A method and system for intelligent early warning of ground fault for high-voltage cable

Abstract

The application discloses a kind of ground fault intelligent early warning method and system for high-voltage cable, it is related to power failure early warning technical field;The method comprises: in high-voltage cable along line each measuring point synchronous acquisition voltage value and current value the voltage value and current value of same sampling time constitute a point in two-dimensional phase space, obtain original curve by connecting in time sequence, after being divided, then, state structure analysis is carried out to obtain the comprehensive fault value of target measuring point, below threshold value then determine high impedance ground fault;Based on fault point and adjacent point curvature sequence, through current direction reversal determination positioning fault section, realize high-voltage cable ground fault intelligent early warning.The application not only amplifies weak fault signal through phase space feature, improves identification sensitivity, but also realizes accurate positioning of fault section using polarity reversal determination, effectively overcomes the problem that high impedance ground fault signal is easily disturbed and submerged, significantly enhances the reliability of high-voltage cable ground fault intelligent early warning.

Description

Technical Field

[0001] This invention belongs to the field of power fault early warning technology, specifically relating to an intelligent early warning method and system for grounding faults in high-voltage cables. Background Technology

[0002] With the widespread application of high-voltage cables in urban power grids and industrial power supply systems, their operational status monitoring and fault early warning have become crucial for ensuring power supply reliability. High-voltage cable grounding faults are characterized by their high degree of concealment and wide-ranging impact, directly affecting the safe and stable operation of the power grid. Therefore, higher demands are placed on intelligent early warning technology for high-voltage cable grounding faults.

[0003] Existing high-voltage cable grounding fault early warning schemes typically collect operational data such as cable sheath current, grounding current, and partial discharge signals. These data are then processed through feature extraction, threshold determination, and trend analysis to output corresponding fault warning results. Furthermore, some schemes incorporate multi-source data fusion, anomaly identification algorithms, or auxiliary correction mechanisms to further improve the accuracy and timeliness of fault warnings. This approach offers advantages such as clear implementation paths, rapid monitoring response, and ease of integration into existing cable monitoring systems, and is therefore widely used in current high-voltage cable grounding fault early warning systems.

[0004] However, in the practical application of the above-mentioned fault early warning methods, due to the extremely small fault current when a ground fault occurs, especially in the case of a high-impedance ground fault, the fault signal is more easily overwhelmed by load current, capacitive current, and background noise, leading to problems such as characteristic distortion and feature loss in the fault signal. This signal anomaly directly caused by the characteristics of ground faults can lead to misjudgments or even missed judgments in the fault early warning algorithm, thereby affecting the accuracy of fault early warning and the effectiveness of power grid operation and maintenance decisions. Summary of the Invention

[0005] The purpose of this invention is to solve the problem that in existing high-voltage cable grounding fault early warning methods, the grounding fault signal is easily submerged by load current, capacitive current and background noise, resulting in insufficient early warning accuracy. Therefore, this invention proposes an intelligent early warning method and system for high-voltage cable grounding faults.

[0006] In a first aspect of this invention, a method for intelligent early warning of grounding faults in high-voltage cables is first proposed, the method comprising: Voltage and current values ​​are simultaneously collected at each measurement point along the high-voltage cable at a preset fixed sampling rate. For the target measurement point, with voltage as the abscissa and current as the ordinate, the voltage and current values ​​at the same sampling time are used to form a point in a two-dimensional phase space, and the points are connected in time sequence to obtain the original curve of the measurement point; the target measurement point is any one of the measurement points. The original curve is segmented according to a preset period to obtain a set of segmented curves; The comprehensive fault value of the target measurement point is obtained by performing state structure analysis on the piecewise curve set in the two-dimensional phase space. If the overall fault value is less than a preset first threshold, it is determined that a high impedance grounding fault has occurred at the location of the target measurement point, and the target measurement point is taken as the fault measurement point. A curvature sequence set is obtained by selecting the curvature sequence corresponding to the fault measurement point and a preset number of adjacent measurement points; The fault section is obtained by determining the direction reversal of the fault current based on the curvature sequence set, and intelligent fault warning is performed based on the fault section.

[0007] This scheme achieves sensitive identification and location of high-impedance grounding faults by constructing a two-dimensional phase space of voltage and current and performing state structure analysis on the curves. It effectively amplifies weak fault signals by utilizing phase space characteristics and accurately divides fault sections by determining the direction reversal of the curvature sequence, thus significantly improving the accuracy, anti-interference ability and reliability of high-voltage cable high-impedance grounding fault early warning.

