A method and system for monitoring medium-voltage cable faults

Abstract

This invention relates to the field of online monitoring technology for power cables, specifically a method and system for monitoring faults in medium-voltage cables. The method includes: continuously collecting current, voltage, and vibration signals from operating medium-voltage cables, generating corresponding time-frequency feature matrices, and fusing them into a comprehensive cable condition feature map to initially identify suspected fault points. For each suspected point, a high-precision simulation model is used to perform a physical field simulation, calculate its theoretical parameter distribution, and compare it with the actual features in the map to eliminate false suspected points caused by environmental interference, thus obtaining a set of precisely located fault points. For each fault point, combined with its historical data and cable topology, the risk level of fault propagation and the expected impact range are assessed, and a monitoring report containing location information and tiered handling recommendations is output. This method can effectively distinguish between real faults and environmental interference, and proactively assess fault risks, improving the accuracy of monitoring results and the initiative in operation and maintenance decisions.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for power cables, and in particular to a method and system for monitoring faults in medium-voltage cables. Background Technology

[0002] Current online monitoring technology for medium-voltage cables mainly relies on the acquisition and analysis of single or a few signals such as vibration, current, partial discharge, or temperature. Common practices include setting fixed thresholds or using pattern recognition algorithms to directly determine signal characteristics, thereby locating potential fault points. These methods are essentially direct responses to the surface appearance of the monitored signals. Furthermore, existing fault location reports typically only provide the physical location or type of the fault, lacking a forward-looking assessment of the fault's subsequent development.

[0003] Existing technical solutions have shortcomings. Direct judgment methods based on thresholds or feature matching are highly susceptible to various random interferences in complex underground or utility tunnel environments. The signal characteristics generated by these interferences may resemble those of actual insulation degradation, partial discharge, and other faults, leading to a large number of false alarms. This severely impacts maintenance efficiency and may cause unnecessary power outages for repairs. Furthermore, reports that only provide static fault location information fail to inform maintenance personnel of the urgency, rate of deterioration, and potential impact range of the fault. This results in a lack of prioritization and foresight in maintenance strategy development, hindering the shift from "passive maintenance" to "proactive early warning and precise intervention."

[0004] A technology is needed to verify the authenticity of faults at the physical mechanism level, thereby greatly suppressing false alarms caused by environmental interference. Simultaneously, a technology is also needed to further quantify and assess the dynamic risks and impact range of faults after precise location, in order to support differentiated operation and maintenance decisions. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a method and system for monitoring medium-voltage cable faults.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a medium-voltage cable fault monitoring method, comprising: Continuously collect the mixed signal stream generated during the operation of medium-voltage cables laid underground or in utility tunnels. And generate the time-frequency characteristic matrix of current, the time-frequency characteristic matrix of voltage, and the time-frequency characteristic matrix of vibration; The current time-frequency feature matrix, voltage time-frequency feature matrix, and vibration time-frequency feature matrix are input into the multi-modal feature fusion engine to construct a unified comprehensive feature map of cable status; Based on the comprehensive feature map of the cable status, the fault mode matching network identifies the abnormal feature areas and generates a preliminary set of suspected fault points. For each suspected fault point in the preliminary set of suspected fault points, a high-precision simulation model is used to simulate the physical field of the cable, and the theoretical distribution of the electrical and thermodynamic parameters corresponding to the suspected fault point is calculated. The theoretical distribution of the electrical and thermodynamic parameters is compared and analyzed with the actual characteristics of the corresponding areas in the comprehensive characteristic map of the cable condition to eliminate false suspected points caused by environmental interference and generate a refined set of fault points. For each fault point in the refined set of fault points, the risk level of fault propagation and the expected scope of impact are assessed by combining its historical data with the cable topology. Based on the aforementioned diffusion risk level and expected impact range, a monitoring report containing fault location information and handling recommendations is formulated and output.

[0007] As a further aspect of the present invention, the generation of the current time-frequency characteristic matrix, the voltage time-frequency characteristic matrix, and the vibration time-frequency characteristic matrix includes: The mixed signal stream includes current signals, voltage signals, and vibration signals; The mixed signal stream is subjected to signal type separation and parallel preprocessing operations to obtain preprocessed current signal components, voltage signal components and vibration signal components; Time-frequency domain transformation and feature extraction operations are performed on the preprocessed current signal component, voltage signal component and vibration signal component respectively to generate current time-frequency feature matrix, voltage time-frequency feature matrix and vibration time-frequency feature matrix; The step of performing signal type separation and parallel preprocessing on the mixed signal stream to obtain preprocessed current signal components, voltage signal components, and vibration signal components includes: The mixed signal stream is input into a signal separation device equipped with multiple bandpass filters. The signal separation device separates the mixed signal stream into independent original current signals, original voltage signals, and original vibration signals based on the inherent frequency band range of the current, voltage, and vibration signals. The original current signal is subjected to power frequency component removal and harmonic enhancement processing to eliminate power grid fundamental frequency interference and highlight fault harmonic components, forming a preprocessed current signal component. The original voltage signal is subjected to period normalization and transient event calibration processing to unify the starting point of the voltage waveform period and mark the voltage change time to form a preprocessed voltage signal component. The original vibration signal is subjected to background noise suppression and event trigger level adjustment processing to filter out environmental vibration noise and set an effective vibration event trigger threshold to form a preprocessed vibration signal component.

[0008] As a further aspect of the present invention, the step of performing time-frequency domain transformation and feature extraction operations on the preprocessed current signal component, voltage signal component, and vibration signal component respectively to generate a current time-frequency feature matrix, a voltage time-frequency feature matrix, and a vibration time-frequency feature matrix includes: Short-time Fourier transform is applied to the preprocessed current signal components to obtain the time-frequency spectrum of the current signal, and specific frequency band energy, harmonic distortion rate and phase jump characteristics are extracted from the time-frequency spectrum and combined into a current time-frequency feature matrix. Wavelet packet transform is applied to the preprocessed voltage signal components to decompose the sub-band energy distribution of the voltage signal at different resolutions, and the energy entropy, waveform steepness and zero-crossing interval features of each sub-band are extracted and combined into a voltage time-frequency feature matrix. The Hilbert-Huang transform is applied to the preprocessed vibration signal components to obtain the instantaneous frequency and marginal spectrum of the vibration signal. The instantaneous frequency variance, marginal spectrum peak frequency and energy concentration characteristics are extracted and combined into a vibration time-frequency feature matrix.

