Cable testing and diagnostic methods and systems based on PDC signal feature fusion

Abstract

This invention relates to a cable testing and diagnostic method and system based on PDC signal feature fusion, comprising: establishing a digital twin model of the cable according to the structural parameters and material properties of the cable under test; collecting PDC signals, thermal imaging data, vibration monitoring data, and environmental data of the cable under test, and performing data standardization processing to obtain a standardized basic feature set; based on the basic feature set, extracting features by mode to generate independent modal features; using a cross-modal attention mechanism to perform weighted fusion of the independent modal features to obtain joint features for predicting the aging level and remaining life of the cable, thus obtaining a diagnostic result; adjusting the state parameters of the cable digital twin model according to the cable diagnostic result, resimulating the internal state of the cable, and visualizing the result in conjunction with the cable diagnostic result. Compared with the prior art, this invention achieves real-time dynamic monitoring, accurate diagnosis, and trend prediction of the internal state of the cable.

Description

Technical Field

[0001] This invention relates to the field of cable diagnostic technology, and in particular to a cable testing and diagnostic method and system based on PDC signal feature fusion. Background Technology

[0002] With the rapid development of power systems and smart grids, cables, as the core medium for power transmission, have become increasingly important in terms of safety and reliability. Modern cable systems, especially high-voltage and ultra-high-voltage cables, face increasingly complex operating environments and more stringent performance requirements. The insulation condition of a cable directly affects its operational stability and lifespan, and aging of insulation materials, partial discharge, and other potential faults are the main causes of cable system failure. Once cable insulation deteriorates or fails, it can lead to widespread power outages or even safety accidents. Therefore, accurate diagnosis and trend prediction of the internal condition of cables have become key issues in current research and engineering applications.

[0003] Currently, polarization-depolarization current (PDC) testing technology is one of the mainstream methods for monitoring the condition of cable insulation. This technology measures the current changes in cable insulation materials during polarization and depolarization processes, reflecting the dielectric properties, energy storage capacity, and aging degree of the material. By extracting and analyzing features of the PDC signal, such as polarization time constant, depolarization time constant, conductivity, and residual charge density, the health status of cable insulation can be effectively assessed. However, existing PDC technology still has many limitations in practical applications: on the one hand, PDC signal acquisition is easily affected by external noise and electromagnetic interference, leading to a decrease in data reliability and accuracy; on the other hand, traditional PDC signal analysis methods are usually static, lacking a comprehensive analysis of the dynamic changes of the signal over time and in the environment, making it difficult to capture the trend changes in the cable during long-term operation. Furthermore, single PDC mode data cannot fully reflect the multi-physics coupling effects inside the cable, such as the combined influence of electric field distribution, temperature gradient, and humidity on insulation performance.

[0004] In recent years, finite element analysis (FEA), as a precise physical field simulation tool, has been increasingly applied in the analysis of the internal state of cables. FEA technology can simulate the electric field distribution, dielectric properties, and thermal effects inside cables, providing detailed data support for evaluating insulation performance. However, existing finite element analysis methods mainly focus on static simulations during the design phase. In actual operation, due to the lack of real-time monitoring data input, the model struggles to dynamically reflect the true state of the cable under different operating conditions. Furthermore, FEA results typically require manual interpretation, making the analysis process cumbersome and reliant on expert experience, which fails to meet the real-time requirements of complex cable operating environments. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology in terms of low reliability and accuracy of cable analysis based on single-mode data, and to provide a cable testing and diagnosis method and system based on PDC signal feature fusion.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A cable testing and diagnostic method based on PDC signal feature fusion includes the following steps: establishing a digital twin model of the cable based on the structural parameters and material properties of the cable under test;

[0008] The PDC signal, thermal imaging data, vibration monitoring data and environmental data of the cable under test are collected and the data are standardized to obtain a standardized basic feature set.

[0009] Based on the aforementioned basic feature set, feature extraction is performed according to modality to generate corresponding independent modal features;

[0010] A cross-modal attention mechanism is used to weight and fuse the features of each independent modality to obtain joint features;

[0011] Based on the combined characteristics, the aging level and remaining life of the cable are predicted, and diagnostic results are obtained.

