Intelligent fault early warning and diagnosis system and method for high-temperature superconducting cable based on multi-mode sensing

Abstract

The invention discloses a high-temperature superconducting cable intelligent fault early warning and diagnosis system based on multi-modal sensing, and the system comprises a multi-modal sensing module, a data processing module, an intelligent fault diagnosis module and an early warning output module which are in communication connection. Wherein the multi-mode sensing module is used for collecting multi-type operation state data of the high-temperature superconducting cable; the data processing module is used for carrying out standardization and multi-modal feature fusion on the acquired data; the fusion diagnosis module extracts spatial features through a convolutional neural network (CNN), weights key features through an attention mechanism, realizes fault classification through a support vector machine (SVM), and outputs a fault early warning result and diagnosis information; and the early warning output module automatically generates an early warning diagnosis report containing the fault type, the risk level and the processing suggestion. According to the invention, fault diagnosis and evaluation can be carried out on the operation state of the high-temperature superconducting cable, and the accuracy of fault prediction is improved.

Description

Technical Field

[0001] This invention relates to the field of power system condition monitoring and fault diagnosis technology, and in particular to an intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing. Technical Background

[0002] Superconducting cables, as a new technology in the field of power transmission, have become a core solution for addressing challenges such as surging electricity loads, scarce land resources, and significant pressure for energy conservation and emission reduction in medium and large cities, thanks to their zero-resistance characteristics leading to low loss and high current-carrying capacity, as well as their significant advantages such as saving urban underground space and being environmentally friendly. They have demonstrated irreplaceable application value in scenarios such as urban power grid backbones and centralized transmission of new energy sources. By maintaining a superconducting state in extremely low-temperature environments, they achieve transmission efficiency and capacity that are difficult for traditional cables to match, providing crucial support for the upgrading of high-density urban power systems.

[0003] However, existing superconducting cable monitoring technologies have significant limitations: First, the sensing methods are limited, relying mainly on single-dimensional monitoring of electrical parameters or temperature, making it difficult to capture early faults caused by multi-physical field coupling (such as insulation degradation often accompanied by the coordinated changes of local temperature rise and vibration abnormalities); second, fault diagnosis models are mostly based on single algorithms, which lack the ability to fuse multi-modal data (such as electrical signals, temperature fields, vibration spectra, and image features), leading to delayed early warnings or misjudgments.

[0004] Furthermore, traditional monitoring methods lack the accuracy to identify minute defects (such as microcracks in joints and localized aging of insulation layers) and lack spatiotemporal correlation analysis of multi-dimensional data, making it difficult to meet the "early detection, early diagnosis, and early handling" operation and maintenance requirements of superconducting cables. Therefore, developing a fault early warning and diagnosis system that integrates multimodal sensor data and combines intelligent algorithms has become a key technological breakthrough direction for ensuring the safe operation of high-temperature superconducting cables. Summary of the Invention

[0005] The purpose of this invention is to address the problems of incomplete perception of multi-physics coupling state, delayed fault early warning, and insufficient diagnostic accuracy in the existing monitoring of high-temperature superconducting cables. It provides a smart fault early warning and diagnosis system and method for high-temperature superconducting cables based on multi-modal sensing, so as to achieve accurate monitoring of the operating status of superconducting cables throughout their entire life cycle, timely early warning of early faults, and rapid location of fault types.

[0006] One technical solution to achieve the above objectives is a high-temperature superconducting cable intelligent fault early warning and diagnosis system based on multimodal sensing. The system comprises a multimodal sensing module, a data processing module, an intelligent fault diagnosis module, and an early warning output module connected by communication. Specifically, the multimodal sensing module collects various types of operating status data of the high-temperature superconducting cable; the data processing module standardizes and fuses multimodal features in the collected data; the fusion diagnosis module extracts spatial features using a convolutional neural network (CNN), weights key features using an attention mechanism, and performs fault classification using a support vector machine (SVM), outputting fault early warning results and diagnostic information; and the early warning output module automatically generates an early warning and diagnostic report containing the fault type, risk level, and handling suggestions.

[0007] In the aforementioned intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing, the multimodal sensing module includes a current sensor, a voltage sensor, a temperature monitoring system, a vibration sensing subsystem, and a high-definition industrial camera, which are used to collect the real-time current value, dynamic voltage signal, temperature distribution data along the cable, multi-frequency vibration signal, and high-definition image of the equipment appearance, respectively.