[0008] Optionally, the comprehensive fault value of the target measurement point is obtained by performing state structure analysis on the piecewise curve set in the two-dimensional phase space, including: Within the two-dimensional phase space, the target piecewise curve is discretized into a finite number of state nodes to obtain a sequence of state nodes; the target piecewise curve is any one of the piecewise curves in the set of piecewise curves. For each state node in the state node sequence, calculate its local curvature at the corresponding position on the target piecewise curve to obtain a curvature sequence; The curvature entropy value is obtained by statistically analyzing the probability distribution of curvature values ​​in the curvature sequence. The comprehensive fault value of the target measurement point is calculated based on the curvature entropy values ​​corresponding to all piecewise curves.

[0009] This scheme obtains a sequence of state nodes by adaptively discretizing the piecewise curves, then calculates the local curvature of each node and statistically analyzes its probability distribution to obtain the curvature entropy value. Finally, it calculates the comprehensive fault value by combining the curvature entropy values ​​of each piecewise curve. This not only achieves fine quantitative extraction of the features of the two-dimensional phase space curve, but also effectively amplifies the weak waveform distortion features caused by high-impedance grounding faults through statistical analysis of curvature entropy values, significantly improving the sensitivity and anti-interference capability of fault identification.

[0010] Optionally, discretizing the target piecewise curve into a finite number of state nodes in the two-dimensional phase space to obtain a state node sequence includes: For each sampling point on the target segmented curve, calculate the angle between the tangent to the curve at that point and the horizontal axis to obtain the tangent direction angle sequence; The direction angle change rate sequence is obtained by subtracting the tangent direction angles of two adjacent sampling points in the tangent direction angle sequence and taking the absolute value. Identify the extreme points in the direction angle change rate sequence and sort them according to their positions on the target piecewise curve to obtain a set of candidate segmentation points; The target segmented curve is divided into intervals based on the candidate segmentation point set to obtain an initial interval set; For each initial interval in the initial interval set, calculate the cumulative sum of the rate of change of the direction angle within the initial interval and normalize it to obtain the interval direction change weight set; The target interval set is obtained by filtering and dividing the initial interval set according to the interval direction change weight set; Each interval in the target interval set is defined as a state node, and all state nodes are arranged in order of position to obtain a state node sequence.

[0011] This scheme achieves adaptive discretization of piecewise curves in two-dimensional phase space by calculating the tangent direction angle and its rate of change sequence, identifying extreme points as segmentation boundaries, and then filtering and dividing the initial interval by combining the direction change weight. This avoids the shortcomings of traditional equal-spacing division in capturing local small curvature changes. Furthermore, by automatically increasing the number of nodes in severely curved areas and reducing the number of nodes in straight areas, it significantly improves the sensitivity of subsequent local curvature and curvature entropy analysis for detecting weak fault distortions while ensuring computational efficiency.

[0012] Optionally, the initial interval set is filtered and divided according to the interval direction change weight set to obtain the target interval set, including: Step 1: Initialize an empty target interval set; Step 2: Select the initial intervals whose interval direction change weights are greater than a preset change threshold from the interval direction change weight set as the intervals to be divided to obtain the interval set to be divided; Step 3: Divide the target interval to be divided into equal parts according to the arc length to obtain the first and second equal parts; the target interval to be divided is any one of the intervals to be divided in the set of intervals to be divided. Step 4: Calculate the cumulative sum of the rates of change of the direction angle for the first and second equally divided intervals, respectively; Step 5: If the cumulative sum of the rate of change of the direction angle is greater than the preset percentage threshold of the total cumulative sum of the target intervals to be divided, return to step 2 and add the equal intervals that do not meet the conditions to the set of intervals to be divided; otherwise, add the first equal interval and the second equal interval to the target interval set, and finally obtain the target interval set after division.

[0013] This scheme achieves adaptive and refined division of the initial interval by filtering the interval to be divided based on a preset change threshold, recursively dividing it into equal parts, and combining cumulative and proportion judgments. This ensures that the node density in the severely curved area is sufficient to capture the distortion characteristics of minor faults, while avoiding excessive subdivision of the straight area that would cause computational redundancy. This keeps the total change in the direction angle of the final interval uniform and moderate, providing a more balanced and efficient discretization foundation for subsequent curvature calculation and fault feature analysis.

[0014] Optionally, calculating the comprehensive fault value of the target measurement point based on the curvature entropy values ​​corresponding to all piecewise curves includes: pass Calculate the comprehensive fault value of the target measurement point; where N is the number of piecewise curves. Let be the curvature entropy value of the i-th piecewise curve. The adaptive threshold for the i-th piecewise curve. Let be the curvature entropy value of the j-th piecewise curve. The adaptive threshold for the j-th piecewise curve. Let be the forgetting coefficient, and 0 < 0. <1.

[0015] This solution constructs a comprehensive fault value calculation model that includes summation and product terms. It can quantify the deviation of the curvature entropy value of a single piecewise curve from the adaptive threshold, and reflect the continuous impact of fault characteristics through the product relationship of subsequent terms. It realizes the weighted fusion and correlation analysis of curve distortion characteristics in different time periods, effectively suppresses single abnormal fluctuations caused by random interference, and significantly improves the accuracy and reliability of high impedance grounding fault identification.

