A low voltage cable fault monitoring method and system

Abstract

The application provides a low-voltage cable fault monitoring method and system, and relates to the technical field of cable fault monitoring. The method comprises the following steps: dividing a low-voltage cable to be monitored into multiple monitoring sections, and collecting voltage signals of the monitoring sections; performing preliminary fault monitoring on the monitoring sections according to the voltage signals of the monitoring sections, obtaining preliminary fault monitoring results of the monitoring sections, and the preliminary fault monitoring results comprising normal sections, suspected fault sections and fault sections; when a monitoring section is determined as a suspected fault section, calculating a fault risk index of each suspected fault section according to the voltage signals of the suspected fault sections, determining whether the fault risk index is less than a preset fault risk index; if yes, determining that the suspected fault section is a normal section; otherwise, determining that the suspected fault section is a fault section; and performing fine fault monitoring on each fault section through a fine fault monitoring model, and outputting a fault category of each fault section.

Description

Technical Field

[0001] This invention relates to the field of cable fault monitoring technology, and in particular to a method and system for monitoring low-voltage cable faults. Background Technology

[0002] Low-voltage cables operate in complex environments for extended periods, influenced by factors such as load variations, environmental conditions, and equipment aging. Their operational status exhibits significant dynamism and uncertainty, making them prone to various anomalies and malfunctions. Failure to monitor and identify the operational status of low-voltage cables in a timely and accurate manner not only reduces power supply reliability but may also pose safety hazards.

[0003] In existing technologies, traditional low-voltage cable fault monitoring methods are mostly based on overall line analysis or single electrical characteristics for judgment. They usually judge voltage signals by fixed thresholds or simple rules, which makes it difficult to distinguish the operating status of different sections of the cable in a fine-grained manner. They lack the ability to accurately identify local anomalies and are easily affected by signal fluctuations under complex operating conditions, resulting in insufficient accuracy and stability of the preliminary fault monitoring results.

[0004] Furthermore, traditional low-voltage cable fault monitoring methods often lack an effective judgment mechanism to further confirm the abnormal state when further processing monitoring sections initially judged to be abnormal. This makes it difficult to distinguish between transient disturbances and real faults, resulting in unstable judgment results for suspected abnormal sections and thus affecting the accuracy of fault identification. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a low-voltage cable fault monitoring method, which can solve the problems of traditional low-voltage cable fault monitoring methods, which are mostly based on overall line analysis or single electrical characteristics for judgment. They usually judge voltage signals by fixed thresholds or simple rules, which makes it difficult to distinguish the operating status of different sections of the cable in a fine-grained manner, lacks the ability to accurately identify local anomalies, and is easily affected by signal fluctuations under complex operating conditions, resulting in insufficient accuracy and stability of the initial fault monitoring results. Furthermore, when further processing the monitoring section initially judged to be abnormal, there is usually a lack of an effective judgment mechanism to further confirm the abnormal state, making it difficult to distinguish between transient disturbances and real faults, resulting in unstable judgment results of suspected abnormal sections, and thus affecting the accuracy of fault identification.

[0006] A first aspect of this invention provides a low-voltage cable fault monitoring method, comprising: S1: Divide the low-voltage cable to be monitored into multiple monitoring sections and collect the voltage signal of each monitoring section; S2: Based on the voltage signals of each monitoring section, perform preliminary fault monitoring on each monitoring section to obtain the preliminary fault monitoring results of each monitoring section. The preliminary fault monitoring results include normal sections, suspected fault sections, and fault sections. S3: When the monitored section is determined to be a normal section, return to S1; when the monitored section is determined to be a suspected fault section, proceed to S4; when the monitored section is determined to be a fault section, proceed to S6. S4: Calculate the fault risk index of each suspected fault section based on the voltage signal of each suspected fault section. S5: Determine if the fault risk index is less than the preset fault risk index; if so, determine the suspected fault section as a normal section and return to S1; otherwise, determine the suspected fault section as a fault section and proceed to S6. S6: Construct a fine-grained fault monitoring model based on deep learning algorithms; S7: Based on the voltage signal of each fault section, the fault fine monitoring model is used to perform fine fault monitoring on each fault section and output the fault category of each fault section.

[0007] A second aspect of the present invention provides a low-voltage cable fault monitoring system, comprising: a processor and a memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the low-voltage cable fault monitoring method as described in the first aspect.

[0008] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, the low-voltage cable to be monitored is divided into multiple monitoring sections. Instead of relying on overall line analysis or single electrical characteristics for judgment, voltage signals from each monitoring section are collected for preliminary fault monitoring. This eliminates the need for fixed thresholds or simple rules to judge voltage signals, allowing for fine-grained differentiation of the operating status of different cable sections. It possesses the ability to accurately identify local anomalies and is less susceptible to signal fluctuations under complex operating conditions, improving the accuracy and stability of preliminary fault monitoring results. By calculating the fault risk index of each suspected fault section, an effective judgment mechanism is in place to further confirm the abnormal state when further processing monitoring sections initially determined to be abnormal. This distinguishes between transient disturbances and actual faults, making the judgment results of suspected abnormal sections more stable and improving the accuracy of fault identification. Attached Figure Description

[0009] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0010] Figure 1 This is a flowchart illustrating a low-voltage cable fault monitoring method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a low-voltage cable fault monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0012] The low-voltage cable fault monitoring method provided by the present invention will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0013] Reference manual attached Figure 1 The diagram shows a flowchart of a low-voltage cable fault monitoring method provided by an embodiment of the present invention.

[0014] This invention provides a low-voltage cable fault monitoring method, which may include the following steps: S1: Divide the low-voltage cable to be monitored into multiple monitoring sections and collect the voltage signal of each monitoring section.

[0015] In one possible implementation, S1 specifically refers to: The low-voltage cable to be monitored is divided into multiple monitoring sections according to a preset division rule, and the voltage signal of the corresponding monitoring section is periodically collected by voltage signal acquisition devices deployed at the beginning and / or end of each monitoring section.