[0012] Based on the cable diagnostic results, the state parameters of the cable digital twin model are adjusted to re-simulate the internal state of the cable, and the results are visualized in conjunction with the cable diagnostic results, while high-risk areas are marked.

[0013] Furthermore, the process of establishing the digital twin model of the cable includes:

[0014] Based on the design parameters of the cable under test, a three-dimensional digital model of the cable is constructed. This three-dimensional digital model describes the geometric structure of the cable's insulation layer, electrical conductor, and shielding layer in different regions.

[0015] Assign material properties to the three-dimensional digital model of the cable;

[0016] The three-dimensional digital model of the cable after assigning material properties is meshed to obtain a discretized mesh model.

[0017] Set boundary conditions and excitation conditions for each level based on the mesh model;

[0018] Multiphysics simulations were performed based on a mesh model with boundary and excitation conditions set, and simulation results were obtained.

[0019] The simulation results are compared with the actual cable test results to verify the accuracy of the cable's three-dimensional digital model. If the verification conditions are met, a well-established cable digital twin model is obtained.

[0020] Furthermore, the process of assigning material properties includes setting the dielectric constant and thermal conductivity for the insulating layer, setting the conductivity for the electrical conductor, and defining the electromagnetic shielding characteristics for the shielding layer.

[0021] The mesh generation process also includes determining the mesh density based on the physical characteristics and geometric complexity of the cable's three-dimensional digital model;

[0022] The setting of the boundary conditions and excitation conditions includes: applying a working voltage to the electrical conductor, setting a grounding condition for the shielding layer, and setting an external boundary condition for the insulating layer.

[0023] Furthermore, the extraction process of the independent modal features specifically includes:

[0024] For PDC signal data, a bidirectional long short-term memory network is used for processing to obtain the forward hidden state and the backward hidden state, and then splices them to obtain the time series features of the PDC signal.

[0025] Spatial features were extracted from thermal imaging data and vibration monitoring data using a three-dimensional convolutional neural network, respectively.

[0026] The extracted time-series and spatial features were organized into independent modal features and then normalized.

[0027] Furthermore, the process of obtaining the joint features specifically includes:

[0028] Cross-modal correlation analysis is performed on each independent modal feature to calculate the degree of mutual influence between different independent modal features, so as to determine the correlation weight between different independent modal features and generate a modal correlation matrix;

[0029] Based on the modality correlation matrix, dynamic weights are assigned to each independent modality feature according to its degree of influence on the target task, generating a dynamic weight set.

[0030] Each independent modal feature is weighted based on the dynamic weight set to generate a weighted modal feature;

[0031] The weighted modal features are deeply fused through aggregation operations to generate joint features that comprehensively represent all modal information.

[0032] Furthermore, the method uses a classification model based on the obtained joint features to predict the aging level of the cable;

[0033] By combining the predicted aging level and joint characteristics of the cable, a regression model is used to predict the remaining life of the cable.

[0034] Furthermore, the method also utilizes a deep classification model to predict the fault type based on the fused joint features;

[0035] Fault location prediction is performed using a location regression model based on joint features.

[0036] Furthermore, the process of adjusting the state parameters of the cable digital twin model includes:

[0037] Update the local parameters of the cable digital twin model based on the predicted cable fault type and fault location;

[0038] The global state parameters of the cable digital twin model are adjusted based on the predicted aging level and remaining life of the cable.

[0039] The cable digital twin model is re-simulated using multiphysics based on the updated local and global state parameters to obtain an updated cable digital twin model. The expression of the updated cable digital twin model is as follows:

[0040] ;

[0041] In the formula, The integrated field distribution function of the updated cable digital twin model. For electric field distribution, For temperature field distribution, For humidity field distribution, , and These are the weighting factors.

[0042] Furthermore, the process of cable diagnosis and visualization using the cable digital twin model includes:

[0043] Data analysis is performed on the cable digital twin model after parameter adjustment to extract multiphysics information and labeled data;

[0044] Based on the extracted multiphysics information, unified spatial interpolation and normalization are performed to generate a unified three-dimensional mesh set and standardized heatmap color values.