[0008] In the aforementioned intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing, the data processing module's processing steps include: removing noisy data using an outlier detection algorithm; and unifying multimodal data of different dimensions to a standardized form using a standardization algorithm. The interval; the feature fusion algorithm is used to integrate electrical features, temperature features, vibration features and image features into a multi-dimensional feature matrix.

[0009] In the aforementioned intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing, the workflow of the fusion diagnosis module is as follows: The CNN layer extracts spatial features from the image features and vibration waveform features in the multimodal feature matrix and outputs a high-dimensional feature vector. The attention mechanism enhances key features in a high-dimensional feature vector by calculating feature weight coefficients. The SVM classifier compares the weighted feature vectors with a pre-set fault feature library to classify fault types such as short circuit, quench, and insulation degradation, and outputs the probability of fault occurrence.

[0010] In the above-mentioned intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing, the CNN layer adopts a two-layer "convolution-pooling" structure. The kernel size of the first layer is 5×5 and the kernel size of the second layer is 3×3, which are used to extract local spatial correlation features in multimodal data step by step. The attention mechanism calculates the importance of features by scaling the dot product attention, and the weight coefficients are dynamically updated through backpropagation.

[0011] In the aforementioned intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing, the output information of the early warning report module includes: fault type (such as current overload fault, cooling system failure, insulation layer damage), fault location, risk level, and targeted maintenance suggestions (such as emergency shutdown, partial maintenance, and cooling system parameter adjustment).

[0012] The second technical solution to achieve the above objectives is: a method for intelligent fault early warning and diagnosis of high-temperature superconducting cables as described above, which includes the following steps: Step ①: The multimodal sensing module collects full-state data to form the raw dataset; Step 2: The data processing module standardizes and fuses features in the original dataset; Step 3: The CNN layer extracts spatial features from the feature matrix and outputs a high-dimensional feature vector; Step 4: The attention mechanism weights the high-dimensional feature vector to enhance the representation ability of key fault features; Step 5: The SVM classifier identifies the fault type based on weighted features; Step 6: Output diagnostic and early warning reports.

[0013] The intelligent fault early warning and diagnosis method for high-temperature superconducting cables in the above-described system is characterized in that the feature fusion algorithm in step ② adopts a weighted fusion strategy, wherein the weight coefficients of temperature features and vibration features are higher than those of electrical features, and the weight values ​​are dynamically adjusted based on the influence factors of historical fault data.

[0014] The intelligent fault early warning and diagnosis method for high-temperature superconducting cables in the above system is characterized in that the kernel function of the SVM classifier in step ⑤ adopts the radial basis function (RBF) to make the fault classification accuracy ≥95%.

[0015] By adopting the above technical solution, this application possesses the following significant advantages: This invention, by constructing a multimodal sensing module to simultaneously collect multi-dimensional data, breaks through the limitations of traditional single-parameter monitoring, achieves comprehensive perception of the coupling state of multiple physical fields, and effectively solves the problem of difficulty in identifying early weak fault signals.

[0016] This invention maintains high classification accuracy even in scenarios where fault samples are scarce by optimizing feature fusion strategies and fusion algorithms, significantly reducing the risk of misclassification and missed classification, and ensuring the reliability of diagnostic results.

[0017] This invention significantly shortens fault response time through an intelligent fault early warning and diagnosis system, providing accurate fault types, locations, and handling suggestions for safe maintenance, thus assisting in efficient decision-making. Attached Figure Description

[0018] Figure 1 This is a block diagram of the overall structure of the intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing according to the present invention. Figure 2 This is a schematic diagram of the overall process of the intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing in this invention. Figure 3 This is a schematic diagram of the composition of the multimodal sensing module in the system of the present invention; Figure 4 This is a logical block diagram of the intelligent fusion diagnostic module algorithm (CNN-attention-SVM) of the present invention; Figure 5 This is a support vector machine network diagram of the intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing in this invention. Figure 6 This is a schematic diagram of the physical connection of the intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing in this invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0020] This implementation case provides an intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing. By integrating multi-dimensional sensor data and algorithm fusion technology, it can achieve accurate monitoring, early fault warning and rapid diagnosis of the entire life cycle operation status of high-temperature superconducting cables. It solves the problems of incomplete perception of multi-physical field coupling state and delayed fault identification in traditional monitoring methods, and provides scientific support for the safe maintenance of high-temperature superconducting cable systems.