[0009] As a further aspect of the present invention, the step of inputting the current time-frequency feature matrix, voltage time-frequency feature matrix, and vibration time-frequency feature matrix into a multi-modal feature fusion engine to construct a unified comprehensive feature map of cable status includes: The multimodal feature fusion engine includes a spatiotemporal registration layer for feature alignment. The spatiotemporal registration layer uses a unified sampling timestamp and cable physical location coordinates as a reference to perform time synchronization and spatial registration of the current time-frequency feature matrix, voltage time-frequency feature matrix, and vibration time-frequency feature matrix. The spatiotemporally registered feature matrix is ​​projected into a shared hidden feature space, and an attention mechanism is used in the hidden feature space to calculate the dynamic weights between current features, voltage features and vibration features. Based on the dynamic weights, features from different modalities are weighted, concatenated, and dimensionality-reduced to generate a set of fused feature vectors; The fused feature vectors are spatially reorganized according to their corresponding cable positions to form a comprehensive feature map of cable status that includes multi-dimensional fused feature vectors and is distributed along the cable length.

[0010] As a further aspect of the present invention, the step of identifying characteristic abnormal regions based on the comprehensive feature map of cable status using a fault mode matching network and generating a preliminary set of suspected fault points includes: The fault mode matching network pre-stores reference feature templates for various typical cable fault modes, including partial discharge features, insulation degradation features, and mechanical damage features. The cable condition integrated feature map is divided into multiple consecutive analysis windows, and the fused feature vector sequence in each analysis window is sequentially subjected to sliding correlation calculation with all the reference feature templates; When the similarity between a certain analysis window and any of the baseline feature templates exceeds a preset matching threshold, the analysis window region is determined to be a feature abnormal region. Record the center coordinates of all the aforementioned abnormal feature regions, and add each center coordinate and its corresponding matching fault mode type as a record to the preliminary set of suspected fault points.

[0011] As a further aspect of the present invention, for each suspected fault point in the preliminary set of suspected fault points, a high-precision simulation model is invoked to perform cable physical field simulation, and the theoretical distribution of the electrical and thermodynamic parameters corresponding to the suspected fault point is calculated, including: Based on the cable's model, structural parameters, and material properties, a high-precision simulation model is constructed, which includes an electromagnetic field simulation module and a thermal field simulation module for the cable. Using the location and fault mode type of the suspected points in the preliminary set of suspected fault points as input conditions, the electromagnetic field simulation module is driven to calculate the theoretical values ​​of electric field intensity distribution and current density distribution along the cable length when the corresponding fault occurs at the suspected point. Simultaneously, the thermal field simulation module is driven to calculate the theoretical value of the temperature field distribution along the cable length direction due to increased losses when a corresponding fault occurs at the suspected point. The calculated theoretical values ​​of electric field intensity distribution, current density distribution, and temperature field distribution are output as the theoretical distributions of electrical and thermodynamic parameters corresponding to the suspected point.

[0012] As a further aspect of the present invention, the theoretical distribution of the electrical and thermodynamic parameters is compared and analyzed with the actual characteristics of the corresponding regions in the comprehensive characteristic map of the cable condition to eliminate false positives caused by environmental interference and generate a refined set of fault points, including: Extract the actual fused feature vector of the region adjacent to the suspected point location from the comprehensive feature map of the cable status; The actual feature components reflecting changes in electric field strength, current density, and temperature are extracted from the actual fused feature vector. The theoretical values ​​in the theoretical distribution of the electrical and thermodynamic parameters are compared point by point with the measured values ​​in the actual characteristic components, and the deviation rate is calculated. If the deviation rate is lower than a preset consistency threshold, the suspected point is determined to be a real fault point; if the deviation rate is higher than the preset consistency threshold, the suspected point is determined to be a false suspected point caused by environmental interference. All suspected fault points identified as actual fault points, including their location coordinates, fault mode type, and deviation rate, are summarized to generate the refined fault point set.

[0013] As a further aspect of the present invention, for each fault point in the refined set of fault points, combined with its historical data and cable topology, the risk level of fault propagation and the expected scope of impact are assessed, including: From the cable historical operation database, retrieve the load rate change curve, insulation aging index and maintenance records of the cable segment where each fault point is located in the refined fault point set over a period of time. From the cable network topology file, obtain the connection relationship of the cable segment where the fault point is located in the power grid, the upstream and downstream load conditions, and whether it is a critical power supply node; Based on the load rate change curve, insulation aging index, maintenance records, connection relationships, upstream and downstream load conditions, and key node information, a risk assessment matrix is ​​used to calculate the probability that the fault point will develop into a permanent fault or trigger a chain of power outages under the current operating conditions, and the probability is mapped to a diffusion risk level. Based on the cable network topology file, after simulating the failure of the fault point, the power outage area is defined as the expected impact range according to the power outage area caused by the power grid protection logic.

[0014] As a further aspect of the present invention, the step of formulating and outputting a monitoring report containing fault location information and handling recommendations based on the diffusion risk level and the expected impact range includes: Based on the location coordinates of each fault point in the refined fault point set, generate cable fault geolocation information and cable well or junction box number information. Based on the spread risk level of each fault point, a corresponding handling suggestion is matched from a preset handling strategy library. The handling suggestions include immediate power outage for maintenance, planned maintenance, or enhanced online monitoring. Based on the estimated impact range of each failure point, generate a list of users and a list of critical facilities within the impact range; The cable fault location information, cable well or junction box number information, matching handling suggestions, user list and key facility list within the affected area are formatted and integrated according to the preset report template to generate the final readable monitoring report file.

[0015] As a further aspect of the present invention, the present invention also includes a medium-voltage cable fault monitoring system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, it implements the steps of the medium-voltage cable fault monitoring method described above.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: For initially identified suspected fault points, a built-in high-precision cable multiphysics simulation model is used to numerically simulate the current, electromagnetic, and temperature fields of each suspected point, calculating the theoretical distribution of the corresponding electrical and thermodynamic parameters. This theoretical distribution is then compared with the actual characteristics of the corresponding area in the real-time monitoring integrated feature map. Only when the two show a clear physical mechanism consistency in the morphology and evolution of the abnormal mode are they confirmed as a real fault. This process effectively filters out signals generated by external electromagnetic interference, mechanical vibration, and other environmental factors that do not possess corresponding internal physical field distortion characteristics, elevating the fault judgment criterion from apparent signal correlation to the consistency of the internal physical mechanism.