[0045] A three-dimensional dynamic heatmap is constructed based on standardized heatmap color values, mapped onto voxels of a three-dimensional mesh set, and high-risk areas are displayed with weighting to generate a dynamic three-dimensional heatmap model file.

[0046] The dynamic 3D thermal model file is loaded using a 3D visualization engine to display the distribution of multiple physical fields and labeled data inside the cable.

[0047] The present invention also provides a cable testing and diagnostic system for implementing the cable testing and diagnostic method based on PDC signal feature fusion as described above, comprising:

[0048] The data acquisition module is used to collect PDC signals, thermal imaging data, vibration monitoring data and environmental data of the cable under test, and to perform data standardization processing to obtain a standardized basic feature set;

[0049] The feature extraction module is used to extract features according to the modality based on the basic feature set, and generate corresponding independent modal features;

[0050] The data fusion module is used to perform weighted fusion of features from each independent modality using a cross-modal attention mechanism to obtain joint features;

[0051] The diagnosis and prediction module is used to predict the aging level and remaining life of cables based on joint characteristics, and obtain diagnostic results.

[0052] The cable digital twin model building module is used to build a cable digital twin model based on the structural parameters and material properties of the cable under test.

[0053] The cable digital twin model dynamic update module is used to adjust the state parameters of the cable digital twin model according to the cable diagnosis results and re-simulate the internal state of the cable.

[0054] The 3D visualization module is used to dynamically display the diagnostic results and the output of the cable digital twin model in 3D form, while marking high-risk areas.

[0055] Compared with the prior art, the present invention has the following advantages:

[0056] (1) This invention extracts and fuses multimodal features by combining PDC signals, thermal imaging data, vibration monitoring data and environmental data. The modal data have different temporal and spatial resolutions and may have highly nonlinear relationships with each other. In the fusion process, cross-modal correlation analysis is performed, and each modality is assigned different weights according to its impact on the target task for weighted fusion, generating joint features that comprehensively express all modal information, thereby realizing the fusion of multimodal data. This is more conducive to capturing the implicit correlation between modes for more accurate cable diagnosis.

[0057] Based on the cable diagnostic results of joint features, the state parameters of the established cable digital twin model are adjusted to realize the dynamic updating and feedback of the model based on cable operation data, and the multi-physics field is re-simulated. This can reflect the impact of the multi-physics field coupling effect inside the cable on the cable performance, improve the accuracy of the cable digital twin model, and realize real-time dynamic monitoring, accurate diagnosis and trend prediction of the internal state of the cable.

[0058] (2) This invention solves the limitations of existing technologies in cable fault diagnosis and life prediction by multimodal data fusion and the construction and updating of cable digital twin models, and improves the comprehensiveness, accuracy and intelligence of cable health management. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating a cable testing and diagnostic method based on PDC signal feature fusion provided in an embodiment of the present invention.

[0060] Figure 2 This is a schematic diagram illustrating the process of establishing a cable digital twin model provided in an embodiment of the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments 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, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0062] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0063] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0064] Example 1

[0065] like Figure 1 As shown, this embodiment provides a cable testing and diagnostic method based on PDC signal feature fusion, including the following steps:

[0066] S1: Establish a digital twin model of the cable based on its structural parameters and material properties;

[0067] S2: Collect PDC signals, thermal imaging data, vibration monitoring data, and environmental data of the cable under test, and perform data standardization processing to obtain a standardized basic feature set;

[0068] S3: Based on the basic feature set, feature extraction is performed according to modality to generate corresponding independent modal features;

[0069] S4: A cross-modal attention mechanism is used to weight and fuse the features of each independent modality to obtain joint features;

[0070] S5: Based on the combined characteristics, predict the aging level and remaining life of the cable to obtain diagnostic results;

[0071] S6: Adjust the state parameters of the cable digital twin model based on the cable diagnosis results and re-simulate the internal state of the cable;

[0072] S7: Visualize the results generated from the cable digital twin model in the form of a three-dimensional dynamic heat map, combined with the cable diagnosis results, while marking high-risk areas;

[0073] S8: Combine diagnostic results with cable operation history data to generate personalized cable maintenance recommendations.