[0021] like Figure 1The diagram illustrates the overall architecture of the intelligent fault early warning and diagnosis system for high-temperature superconducting cables of this invention. The system adopts a modular design, comprising four integrated modules: a multimodal sensing module, a data processing module, an intelligent fault diagnosis module, and an early warning output module. The multimodal sensing module, serving as the system's perception layer, is responsible for collecting operational status data of the superconducting cable itself. The data processing module standardizes and fuses the various data collected by the multimodal sensing module, optimizes data quality through outlier detection and interpolation, unifies the scale of data with different dimensions, and integrates them into a multi-dimensional feature matrix using a weighted approach, dynamically adjusting weights to highlight key features. The intelligent fault diagnosis module, as the core, combines a fusion algorithm, extracts spatial features through CNN, strengthens key information through an attention mechanism, completes fault classification using SVM, and outputs probabilities. It then performs in-depth analysis and pattern recognition on the processed features to achieve accurate fault classification. Finally, the early warning output module generates a visualized report based on the diagnostic results, including fault type, location, risk level, and maintenance recommendations, providing rapid feedback on fault information, intuitive support for safe maintenance, and ensuring efficient system response. The diagram clearly illustrates the interconnected relationship and functional division of the four modules, and clearly explains the complete information flow from data acquisition to diagnosis and early warning, reflecting the system's integrated and intelligent characteristics.

[0022] like Figure 2 The diagram shows a detailed framework of the multimodal sensing module, which serves as the sensing front-end of the entire intelligent diagnostic system. It is not a simple collection of single sensors, but rather an integrated system comprised of several high-precision monitoring sub-units. The multimodal sensing module is used to simultaneously acquire multi-dimensional operational status data of the high-temperature superconducting cable. The composition and function of each monitoring sub-unit are as follows: Current and voltage sensors are used to collect real-time current values ​​and dynamic voltage signals of the high-temperature superconducting cable, respectively. The temperature monitoring system acquires axial temperature distribution data of the high-temperature superconducting cable body and its surrounding environment through fiber optic grating arrays deployed along the line. The vibration sensing subsystem consists of a grating sensor and a piezoelectric sensor, which monitors the vibration frequency, vibration acceleration and external impact intensity of the high-temperature superconducting cable body, and captures the subtle vibration characteristics caused by mechanical stress. The high-definition industrial camera is equipped with a telephoto lens to capture images of the high-temperature superconducting cable body, joints and terminals (resolution ≥ 20 million pixels) and identify appearance defects (such as insulation layer damage, joint leakage, etc.). like Figure 3As shown, the physical structure of the system and the layout of the sensor network are clearly defined, directly presenting the actual deployment scheme of multiple monitoring sub-units along the high-temperature superconducting cable line. The core of the diagram is the high-temperature superconducting cable itself, along which various sensors are deployed at key nodes: current and voltage sensors are directly connected to monitor electrical parameters; the temperature monitoring system achieves distributed temperature measurement through a sensor array deployed along the line; vibration sensing units are attached to the surface of the high-temperature superconducting cable to capture mechanical vibration signals; and high-definition industrial cameras are aimed at key areas such as cable joints for visual monitoring. All sensor data is ultimately transmitted to the terminal for integration and processing, generating a diagnostic report.

[0023] like Figure 4 As shown, the specific operation flow of this system is as follows: Step ①: The multimodal sensing module collects full-state data to form the raw dataset. When a user needs to predict faults in a high-temperature superconducting cable and issues a prediction command, the server receives the command and, through the multimodal sensing module, acquires the complete dataset of the high-temperature superconducting cable. This dataset contains multiple samples, each corresponding to a time point or state point, and includes multimodal raw data reflecting the cable's operating status and environmental conditions. (For example, information such as current, voltage, temperature, and pressure).

[0024] Step 2: The data processing module standardizes and fuses features of the original dataset. To eliminate interference caused by differences in sensor type and sampling frequency among different monitoring subsystems, the raw monitoring data needs to be systematically processed first, and then a standardized feature matrix needs to be constructed through feature fusion. Subsequently, standardization and integration were performed to address the compatibility issue of multimodal data with both discrete and continuous characteristics: For discrete binary state data, a unified mapping to the [0,1] numerical form was used to achieve consistent expression of switching characteristics; for continuous monitoring indicators (such as analog quantities like current, voltage, and temperature), a min-max standardization algorithm was used to compress the data to the [0,1] interval, with the calculation formula as follows:

[0025] in The original value, These are the historical minimum and maximum values ​​of the indicator, respectively. These are normalized multimodal standardized feature vectors. Next, the system fuses these standardized features to construct a joint feature matrix. .