[0017] After confirming the actual fault location, the system automatically integrates the historical status data sequence of that point with the topology connection information of the cable network. By analyzing the degradation trend parameters of the historical data, the system predicts the rate of fault development. Combining the topology analysis with the electrical location importance and physical spatial relationships of the fault point, the system simulates and assesses the potential electrical cascading effects and physical impact range caused by the fault. Ultimately, a quantitative assessment of the diffusion risk level and expected impact range is generated, expanding the monitoring results from static location information to decision-making information that includes dynamic risk prediction. This provides a direct basis for developing differentiated maintenance priorities and emergency plans. Attached Figure Description

[0018] Figure 1 This is a flowchart of the medium-voltage cable fault monitoring method described in this invention; Figure 2 This is a flowchart of time-frequency domain transformation and feature extraction; Figure 3 A comparison of the theoretical distribution of electrical parameters for different fault types; Figure 4 The thermodynamic parameter comparison curves for the theoretical and actual values ​​of the partial discharge fault point at 215 meters are shown. Figure 5 This is the time-frequency characteristic matrix of fault monitoring current for medium-voltage cables. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 The system collects mixed signal flows generated by medium-voltage cables laid underground or in utility tunnels during operation, generating time-frequency characteristic matrices for current, voltage, and vibration. These matrices are then input into a multi-modal feature fusion engine to construct a unified comprehensive cable condition feature map. Based on this map, a fault mode matching network identifies anomalous regions and generates a preliminary set of suspected fault points. For each suspected point in this preliminary set, a high-precision simulation model is used to simulate the cable's physical field, calculating the theoretical distribution of electrical and thermodynamic parameters. The theoretical distributions of these parameters are compared with the actual characteristics of the corresponding regions in the comprehensive cable condition feature map to eliminate false suspected points caused by environmental interference, generating a refined set of fault points. For each fault point in this refined set, its historical data and cable topology are used to assess the risk level and expected impact range of the fault. Based on the risk level and expected impact range, a monitoring report containing fault location information and handling recommendations is developed and output.

[0022] In one embodiment of the invention, the mixed signal stream is acquired by a sensor array deployed at the cable joint. The sensor array includes a current transformer, a voltage divider, and a vibration accelerometer. In an example scenario, the frequency band of the current signal is concentrated near the 50 Hz power frequency and its harmonics, the frequency band of the voltage signal covers transient components from the power frequency to several kilohertz, and the frequency band of the vibration signal is distributed in the mechanical vibration frequency band from tens of kilohertz to several kilohertz. Data comparison shows that the original mixed signal stream exhibits a superimposed waveform in the time domain. After signal type separation and parallel preprocessing, each signal component exhibits clear separation characteristics in the frequency domain. In some embodiments, the signal separation device is configured with three independent bandpass filters, corresponding to the inherent frequency bands of the current signal, voltage signal, and vibration signal, respectively. The frequency response function of the bandpass filter is:

[0023]

[0024] in: This represents the frequency response of the k-th bandpass filter. Represents frequency variables. This represents the center frequency of the k-th bandpass filter. It is the imaginary unit. It is the quality factor of the filter; for a current signal bandpass filter, the center frequency is... Set to 50 Hz; for voltage signal bandpass filters, the center frequency is... Set to power frequency and extend to high frequencies; for vibration signal bandpass filters, the center frequency... The typical mechanical vibration frequency is set. After the mixed signal stream is input into the signal separation device, the mixed signal stream is separated into independent original current signal, original voltage signal, and original vibration signal according to these inherent frequency bands. In the data recording of the example scenario, the separated original current signal mainly contains a 50 Hz component, the original voltage signal shows a periodic waveform, and the original vibration signal presents a pulse event against a background of random noise.

[0025] In practical implementation, the raw current signal undergoes power frequency component removal and harmonic enhancement processing. Power frequency component removal is achieved through a notch filter with deep attenuation at 50 Hz. Harmonic enhancement is achieved through a high-pass filter to highlight frequency components higher than the power frequency, thereby eliminating grid fundamental interference and highlighting fault harmonic components, forming the preprocessed current signal component. Data comparison shows that the amplitude of the 50 Hz component in the preprocessed current signal component is reduced, while the relative amplitude of harmonic components at 100 Hz and above is increased. Similarly, the raw voltage signal undergoes period normalization and transient event calibration processing. Period normalization uses a phase-locked loop circuit to lock the zero-crossing point of the voltage waveform, unifying the starting point of the voltage waveform period. Transient event calibration marks voltage abrupt changes by monitoring the voltage change rate exceeding a preset threshold, forming the preprocessed voltage signal component. In the example scenario, the preset threshold is set to 5 volts per millisecond, and voltage abrupt changes are recorded as a timestamp sequence. Optionally, background noise suppression and event trigger level adjustment are performed on the original vibration signal. Background noise suppression uses an adaptive filter to filter out environmental vibration noise. Event trigger level adjustment sets an effective vibration event trigger threshold based on the historical statistical characteristics of the vibration signal, forming a preprocessed vibration signal component. The historical statistical characteristics include the average amplitude and standard deviation of the vibration signal. The trigger threshold is set to the average value plus three times the standard deviation.

[0026] In some embodiments, the preprocessed current signal components, voltage signal components, and vibration signal components are stored as digital sequences for subsequent time-frequency domain transformation and feature extraction operations. The sampling rate of the digital sequences is consistent with that of the original mixed signal stream, for example, 10,000 samples per second. In a specific implementation, time-frequency domain transformation and feature extraction operations are performed on the preprocessed current signal components, voltage signal components, and vibration signal components respectively to generate current time-frequency feature matrices, voltage time-frequency feature matrices, and vibration time-frequency feature matrices. For the preprocessed current signal components, a short-time Fourier transform is applied to obtain the time-frequency spectrum of the current signal, and specific frequency band energy, harmonic distortion rate, and phase jump features are extracted from the time-frequency spectrum and combined to form the current time-frequency feature matrix. The specific frequency band energy refers to the integrated energy in the 100 Hz to 1000 Hz frequency band. It is understandable that for the preprocessed voltage signal component, wavelet packet transform is applied to decompose the voltage signal into sub-band energy distributions at different resolutions, and the energy entropy, waveform steepness, and zero-crossing interval features of each sub-band are extracted and combined into a voltage time-frequency feature matrix. The sub-band energy distribution is calculated based on the Daubechies wavelet basis function. Optionally, for the preprocessed vibration signal component, Hilbert-Huang transform is applied to obtain the instantaneous frequency and marginal spectrum of the vibration signal, and the instantaneous frequency variance, marginal spectrum peak frequency, and energy concentration features are extracted and combined into a vibration time-frequency feature matrix. The energy concentration is defined as the proportion of energy within a 10 Hz bandwidth around the peak frequency in the marginal spectrum. In specific implementations, the current time-frequency feature matrix, voltage time-frequency feature matrix, and vibration time-frequency feature matrix are stored in the form of two-dimensional arrays, where rows represent time points and columns represent feature dimensions, used as input to the multimodal feature fusion engine. In the data records of the example scenario, each feature matrix contains 1000 time points and 6 feature dimensions.