[0016] Optionally, determining the fault segment by reversing the fault current direction based on the curvature sequence set includes: For each curvature sequence in the curvature sequence set, after removing the times when the curvature value is 0, the positive and negative signs of all remaining curvature values ​​are extracted and arranged in chronological order to obtain the curvature polarity spectrum of each measurement point in each preset period; For each measurement point, the symbol values ​​at the same sampling number position in multiple consecutive preset periods are integrated to obtain a symbol set. The number of times the positive and negative signs appear in the symbol set is counted. The symbol with the most occurrences is taken as the fused symbol value at that sampling number position. All the fused symbol values ​​at all sampling number positions are arranged in the order of sampling number to obtain the fused polarity spectrum of that measurement point. The polarity reversal value is obtained by performing difference mapping on the fused polarity spectrum of each pair of adjacent measurement points; Compare the polarity reversal value with a preset second threshold; If the polarity reversal value of adjacent measurement points is lower than the preset second threshold, it is determined that the fault current direction has reversed in the high-voltage cable section between the adjacent measurement points, and the high-voltage cable section is located as the fault section.

[0017] This scheme constructs a curvature polarity map and performs multi-cycle fusion. Then, it determines the reversal of the fault current direction based on the polarity reversal value of adjacent measurement points. This not only effectively filters out the influence of random interference on polarity characteristics, but also accurately captures the changing characteristics of the fault current direction. This enables reliable location of high-impedance grounding fault sections of high-voltage cables and significantly improves the anti-interference capability and location accuracy of fault section determination.

[0018] Optionally, the polarity reversal value can be obtained by performing difference mapping on the fused polarity spectrum of each pair of adjacent measurement points, including: pass Calculate the polarity reversal value between adjacent measurement points; where L is the length of the fused polarity map. and These are the sign values ​​of the fused polarity spectra of adjacent measurement points A and B at the kth sampling sequence, with the sign value being either 0 or 1.

[0019] This scheme calculates the point-by-point differences in the fused polarity spectrum of adjacent measurement points and takes the average value to obtain the polarity reversal value. It can intuitively reflect the overall difference in curvature polarity characteristics between two points in a quantitative form. This simplifies the judgment logic of fault current direction reversal characteristics and effectively suppresses the influence of local interference on polarity judgment, providing a stable and reliable quantitative basis for the accurate location of high-voltage cable grounding fault sections.

[0020] In a second aspect of this invention, a grounding fault intelligent early warning system for high-voltage cables is provided, comprising: The acquisition module is used to synchronously acquire voltage and current values ​​at a preset fixed sampling rate at various measurement points along the high-voltage cable. The curve generation module is used to construct a point in a two-dimensional phase space for the target measurement point, with voltage as the abscissa and current as the ordinate. The voltage and current values ​​at the same sampling time are used to form a point, and the points are connected in time sequence to obtain the original curve of the measurement point. The target measurement point can be any measurement point among all measurement points. The segmentation module is used to segment the original curve according to a preset period to obtain a segmented curve set; The fault value module is used to perform state structure analysis on the piecewise curve set in the two-dimensional phase space to obtain the comprehensive fault value of the target measurement point; The condition module is used to determine that a high impedance grounding fault has occurred at the location of the target measurement point if the comprehensive fault value is less than a preset first threshold, and to designate the target measurement point as the fault measurement point. The sequence selection module is used to select the curvature sequence corresponding to the fault measurement point and a preset number of adjacent measurement points to obtain a curvature sequence set. The early warning module is used to determine the fault section by reversing the direction of the fault current based on the curvature sequence set, and to provide intelligent early warning of the fault based on the fault section. Attached Figure Description

[0021] The present invention will now be further described with reference to the accompanying drawings.

[0022] Figure 1 A flowchart illustrating an intelligent early warning method for grounding faults in high-voltage cables, provided as an embodiment of the present invention; Figure 2 This is a flowchart for generating a target interval set, provided as an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0024] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] This invention provides an intelligent early warning method for grounding faults in high-voltage cables. See also... Figure 1 The method includes the following steps: S101, synchronously collects voltage and current values ​​at each measurement point along the high-voltage cable at a preset fixed sampling rate; S102, for the target measurement point, with voltage as the abscissa and current as the ordinate, the voltage and current values ​​at the same sampling time are used to form a point in a two-dimensional phase space, and the points are connected in time sequence to obtain the original curve of the measurement point. S103, the original curve is segmented according to a preset period to obtain a set of segmented curves; S104, perform state structure analysis on the piecewise curve set in two-dimensional phase space to obtain the comprehensive fault value of the target measurement point; S105, if the comprehensive fault value is less than the preset first threshold, it is determined that a high impedance grounding fault has occurred at the location of the target measurement point, and the target measurement point is taken as the fault measurement point; S106, Select the curvature sequence set by selecting the curvature sequence corresponding to the fault measurement point and its adjacent preset number of measurement points; S107, based on the curvature sequence set, the fault current direction is reversed to determine the fault section, and the fault section is used for intelligent fault warning. The target measurement point is any one of the measurement points.