[0016] Optionally, the preset division rules include at least one or more of the following: division by cable length, division by joint location, division by branch node, and division by laying environment area. When dividing by cable length, the cable is divided into multiple monitoring sections according to preset length intervals. When dividing by joint location or branch node, the cable is divided at each joint or branch as the section boundary. When dividing by laying environment area, the location where environmental properties change is used as the section boundary point, thus obtaining multiple monitoring sections with clear spatial boundaries and attribute characteristics.

[0017] In this embodiment of the invention, the low-voltage cable to be monitored is divided into multiple monitoring sections, and voltage signals are collected from each monitoring section separately. This helps to refine the overall operating state of the entire cable into multiple independently analyzable local section states, thereby improving the spatial resolution of fault monitoring and the accuracy of anomaly identification. Simultaneously, by dividing the sections according to one or more rules based on cable length, joint location, branch nodes, and the laying environment, the monitoring sections can be matched with the actual structural characteristics, connection relationships, and external operating environment of the cable, facilitating a more accurate reflection of operational differences and anomaly characteristics at different locations. Furthermore, deploying voltage signal acquisition devices at the beginning and / or end of each monitoring section and periodically collecting data helps to continuously obtain section-level dynamic voltage information, providing a reliable data foundation for subsequent preliminary fault monitoring, risk assessment, fine-grained fault identification, and fault location, thereby improving the real-time performance, relevance, and accuracy of low-voltage cable fault monitoring.

[0018] S2: Based on the voltage signals of each monitoring section, perform preliminary fault monitoring on each monitoring section to obtain preliminary fault monitoring results for each monitoring section. The preliminary fault monitoring results include normal sections, suspected fault sections, and fault sections.

[0019] In one possible implementation, S2 specifically includes sub-steps S201 to S206: S201: Calculate the lower and upper amplitude thresholds of the voltage signal within the allowable error range, and determine whether the voltage signal is greater than or equal to the lower amplitude threshold and less than or equal to the upper amplitude threshold. If so, determine the monitoring section as a normal section. Otherwise, proceed to S202.

[0020] Optionally, the lower threshold and the upper threshold of amplitude are respectively: Where V1 represents the lower limit threshold of amplitude, and α represents the allowable error. This represents a historical sampling point that is half a cycle out of phase with the current sampling point n. The corresponding voltage value, This indicates taking the absolute value, N represents the total number of sampling points in a complete voltage cycle, and V2 represents the upper limit threshold of the amplitude.

[0021] S202: Select the first half-cycle voltage sequence and the second half-cycle voltage sequence corresponding to the sampling point of the voltage signal, calculate the cross-correlation coefficient between the first half-cycle voltage sequence and the second half-cycle voltage sequence, and determine whether the cross-correlation coefficient is greater than the preset cross-correlation coefficient. If so, determine that the monitoring segment is a normal segment. Otherwise, proceed to S203.

[0022] Optionally, the cross-correlation coefficient is specifically: Where C represents the cross-correlation coefficient, S1 represents the voltage sequence of the first half-cycle, S2 represents the voltage sequence of the second half-cycle, and S1(k) represents the k-th element in the voltage sequence of the first half-cycle. Sk represents the average voltage of the voltage sequence in the first half-cycle, and S2(k) represents the k-th element in the voltage sequence in the second half-cycle. This represents the voltage mean of the voltage sequence in the second half of the cycle. This represents the variance of the absolute value of the voltage in the first half-cycle voltage sequence. This represents the variance of the absolute value of the voltage in the second half-cycle voltage sequence. Indicates the first Voltage values ​​at each sampling point This represents the voltage value at the (n-1)th sampling point. This represents the voltage value at the nth sampling point. This represents the voltage value at the (n+1)th sampling point. Indicates the first Voltage values ​​at each sampling point.

[0023] S203: Using the sampling point of the voltage signal as the starting point, select multiple half-cycle voltage sequences within a preset time period, and perform sine fitting on each half-cycle voltage sequence to obtain the fitted voltage estimate under each half-cycle voltage sequence.

[0024] S204: Calculate the voltage mean of each half-cycle voltage sequence.

[0025] S205: Calculate the goodness of fit of each half-cycle voltage sequence based on the fitted voltage estimate and the voltage mean, and record the goodness of fit with a counter if the goodness of fit is less than or equal to the preset goodness of fit.

[0026] Optionally, the goodness of fit is specifically: in, This represents the goodness of fit of the voltage sequence in the m-th half-cycle. This represents the fitted voltage estimate at the i-th sampling point in the m-th half-cycle voltage sequence. This represents the DC bias of the voltage sequence in the m-th half-cycle. Let cos represent the amplitude of the cosine component of the voltage sequence in the m-th half-cycle, where cos represents the cosine function and ω represents the angular frequency. Let f represent the amplitude of the sinusoidal component of the voltage sequence in the m-th half-cycle, where sin represents the sine function, π represents pi, and f represents the voltage frequency. This represents the average voltage value of the voltage sequence in the m-th half-cycle. This represents the actual voltage value at the i-th sampling point in the m-th half-cycle voltage sequence.

[0027] S206: When the counter's recorded value is greater than the first preset recorded value, the monitoring section is determined to be a normal section. When the counter's recorded value is less than the second preset recorded value, the monitoring section is determined to be a suspected fault section. When the counter's recorded value is less than or equal to the first preset recorded value and greater than or equal to the second preset recorded value, the monitoring section is determined to be a fault section.