[0027] See Figure 2In one embodiment of the present invention, a short-time Fourier transform is applied to the preprocessed current signal components. The short-time Fourier transform uses a Hanning window with a length of 256 sampling points. The window function slides at intervals of 128 sampling points to obtain the time-frequency spectrum of the current signal. Specific frequency band energy, harmonic distortion rate, and phase jump characteristics are extracted from the time-frequency spectrum and combined to form a current time-frequency feature matrix. In the example scenario, the specific frequency band energy refers to the energy value in the 100 Hz to 1500 Hz frequency band calculated by integration from the time-frequency spectrum. The harmonic distortion rate is obtained by calculating the sum of the ratios of the amplitudes of the 3rd, 5th, and 7th harmonics to the fundamental amplitude. The phase jump characteristics are obtained by analyzing the number of times the change in the fundamental phase angle of the current signal exceeds 10 degrees in a continuous period. Data comparison shows that the harmonic distortion rate of the current time-frequency feature matrix is ​​less than 0.05 during normal operation, while this value may rise to above 0.15 when there is a potential fault. In some embodiments, wavelet packet transform is applied to the preprocessed voltage signal components. The wavelet packet transform employs a 4-level decomposition using the db4 wavelet basis function to obtain the sub-band energy distribution of the voltage signal at different resolutions. The energy entropy, waveform steepness, and zero-crossing interval features of each sub-band are extracted and combined into a voltage time-frequency feature matrix. The energy entropy is calculated based on the probability of the sub-band energy distribution. The waveform steepness is defined as the maximum rate of change of the sub-band signal waveform. The zero-crossing interval is the average time interval between adjacent zero-crossing points of the sub-band signal. Similarly, Hilbert-Huang transform is applied to the preprocessed vibration signal components. The Hilbert-Huang transform first decomposes the vibration signal into multiple intrinsic mode functions (EMFs) through empirical mode decomposition. Then, a Hilbert transform is performed on each EMF to obtain the instantaneous frequency and marginal spectrum of the vibration signal. The instantaneous frequency variance, marginal spectrum peak frequency, and energy concentration features are extracted and combined into a vibration time-frequency feature matrix. The energy concentration is obtained by calculating the percentage of the amplitude corresponding to the peak frequency in the marginal spectrum relative to the total amplitude of the entire spectrum.

[0028] In its implementation, the multimodal feature fusion engine includes a spatiotemporal registration layer for feature alignment. This layer uses a unified sampling timestamp and cable physical location coordinates as a reference to perform time synchronization and spatial registration on the current, voltage, and vibration time-frequency feature matrices. Time synchronization aligns data streams from different sensors to the same millisecond-level time reference. Spatial registration maps the feature matrices to a unified cable length coordinate system based on the physical location of the sensors. The zero point of the coordinate system is defined as the cable's starting point. In the example scenario, five sets of sensors are deployed on a 500-meter-long cable. The features collected by each set of sensors are registered to their corresponding coordinate points at 100-meter intervals. Essentially, the spatiotemporally registered feature matrices are projected into a shared hidden feature space. This projection operation is implemented through a fully connected neural network layer. The weight matrix of this layer is learned during the training phase. An attention mechanism is used within the hidden feature space to calculate the dynamic weights between the current, voltage, and vibration features. The dynamic weights are calculated as follows:

[0029]

[0030] in: This represents the dynamic weight scalar of the m-th modal feature. Let represent the feature vector of the m-th mode projected into the hidden layer feature space. It is a trainable weight matrix. It is a trainable bias vector. It is a trainable weight vector. This indicates that the total number of modes is 3. It is the hyperbolic tangent activation function. Optionally, features from different modes can be weighted, concatenated, and dimensionality-reduced based on dynamic weights. The weighted concatenation multiplies the current feature vector, voltage feature vector, and vibration feature vector by their respective dynamic weights. , , Then, they are connected to form an extended fusion feature vector. Dimensionality reduction fusion is performed by another fully connected neural network layer to compress the extended fusion feature vector to a preset dimension, generating a set of fusion feature vectors with uniform dimensions.

[0031] In some embodiments, the fused feature vectors are spatially reorganized according to their corresponding cable locations. This spatial reorganization is based on the coordinate mapping relationship established by the spatiotemporal registration layer. The fused feature vectors at each spatial coordinate point are arranged in ascending order of cable length, forming a comprehensive cable state feature map distributed along the cable length direction, containing multi-dimensional fused feature vectors. The comprehensive cable state feature map is structurally a two-dimensional matrix. The row indices of the matrix correspond to the cable length coordinates, and the columns correspond to the dimensions of the fused feature vectors. In the example scenario, for a 500-meter cable with a resolution of 1 meter, the comprehensive cable state feature map contains 500 rows, each row being an 8-dimensional fused feature vector. In specific implementations, the comprehensive cable state feature map, as a unified feature representation, is fed into the subsequent fault mode matching network for analysis. The construction of the comprehensive cable state feature map completes the transformation from multi-source heterogeneous signals to a unified representation of cable state.

[0032] In one embodiment of the present invention, the fault mode matching network pre-stores reference feature templates for various typical cable fault modes. These reference feature templates include partial discharge features, insulation degradation features, and mechanical damage features. These templates are obtained offline through feature extraction and clustering analysis of historical fault case data. Each reference feature template is a sequence of feature vectors representing the typical manifestation of the corresponding fault mode in the comprehensive cable condition feature map. For example, the partial discharge feature template exhibits a pattern of periodic abrupt changes in high-frequency components, while the insulation degradation feature template shows a trend of slow growth in low-frequency energy. In some embodiments, the comprehensive cable condition feature map is divided into multiple consecutive analysis windows. The length of each analysis window is set to 10 meters, corresponding to the physical length of the cable. In an example scenario, a 1000-meter cable is divided into 100 consecutive analysis windows. Each analysis window contains 10 fused feature vectors at consecutive spatial locations. The fused feature vector sequence within each analysis window is sequentially subjected to sliding correlation calculation with all reference feature templates. The sliding correlation calculation measures the similarity between the feature sequence within the analysis window and the reference template sequence. The calculation formula is as follows:

[0033]

[0034] in: This represents the similarity coefficient between the i-th analysis window and the k-th baseline feature template. This represents the fused feature vector of the p-th location point in the i-th analysis window. Let represent the mean vector of all fused feature vectors within the i-th analysis window. This represents the feature vector of the p-th position point in the k-th reference feature template. This represents the mean vector of the k-th baseline feature template. This indicates the total number of location points within each analysis window, which is 10 in this case. Represents the vector dot product, i.e. The sum of the products of the corresponding components of two vectors. The L2 norm of a vector, i.e., the Euclidean length of the vector, is defined for a given vector. Its L2 norm is the square root of the sum of squares of its components. When the similarity between an analysis window and any baseline feature template exceeds a preset matching threshold, the analysis window region is determined to be a feature anomaly region. The preset matching threshold is set to 0.85. Data comparison shows that the similarity coefficient between the analysis window in the normal operating area of ​​the cable and all baseline feature templates is usually below 0.3, while the similarity coefficient in areas with potential faults may exceed 0.9. It can be understood that the center coordinates of all feature anomaly regions are recorded. The center coordinates are obtained by calculating the midpoint coordinates of the analysis window corresponding to the feature anomaly region along the cable length direction. Each center coordinate and its corresponding matching fault mode type are added as a record to the preliminary set of suspected fault points. The preliminary set of suspected fault points is a list containing location, fault mode type, and similarity coefficient.

