced current in the multi-band shielding layer current data is separated by using the electromagnetic coupling coefficient matrix to obtain the real current signal, which includes: Performing adaptive wavelet packet decomposition on the multi-band shielding layer current data to obtain a time-frequency two-dimensional feature of each measuring point; Perform spatial correlation analysis on the time-frequency two-dimensional features of multiple measuring points, and construct a frequency-time-space three-dimensional feature map based on the location information of the measuring points; According to the three-dimensional characteristic spectrum, time domain characteristic parameters and spatial distribution characteristics of signals in different frequency bands are extracted, and based on the time domain characteristic parameters and spatial distribution characteristics, combined with the relative position relationship of the cables, a matrix equation group is constructed to solve the electromagnetic coupling coefficient matrix; Substituting the electromagnetic coupling coefficient matrix into a filter model pre-trained offline to obtain corresponding filter coefficients; The multi-band shielding layer current data is filtered using the filter coefficients to separate the near-end cross-induced current component and obtain a real current signal.

4. The real-time current detection method of power grid cable according to claim 3 is characterized in that: According to the three-dimensional feature spectrum, the time domain characteristic parameters and spatial distribution characteristics of signals in different frequency bands are extracted, and based on the time domain characteristic parameters and spatial distribution characteristics, combined with the relative position relationship of the cables, a matrix equation group is constructed to solve the electromagnetic coupling coefficient matrix, including: Performing peak detection on the three-dimensional feature spectrum in each frequency band, extracting time domain feature parameters, including signal rise time, peak time, decay rate and oscillation frequency, to form a time domain feature vector; According to the phase difference and amplitude ratio of the signals at different measuring points, the propagation characteristics of the current wave are calculated to obtain the spatial distribution characteristics including phase delay, amplitude attenuation and propagation speed; Based on the time domain feature vector and spatial distribution characteristics, combined with the physical laying position and spacing information of the cable, a linear equation group of the electromagnetic coupling relationship between each loop is established; The linear equations are solved by applying singular value decomposition method and weighted least square method to obtain initial electromagnetic coupling coefficients, and physical constraint check is performed on the initial electromagnetic coupling coefficients to obtain an electromagnetic coupling coefficient matrix.

5. The real-time current detection method of power grid cable according to claim 1 is characterized in that: According to the real current signal, a cable dispersion model including group velocity estimation parameters is established, and based on the cable dispersion model, fault current characteristics are identified by combining time domain criterion and frequency domain criterion with time-frequency joint entropy analysis, and preliminary fault judgment results are obtained, including: Performing spectrum analysis on the real current signal at multiple measuring points, calculating the arrival time difference of different frequency components of the same wavefront, and generating frequency-velocity relationship data; Using the frequency-velocity relationship data to perform polynomial fitting, a functional relationship between group velocity and frequency is established to obtain a group velocity estimation parameter that characterizes the cable dispersion characteristics; Constructing a cable dispersion model according to the group velocity estimation parameters and the cable structure parameters, and performing dispersion compensation processing on the real current signal to obtain a compensated current signal; The time domain criterion, the frequency domain criterion and the time-frequency joint entropy are respectively calculated for the compensated current signal to form a multi-dimensional feature vector; According to a preset fault type feature library, pattern matching is performed on the multi-dimensional feature vector to obtain a fault type judgment result and a fault location estimation value, thereby forming a preliminary fault judgment result.

6. The real-time current detection method of power grid cable according to claim 5 is characterized in that: The method of performing pattern matching on the multidimensional feature vector according to the preset fault type feature library to obtain a fault type judgment result and a fault location estimation value to form a preliminary fault judgment result includes: Normalizing the multidimensional feature vector so that the feature value of each dimension is mapped to a unified interval, and calculating the Mahalanobis distance between the normalized feature vector and each fault template in the fault type feature library to obtain a distance matrix; Applying a weighted voting mechanism to the distance matrix, assigning different weights to candidate fault types according to the distance, calculating the cumulative score, determining the corresponding fault type, and obtaining a fault type judgment result; The group velocity parameters in the cable dispersion model are used in combination with the arrival time difference of the current waves at different measuring points to construct a fault location equation, and the fault location estimate is obtained by solving it. The fault type judgment result and the fault location estimate are combined to form a preliminary fault judgment result.

7. The real-time current detection method of power grid cable according to claim 1 is characterized in that: The calculation of the confidence index of the fault judgment based on the preliminary fault judgment result and the real-time monitored temperature gradient and electromagnetic environment parameters, the correction of the preliminary fault judgment result based on the confidence index, and the generation of the cable fault judgment result and the corresponding protection control signal include: Obtain temperature distribution data and electromagnetic environmental noise level along the cable and use them as input parameters to build an environmental impact assessment model; Associating the environmental impact assessment model with the characteristic parameters in the preliminary fault judgment result, and calculating the theoretical deviation degree of the characteristic parameters under the current environmental conditions; According to the theoretical deviation degree, the preliminary fault judgment result is used as a priori probability, the posterior probability of the fault occurrence is calculated, and a confidence index of the fault judgment is generated; A confidence threshold is set according to the operating status of the cable, the confidence index is compared with the threshold, the preliminary fault judgment result is corrected to obtain the cable fault judgment result, and a protection control signal corresponding to the cable fault judgment result is generated according to a preset protection strategy.

8. A real-time current detection system for power grid cables, characterized in that: The grid cable real-time current detection system comprises: The data acquisition module is used to synchronously collect the shielding layer current of the cable of the multi-circuit high-voltage direct current transmission system in parallel, and add a timestamp and a position mark to the collected data of each measuring point during the collection process to obtain the shielding layer current data of multiple frequencies containing time and space information; An induction separation module is used to calculate the electromagnetic coupling transfer function between the multi-loop cables according to the multi-band shielding layer current data, construct an electromagnetic coupling coefficient matrix, and use the electromagnetic coupling coefficient matrix to separate the proximal cross-induction current in the multi-band shielding layer current data to obtain a real current signal; A feature recognition module is used to establish a cable dispersion model including group velocity estimation parameters according to the real current signal, and to perform fault current feature recognition based on the cable dispersion model by combining time domain criterion and frequency domain criterion with time-frequency joint entropy analysis to obtain a preliminary fault judgment result; The judgment and correction module is used to calculate the confidence index of the fault judgment according to the preliminary fault judgment result in combination with the real-time monitored temperature gradient and electromagnetic environment parameters, correct the preliminary fault judgment result based on the confidence index, and generate a cable fault judgment result and a corresponding protection control signal.

Images