[0028] Specifically, in the actual operating environment of low-voltage cables, the voltage waveforms of each monitoring section are affected by multiple factors, resulting in diverse and complex sources of distortion. These include global non-fault disturbances such as inverter modulation ripple, illumination fluctuations, and capacitor switching, as well as local fault factors such as poor cable contact, insulation aging, partial discharge, and initial arcing. Therefore, based on the counter records obtained by sinusoidal fitting of the voltage waveforms of each monitoring section, a graded judgment mechanism for the degree of waveform distortion at the section level is constructed: When the counter record value is greater than the first preset record value, it indicates that the distortion mainly originates from stable harmonic components of power supply-side modulation or grid-connected operation, belonging to the inherent characteristics of the system rather than cable body abnormalities, and is judged as a normal section. When the counter record value is less than the second preset record value, it indicates that the abnormal waveform occurs less frequently and is discretely distributed, belonging to early abnormal signs, and is judged as a suspected fault section. When the counter record value is less than or equal to the first preset record value and greater than or equal to the second preset record value, it indicates that the abnormal waveform appears continuously within multiple half-cycles and the structural consistency has significantly decreased, deviating from the normal operating state, and is judged as a fault section. Those skilled in the art can set the allowable error, preset cross-correlation coefficient, preset duration, preset goodness of fit, first preset record value, and second preset record value according to actual needs; the present invention does not limit these settings.

[0029] In this embodiment of the invention, by sequentially performing amplitude threshold determination, cross-correlation analysis of the preceding and following half-cycles, and sine fit goodness statistics on the voltage signals of each monitoring section, and combining this with a counter to construct a multi-level determination mechanism, it is beneficial to comprehensively evaluate the voltage waveform from multiple dimensions such as amplitude consistency, waveform symmetry, and structural fit degree. This effectively distinguishes between global non-fault disturbances caused by power supply-side modulation and load fluctuations and local anomalies caused by cable defects. Simultaneously, by incorporating the frequency and continuity of waveform distortion into the determination criteria, it achieves graded identification of normal states, early anomalies, and developed faults. This helps reduce false alarm and false negative rates, improves the accuracy and robustness of initial fault monitoring, and provides a reliable screening basis for subsequent fault risk assessment and refined identification.

[0030] In one possible implementation, after S2 and before S3, steps S2A to S2E are also included: S2A: Screen monitoring segments whose preliminary fault monitoring results indicate faulty sections and / or suspected faulty sections as abnormal monitoring segments, and construct an abnormal monitoring segment set.

[0031] S2B: Determine whether the anomaly monitoring segments in the set of anomaly monitoring segments meet the preset association conditions. If so, determine the anomaly monitoring segments as continuous anomaly monitoring segments, and group the continuous anomaly monitoring segments according to spatial continuity to obtain one or more groups of continuous anomaly monitoring segments, then proceed to S2C. Otherwise, determine the anomaly monitoring segments as independent anomaly monitoring segments, then proceed to S2E.

[0032] Optionally, the preset association conditions include at least one or more of the following: the anomaly monitoring segments are physically adjacent, the time difference between the occurrence of anomalies in adjacent anomaly monitoring segments is less than a preset time difference, and the similarity of anomaly features between adjacent anomaly monitoring segments is greater than a preset similarity threshold. The anomaly feature similarity is used to characterize the degree of consistency in the anomaly behavior of adjacent anomaly monitoring segments.

[0033] Specifically, feature vectors describing the abnormal state are extracted from each anomaly monitoring segment. These feature vectors include at least one or more of the following: voltage over-limit amplitude, cross-correlation coefficient, sinusoidal fit goodness of fit, number of distortion half-cycles, and distortion duration. After normalizing the feature vectors, similarity is calculated based on one or more of the following: cosine similarity, reciprocal Euclidean distance, or correlation coefficient. In determining correlation, it is preferable to perform pairwise analysis on spatially adjacent anomaly monitoring segments. When any preset correlation condition is met, adjacent segments are determined to have a correlation and are identified as continuous anomaly monitoring segments; otherwise, they are determined to be independent anomaly monitoring segments. For multiple anomaly monitoring segments with correlation, they are continuously grouped according to their spatial adjacency, dividing spatially continuous and interconnected anomaly monitoring segments into the same continuous anomaly monitoring segment group, thus obtaining one or more continuous anomaly monitoring segment groups.

[0034] S2C: Merge each continuous anomaly monitoring segment in the continuous anomaly monitoring segment group to obtain the merged anomaly monitoring segment.

[0035] S2D: When a merged anomaly monitoring segment contains a faulty segment, the merged anomaly monitoring segment is identified as a faulty segment. When a merged anomaly monitoring segment contains only a suspected faulty segment, the merged anomaly monitoring segment is identified as a suspected faulty segment to correct the initial fault monitoring results.

[0036] It should be noted that during the operation of low-voltage cables, faults typically originate from local defects and gradually expand along conductor connections, insulation defect areas, or adjacent sections, thus spatially manifesting as simultaneous anomalies in multiple consecutive monitoring sections. Furthermore, due to differences in the intensity, duration, and waveform distortion between the fault center section and the affected peripheral sections, the fault center section may be identified as a faulty section, while adjacent sections may only be identified as suspected faulty sections. Therefore, when a merged abnormal monitoring section contains a faulty section, it is classified as a faulty section as a whole. When it only contains suspected faulty sections, it is classified as a suspected faulty section. This method ensures that the merged classification matches the actual fault impact range, avoiding the fragmentation of the same fault into multiple isolated abnormal sections due to overly fine segmentation, and preventing the overclassification of early continuous anomalies as faults. This improves the rationality and accuracy of the initial fault monitoring result correction and provides a reliable basis for subsequent refined identification and fault location.

[0037] S2E: Keep the preliminary fault monitoring results of the independent anomaly monitoring section unchanged.