[0035] In practical implementation, a high-precision simulation model is constructed based on the cable's model, structural parameters, and material properties. The cable model determines geometric parameters such as conductor cross-sectional area and insulation layer thickness. Structural parameters include the cable's layered structure information, and material properties include conductor resistivity, insulation dielectric constant, and thermal conductivity. The high-precision simulation model includes an electromagnetic field simulation module and a thermal field simulation module. The electromagnetic field simulation module solves Maxwell's equations using the finite element method, while the thermal field simulation module solves the heat conduction equation using the finite volume method. In some embodiments, the location of suspected fault points and the fault mode type in a preliminary set of suspected fault points are used as input conditions to drive the electromagnetic field simulation module to calculate the theoretical values ​​of electric field intensity distribution and current density distribution along the cable length when a corresponding fault occurs at a suspected fault point. For example, for a partial discharge fault mode, the input conditions include setting a small insulation air gap defect model at the suspected fault point location. The electromagnetic field simulation module calculates the electric field redistribution of the entire cable model under the operating voltage and outputs the electric field intensity distribution curve and current density distribution curve within a 20-meter range before and after the suspected fault point. Optionally, the thermal simulation module can be driven to calculate the theoretical temperature field distribution along the cable length due to increased losses when a corresponding fault occurs at a suspected point. For partial discharge fault modes, the thermal simulation module uses the local loss density calculated by the electromagnetic field simulation module as a heat source, and combines it with the environmental temperature boundary conditions of the soil or pipe gallery around the cable to solve the steady-state temperature field, outputting the temperature distribution curve within the same length range. In specific implementation, the calculated theoretical values ​​of electric field strength distribution, current density distribution, and temperature field distribution are used as the theoretical distribution of electrical and thermodynamic parameters corresponding to the suspected point for output. The theoretical distribution is stored in the form of a data file, containing a sequence of electric field strength values, current density values, and temperature values ​​along the cable length coordinate sequence. In the example scenario, the coordinate sequence is spaced at 0.1-meter intervals, covering a 40-meter range around the suspected point. It can be understood that the construction and use of the high-precision simulation model are performed independently for each suspected point in the initial set of suspected fault points. The simulation calculation for each suspected point uses its specific location and fault mode as input to generate corresponding theoretical distribution data for subsequent comparative analysis.

[0036] See Figure 3This study presents the theoretical temperature field distribution characteristics along the cable length under three typical fault modes in medium-voltage cable fault monitoring: partial discharge, insulation degradation, and mechanical damage. The fault location is taken as the origin (0 meters), and cable sections 20 meters before and after the fault point are selected as the analysis domain. Temperature is the core thermodynamic parameter, expressed in degrees Celsius (°C). The temperature distribution under all three fault modes follows the pattern of "the highest temperature peak at the fault point, gradually decreasing towards both sides to the ambient reference temperature," and the attenuation process conforms to the steady-state solution characteristics of the heat conduction equation. The ambient reference temperature is approximately 25°C. The specific distribution characteristics differ significantly. Mechanical damage faults have the highest temperature peak, exceeding 34℃, and their temperature field distribution curve is the steepest, indicating that this fault causes the highest local loss density and the strongest heat concentration effect, with a significant temperature increase only within ±5 meters of the fault point. Partial discharge faults have the second highest temperature peak, at approximately 33℃, and their temperature field influence range is slightly larger than that of mechanical damage, maintaining a relatively high temperature within ±8 meters of the fault point. Insulation degradation faults have the lowest temperature peak, at approximately 31℃, but their temperature field distribution curve is the flattest, with the slowest decay rate and the widest influence range. Within ±15 meters of the fault point, the temperature is still higher than the environmental benchmark, reflecting that insulation degradation is a progressive fault with slow and diffuse loss growth characteristics.

[0037] In one embodiment of the present invention, actual fused feature vectors of the region adjacent to the suspected point location are extracted from the cable condition comprehensive feature map. The adjacent region is defined as a range of 10 meters before and after the suspected point's coordinates, corresponding to 21 consecutive rows of data in the cable condition comprehensive feature map. In the example scenario, for a suspected point located at 215 meters along the cable length, the extracted actual fused feature vector covers all fused feature vectors between 205 meters and 225 meters. Actual feature components reflecting electric field strength, current density, and temperature changes are parsed from the actual fused feature vectors. These actual feature components are encoded during the construction stage of the fused feature vector; for example, the third dimension of the actual fused feature vector corresponds to the electric field strength feature, the fifth dimension to the current density feature, and the seventh dimension to the temperature change feature. The values ​​of these feature components originate from the mapping of the original signal by the multimodal feature fusion engine. The theoretical values ​​of the electrical and thermodynamic parameters are compared point-by-point with the measured values ​​of the actual characteristic components. The theoretical values ​​are derived from the electric field intensity, current density, and temperature sequences along the cable length coordinate sequence output by the high-precision simulation model. The measured values ​​are derived from the characteristic component values ​​resolved at the corresponding coordinate positions of the actual fused characteristic vector sequence. The comparison is performed on the theoretical and measured values ​​for the same coordinate point along the same length, and the deviation rate is calculated. The deviation rate is calculated as follows:

[0038]

[0039] in: Represented in cable length coordinates Overall deviation rate at the location and Representing coordinates The theoretical value and the analytical value of the actual characteristic components of the electric field strength at that location. and Representing coordinates The theoretical value and the analytical value of the actual characteristic component of the current density at the location, and Representing coordinates The theoretical and actual characteristic component analytical values ​​of the temperature were compared using a formula to calculate the normalized Euclidean distance, which measures the overall deviation between the theoretical distribution and the actual characteristics in terms of multidimensional parameters. If the deviation rate is lower than a preset consistency threshold, the suspected point is determined to be a real fault point; if the deviation rate is higher than the preset consistency threshold, the suspected point is determined to be a false suspected point caused by environmental interference. The preset consistency threshold is set to 0.25. See Table 1, which shows the analysis results of suspected points at three different locations.