[0074] Specifically, such as Figure 2 As shown, the process of establishing the cable digital twin model in step S1 includes:

[0075] S11: Based on the design parameters of the cable under test, construct a three-dimensional digital model of the cable. This three-dimensional digital model accurately describes the geometric structure of the cable's insulation layer, electrical conductor, and shielding layer in different regions.

[0076] S12: Assign material properties to the three-dimensional digital model of the cable. The process of assigning material properties includes combining the material characteristics used in the actual cable, defining the physical parameters of each region in the model, setting the dielectric constant and thermal conductivity for the insulation layer, setting the conductivity for the electrical conductor, and defining the electromagnetic shielding characteristics for the shielding layer.

[0077] S13: Mesh the cable 3D digital model after assigning material properties to obtain a discretized mesh model;

[0078] The mesh generation process also includes determining the mesh density based on the physical characteristics and geometric complexity of the cable's three-dimensional digital model, using a higher density mesh in regions where the electric field intensity changes significantly, generating a discretized model, and outputting the discretized mesh model.

[0079] S14: Set the boundary conditions and excitation conditions for each level based on the discretized mesh model;

[0080] The setting of boundary conditions and excitation conditions includes: applying a working voltage to the electrical conductor, setting a grounding condition for the shielding layer, and setting an external boundary condition for the insulation layer;

[0081] S15: Based on the mesh model with boundary conditions and excitation conditions set, multiphysics simulation is performed using finite element analysis software to obtain simulation results;

[0082] S16: Compare the simulation results with the actual cable test results to verify the accuracy of the cable's three-dimensional digital model. If the verification conditions are met, a well-established cable digital twin model is obtained.

[0083] In step S2, the data standardization process performed in this embodiment includes removing noise and extracting key features, using a cable digital twin model to correct the data, and outputting a standardized basic feature set.

[0084] In step S3, the extraction process of independent modal features specifically includes:

[0085] S31: Based on a standardized basic feature set, the data is input into the corresponding feature extraction network according to the modality;

[0086] S32: For PDC signal data, a bidirectional long short-term memory network is used for processing to obtain the forward hidden state and the backward hidden state, and then splices them together to obtain the time series features of the PDC signal.

[0087] The computational expression for a bidirectional long short-term memory network is:

[0088] ;

[0089] ;

[0090] In the formula, It is a time series. Indicates the first The PDC signal value at time 10:00. This is a forward hidden state. It is a backward hidden state. , For bias vectors, For activation functions;

[0091] The expression for concatenating the forward and backward hidden states is:

[0092] ;

[0093] In the formula, This represents the features of the time series.

[0094] S33: Spatial features are extracted from thermal imaging data and vibration monitoring data using a three-dimensional convolutional neural network, respectively.

[0095] The computational expression for a three-dimensional convolutional neural network is:

[0096] ;

[0097] in, For thermal imaging data, for , The kernel size is the convolution kernel size. The first feature map is the one whose output feature map is represented by the first feature map. The value of the point;

[0098] S34: Organize the extracted time series features and spatial features into independent modal features respectively;

[0099] S35: Normalize the representations of all independent modal features.

[0100] In step S4, the process of obtaining joint features specifically includes:

[0101] S41: Perform cross-modal correlation analysis on each independent modal feature, calculate the degree of mutual influence between different independent modal features, determine the correlation weight between different independent modal features, and generate a modal correlation matrix;

[0102] S42: Based on the modality correlation matrix, assign dynamic weights to each independent modality feature according to its degree of influence on the target task, and generate a dynamic weight set; specifically: calculate the dynamic weight of each modality feature, highlight the modalities that have a greater impact on the target task, and at the same time reduce the weights of the modalities that contribute less to the target task, and generate a dynamic weight set.

[0103] S43: Weight each independent modal feature based on the dynamic weight set to generate weighted modal features;

[0104] That is, the dynamic weight set is applied to the input modal feature set, and each modal feature is weighted so that the numerical distribution of the feature matches its importance, thus generating weighted modal features;

[0105] S44: The weighted modal features are deeply fused through aggregation operations to generate joint features that comprehensively represent all modal information.