[0038] In this embodiment of the invention, by introducing a correlation analysis and grouping correction mechanism for abnormal monitoring segments after initial fault monitoring, it is beneficial to comprehensively identify the inherent correlation between abnormal segments from multiple dimensions such as spatial continuity, temporal consistency, and similarity of abnormal features. This effectively integrates multiple adjacent abnormal segments caused by the same fault source, avoiding the fragmentation of the fault into multiple isolated judgment results due to overly fine segmentation. Simultaneously, by merging and uniformly judging continuous abnormal segments, the fault identification results are closer to the spatial expansion characteristics of actual faults, and can distinguish between the fault center and the peripheral influence area. This prevents misjudging local early anomalies as overall faults or underestimating developed faults as discrete anomalies, thereby improving the overall consistency, accuracy, and engineering applicability of the initial fault monitoring results, and providing a more reliable input basis for subsequent risk assessment and refined identification.

[0039] S3: When the monitored section is determined to be a normal section, return to S1. When the monitored section is determined to be a suspected fault section, proceed to S4. When the monitored section is determined to be a fault section, proceed to S6.

[0040] S4: Calculate the fault risk index of each suspected fault section based on the voltage signal of each suspected fault section.

[0041] In one possible implementation, S4 specifically includes sub-steps S401 to S406: S401: Extract the section-level anomaly characterization parameters for each suspected fault section based on the voltage signal.

[0042] Optionally, the segment-level anomaly characterization parameters include voltage over-limit amplitude, cross-correlation coefficient deviation, goodness-of-fit decrease, and the proportion of distorted half-cycles. First, the sampling point range for a single cycle and half-cycle is determined based on the sampling frequency and grid frequency, and the voltage sequence is segmented. Then, the voltage over-limit amplitude is extracted by comparing the amplitude difference between the current sampling point and the corresponding sampling point in the half-cycle. Further, the voltage sequences of the first and second half-cycles are normalized, and the cross-correlation coefficient is calculated, with its deviation from the ideal symmetry value characterizing the waveform symmetry change. Based on this, a sine fit is performed on each half-cycle voltage sequence, the goodness-of-fit is calculated, and its decrease relative to a preset threshold is statistically analyzed to characterize the degree of waveform structural distortion. Simultaneously, within a preset time window, the number of half-cycles with a goodness-of-fit below the threshold is counted, and the ratio is calculated with the total number of half-cycles to obtain the proportion of distorted half-cycles.

[0043] S402: Calculate the basic anomaly intensity value for each suspected faulty section based on the section-level anomaly characterization parameters.

[0044] S403: Obtain the segment attribute information of each suspected faulty segment.

[0045] Optionally, the section attribute information includes commissioning duration, number of joints, laying environment risk level, historical fault frequency, and current carrying capacity. Based on the basic line files, operation management data, and on-site environmental monitoring data corresponding to each suspected fault section, the section attribute information for each suspected fault section is obtained. Specifically, the commissioning time, joint distribution, and historical fault records for each suspected fault section are extracted from the cable operation management system, equipment ledger, or line files, and the commissioning duration, number of joints, and historical fault frequency are determined accordingly. Combining the laying location, laying method, and on-site environmental monitoring results for each suspected fault section, temperature, humidity, water accumulation, and exposure to corrosive media are obtained, and the laying environment risk level is determined according to preset environmental risk classification rules. Simultaneously, the current carrying capacity is calculated based on the real-time current data, rated current carrying capacity, and historical load operation data corresponding to each suspected fault section.

[0046] S404: Calculate the vulnerability coefficient of each suspected faulty section based on the section attribute information.

[0047] S405: Calculate the spatial coupling correction coefficient for each suspected fault section by combining the abnormal propagation relationship between adjacent suspected fault sections.

[0048] S406: The basic anomaly intensity value, the section vulnerability coefficient, and the spatial coupling correction coefficient are fused to obtain the original risk fusion value of each suspected fault section. The original risk fusion value is then normalized to obtain the fault risk index of each suspected fault section.

[0049] Optionally, the original risk fusion value is specifically: Among them, F j Let A represent the fusion value of the j-th suspected fault segment, λ represent the balance coefficient between the anomaly intensity term and the attribute correction term, and A' represent the fusion value of the j-th suspected fault segment. j Let μ1 represent the basic anomaly intensity value of the j-th suspected fault section, μ1 represent the weighting coefficient of the voltage over-limit amplitude, and ΔV represent the voltage over-limit amplitude. j V represents the voltage exceedance of the j-th suspected fault section. ref,j Let μ2 represent the reference voltage amplitude of the j-th suspected fault section, and C represent the weighting coefficient of the deviation of the cross-correlation coefficient. j Let μ represent the cross-correlation coefficient of the j-th suspected fault segment, and μ3 represent the weighting coefficient for the decrease in goodness of fit. M represents the average goodness of fit of the voltage sequences of multiple half-cycles within a preset time period for the j-th suspected fault segment, μ4 represents the weighting coefficient of the proportion of distorted half-cycles, and M represents the average goodness of fit of the j-th suspected fault segment. j K represents the number of half-cycles in the j-th suspected fault segment where the goodness of fit is less than or equal to the preset goodness of fit. j B represents the total number of half-cycles within a preset time period for the j-th suspected faulty segment. j ω1 represents the segment vulnerability coefficient of the j-th suspected faulty segment, and T represents the weighting coefficient of the commissioning duration. j T represents the operational duration of the j-th suspected faulty section. max N represents the maximum commissioning time within the preset reference period, ω2 represents the weighting coefficient for the number of joints, and N represents the maximum commissioning time within the preset reference period. j N represents the number of joints in the j-th suspected faulty section. max This represents the maximum number of joints in all suspected fault sections, ω3 represents the weighting coefficient for the laying environment risk level, and E j ω4 represents the laying environment risk level of the j-th suspected fault section, H represents the weighting coefficient of historical fault frequency, and H represents the risk level of the laying environment. j H represents the historical fault frequency of the j-th suspected faulty section. max This represents the maximum historical fault frequency among all suspected faulty sections, ω5 represents the weighting coefficient of the current carrying rate, and L j L represents the current carrying capacity of the j-th suspected faulty section. max S represents the maximum current carrying capacity under preset operating conditions. j Ω represents the spatial coupling correction coefficient for the j-th suspected faulty section, ρ represents the spatial propagation enhancement coefficient, and Ω represents the spatial propagation enhancement coefficient. j Let η represent the set of suspected fault sections adjacent to the j-th suspected fault section.jp A represents the association weight between the j-th suspected faulty section and the p-th suspected faulty section. p ε represents the basic anomaly intensity value of the p-th suspected fault section, and ε represents the minimum value to prevent the denominator from being zero.