[0026] In one implementation, the preset period and the preset first threshold are set by technicians, specifically 0.02s and 0.8s, and the preset quantity is 3.

[0027] In one embodiment, the comprehensive fault value of the target measurement point is obtained by performing state structure analysis on a piecewise curve set in a two-dimensional phase space, including: In a two-dimensional phase space, the target piecewise curve is discretized into a finite number of state nodes to obtain a sequence of state nodes; the target piecewise curve is any piecewise curve in the set of piecewise curves. For each state node in the state node sequence, calculate its local curvature at the corresponding position on the target piecewise curve to obtain the curvature sequence; The curvature entropy value is obtained by statistically analyzing the probability distribution of curvature values ​​in a curvature sequence. The comprehensive fault value of the target measurement point is calculated based on the curvature entropy values ​​corresponding to all segmented curves.

[0028] In one implementation, the key data of loop resistance and sheath insulation can be collected to reflect whether there are connection defects such as oxidation, loosening, or breakage in the sheath, grounding wire, and grounding box joints, as well as whether the outer sheath itself has fundamental deterioration problems such as damage, moisture, or aging. This can build a multi-dimensional monitoring system to provide intelligent early warning for grounding faults in high-voltage cables.

[0029] In one implementation, the curvature sequence is generated as follows: For each state node in the state node sequence, firstly, the range of sampling points on the target piecewise curve corresponding to that node is determined; for each sampling point within the range, the first and second derivatives of the internal points are calculated using the central difference method, with forward differencing for the first point and backward differencing for the last point, and then the curvature formula is used. Calculate the local curvature at this point and use this curvature value as the curvature of its corresponding state node at that location; then calculate the average of all curvatures within each node. The curvature value of this state node is used as the curvature value. The curvature values ​​of all state nodes are then summed to obtain a curvature sequence; where, and These are the first and second derivatives of the voltage, respectively. and These are the first and second derivatives of the current, respectively. This is a preset positive number to prevent calculation overflow caused by a denominator of zero. Let be the starting index number among the sampling points contained in the p-th state node. Let be the end index number among the sampling points contained in the p-th state node. Let be the local curvature of the nth sampling point.

[0030] In one implementation, the curvature entropy value is generated by determining the minimum and maximum curvature values ​​based on the curvature sequence of the target piecewise curve, which contains curvature values ​​at A sampling points. This minimum and maximum curvature values ​​are then used to define the distribution interval, which is adaptively divided into M continuous and non-overlapping sub-intervals. The division employs an equal probability principle to ensure that the number of curvature values ​​in each sub-interval is as equal as possible; typically, a value of M is used. ; Iterate through each curvature value, determine its sub-interval, and accumulate the frequencies to obtain the frequency of each sub-interval. Then calculate the probability of each subinterval. Finally, substitute the formula into Shannon's entropy formula. The curvature entropy value is obtained, which characterizes the uniformity of the curvature distribution of the target piecewise curve.

[0031] In one implementation, grounding faults, especially high-impedance grounding faults, can cause changes in the electric and magnetic fields along the entire line. Relying solely on single-point monitoring data cannot fully reflect the fault propagation pattern. Therefore, this invention deploys multiple sets of measurement points along the entire high-voltage cable and uses a fixed sampling rate to perform high-precision time-synchronous acquisition of the instantaneous values ​​of line voltage and current.

[0032] Considering that short-term operating fluctuations and instantaneous interference can easily lead to misjudgments and missed judgments, this invention integrates multi-period operating characteristics for comprehensive evaluation and combines dynamic adaptive judgment thresholds to achieve reliable identification of high-impedance grounding faults, effectively distinguishing between accidental disturbances and persistent real faults. To avoid judgment bias caused by the instability of single-cycle characteristics, the influence of random interference is eliminated through multi-cycle feature fusion, forming a stable and reliable state feature spectrum. Then, taking advantage of the inherent reverse difference in electrical flow direction on both sides of the fault area, the current flow reversal interval is accurately identified by comparing the state feature differences of adjacent monitoring points, realizing accurate location of high-voltage cable grounding fault sections, and finally completing the intelligent early warning and judgment of line faults.