[0106] S45: Outputs the fused joint features, which include comprehensive information on time series and spatial features.

[0107] In step S5, the method uses a classification model based on the obtained joint features to predict the aging level of the cable.

[0108] By combining the predicted aging level and joint characteristics of the cable, a regression model is used to predict the remaining life of the cable.

[0109] Preferably, the method also utilizes a deep classification model to predict the fault type based on the fused joint features;

[0110] Fault location prediction is performed using a location regression model based on joint features.

[0111] In this embodiment, step S5 includes the following specific steps:

[0112] S51: Using a deep classification model to predict fault types based on the fused joint feature representation:

[0113] ;

[0114] in, For the combined features after fusion, Indicates the first Predicted probability of class of faults Labels for predicted fault types;

[0115] S52: Use a location regression model to predict the fault location based on the joint features. The output of the regression model is the three-dimensional coordinates of the fault location.

[0116] S53: Based on joint feature representation, a classification model is used to predict the overall aging level of the cable, and the output result is the aging level of the cable.

[0117] S54: Combining aging levels and joint characteristic representations, a regression model is used to predict the remaining life of the cable;

[0118] S55: Quantify the uncertainty of the prediction results of cable aging level and remaining life using Bayesian neural network.

[0119] Step S6, the process of adjusting the state parameters of the cable digital twin model, includes:

[0120] S61: Update the local parameters of the cable digital twin model based on the predicted cable fault type and fault location;

[0121] S62: Adjust the global state parameters of the cable digital twin model based on the predicted aging level and remaining life of the cable;

[0122] S63: The cable digital twin model is re-simulated using multiphysics based on the updated local and global state parameters to obtain an updated cable digital twin model. The expression for this updated cable digital twin model is as follows:

[0123] ;

[0124] In the formula, The integrated field distribution function of the updated cable digital twin model. For electric field distribution, For temperature field distribution, For humidity field distribution, , and These are the weighting factors;

[0125] S64: Generate 3D model files and annotation data based on the re-simulated multiphysics distribution;

[0126] S65: Integrates the re-simulated multiphysics distribution, the generated 3D model file, and the annotation data to output an updated twin model.

[0127] In step S7, the process of cable diagnosis and visualization using the cable digital twin model includes:

[0128] S71: Receive the updated cable digital twin model data, perform data parsing based on the cable digital twin model with adjusted parameters, and extract multiphysics information and annotation data;

[0129] S72: Based on the extracted multiphysics information, perform unified spatial interpolation and normalization to generate a unified three-dimensional mesh set and standardized heat map color values;

[0130] S73: Construct a three-dimensional dynamic heat map based on standardized heat map color values, map it onto voxels of a three-dimensional mesh set, and display high-risk areas with weighted values ​​to generate a dynamic three-dimensional heat map model file;

[0131] S74: Loads a 3D dynamic heat map model file, displays the distribution of multiple physical fields inside the cable through a 3D visualization engine, generates a visualization interface, supports users to switch the display of physical field layers and adjust the dynamic transparency, and intuitively displays the dynamic evolution process of high-risk areas.

[0132] S75: In the visual interface, users can interact with high-risk areas, revise the risk level or add notes, and integrate the user-revised data with the system's labeled data to generate a revised labeled dataset.

[0133] S76: Integrate the generated revised annotation dataset and 3D dynamic heat map model file and export them in a standardized format.

[0134] Example 2

[0135] This embodiment provides a cable testing and diagnostic system that implements the cable testing and diagnostic method based on PDC signal feature fusion as described in Embodiment 1, including:

[0136] The data acquisition module is used to collect PDC signals, thermal imaging data, vibration monitoring data and environmental data of the cable under test, and to perform data standardization processing to obtain a standardized basic feature set;

[0137] The feature extraction module is used to extract features based on the basic feature set and by modality, generating corresponding independent modal features.

[0138] The data fusion module is used to perform weighted fusion of features from each independent modality using a cross-modal attention mechanism to obtain joint features;

[0139] The diagnosis and prediction module is used to predict the aging level and remaining life of cables based on joint characteristics, and obtain diagnostic results.