[0050] In this embodiment of the invention, by extracting segment-level anomaly characterization parameters and introducing segment attribute information and spatial coupling relationships between adjacent segments, a multi-dimensional fusion fault risk index is constructed for suspected fault segments. This facilitates a comprehensive assessment of the immediate anomaly characteristics reflected in the voltage signal, the structural vulnerability of the cable segment itself, and the propagation impact of surrounding anomalies, thereby avoiding the one-sidedness caused by relying solely on a single electrical characteristic for judgment. Simultaneously, by weighting, fusing, and normalizing the anomaly intensity, inherent segment risk, and spatial correlation effects, the true risk level of each suspected fault segment can be more accurately characterized. This enables effective differentiation between early hidden faults and transient disturbances, reduces the false positive rate, and improves the stability and reliability of risk assessment, providing a more scientific basis for subsequent refined fault identification and handling decisions.

[0051] S5: Determine if the fault risk index is less than the preset fault risk index. If yes, determine the suspected faulty section as a normal section and return to S1. Otherwise, determine the suspected faulty section as a faulty section and proceed to S6.

[0052] It should be noted that those skilled in the art can set the preset fault risk index according to actual needs, and this invention does not limit this.

[0053] In this embodiment of the invention, by comparing the fault risk index of a suspected faulty section with a preset threshold, further screening and confirmation of suspected anomalies are achieved. This helps to reclassify non-persistent anomalies caused by occasional disturbances as normal sections, thereby reducing the false alarm rate. Simultaneously, timely confirmation of sections with high risk indices as faulty sections helps improve the accuracy of fault determination and the timeliness of response.

[0054] S6: Construct a fine-grained fault monitoring model based on deep learning algorithms.

[0055] Optionally, the deep learning algorithm is specifically: a multi-branch deep neural network, which includes a cascaded input layer, a feature extraction layer, a feature fusion layer, an attention mechanism layer, a fully connected layer, a softmax layer, and an output layer. The feature extraction layer includes parallel time-domain feature extraction branches based on CBP modules, time-frequency domain feature extraction branches based on CBP modules, statistical feature extraction branches based on multilayer perceptron structures, and attribute feature extraction branches based on multilayer perceptron structures. The CBP module includes cascaded convolutional layers, batch normalization layers, and pooling layers.

[0056] In this embodiment of the invention, by constructing a multi-branch deep neural network, time-domain features, time-frequency domain features, statistical features, and segment attribute features are extracted and fused in parallel. This facilitates a comprehensive characterization of the voltage signal features and operational background information of the fault segment from multiple dimensions, thereby improving the completeness and discriminative ability of feature representation. Simultaneously, the introduction of an attention mechanism for adaptive weighting of the fused features highlights key abnormal features, suppresses redundant information, improves the accuracy and robustness of fault identification, and enables refined classification of complex fault types.

[0057] S7: Based on the voltage signal of each fault section, the fault fine monitoring model is used to perform fine fault monitoring on each fault section and output the fault category of each fault section.

[0058] In one possible implementation, S7 specifically includes sub-steps S701 to S709: S701: Based on the voltage signal, generate time-domain image and time-frequency domain image respectively.

[0059] Specifically, the voltage signals collected from each fault section are used as the original input signals. First, the original input signals are length-standardized, truncated or padded to a voltage sequence of a preset length. Then, the standardized voltage sequence is rearranged row-wise into a two-dimensional matrix according to a preset matrix dimension, and the two-dimensional matrix is ​​subjected to grayscale normalization to obtain the time-domain image of the corresponding fault section. Simultaneously, continuous wavelet transform is performed on the original input signals to obtain wavelet coefficients at different time positions and scale parameters. A time-frequency energy matrix is ​​constructed based on the magnitude or energy distribution of the wavelet coefficients. The time-frequency energy matrix is ​​then subjected to size unification and normalization to obtain the time-frequency domain image of the corresponding fault section. The time-domain image is used to characterize the morphological distribution characteristics of the voltage waveform in the time domain, while the time-frequency domain image is used to characterize the energy evolution characteristics of the voltage signal in the time-frequency joint domain.

[0060] S702: Combine the anomaly representation parameters of each segment level to obtain a segment-level anomaly representation parameter vector. Combine the attribute information of each segment to obtain a segment attribute information vector.

[0061] Specifically, the voltage exceedance amplitude, cross-correlation coefficient deviation, goodness-of-fit decrease, and distortion half-cycle ratio extracted from each fault section are combined in a preset order to obtain a section-level anomaly characterization parameter vector. The commissioning duration, number of joints, laying environment risk level, historical fault frequency, and current carrying rate corresponding to each fault section are normalized and then combined in a preset order to obtain a section attribute information vector. The section-level anomaly characterization parameter vector is used to characterize the degree of anomaly and waveform distortion characteristics of the current voltage signal in the fault section, while the section attribute information vector is used to characterize the inherent vulnerability and operational background characteristics of the fault section.

[0062] S703: In the input layer, receive time-domain image, time-frequency domain image, segment-level anomaly representation parameter vector, and segment attribute information vector.

[0063] S704: In the feature extraction layer, features are extracted from the time-domain image through the time-domain feature extraction branch. Features are extracted from the time-frequency domain image through the time-frequency domain feature extraction branch. Features are extracted from the segment-level anomaly representation parameter vector through the statistical feature extraction branch. Features are extracted from the segment attribute information vector through the attribute feature extraction branch.