[0026] in: This represents the number of samples in the dataset (i.e., the number of time points or state points). This represents the total dimension of the features after fusion processing.

[0027] This joint feature matrix diagram This constitutes a feature dataset corresponding to the original high-temperature superconducting cable dataset. This feature dataset contains multiple feature data points, each represented as a d-dimensional feature vector. It contains key information about the current state of the high-temperature superconducting cable after standardization and fusion, providing input for subsequent fault prediction models.

[0028] like Figure 5 As shown, it details the algorithm flow used by the intelligent fault diagnosis module, which is an end-to-end deep learning architecture integrating feature extraction, enhancement, and decision-making. The complete workflow begins with the input of the original multimodal feature matrix. First, it is processed hierarchically through a deep sequence model consisting of two "convolution-pooling" layers—that is, the first dashed box diagram. Then, the data first enters the convolutional layer 1, where it undergoes local perception and nonlinear transformation through a 5×5 convolutional kernel to capture primary local patterns in data such as current, temperature, and vibration. These features are then fed into the pooling layer 1 for max pooling downsampling. The condensed features are then further processed by the convolutional layer 2, which uses a 3×3 convolutional kernel to combine the primary features to form deep abstract features that can characterize complex fault modes. These high-level features are then further refined and compressed by the pooling layer 2, outputting a high-dimensional feature map containing deep abstract information.

[0029] Step 3: The CNN layer extracts spatial features from the feature matrix and outputs a high-dimensional feature map. Feature matrix diagram The input is processed by a convolutional neural network for feature extraction. The formula for the convolution operation is:

[0030] in, For the input feature map, For the convolution kernel at position ( The weight, For item, For non-linear activation functions (using) ). To output feature map at location Value, index This is a spatial location index on the feature map. (Subscript) This refers to the weight index inside the convolution kernel.

[0031] use Activation functions and max pooling operations extract spatial or temporal local features, resulting in high-dimensional feature maps.

[0032] Next, the model flattens the high-dimensional feature map and feeds it into the attention mechanism for adaptive weighting, as shown in the second dashed box diagram. Figure 5 This is to highlight the key channel information most relevant to the fault.

[0033] Step 4: The attention mechanism weights the high-dimensional feature vector to enhance the representation ability of key fault features. Output of the above steps An attention mechanism is introduced to dynamically learn and assign different weights to different channels or regions, thereby increasing the weights of key dimensions and making the model focus more on key features related to faults. A lightweight channel attention mechanism is used, which first performs global compression on the feature map of each channel, then predicts the importance weight of each channel through two fully connected layers, and performs channel-wise weighted fusion with the original feature map.

[0034] The core calculation formula for this step is:

[0035] in, Indicates attention weights. The feature map is after feature extraction and weighting. (Subscript) This is the spatial location index on the feature map. The weighted feature map is flattened into a vector and then input into the next classification module to achieve intelligent judgment of the fault type.

[0036] Finally, as Figure 5 After the high-dimensional feature map is flattened by two fully connected layers, it is input into the support vector machine (SVM) classifier. By utilizing its powerful structural risk minimization principle, the optimal classification hyperplane is constructed in the feature space, thereby achieving accurate classification of fault types in high-temperature superconducting cables.

[0037] Step 5: The SVM classifier identifies the fault type based on weighted features. After feature extraction and enhancement, a Support Vector Machine (SVM) is introduced as the final classifier. Specifically, the attention-weighted feature map... The feature vectors are flattened and used as input features for the support vector machine. Then, the feature vectors are classified according to the optimal hyperplane obtained during training, and the corresponding fault criteria are output.

[0038] like Figure 6 As shown, the classification principle of Support Vector Machines is presented. Its core idea is to transform the input samples into a high-dimensional feature space through a nonlinear mapping, and then find an optimal classification hyperplane with the maximum margin in this space to achieve robust classification decisions.

[0039] Its core discrimination formula is as follows:

[0040] in, The flattened feature vector to be classified For the first The feature vectors of the support vectors The weight vectors are learned by the support vector machine. For bias terms, A function to output the sign of the final category label based on the sign of the calculated value within the parentheses. For the first The Lagrange multipliers corresponding to the support vectors are the optimal solutions to the dual problem. Radial basis ( Kernel function, The classification results correspond to different types of high-temperature superconducting cable faults.

[0041] Step 6: Output diagnostic and early warning reports The system ultimately generates a fault warning report, which includes the fault type, location, risk level, and diagnostic and handling recommendations. Based on the classification results, combined with historical data and the degree of fault impact, the system assesses the current fault risk level.