[0033] In one embodiment, discretizing the target piecewise curve into a finite number of state nodes in a two-dimensional phase space to obtain a sequence of state nodes includes: For each sampling point on the target piecewise curve, calculate the angle between the tangent to the curve at that point and the horizontal axis to obtain the tangent direction angle sequence; The direction angle change rate sequence is obtained by subtracting the tangent direction angles of two adjacent sampling points in the tangent direction angle sequence and taking the absolute value. Identify the extreme points in the direction angle change rate sequence and sort them according to their positions on the target piecewise curve to obtain a set of candidate segmentation points; The target piecewise curve is divided into intervals based on the candidate segmentation point set to obtain the initial interval set; For each initial interval in the initial interval set, calculate the cumulative sum of the rate of change of the direction angle within the initial interval and normalize it to obtain the interval direction change weight set; The target interval set is obtained by filtering and dividing the initial interval set according to the interval direction change weight set; Each interval in the target interval set is defined as a state node, and all state nodes are arranged in order of position to obtain a state node sequence.

[0034] In one implementation, the process of identifying extreme points involves obtaining a sequence of direction angle change rates. This sequence reflects the curvature of the target piecewise curve at each local location and contains multiple values, each corresponding to a sampling point (excluding the first and last points). Starting from the second sampling point in the sequence, the process is repeated until the second-to-last sampling point. For each currently scanned interior point, its value is compared with the values ​​of its preceding and following points. If the value is greater than both, the point is marked as a local maximum. If the value is less than both, the point is marked as a local minimum. If the value is equal to either the preceding or following point, it is not marked as an extreme point. After scanning all interior points, all marked local maxima and minima are collected, sorted according to their position on the target piecewise curve, and points with duplicate positions are removed, ultimately yielding a set of candidate segmentation points.

[0035] In one implementation, the generation process of the interval direction change weight set is as follows: First, each initial interval in the initial interval set is sequentially extracted. For the currently extracted initial interval, the direction angle change rate values ​​corresponding to all sampling points within the interval are obtained. These values ​​are added one by one according to the sampling point order to obtain the cumulative sum of the direction angle change rates of the initial interval. Then, after the cumulative sum calculation is completed for all initial intervals, the cumulative sum values ​​of all initial intervals are added together to obtain the sum of the cumulative sums of all initial intervals. Finally, for each initial interval, its own cumulative sum of direction angle change rates is divided by the sum of the cumulative sums, and the quotient obtained is the direction change weight of the initial interval. The direction change weights of all initial intervals are arranged according to their order on the target piecewise curve to obtain the interval direction change weight set.

[0036] In one implementation, the process adaptively discretizes the continuous piecewise curve into a non-uniformly distributed sequence of state nodes. Since track distortion caused by high-impedance grounding faults often manifests as small but continuous changes in local curvature, traditional equal-interval partitioning cannot effectively capture such features. By calculating the tangent direction angle and its rate of change, the severity of local curvature of the track geometry can be quantified. Identifying extreme points as natural partitioning boundaries allows for initial partitioning based on the track's own undulating structure, avoiding feature misalignment caused by artificially setting fixed intervals. Further, the cumulative sum of the direction angle change rates of each initial interval is calculated and normalized to obtain the direction change weight. Based on this, the initial intervals are filtered and subdivided, ensuring that the total curvature borne by each final state node is basically consistent. The node density is adaptively adjusted based on the geometric characteristics of the current piecewise curve, automatically increasing the number of nodes in areas of severe curvature and decreasing the number of nodes in flat areas. This significantly improves the sensitivity and accuracy of subsequent local curvature calculation and curvature entropy analysis in detecting and locating weak fault distortions while ensuring computational efficiency.

[0037] In one embodiment, reference Figure 2 The target interval set is obtained by filtering and dividing the initial interval set according to the interval direction change weight set, including: Step 1: Initialize an empty target interval set; Step 2: Select the initial intervals whose interval direction change weights are greater than the preset change threshold as the intervals to be divided to obtain the set of intervals to be divided; Step 3: Divide the target interval to be divided into equal parts according to the arc length to obtain the first and second equal parts; the target interval to be divided is any one of the intervals to be divided in the set of intervals to be divided. Step 4: Calculate the cumulative sum of the rates of change of the direction angle in the first and second equally divided intervals, respectively; Step 5: If the cumulative sum of the angle change rate is greater than the preset percentage threshold of the total cumulative sum of the target intervals to be divided, return to Step 2 and add the equal intervals that do not meet the conditions to the set of intervals to be divided; otherwise, add the first and second equal intervals to the target interval set, and finally obtain the target interval set after division.

[0038] In one implementation, the preset change threshold and the preset proportion threshold are set by technical personnel, and their values ​​can be 0.3 and two-thirds, respectively. The equally divided intervals that do not meet the conditions refer to equally divided intervals where the cumulative sum of the direction angle change rate is greater than the preset proportion threshold of the total cumulative sum of the target intervals to be divided.