[0064] Specifically, the time-domain image is input into the time-domain feature extraction branch based on the CBP module. Multiple concatenated convolutional layers, batch normalization layers, and pooling layers are used to extract local texture features, waveform morphology distribution features, and anomalous distortion features from the time-domain image layer by layer, yielding time-domain features. Similarly, the time-frequency domain image is input into the time-frequency domain feature extraction branch based on the CBP module. Multiple concatenated convolutional layers, batch normalization layers, and pooling layers are used to extract energy accumulation features, spectral evolution features, and transient perturbation features from the time-frequency domain image layer by layer, yielding time-frequency domain features. The segment-level anomaly representation parameter vector is input into the statistical feature extraction branch based on a multilayer perceptron structure. At least one fully connected mapping layer and nonlinear activation operation are used to perform deep mapping of anomaly intensity information and waveform distortion information, yielding statistical features. Finally, the segment attribute information vector is input into the attribute feature extraction branch based on a multilayer perceptron structure. At least one fully connected mapping layer and nonlinear activation operation are used to perform deep mapping of segment vulnerability information and operational background information, yielding attribute features.

[0065] S705: In the feature fusion layer, the features extracted from each branch are concatenated and fused to obtain a fused feature vector.

[0066] S706: In the attention mechanism layer, the fused feature vector is weighted to obtain a weighted fused feature vector. in, Indicates the first The weighted fusion feature vector of each fault segment Indicates the first The attention weight vector corresponding to each faulty segment This represents element-wise multiplication. Indicates the first The fused feature vector of each fault segment, where Softmax represents the Softmax activation function, W att b represents the weight matrix of the attention mechanism layer. att This represents the bias vector of the attention mechanism layer.

[0067] S707: In the fully connected layer, the weighted fusion feature vector is nonlinearly mapped to obtain the fault discrimination feature vector.

[0068] S708: In the Softmax layer, the fault discrimination feature vector is classified to obtain the fault category of the fault segment. in, Indicates the first The fault category of each fault section This indicates selecting the predicted probability from all fault categories. The largest category of failure, Indicates the first The predicted probability that a faulty section belongs to the r-th type of fault, where exp represents an exponential function, w r This represents the classification weight vector corresponding to the r-th type of fault, with the subscript T indicating the transpose operation. Indicates the first The fault discrimination feature vector of each faulty section, where R represents the total number of fault categories.

[0069] S709: In the output layer, output the fault category.

[0070] In this embodiment of the invention, by simultaneously constructing voltage signals into time-domain and time-frequency domain images, and combining them with segment-level anomaly characterization parameter vectors and segment attribute information vectors for multi-source input, it is beneficial to comprehensively characterize fault features from multiple dimensions such as waveform morphology, spectral evolution, anomaly degree, and operational background. Simultaneously, through multi-branch feature extraction, feature fusion, and attention-based weighted processing, key discrimination information can be adaptively highlighted and redundant features suppressed, thereby improving the effectiveness of feature representation and discrimination accuracy. Ultimately, this achieves accurate classification of different types of faults, enhancing the reliability and generalization capability of fine-grained fault monitoring.

[0071] In one possible implementation, after S7, step S7A is further included: S7A: Based on the voltage and current signals at the beginning and end of each fault section, the fault point of each fault section is located, and the location of the fault point corresponding to each fault section is obtained.

[0072] In one possible implementation, S7A specifically includes sub-steps S7A1 to S7A6: S7A1: Construct distributed parameter models for each faulty section based on the line parameters of each faulty section.

[0073] Optionally, the line parameters include: resistance per unit length, inductance per unit length, capacitance per unit length, conductance per unit length, segment length, system operating frequency, and conductor structure parameters. Among these, resistance per unit length, inductance per unit length, capacitance per unit length, and conductance per unit length characterize the distributed electrical characteristics of the fault segment; segment length determines the spatial range; system operating frequency determines the angular frequency; and conductor structure parameters reflect differences in line type.

[0074] Specifically, for each faulty section, the resistance, inductance, capacitance, and conductance per unit length of the section are first obtained based on line archive data or operational monitoring data. Combined with the section length and system operating frequency, the corresponding angular frequency is determined. Then, using the section length direction as a spatial variable, the faulty section is divided into multiple micro-units. Each micro-unit is equivalent to a basic electrical unit composed of series resistance and inductance, and parallel conductance and capacitance, thus forming an equivalent transmission structure continuously distributed along the section. Based on this, according to the coupling relationship between voltage and current along the spatial position and time changes in the transmission line, a first-class equation describing the relationship between voltage and current changes with space, and a second-class equation describing the relationship between current and voltage changes with space, are established. These two types of equations uniformly constrain the spatial propagation characteristics of voltage and current, thereby constructing a distributed parameter model for each faulty section.

[0075] S7A2: Solve the distributed parameter model to obtain the wave impedance and propagation constant of each fault section: in, Indicates the first The wave impedance of each fault section, Indicates the first Resistance per unit length of each fault section Represents the imaginary unit. Indicates the angular frequency of the system operation. Indicates the first Inductance per unit length of each fault section, Indicates the first Conductivity per unit length of each fault segment Indicates the first Capacitance per unit length of each fault section Indicates the first The propagation constant of each faulty section.

[0076] S7A3: Based on the voltage and current signals, wave impedance, and propagation constant at the beginning and end of each fault section, construct the two-terminal expression for the fault point voltage.

[0077] Optionally, the two-ended expression for the fault point voltage includes: a fault point voltage expression derived from the starting end and a fault point voltage expression derived from the ending end.

[0078] The specific expression for the fault point voltage derived from the starting end is as follows: in, Indicates the first The voltage value at the fault point in each fault section, where cosh represents the hyperbolic cosine function. Indicates the first The length of each faulty section. Indicates the first The distance from the fault point in each faulty section to the end. Indicates the first The starting voltage signal of each fault segment, where sinh represents a hyperbolic sine function. Indicates the first The starting current signal of each fault section.