[0140] The cable digital twin model building module is used to build a cable digital twin model based on the structural parameters and material properties of the cable under test.

[0141] The dynamic update module for the cable digital twin model is used to adjust the state parameters of the cable digital twin model based on the cable's diagnostic results, and to re-simulate the internal state of the cable.

[0142] The 3D visualization module is used to dynamically display the diagnostic results and the output of the cable digital twin model in 3D form, while marking high-risk areas.

[0143] Preferably, this system also includes:

[0144] The maintenance suggestion generation module is used to generate personalized maintenance strategies based on diagnostic results and historical data.

[0145] The system integration module is used to combine edge computing and cloud platforms, supporting remote monitoring and long-term data storage.

[0146] To verify the feasibility of this invention in practice, it was applied to a high-voltage transmission line project. This transmission line operates in a coastal area with a hot and humid climate, resulting in a complex cable operating environment and frequent insulation aging and partial discharge faults. Traditional cable insulation condition monitoring mainly relies on periodic PDC testing and manual maintenance. However, due to insufficient sampling accuracy and limited analytical methods, it is often difficult to detect potential faults in a timely manner, leading to multiple sudden faults during operation and seriously affecting the reliability and stability of the transmission line.

[0147] In this scenario, the PDC-based cable testing and diagnostic system provided by this invention is deployed at critical nodes of transmission lines. Through a PDC signal acquisition device, combined with temperature and humidity sensors and vibration monitoring equipment, real-time acquisition of multimodal data is achieved. The acquired raw data includes the time-varying trends of polarization and depolarization currents in the cable insulation layer, temperature and humidity data of the operating environment, and dynamic signals of the cable structure captured by vibration sensors. After preprocessing by the system, this data is input into a digital twin model to generate multiphysics distribution and health status assessment results reflecting the internal state of the cable.

[0148] During application, the system continuously monitored coastal high-voltage transmission cables for three months. Diagnostic data from a section of cable (number: A-01) indicated significant aging characteristics in its insulation layer during operation. PDC signal analysis extracted features such as the polarization time constant, conductivity, and residual charge density of the insulation layer, revealing that the conductivity increased by approximately 25% compared to normal values, while the polarization time constant decreased by 20%. Temperature field distribution generated from thermal imaging data showed that hotspots in the cable insulation layer were concentrated in the joint area, with a temperature rise reaching 65°C, significantly higher than the normal operating value of 45°C.

[0149] Furthermore, vibration monitoring data revealed that the cable was subjected to periodic mechanical vibrations during operation, leading to localized stress concentration within the insulation layer. Finite element analysis simulations further confirmed this: the electric field strength in the cable joint area was approximately 40% higher than the average, significantly increasing the risk of partial discharge. Through dynamic verification using a digital twin model and actual data, the remaining lifespan of this cable section was ultimately predicted to be approximately 6 months, recommending prompt repair or replacement.

[0150] To further verify the system's diagnostic capabilities, a comparative test was conducted on another cable (No. A-02) with similar operating environment and load, comparing it with traditional methods. Traditional methods, through manual inspection and periodic PDC testing, failed to detect any obvious anomalies; however, this system, through multimodal data fusion and trend prediction, successfully detected early signs of insulation aging in this cable section. The comparative results show that the diagnostic accuracy of this system reached 92%, significantly higher than the 72% of the traditional method. Furthermore, the multimodal data fusion and intelligent analysis functions of this system reduce the frequency of manual intervention, significantly lowering maintenance costs.

[0151] Table 1. Results of PDC signal characteristic analysis for cable A-01

[0152]

[0153] Table 2 Comparison of Diagnostic Results of Cable A-02 Using Traditional Methods and This System

[0154]

[0155] Table 3. Economic Benefit Analysis Based on This System

[0156]

[0157] As can be observed from the data in Table 1, the polarization time constant of cable A-01 gradually shortened over time, decreasing from 1.2 ms initially to 0.7 ms after 60 days, while the conductivity increased from 0.8 μS / cm to 1.2 μS / cm. These changes indicate a significant decline in the cable's insulation performance, manifested as increased conductivity and faster polarization response. Furthermore, the continuous increase in residual charge density also indicates a degradation in the energy storage characteristics of the insulation layer. Temperature rise data further validates this trend, gradually increasing from 45°C to 65°C, with hotspots concentrated in the joint area, which may be a direct manifestation of partial discharge or insulation aging. By combining PDC signal characteristics with hotspot location, the system of this invention can quickly identify key areas of cable aging, providing a clear basis for subsequent maintenance.