[0039] In one implementation, Figure 2 The cumulative sum threshold is the preset percentage threshold of the total cumulative sum of the target interval to be divided. Y represents that the direction angle change rate is greater than the cumulative sum threshold, and N represents that the direction angle change rate is not greater than the cumulative sum threshold.

[0040] In one implementation, the total directional angle changes borne by each interval in the initial interval set vary significantly. Directly using these intervals as state nodes would result in severely curved intervals being coarsely represented by a single node, losing detail, while gently curved intervals might be over-subdivided, causing computational redundancy. To address this, a preset change threshold is set to filter intervals with directional change weights greater than the threshold as the intervals to be divided, and these are recursively divided equally. Simultaneously, it is checked whether the cumulative sum of directional change rates of each sub-interval still exceeds a preset percentage threshold of the total cumulative sum for that interval. Sub-segments with unevenly distributed bending energy are continuously subdivided until the percentage of cumulative bending energy in all intervals falls below the threshold. This process ensures that the total directional angle change within each state node in the final target interval set remains uniform and moderate. It guarantees sufficient node density in areas of drastic track curvature changes to capture minute distortions, while avoiding unnecessary subdivisions in flat areas. This provides a set of geometrically consistent discrete state nodes that adapt to the actual curvature of the track for subsequent local curvature calculations and curvature entropy analysis, thereby improving the sensitivity and reliability of high-impedance grounding fault detection for weak waveform distortions.

[0041] In one embodiment, calculating the comprehensive fault value of the target measurement point based on the curvature entropy values ​​corresponding to all piecewise curves includes: pass Calculate the comprehensive fault value of the target measurement point; where N is the number of piecewise curves. Let be the curvature entropy value of the i-th piecewise curve. The adaptive threshold for the i-th piecewise curve. Let be the curvature entropy value of the j-th piecewise curve. The adaptive threshold for the j-th piecewise curve. Let be the forgetting coefficient, and 0 < 0. <1.

[0042] In one implementation, the curvature entropy value of a single piecewise curve only reflects the uniformity of track curvature distribution within that power frequency cycle. However, the weak distortion characteristics of high-impedance grounding faults often exhibit multi-cycle persistence and may be disturbed by random noise or load fluctuations in individual cycles. To fully utilize the temporal continuity of the fault and suppress accidental disturbances, it is necessary to fuse the curvature entropy values ​​of multiple consecutive piecewise curves. The normalized deviation depth representing the i-th piecewise curve is determined only if the curvature entropy value Below the adaptive threshold It produces a positive contribution if it is positive, otherwise it is zero. This implements an exponential decay mechanism based on the depth of deviation in subsequent cycles, ensuring that the contribution of earlier cycles is factored by each subsequent cycle that still exhibits deviation. The structure weakens gradually, ensuring that when the entropy value is low for multiple consecutive cycles, the early cycles are attenuated but still contribute cumulatively. When there is only a single isolated cycle with a deviation, Z is unlikely to exceed the threshold. It can sensitively reflect the fault characteristics of multiple consecutive cycles, while effectively resisting transient disturbances and achieving highly reliable high-impedance grounding fault detection.

[0043] In one implementation, the adaptive threshold of the piecewise curve is dynamically calculated using an exponentially weighted moving average method based on the curvature entropy value sequence during the historical normal operation period of the measurement point: for the current i-th piecewise curve, its adaptive threshold... From the threshold of the previous time step The mean of recent curvature entropy values The weighted update yields, i.e. ,in This is a smoothing factor, with values ​​ranging from 0.05 to 0.2. The value is the arithmetic mean of the curvature entropy values ​​over the previous few cycles. This update method allows the threshold to be slowly adjusted to follow changes in the normal operating state of the system, avoiding misjudgments caused by environmental or line parameter drift due to a fixed threshold. At the same time, it maintains moderate sensitivity to recent normal fluctuations, ensuring that the curvature entropy value is significantly lower than the current adaptive threshold when a high-impedance grounding fault occurs.

[0044] In one embodiment, determining the fault segment by reversing the fault current direction based on the curvature sequence set includes: For each curvature sequence in the curvature sequence set, after removing the moments when the curvature value is 0, extract the positive and negative signs of all remaining curvature values, arrange them in chronological order, and obtain the curvature polarity spectrum of each measurement point in each preset period. For each measurement point, the symbol values ​​at the same sampling number position in multiple consecutive preset periods are integrated to obtain a symbol set. The number of times the positive and negative signs appear in the symbol set is counted. The symbol with the most occurrences is taken as the fused symbol value at that sampling number position. All the fused symbol values ​​at all sampling number positions are arranged in the order of sampling number to obtain the fused polarity spectrum of that measurement point. The polarity reversal value is obtained by performing difference mapping on the fused polarity spectrum of each pair of adjacent measurement points; Compare the polarity reversal value with a preset second threshold; If the polarity reversal value of adjacent measurement points is lower than the preset second threshold, it is determined that the fault current direction has reversed in the high-voltage cable section between the adjacent measurement points, and the high-voltage cable section is located as the fault section.