[0040] Table 1: Suspected Point Theory - Actual Feature Comparison and Judgment Results

[0041] It is understandable that the average deviation rate in Table 1... It is the deviation rate of all coordinate points in the adjacent area of ​​the suspected point. The arithmetic mean of the values ​​was used. The average deviation rate at the suspected point coordinates of 478 meters exceeded the consistency threshold of 0.25, and it was judged as a false suspected point and excluded. Information on all suspected points judged as real fault points, including their location coordinates, fault mode type, and deviation rate, was compiled to generate a refined set of fault points. The refined set of fault points is a subset of the initial set of suspected fault points and only contains entries that have passed the theoretical-actual comparison.

[0042] In some embodiments, the load rate variation curve, insulation aging index, and maintenance records of each cable segment containing a fault point are retrieved from a refined set of fault points in the cable historical operation database. The past period is defined as the last six months. The load rate variation curve is stored in a format with 96 sampling points per day. The insulation aging index is a scalar value calculated through periodic offline dielectric response testing. The maintenance records contain a summary of the time, type, and content of each maintenance performed on the cable segment. From the cable network topology file, the connection relationships of the cable segment containing the fault point in the power grid, upstream and downstream load conditions, and whether it is a critical power supply node are obtained. The connection relationships describe the feeder to which the cable segment belongs and its connected substations and downstream distribution transformers. The upstream and downstream load conditions record the peak current and total capacity of the load supplied by the cable segment. The critical power supply node flag indicates whether the cable segment is connected to uninterruptible loads such as hospitals and data centers. By combining load rate change curves, insulation aging index, maintenance records, connection relationships, upstream and downstream load conditions, and key node information, a risk assessment matrix is ​​used to calculate the probability that a fault point will develop into a permanent fault or trigger a chain of power outages under the current operating conditions. The probability is then mapped to a diffusion risk level. The risk assessment matrix is ​​a pre-set lookup table. The input includes discretized conditions such as the number of hours with a load rate exceeding 90%, the recent growth rate of the insulation aging index, whether there are recent maintenance records, and whether key nodes are connected. The output is a probability value, which is then mapped to a three-level diffusion risk level: "high", "medium", and "low".

[0043] In practical implementation, based on the cable network topology file, after simulating the failure of a fault point, the outage area caused by the power grid protection logic is defined as the expected impact range. The simulation process uses a breadth-first search algorithm, starting from the cable segment where the fault point is located, traversing the topology network until encountering the boundaries set by protection devices such as circuit breakers and sectionalizing switches. The set of all downstream power supply nodes traversed is the expected impact range. In the example scenario, a partial discharge fault point located at a non-critical node with a stable load history and recently repaired has a probability of developing into a permanent fault of 0.15 calculated by the risk assessment matrix, and is mapped to a "low" diffusion risk level. Its expected impact range simulation result is the block supplied by the two downstream distribution transformers. Optionally, another mechanical damage fault point located at a critical power supply node with a high load rate and continuously increasing insulation aging index has a probability of triggering a cascading power outage accident of 0.65 calculated by the risk assessment matrix, and is mapped to a "high" diffusion risk level. Its expected impact range simulation result includes not only the direct downstream load but also some adjacent feeder loads that need to be transferred through this node.

[0044] See Figure 4This study presents a comparison between theoretical and actual temperature values ​​along the cable length (205 meters to 225 meters) at the 215-meter partial discharge fault point stage of a medium-voltage cable. It serves as a core visualization tool for verifying thermodynamic parameters in fault point authenticity determination. The vertical axis represents temperature (°C), reflecting the thermodynamic state of the cable operation; the horizontal axis represents cable length (meters), centered on the suspected fault point at 215 meters, covering a 10-meter feature analysis area before and after it, consistent with the spatial range of the actual feature components extracted after multi-modal feature fusion. On the curve level, the solid line represents the theoretical temperature distribution value calculated by the high-precision simulation model. Starting at 60°C at 205 meters, it smoothly increases as the coordinate approaches the fault core area at 215 meters, reaching a peak of 65°C near 215 meters, and then gradually decreases as the coordinate moves away from the fault area, forming a symmetrical and continuous single-peak curve.

[0045] The dashed line represents the actual temperature characteristic component value analyzed from the comprehensive characteristic spectrum of cable condition. Its overall trend is highly consistent with the theoretical value, and it also forms a peak in the 215-meter region. However, it shows different characteristics in terms of value and fluctuation: In the range of 205 meters to 212.5 meters, the actual value is lower than the theoretical value, and the lowest value occurs at 205 meters (about 59.2℃); In the range of 212.5 meters to 220 meters, the actual value is generally higher than the theoretical value, and the peak occurs near 213.5 meters to 214 meters (over 65.5℃), and it shows a small fluctuation in the peak range, reflecting the impact of environmental interference and signal noise on the monitoring data in actual operation; After 220 meters, the actual value and the theoretical value tend to be consistent again, and gradually decrease synchronously. The core value of the comparison curve lies in its quantitative support for fault point determination: by calculating the deviation rate of temperature dimension point by point in the 205-225 meter range, and combining the theoretical-actual value deviation of electric field strength and current density, the comprehensive average deviation rate of the suspected point at 215 meters (0.12) was finally obtained, which is lower than the preset consistency threshold (0.25), verifying that the point is a real partial discharge fault point and eliminating the possibility of a false suspected point. At the same time, the deviation distribution characteristics of the curve also provide measured data basis for optimizing the parameters of the high-precision simulation model (such as heat loss coefficient and environmental thermal resistance), realizing closed-loop verification between theoretical simulation and actual monitoring.

[0046] In one embodiment of the present invention, based on the location coordinates of each fault point in the refined fault point set, cable fault geographic location information and cable well or junction box number information are generated. The location coordinates are derived from the spatial registration of the cable status comprehensive feature map, with the cable starting end as the origin and the length coordinates in meters. The geographic location information is obtained by querying the geographic information system layer of the cable laying path, converting the length coordinates into latitude and longitude coordinates and street address descriptions. The cable well or junction box number information is obtained by querying the cable line ledger to determine the location of the fault point or the nearest maintainable facility number. In the example scenario, a fault point located at a length coordinate of 215 meters has its geographic location information converted into "longitude X.XXXX, latitude Y.YYYY, located under the sidewalk on the southeast side of the intersection of XX Road and YY Road", and its nearest junction box number is determined to be "JHB-12-07".