[0079] The specific expression for the fault point voltage derived from the end is as follows: in, Indicates the first The voltage signal at the end of each fault section, Indicates the first The end current signal of each fault section.

[0080] S7A4: Construct a fault distance solution model based on the two-terminal expression of the fault point voltage.

[0081] Specifically, the voltage at the same fault point corresponding to both expressions is subjected to consistency constraints. This involves establishing an equivalent correlation between the fault point voltage derived from the starting point and the fault point voltage derived from the ending point, thereby establishing a constraint relationship with the fault point distance as the unknown quantity. Further, the equivalent correlation is organized, unifying and merging the exponential variation terms, hyperbolic variation terms, and voltage and current coefficient terms related to the fault point distance to construct a nonlinear relationship model for the fault point distance. In this process, the combination term formed by the electrical quantities at the starting point and the section length is defined as the first intermediate quantity, and the combination term formed by the electrical quantities at the ending point and the section parameters is defined as the second intermediate quantity. Through the ratio between the first and second intermediate quantities, a mapping model between the fault point distance and the section parameters and electrical quantities is established, thus completing the construction of the fault distance solution model.

[0082] S7A5: Solve the fault distance calculation model to obtain the fault distance values ​​for each fault segment: Among them, tanh -1 Represents the inverse hyperbolic tangent function. Indicates the first The first intermediate quantity of the faulty section Indicates the first The second intermediate quantity of the faulty section.

[0083] S7A6: Calculate the location of the fault point corresponding to each fault section based on the fault distance value.

[0084] Specifically, firstly, a reference point for the fault distance value is determined, and this reference point is associated with the spatial boundary information of the corresponding fault section. When the fault distance value is based on the starting point of the fault section, the fault point is located along the length of the fault section, starting from the starting point and following the distance corresponding to the fault distance value, thus determining the specific location of the fault point within that fault section. When the fault distance value is based on the ending point of the fault section, the fault point is located in reverse, starting from the ending point and following the distance corresponding to the fault distance value, thus obtaining the location of the fault point. Further, the relative position of the fault point within the section is mapped to the absolute spatial position of the fault section within the entire cable line, thereby obtaining the actual location of the fault point corresponding to each fault section within the entire line.

[0085] In this embodiment of the invention, by further locating the fault point based on a distributed parameter model and two-terminal voltage and current signals after fault classification, it is beneficial to fully utilize the propagation characteristics of voltage and current in the transmission line to achieve accurate fault location inversion. Simultaneously, by constructing two-terminal consistency constraints and solving for the fault distance, the positioning error caused by single-terminal measurement can be effectively reduced, improving positioning accuracy and stability. This extends the function from "fault identification" to "fault location," enhancing the practicality and engineering application value of the overall monitoring system.

[0086] Reference manual attached Figure 2 The diagram shows a structural schematic of a low-voltage cable fault monitoring system provided in an embodiment of the present invention.

[0087] This invention provides a low-voltage cable fault monitoring system 20, including: a processor 201 and a memory 202; The memory 202 stores programs or instructions that can run on the processor 201. When the program or instructions are executed by the processor 201, they implement the steps of the low-voltage cable fault monitoring method described above and achieve the same technical effect. To avoid repetition, the present invention will not elaborate further.

[0088] It should be understood that the processor 201 in this embodiment of the invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0089] It should also be understood that the memory 202 in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments 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 skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring low-voltage cable faults, characterized in that, include: S1: Divide the low-voltage cable to be monitored into multiple monitoring sections and collect the voltage signal of each monitoring section; S2: Based on the voltage signal of each monitoring section, perform preliminary fault monitoring on each monitoring section to obtain preliminary fault monitoring results for each monitoring section. The preliminary fault monitoring results include normal sections, suspected fault sections, and fault sections. S3: When the monitored section is determined to be a normal section, return to S1; when the monitored section is determined to be a suspected fault section, proceed to S4; when the monitored section is determined to be a fault section, proceed to S6. S4: Calculate the fault risk index of each of the suspected fault sections based on the voltage signal of each suspected fault section; S5: Determine whether the fault risk index is less than the preset fault risk index; If so, determine that the suspected faulty section is the normal section, and return to S1; Otherwise, the suspected faulty section is determined to be the faulty section, and the process proceeds to S6; S6: Construct a fine-grained fault monitoring model based on deep learning algorithms; S7: Based on the voltage signal of each fault segment, perform fine fault monitoring on each fault segment using the fine fault monitoring model, and output the fault category of each fault segment.

2. The low-voltage cable fault monitoring method according to claim 1, characterized in that, Specifically, S1 is: The low-voltage cable to be monitored is divided into multiple monitoring sections according to a preset division rule, and the voltage signal of the corresponding monitoring section is periodically collected by voltage signal acquisition devices deployed at the beginning and / or end of each monitoring section.

3. The low-voltage cable fault monitoring method according to claim 1, characterized in that, S2 specifically includes: S201: Calculate the lower amplitude threshold and upper amplitude threshold of the voltage signal under the allowable error, and determine whether the voltage signal is greater than or equal to the lower amplitude threshold and less than or equal to the upper amplitude threshold; if so, determine that the monitoring segment is the normal segment; otherwise, proceed to S202. S202: Select the first half-cycle voltage sequence and the second half-cycle voltage sequence corresponding to the sampling point where the voltage signal is located, calculate the cross-correlation coefficient between the first half-cycle voltage sequence and the second half-cycle voltage sequence, and determine whether the cross-correlation coefficient is greater than the preset cross-correlation coefficient; if so, determine that the monitoring segment is the normal segment; otherwise, proceed to S203. S203: Taking the sampling point where the voltage signal is located as the starting point, select multiple half-cycle voltage sequences within a preset time period, and perform sine fitting on each half-cycle voltage sequence to obtain the fitted voltage estimate value under each half-cycle voltage sequence. S204: Calculate the voltage mean of each of the half-cycle voltage sequences; S205: Calculate the goodness of fit of each half-cycle voltage sequence based on the fitted voltage estimate and the voltage mean, and record the goodness of fit with a counter if the goodness of fit is less than or equal to a preset goodness of fit. S206: When the recorded value of the counter is greater than the first preset recorded value, the monitoring segment is determined to be the normal segment; when the recorded value of the counter is less than the second preset recorded value, the monitoring segment is determined to be the suspected fault segment; when the recorded value of the counter is less than or equal to the first preset recorded value and greater than or equal to the second preset recorded value, the monitoring segment is determined to be the fault segment.