[0158] As shown in Table 2, the diagnostic time of the system of this invention is only 15 days, while the traditional method requires 60 days. This significant time advantage indicates a substantial improvement in fault detection efficiency. Furthermore, in terms of the number of potential hazards detected, the traditional method only identified one hazard, while the system of this invention identified three, demonstrating its more comprehensive diagnostic capabilities. This system can also accurately predict the remaining lifespan of the cable (180 days), an area where the traditional method is completely unsupported. Regarding diagnostic accuracy, the system of this invention achieves 92%, far exceeding the 72% of the traditional method. The number of manual interventions is also reduced from five in the traditional method to one, significantly reducing the workload of maintenance personnel.

[0159] Table 3 shows the significant economic advantages of the system of this invention. The average fault repair time using traditional methods is 8 hours, while this system reduces it to 2 hours, saving 75% of the repair time. Monthly maintenance costs are reduced from 200,000 yuan using traditional methods to 120,000 yuan, a saving of 40%. Regarding annual fault losses, traditional methods resulted in 10 million yuan in economic losses, while this system, through proactive detection and maintenance, reduces the losses to 2 million yuan, a reduction of 80%. Furthermore, this system achieves total economic savings of 7.6 million yuan in annual maintenance management, providing strong support for the efficient operation and maintenance of power systems.

[0160] The analysis of the above tables demonstrates that this invention exhibits significant advantages in terms of diagnostic efficiency, hazard identification capability, remaining life prediction accuracy, and economic benefits for cable operation. The main problems with traditional methods lie in diagnostic lag, insufficient hazard identification, and high maintenance costs. This invention, through the application of PDC signal analysis, multimodal data fusion, and digital twin technology, not only solves these problems but also provides an intelligent solution for the entire lifecycle management of cables. This capability is particularly important for cable systems in complex operating environments, effectively improving the reliability and stability of power systems while significantly reducing operation and maintenance costs, providing solid technical support for the construction and promotion of smart grids.

[0161] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A cable testing and diagnostic method based on PDC signal feature fusion, characterized in that, Includes the following steps: A digital twin model of the cable is established based on its structural parameters and material properties. The PDC signal, thermal imaging data, vibration monitoring data and environmental data of the cable under test are collected and the data are standardized to obtain a standardized basic feature set. Based on the aforementioned basic feature set, feature extraction is performed according to modality to generate corresponding independent modal features; A cross-modal attention mechanism is used to weight and fuse the features of each independent modality to obtain joint features; Based on the combined characteristics, the aging level and remaining life of the cable are predicted, and diagnostic results are obtained. Based on the cable diagnostic results, the state parameters of the cable digital twin model are adjusted to re-simulate the internal state of the cable, and the results are visualized in conjunction with the cable diagnostic results, while high-risk areas are marked.

2. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 1, characterized in that, The process of establishing the digital twin model of the cable includes: Based on the design parameters of the cable under test, a three-dimensional digital model of the cable is constructed. This three-dimensional digital model describes the geometric structure of the cable's insulation layer, electrical conductor, and shielding layer in different regions. Assign material properties to the three-dimensional digital model of the cable; The three-dimensional digital model of the cable after assigning material properties is meshed to obtain a discretized mesh model. Set boundary conditions and excitation conditions for each level based on the mesh model; Multiphysics simulations were performed based on a mesh model with boundary and excitation conditions set, and simulation results were obtained. The simulation results are compared with the actual cable test results to verify the accuracy of the cable's three-dimensional digital model. If the verification conditions are met, a well-established cable digital twin model is obtained.

3. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 2, characterized in that, The process of assigning material properties includes setting the dielectric constant and thermal conductivity for the insulating layer, setting the conductivity for the electrical conductor, and defining the electromagnetic shielding characteristics for the shielding layer. The mesh generation process also includes determining the mesh density based on the physical characteristics and geometric complexity of the cable's three-dimensional digital model; The setting of the boundary conditions and excitation conditions includes: applying a working voltage to the electrical conductor, setting a grounding condition for the shielding layer, and setting an external boundary condition for the insulating layer.

4. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 1, characterized in that, The extraction process of the independent modal features specifically includes: For PDC signal data, a bidirectional long short-term memory network is used for processing to obtain the forward hidden state and the backward hidden state, and then splices them to obtain the time series features of the PDC signal. Spatial features were extracted from thermal imaging data and vibration monitoring data using a three-dimensional convolutional neural network, respectively. The extracted time-series and spatial features were organized into independent modal features and then normalized.

5. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 1, characterized in that, The process of obtaining the joint features specifically includes: Cross-modal correlation analysis is performed on each independent modal feature to calculate the degree of mutual influence between different independent modal features, so as to determine the correlation weight between different independent modal features and generate a modal correlation matrix; Based on the modality correlation matrix, dynamic weights are assigned to each independent modality feature according to its degree of influence on the target task, generating a dynamic weight set. Each independent modal feature is weighted based on the dynamic weight set to generate a weighted modal feature; The weighted modal features are deeply fused through aggregation operations to generate joint features that comprehensively represent all modal information.

6. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 1, characterized in that, The method uses a classification model based on the obtained joint features to predict the aging level of the cable. By combining the predicted aging level and joint characteristics of the cable, a regression model is used to predict the remaining life of the cable.

7. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 6, characterized in that, The method also utilizes a deep classification model to predict the fault type based on the fused joint features; Fault location prediction is performed using a location regression model based on joint features.

8. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 7, characterized in that, The process of adjusting the state parameters of the cable digital twin model includes: Update the local parameters of the cable digital twin model based on the predicted cable fault type and fault location; The global state parameters of the cable digital twin model are adjusted based on the predicted aging level and remaining life of the cable. The cable digital twin model is re-simulated using multiphysics based on the updated local and global state parameters to obtain an updated cable digital twin model. The expression of the updated cable digital twin model is as follows: ； In the formula, The integrated field distribution function of the updated cable digital twin model. For electric field distribution, For temperature field distribution, For humidity field distribution, , and These are the weighting factors.

9. The cable testing and diagnostic method based on PDC signal feature fusion according to claim 1, characterized in that, The process of using the cable digital twin model for cable diagnosis and visualization includes: Data analysis is performed on the cable digital twin model after parameter adjustment to extract multiphysics information and labeled data; Based on the extracted multiphysics information, unified spatial interpolation and normalization are performed to generate a unified three-dimensional mesh set and standardized heatmap color values. A three-dimensional dynamic heatmap is constructed based on standardized heatmap color values, mapped onto voxels of a three-dimensional mesh set, and high-risk areas are displayed with weighting to generate a dynamic three-dimensional heatmap model file. The dynamic 3D thermal model file is loaded using a 3D visualization engine to display the distribution of multiple physical fields and labeled data inside the cable.

10. A cable testing and diagnostic system that implements the cable testing and diagnostic method based on PDC signal feature fusion as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect PDC signals, thermal imaging data, vibration monitoring data and environmental data of the cable under test, and to perform data standardization processing to obtain a standardized basic feature set; The feature extraction module is used to extract features according to the modality based on the basic feature set, and generate corresponding independent modal features; The data fusion module is used to perform weighted fusion of features from each independent modality using a cross-modal attention mechanism to obtain joint features; The diagnosis and prediction module is used to predict the aging level and remaining life of cables based on joint characteristics, and obtain diagnostic results. The cable digital twin model building module is used to build a cable digital twin model based on the structural parameters and material properties of the cable under test. The cable digital twin model dynamic update module is used to adjust the state parameters of the cable digital twin model according to the cable diagnosis results and re-simulate the internal state of the cable. The 3D visualization module is used to dynamically display the diagnostic results and the output of the cable digital twin model in 3D form, while marking high-risk areas.

Images