[0047] In some embodiments, based on the spread risk level of each fault point, corresponding handling suggestions are matched from a preset handling strategy library. The handling strategy library is a predefined mapping table that maps the three spread risk levels of "high," "medium," and "low" to three types of handling suggestions: "immediate power outage for maintenance," "planned maintenance," and "enhanced online monitoring," respectively. For a fault point with a spread risk level of "high" in the example scenario, the system automatically matches and outputs the handling suggestion of "immediate power outage for maintenance." This suggestion includes standardized operation instructions, such as "apply for a power outage window, organize a maintenance team, and prepare insulation testers and fault location instruments." It can be understood that, based on the expected impact range of each fault point, a user list and a critical facility list within the impact range are generated. The expected impact range is a set consisting of distribution transformer numbers or power supply area codes. The user list is generated by querying all user files powered by these transformers or power supply areas in the customer relationship management system, including user account numbers, names, and contact information. The critical facility list is generated by querying a critical user identifier database, filtering out the names and addresses of important facilities such as hospitals, transportation hubs, and data centers within the expected impact range.

[0048] In practice, the geographic location information of cable faults, cable well or junction box number information, matching handling suggestions, and lists of users and critical facilities within the affected area are formatted and integrated according to a preset report template. The report template is a structured document framework containing fixed chapters such as fault summary, fault details, handling suggestions, and impact analysis. The formatting and integration process automatically fills the corresponding fields of the template with the above information, generating a final readable monitoring report file. The fault summary chapter presents an overview of all refined fault points and their risk levels; the fault details chapter lists the location coordinates, geographic location information, number information, fault mode type, and deviation rate for each fault point; the handling suggestions chapter lists specific suggestions and operational points for each fault point; and the impact analysis chapter lists the affected transformer numbers, number of users, and critical facility lists for each fault point. In the example scenario, the generated monitoring report file shows that for a high-risk partial discharge fault at 215 meters, the report clearly indicates that the "JHB-12-07" junction box needs to be repaired, and lists the 85 households and 1 community health service center served by the two affected transformers. Optionally, the final readable monitoring report file can be in Portable Document Format or Extensible Markup Language Format, which is convenient for sending to operation and maintenance management personnel via email or messaging platform. The naming rules of the report file include the line number and the generation timestamp.

[0049] See Figure 5 In the feature extraction stage of medium-voltage cable fault monitoring, the construction of the current time-frequency feature matrix relies on short-time Fourier transform technology. Specifically, after power frequency component removal and harmonic enhancement preprocessing, the original current signal components are mapped into a two-dimensional time-frequency feature space with time as the vertical axis and frequency as the horizontal axis. The color coding of this matrix represents the feature intensity value in the time-frequency domain; the gradient from blue to red corresponds to a continuous change in feature intensity from 0 to 1.2. The red area indicates that the harmonic distortion rate or specific frequency band energy at that time-frequency point is significantly higher than the background level, directly reflecting potential fault characteristics. During matrix generation, the time dimension is divided into analysis windows with fixed sampling intervals, and the frequency dimension covers the typical distribution range of fault harmonics (0~18Hz). By extracting the energy, harmonic distortion rate, and phase jump characteristics within each time-frequency window, a standardized current time-frequency feature matrix is ​​finally formed, providing highly recognizable current mode input for subsequent multi-modal feature fusion.

[0050] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for monitoring faults in medium-voltage cables, characterized in that, The method includes: Continuously collect the mixed signal stream generated during the operation of medium-voltage cables laid underground or in utility tunnels. And generate the time-frequency characteristic matrix of current, the time-frequency characteristic matrix of voltage, and the time-frequency characteristic matrix of vibration; The current time-frequency feature matrix, voltage time-frequency feature matrix, and vibration time-frequency feature matrix are input into the multi-modal feature fusion engine to construct a unified comprehensive feature map of cable status; Based on the comprehensive feature map of the cable status, the fault mode matching network identifies the abnormal feature areas and generates a preliminary set of suspected fault points. For each suspected fault point in the preliminary set of suspected fault points, a high-precision simulation model is used to simulate the physical field of the cable, and the theoretical distribution of the electrical and thermodynamic parameters corresponding to the suspected fault point is calculated. The theoretical distribution of the electrical and thermodynamic parameters is compared and analyzed with the actual characteristics of the corresponding areas in the comprehensive characteristic map of the cable condition to eliminate false suspected points caused by environmental interference and generate a refined set of fault points. For each fault point in the refined set of fault points, the risk level of fault propagation and the expected scope of impact are assessed by combining its historical data with the cable topology. Based on the aforementioned diffusion risk level and expected impact range, a monitoring report containing fault location information and handling recommendations is formulated and output.

2. The method for monitoring medium-voltage cable faults according to claim 1, characterized in that, The generated current time-frequency characteristic matrix, voltage time-frequency characteristic matrix, and vibration time-frequency characteristic matrix include: The mixed signal stream includes current signals, voltage signals, and vibration signals; The mixed signal stream is subjected to signal type separation and parallel preprocessing operations to obtain preprocessed current signal components, voltage signal components and vibration signal components; Time-frequency domain transformation and feature extraction operations are performed on the preprocessed current signal component, voltage signal component and vibration signal component respectively to generate current time-frequency feature matrix, voltage time-frequency feature matrix and vibration time-frequency feature matrix; The step of performing signal type separation and parallel preprocessing on the mixed signal stream to obtain preprocessed current signal components, voltage signal components, and vibration signal components includes: The mixed signal stream is input into a signal separation device equipped with multiple bandpass filters. The signal separation device separates the mixed signal stream into independent original current signals, original voltage signals, and original vibration signals based on the inherent frequency band range of the current, voltage, and vibration signals. The original current signal is subjected to power frequency component removal and harmonic enhancement processing to eliminate power grid fundamental frequency interference and highlight fault harmonic components, forming a preprocessed current signal component. The original voltage signal is subjected to period normalization and transient event calibration processing to unify the starting point of the voltage waveform period and mark the voltage change time to form a preprocessed voltage signal component. The original vibration signal is subjected to background noise suppression and event trigger level adjustment processing to filter out environmental vibration noise and set an effective vibration event trigger threshold to form a preprocessed vibration signal component.