4. The low-voltage cable fault monitoring method according to claim 1, characterized in that, After S2 and before S3, it also includes: S2A: Select monitoring segments whose preliminary fault monitoring results are the faulty segment and / or the suspected faulty segment as abnormal monitoring segments, and construct an abnormal monitoring segment set; S2B: Determine whether the abnormal monitoring segments in the abnormal monitoring segment set meet the preset association conditions; if so, determine the abnormal monitoring segment as a continuous abnormal monitoring segment, and group the continuous abnormal monitoring segments according to spatial continuity to obtain one or more continuous abnormal monitoring segment groups, and proceed to S2C; otherwise, determine the abnormal monitoring segment as an independent abnormal monitoring segment, and proceed to S2E. S2C: Merge each continuous anomaly monitoring segment in the continuous anomaly monitoring segment group to obtain a merged anomaly monitoring segment; S2D: When the merged anomaly monitoring segment contains the faulty segment, the merged anomaly monitoring segment is determined to be the faulty segment; when the merged anomaly monitoring segment only contains the suspected faulty segment, the merged anomaly monitoring segment is determined to be the suspected faulty segment, so as to correct the preliminary fault monitoring result. S2E: Keep the preliminary fault monitoring results of the independent anomaly monitoring section unchanged.

5. The low-voltage cable fault monitoring method according to claim 1, characterized in that, S4 specifically includes: S401: Based on the voltage signal, extract the section-level anomaly characterization parameters of each of the suspected fault sections; S402: Calculate the basic anomaly intensity value of each of the suspected fault sections based on the section-level anomaly characterization parameters; S403: Obtain the segment attribute information of each of the suspected faulty segments; S404: Calculate the segment vulnerability coefficient of each of the suspected faulty segments based on the segment attribute information; S405: Calculate the spatial coupling correction coefficient for each of the suspected fault sections by combining the abnormal propagation relationship between adjacent suspected fault sections; S406: The basic anomaly intensity value, the segment vulnerability coefficient, and the spatial coupling correction coefficient are fused to obtain the original risk fusion value of each suspected fault segment, and the original risk fusion value is normalized to obtain the fault risk index of each suspected fault segment.

6. The low-voltage cable fault monitoring method according to claim 1, characterized in that, The deep learning algorithm is specifically a multi-branch deep neural network, which includes a cascaded input layer, a feature extraction layer, a feature fusion layer, an attention mechanism layer, a fully connected layer, a softmax layer, and an output layer. The feature extraction layer includes a parallel temporal feature extraction branch based on a CBP module, a time-frequency domain feature extraction branch based on the CBP module, a statistical feature extraction branch based on a multilayer perceptron structure, and an attribute feature extraction branch based on the multilayer perceptron structure. The CBP module includes a cascaded convolutional layer, a batch normalization layer, and a pooling layer.

7. The low-voltage cable fault monitoring method according to claim 6, characterized in that, Specifically, S7 includes: S701: Based on the voltage signal, generate a time-domain image and a time-frequency domain image respectively; S702: Combine the anomaly representation parameters of each segment level to obtain the segment level anomaly representation parameter vector; combine the attribute information of each segment to obtain the segment attribute information vector. S703: In the input layer, the time-domain image, the time-frequency domain image, the segment-level anomaly representation parameter vector, and the segment attribute information vector are received; S704: In the feature extraction layer, features are extracted from the time-domain image through the time-domain feature extraction branch; features are extracted from the time-frequency domain image through the time-frequency domain feature extraction branch; features are extracted from the segment-level anomaly representation parameter vector through the statistical feature extraction branch; and features are extracted from the segment attribute information vector through the attribute feature extraction branch. S705: In the feature fusion layer, the features extracted from each branch are concatenated and fused to obtain a fused feature vector; S706: In the attention mechanism layer, the fused feature vector is weighted to obtain a weighted fused feature vector; S707: In the fully connected layer, the weighted fusion feature vector is nonlinearly mapped to obtain the fault discrimination feature vector; S708: In the Softmax layer, the fault discrimination feature vector is classified and calculated to obtain the fault category of the fault segment; S709: In the output layer, output the fault category.

8. The low-voltage cable fault monitoring method according to claim 1, characterized in that, Following S7, it also includes: S7A: Based on the voltage and current signals at the beginning and end of each fault section, locate the fault point in each fault section to obtain the location of the fault point corresponding to each fault section.

9. The low-voltage cable fault monitoring method according to claim 8, characterized in that, The S7A specifically includes: S7A1: Construct a distributed parameter model for each of the fault sections based on the line parameters of each fault section; S7A2: Solve the distributed parameter model to obtain the wave impedance and propagation constant of each fault section; S7A3: Based on the voltage and current signals at the beginning and end of each fault section, the wave impedance, and the propagation constant, construct a two-terminal expression for the fault point voltage; S7A4: Construct a fault distance solution model based on the two-terminal expression of the fault point voltage; S7A5: Solve the fault distance solution model to obtain the fault distance value for each fault segment; S7A6: Calculate the location of the fault point corresponding to each fault segment based on the fault distance value.

10. A low-voltage cable fault monitoring system, characterized in that, include: Processor and memory; The memory stores programs or instructions that can run on the processor, which, when executed by the processor, implement the steps of the low-voltage cable fault monitoring method as described in any one of claims 1 to 9.

Images