[0042] The technical solution of the present invention has been described in detail above with reference to the accompanying drawings and specific embodiments. Those skilled in the art can make various adjustments and modifications to the present invention based on the above description. Therefore, the specific details in the embodiments should not be considered as limitations on the scope of protection of the present invention, and the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A smart fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing, characterized in that: The system includes a multimodal sensing module, a data processing module, an intelligent fault diagnosis module, and an early warning output module, all connected by communication. Specifically: the multimodal sensing module collects various types of operational status data of the high-temperature superconducting cable; the data processing module cleans, standardizes, and fuses multimodal features in the collected data; the intelligent fault diagnosis module extracts spatial features using a convolutional neural network, weights key features using an attention mechanism, and performs fault classification using a support vector machine, outputting fault warning results and diagnostic information; and the early warning output module automatically generates an early warning diagnostic report containing the fault type, risk level, and handling suggestions.

2. The intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing according to claim 1, characterized in that, The multimodal sensing module includes a current sensor, a voltage sensor, a temperature monitoring system, a vibration sensing subsystem, and a high-definition industrial camera, which are used to collect real-time current values, dynamic voltage signals, temperature distribution data along the cable, multi-frequency vibration signals, and high-definition images of the equipment appearance, respectively.

3. The intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing according to claim 2, characterized in that, The vibration sensing subsystem consists of a grating sensor and a piezoelectric sensor, used to monitor the vibration frequency, vibration acceleration and impact intensity signals of the cable body and its surrounding environment; the high-definition industrial camera is equipped with a telephoto lens module, used to capture images of appearance defects of the cable body, joints and terminals.

4. The intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing according to claim 1, characterized in that, The data processing module's processing steps include: removing noisy data using an outlier detection algorithm; unifying multimodal data of different dimensions to the [0,1] interval using a standardization algorithm; and integrating electrical features, temperature features, vibration features, and image features into a multidimensional feature matrix using a feature fusion algorithm.

5. The intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing according to claim 1, characterized in that, The workflow of the intelligent fault diagnosis module is as follows: The convolutional neural network layer extracts spatial features from the image features and vibration waveform features in the multimodal feature matrix and outputs a high-dimensional feature vector; Attention mechanisms enhance key features in high-dimensional feature vectors by calculating feature weight coefficients. The support vector machine classifier compares the weighted feature vectors with a pre-set fault feature library to classify fault types such as short circuit, quench, and insulation degradation, and outputs the probability of fault occurrence.

6. The intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing according to claim 5, characterized in that, The convolutional neural network layer adopts a two-layer "convolution-pooling" structure, with the first layer having a kernel size of 5×5 and the second layer having a kernel size of 3×3, which is used to extract local spatial correlation features in multimodal data step by step; The attention mechanism calculates feature importance by scaling dot product attention, and the weight coefficients are dynamically updated through backpropagation.

7. The intelligent fault early warning and diagnosis system for high-temperature superconducting cables based on multimodal sensing according to claim 1, characterized in that, The output information of the early warning report module includes: fault type, which includes current overload fault, cooling system failure, insulation layer damage, fault location, risk level, and targeted maintenance suggestions, which include emergency shutdown, partial maintenance, and cooling system parameter adjustment.

8. A method for intelligent fault early warning and diagnosis of high-temperature superconducting cables according to any one of claims 1 to 7, comprising the following steps: Step ①: The multimodal sensing module collects current, voltage, temperature, vibration, and image status data to form a raw dataset; Step 2: The data processing module standardizes and fuses features in the original dataset to generate a multi-dimensional feature matrix; Step 3: The convolutional neural network layer extracts spatial features from the feature matrix and outputs a high-dimensional feature vector; Step 4: The attention mechanism weights the high-dimensional feature vectors to enhance the representation ability of key fault features; Step 5: The support vector machine classifier completes fault type identification and probability calculation based on weighted features; Step 6: The early warning report module generates an early warning diagnosis report containing fault information and handling suggestions based on the classification results.

9. The intelligent fault early warning and diagnosis method for high-temperature superconducting cables according to claim 8, characterized in that, The feature fusion algorithm in step ② adopts a weighted fusion strategy, in which the weight coefficients of temperature features and vibration features are higher than those of electrical features, and the weight values ​​are dynamically adjusted based on the influence factors of historical fault data.

10. The method according to claim 8, characterized in that, In step ⑤, the kernel function of the support vector machine classifier adopts the radial basis function to ensure that the fault classification accuracy is ≥95%.

Images