3. The method for monitoring medium-voltage cable faults according to claim 2, characterized in that, The step involves performing time-frequency domain transformation and feature extraction operations on the preprocessed current signal components, voltage signal components, and vibration signal components to generate time-frequency feature matrices for current, voltage, and vibration, respectively. Short-time Fourier transform is applied to the preprocessed current signal components to obtain the time-frequency spectrum of the current signal, and specific frequency band energy, harmonic distortion rate and phase jump characteristics are extracted from the time-frequency spectrum and combined into a current time-frequency feature matrix. Wavelet packet transform is applied to the preprocessed voltage signal components to decompose the sub-band energy distribution of the voltage signal at different resolutions, and the energy entropy, waveform steepness and zero-crossing interval features of each sub-band are extracted and combined into a voltage time-frequency feature matrix. The Hilbert-Huang transform is applied to the preprocessed vibration signal components to obtain the instantaneous frequency and marginal spectrum of the vibration signal. The instantaneous frequency variance, marginal spectrum peak frequency and energy concentration characteristics are extracted and combined into a vibration time-frequency feature matrix.

4. The method for monitoring medium-voltage cable faults according to claim 3, characterized in that, The step of inputting the current time-frequency feature matrix, voltage time-frequency feature matrix, and vibration time-frequency feature matrix into a multi-modal feature fusion engine to construct a unified comprehensive feature map of cable status includes: The multimodal feature fusion engine includes a spatiotemporal registration layer for feature alignment. The spatiotemporal registration layer uses a unified sampling timestamp and cable physical location coordinates as a reference to perform time synchronization and spatial registration of the current time-frequency feature matrix, voltage time-frequency feature matrix, and vibration time-frequency feature matrix. The spatiotemporally registered feature matrix is ​​projected into a shared hidden feature space, and an attention mechanism is used in the hidden feature space to calculate the dynamic weights between current features, voltage features and vibration features. Based on the dynamic weights, features from different modalities are weighted, concatenated, and dimensionality-reduced to generate a set of fused feature vectors; The fused feature vectors are spatially reorganized according to their corresponding cable positions to form a comprehensive feature map of cable status that includes multi-dimensional fused feature vectors and is distributed along the cable length.

5. The method for monitoring medium-voltage cable faults according to claim 4, characterized in that, Based on the comprehensive feature map of the cable condition, a fault mode matching network is used to identify characteristic abnormal regions and generate a preliminary set of suspected fault points, including: The fault mode matching network pre-stores reference feature templates for various typical cable fault modes, including partial discharge features, insulation degradation features, and mechanical damage features. The cable condition integrated feature map is divided into multiple consecutive analysis windows, and the fused feature vector sequence in each analysis window is sequentially subjected to sliding correlation calculation with all the reference feature templates; When the similarity between a certain analysis window and any of the baseline feature templates exceeds a preset matching threshold, the analysis window region is determined to be a feature abnormal region. Record the center coordinates of all the aforementioned abnormal feature regions, and add each center coordinate and its corresponding matching fault mode type as a record to the preliminary set of suspected fault points.

6. The method for monitoring medium-voltage cable faults according to claim 5, characterized in that, For each suspected fault point in the preliminary set of suspected fault points, a high-precision simulation model is invoked to simulate the physical field of the cable, and the theoretical distribution of the electrical and thermodynamic parameters corresponding to the suspected fault point is calculated, including: Based on the cable's model, structural parameters, and material properties, a high-precision simulation model is constructed, which includes an electromagnetic field simulation module and a thermal field simulation module for the cable. Using the location and fault mode type of the suspected points in the preliminary set of suspected fault points as input conditions, the electromagnetic field simulation module is driven to calculate the theoretical values ​​of electric field intensity distribution and current density distribution along the cable length when the corresponding fault occurs at the suspected point. At the same time, the thermal field simulation module is driven to calculate the theoretical value of the temperature field distribution along the cable length direction due to increased loss when the corresponding fault occurs at the suspected point. The calculated theoretical values ​​of electric field intensity distribution, current density distribution, and temperature field distribution are output as the theoretical distributions of electrical and thermodynamic parameters corresponding to the suspected point.

7. The method for monitoring medium-voltage cable faults according to claim 6, characterized in that, The theoretical distribution of the electrical and thermodynamic parameters is compared and analyzed with the actual characteristics of the corresponding areas in the comprehensive characteristic map of the cable condition to eliminate false positives caused by environmental interference and generate a refined set of fault points, including: Extract the actual fused feature vector of the region adjacent to the suspected point location from the comprehensive feature map of the cable status; The actual feature components reflecting changes in electric field strength, current density, and temperature are extracted from the actual fused feature vector. The theoretical values ​​in the theoretical distribution of the electrical and thermodynamic parameters are compared point by point with the measured values ​​in the actual characteristic components, and the deviation rate is calculated. If the deviation rate is lower than a preset consistency threshold, the suspected point is determined to be a real fault point; if the deviation rate is higher than the preset consistency threshold, the suspected point is determined to be a false suspected point caused by environmental interference. All suspected fault points identified as actual fault points, including their location coordinates, fault mode type, and deviation rate, are summarized to generate the refined fault point set.

8. The method for monitoring medium-voltage cable faults according to claim 7, characterized in that, For each fault point in the refined set of fault points, the risk level of fault propagation and the expected scope of impact are assessed by combining its historical data with the cable topology, including: From the cable historical operation database, retrieve the load rate change curve, insulation aging index and maintenance records of the cable segment where each fault point is located in the refined fault point set over a period of time. From the cable network topology file, obtain the connection relationship of the cable segment where the fault point is located in the power grid, the upstream and downstream load conditions, and whether it is a critical power supply node; Based on the load rate change curve, insulation aging index, maintenance records, connection relationships, upstream and downstream load conditions, and key node information, a risk assessment matrix is ​​used to calculate the probability that the fault point will develop into a permanent fault or trigger a chain of power outages under the current operating conditions, and the probability is mapped to a diffusion risk level. Based on the cable network topology file, after simulating the failure of the fault point, the power outage area is defined as the expected impact range according to the power outage area caused by the power grid protection logic.

9. A method for monitoring medium-voltage cable faults according to claim 8, characterized in that, Based on the aforementioned diffusion risk level and expected impact range, a monitoring report is formulated and output, including fault location information and handling recommendations, comprising: Based on the location coordinates of each fault point in the refined fault point set, generate cable fault geolocation information and cable well or junction box number information. Based on the spread risk level of each fault point, a corresponding handling suggestion is matched from a preset handling strategy library. The handling suggestions include immediate power outage for maintenance, planned maintenance, or enhanced online monitoring. Based on the estimated impact range of each failure point, generate a list of users and a list of critical facilities within the impact range; The cable fault location information, cable well or junction box number information, matching handling suggestions, user list and key facility list within the affected area are formatted and integrated according to the preset report template to generate the final readable monitoring report file.

10. A medium-voltage cable fault monitoring system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of a medium-voltage cable fault monitoring method as described in any one of claims 1 to 9.

Images