[0045] In one implementation, the preset second threshold is set by a technician, specifically 0.6.

[0046] In one implementation, when a high-impedance ground fault occurs, the fault current directions on both sides of the fault point are opposite, resulting in significant differences in the curvature polarity spectra extracted from the measurement points before and after the fault point. In contrast, the polarity spectra on both sides of the non-fault section are basically consistent. To eliminate single-cycle random noise interference, the symbol values ​​at the same sampling sequence position of the same measurement point in multiple consecutive cycles are fused by majority voting to obtain a more reliable fused polarity spectra. The fused polarity spectra of adjacent measurement points are differentially mapped, and the proportion of different symbols at corresponding positions is calculated to obtain a polarity reversal value. This value directly quantifies the degree of difference between the spectra on both sides. When the polarity reversal value exceeds a preset second threshold, it indicates that the current direction has reversed between the adjacent measurement points, thereby accurately locating the fault section. This method can achieve section location by only using the difference comparison of symbol sequences. It is simple to calculate and has strong anti-noise capability, effectively solving the problem that traditional directional protection is difficult to reliably identify faults due to weak current under high-impedance faults.

[0047] In one embodiment, obtaining the polarity reversal value by performing difference mapping on the fused polarity spectrum of each pair of adjacent measurement points includes: pass Calculate the polarity reversal value between adjacent measurement points; where L is the length of the fused polarity map. and These are the sign values ​​of the fused polarity spectra of adjacent measurement points A and B at the kth sampling sequence, with the sign value being either 0 or 1.

[0048] In one implementation, the sign value is 1 when the sign is positive and 0 when the sign is negative.

[0049] In one implementation, the fused polarity spectrum of adjacent measurement points is a binary symbol sequence. The formula can be used to directly calculate the proportion of different symbols at corresponding positions, which can quantify the degree of difference between the two in the simplest and most intuitive way. It can quickly and accurately obtain the polarity reversal value. When the value exceeds the preset threshold, it can be determined that the current direction has reversed between adjacent measurement points, thereby locating the fault section. It has both high real-time performance and reliability.

[0050] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the scope of the claims of this invention.

Claims

1. A method for intelligent early warning of grounding faults in high-voltage cables, characterized in that, The method includes: Voltage and current values ​​are simultaneously collected at each measurement point along the high-voltage cable at a preset fixed sampling rate. For the target measurement point, with voltage as the abscissa and current as the ordinate, the voltage and current values ​​at the same sampling time are used to form a point in a two-dimensional phase space, and the points are connected in time sequence to obtain the original curve of the measurement point; the target measurement point is any one of the measurement points. The original curve is segmented according to a preset period to obtain a set of segmented curves; The comprehensive fault value of the target measurement point is obtained by performing state structure analysis on the piecewise curve set in the two-dimensional phase space. If the overall fault value is less than a preset first threshold, it is determined that a grounding fault has occurred at the location of the target measurement point, and the target measurement point is taken as the fault measurement point. A curvature sequence set is obtained by selecting the curvature sequence corresponding to the fault measurement point and a preset number of adjacent measurement points; The fault section is obtained by determining the direction reversal of the fault current based on the curvature sequence set, and intelligent fault warning is performed based on the fault section.

2. The intelligent early warning method for grounding faults in high-voltage cables according to claim 1, characterized in that, The comprehensive fault value of the target measurement point obtained by performing state structure analysis on the piecewise curve set in the two-dimensional phase space includes: Within the two-dimensional phase space, the target piecewise curve is discretized into a finite number of state nodes to obtain a sequence of state nodes; the target piecewise curve is any one of the piecewise curves in the set of piecewise curves. For each state node in the state node sequence, calculate its local curvature at the corresponding position on the target piecewise curve to obtain a curvature sequence; The curvature entropy value is obtained by statistically analyzing the probability distribution of curvature values ​​in the curvature sequence. The comprehensive fault value of the target measurement point is calculated based on the curvature entropy values ​​corresponding to all piecewise curves.

3. The intelligent early warning method for grounding faults in high-voltage cables according to claim 2, characterized in that, Discretizing the target piecewise curve into a finite number of state nodes within the two-dimensional phase space yields a sequence of state nodes, including: For each sampling point on the target segmented curve, calculate the angle between the tangent to the curve at that point and the horizontal axis to obtain the tangent direction angle sequence; The direction angle change rate sequence is obtained by subtracting the tangent direction angles of two adjacent sampling points in the tangent direction angle sequence and taking the absolute value. Identify the extreme points in the direction angle change rate sequence and sort them according to their positions on the target piecewise curve to obtain a set of candidate segmentation points; The target segmented curve is divided into intervals based on the candidate segmentation point set to obtain an initial interval set; For each initial interval in the initial interval set, calculate the cumulative sum of the rate of change of the direction angle within the initial interval and normalize it to obtain the interval direction change weight set; The target interval set is obtained by filtering and dividing the initial interval set according to the interval direction change weight set; Each interval in the target interval set is defined as a state node, and all state nodes are arranged in order of position to obtain a state node sequence.

4. The intelligent early warning method for grounding faults in high-voltage cables according to claim 3, characterized in that, The target interval set is obtained by filtering and dividing the initial interval set according to the interval direction change weight set, including: Step 1: Initialize an empty target interval set; Step 2: Select the initial intervals whose interval direction change weights are greater than a preset change threshold from the interval direction change weight set as the intervals to be divided to obtain the interval set to be divided; Step 3: Divide the target interval to be divided into equal parts according to the arc length to obtain the first and second equal parts; the target interval to be divided is any one of the intervals to be divided in the set of intervals to be divided. Step 4: Calculate the cumulative sum of the rates of change of the direction angle for the first and second equally divided intervals, respectively; Step 5: If the cumulative sum of the rate of change of the direction angle is greater than the preset percentage threshold of the total cumulative sum of the target intervals to be divided, return to step 2 and add the equal intervals that do not meet the conditions to the set of intervals to be divided; otherwise, add the first equal interval and the second equal interval to the target interval set, and finally obtain the target interval set after division.

5. The intelligent early warning method for grounding faults in high-voltage cables according to claim 2, characterized in that, The comprehensive fault value of the target measurement point is calculated based on the curvature entropy values ​​corresponding to all piecewise curves, including: pass Calculate the comprehensive fault value of the target measurement point; where N is the number of piecewise curves. Let be the curvature entropy value of the i-th piecewise curve. The adaptive threshold for the i-th piecewise curve. Let be the curvature entropy value of the j-th piecewise curve. The adaptive threshold for the j-th piecewise curve. Let be the forgetting coefficient, and 0 < 0. <1.

6. The intelligent early warning method for grounding faults in high-voltage cables according to claim 1, characterized in that, Based on the curvature sequence set, the fault current direction reversal determination results in the following fault segments: For each curvature sequence in the curvature sequence set, after removing the times when the curvature value is 0, the positive and negative signs of all remaining curvature values ​​are extracted and arranged in chronological order to obtain the curvature polarity spectrum of each measurement point in each preset period; For each measurement point, the symbol values ​​at the same sampling number position in multiple consecutive preset periods are integrated to obtain a symbol set. The number of times the positive and negative signs appear in the symbol set is counted. The symbol with the most occurrences is taken as the fused symbol value at that sampling number position. All the fused symbol values ​​at all sampling number positions are arranged in the order of sampling number to obtain the fused polarity spectrum of that measurement point. The polarity reversal value is obtained by performing difference mapping on the fused polarity spectrum of each pair of adjacent measurement points; Compare the polarity reversal value with a preset second threshold; If the polarity reversal value of adjacent measurement points is lower than the preset second threshold, it is determined that the fault current direction has reversed in the high-voltage cable section between the adjacent measurement points, and the high-voltage cable section is located as the fault section.

7. The intelligent early warning method for grounding faults in high-voltage cables according to claim 6, characterized in that, Differential mapping of the fused polarity spectrum of each pair of adjacent measurement points yields polarity reversal values, including: pass Calculate the polarity reversal value between adjacent measurement points; where L is the length of the fused polarity map. and These are the sign values ​​of the fused polarity spectra of adjacent measurement points A and B at the kth sampling sequence, with the sign value being either 0 or 1.

8. A grounding fault intelligent early warning system for high-voltage cables, characterized in that, The system includes: The acquisition module is used to synchronously acquire voltage and current values ​​at a preset fixed sampling rate at various measurement points along the high-voltage cable. The curve generation module is used to construct a point in a two-dimensional phase space for the target measurement point, with voltage as the abscissa and current as the ordinate. The voltage and current values ​​at the same sampling time are used to form a point, and the points are connected in time sequence to obtain the original curve of the measurement point. The target measurement point can be any measurement point among all measurement points. The segmentation module is used to segment the original curve according to a preset period to obtain a segmented curve set; The fault value module is used to perform state structure analysis on the piecewise curve set in the two-dimensional phase space to obtain the comprehensive fault value of the target measurement point; The condition module is used to determine that a high impedance grounding fault has occurred at the location of the target measurement point if the comprehensive fault value is less than a preset first threshold, and to designate the target measurement point as the fault measurement point. The sequence selection module is used to select the curvature sequence corresponding to the fault measurement point and a preset number of adjacent measurement points to obtain a curvature sequence set. The early warning module is used to determine the fault section by reversing the direction of the fault current based on the curvature sequence set, and to provide intelligent early warning of the fault based on the fault section.

Images