High-voltage cable insulation fault intelligent diagnosis method and system based on multi-information fusion

Abstract

The invention relates to the field of high-voltage cable fault diagnosis, and discloses a high-voltage cable insulation fault intelligent diagnosis method and system based on multi-information fusion, and the method comprises the following steps: S1, obtaining the operation multi-source data of a high-voltage cable, and constructing a multi-source original data matrix; s2, preprocessing to obtain a time-space aligned standardized data matrix; s3, performing feature extraction to obtain a multi-dimensional feature vector, and learning internal association between features by using a multi-modal deep network to obtain a joint multi-modal feature; s4, constructing a fault type identification model based on Bayesian reasoning and Monte Carlo sampling, and obtaining a fault type classification result; s5, in combination with deep learning and a physical model, obtaining a fault occurrence interval, positioning information and a fault level; and S6, based on a fault type classification result, a fault occurrence interval and positioning information, obtaining a fault level, and carrying out early warning pushing on a generated diagnosis report. According to the invention, high-precision identification, positioning and risk assessment of high-voltage cable insulation faults are realized.

Description

Technical Field

[0001] The present invention relates to the field of high-voltage cable fault diagnosis, and in particular to a high-voltage cable insulation fault intelligent diagnosis method and system based on multi-information fusion. Background Art

[0002] As modern society's demand for stable and reliable power supply continues to rise, the operational safety of high-voltage cables, the core carriers of urban power grids and interregional transmission networks, directly impacts the order of life and production. In recent years, the lengths and voltage levels of high-voltage cables have continued to increase, and their operating environments have become more complex. Long-term exposure to multiple factors, including electrical, thermal, and mechanical stresses and chemical corrosion, has led to an increasing number of potential failure risks, including insulation aging, partial discharge, and overheating. According to statistics, insulation failure has become the leading cause of high-voltage cable failure. Its sudden and hidden nature not only can cause large-scale power outages, but also poses safety risks such as fire, posing a significant threat to power grid security and the social economy.

[0003] Traditional methods for diagnosing high-voltage cable insulation faults primarily rely on single-sensor data or manual inspections. For example, methods based on partial discharge detection only capture electrical signal characteristics and are unable to reflect insulation degradation caused by mechanical damage or environmental factors. Infrared temperature measurement can detect overheating anomalies but cannot precisely locate the root cause of the fault. Manual inspections, limited by subjective experience and workload, suffer from low efficiency and high rates of missed detection. These methods lack diagnostic accuracy and timeliness when faced with complex operating conditions and early, subtle faults, making them unable to meet the smart grid's demand for fault prediction and rapid resolution. Summary of the Invention

[0004] In order to solve the above problems, the purpose of the present invention is to provide an intelligent diagnosis method for high-voltage cable insulation faults based on multi-information fusion, which can effectively realize high-precision identification, positioning and risk assessment of high-voltage cable insulation faults.

[0005] To achieve the above object, the present invention adopts the following technical solutions: An intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion includes the following steps: S1: Obtain multi-source operating data of high-voltage cables and construct a multi-source raw data matrix; S2: Preprocessing the acquired multi-source data of the high-voltage cable operation to obtain a standardized data matrix aligned in time and space; S3: Based on the standardized data matrix aligned in time and space, feature extraction is performed to obtain a multi-dimensional feature vector, and the intrinsic correlation between features is learned using a multimodal deep network to obtain a joint multimodal feature; S4: Build a fault type recognition model based on Bayesian reasoning and Monte Carlo sampling, and perform fault type discrimination based on the joint multimodal features to obtain the fault type classification results; S5: Based on the joint multimodal features, combined with deep learning and physical models, the fault interval and location information are obtained; S6: Based on the fault type classification results and the fault interval and location information, the fault level is obtained and a diagnostic report is generated for early warning push.

[0006] Furthermore, we acquire multi-source data on the operation of high-voltage cables, including partial discharge signals, dielectric loss factor, cable temperature and humidity, environmental parameters, operation and maintenance history data, and fault tags, as follows: High-frequency current transformers are deployed at cable joints to detect electrical pulse signals, and piezoelectric ceramic sensors are installed along the cable body to capture ultrasonic signals. Fiber Bragg grating arrays are embedded in the cable shield to monitor optical pulse signals in real time. Install a digital dielectric loss tester on the cable grounding wire, inject the test signal using the low-frequency superposition method, and obtain the time series data of the dielectric loss factor and capacitance change rate; A distributed optical fiber temperature measurement system is spirally wound along the cable outer sheath to obtain the axial temperature distribution matrix. A three-axis MEMS vibration sensor is installed at the cable joint to monitor the X / Y / Z three-axis vibration acceleration time series signal. Environmental sensors are used to obtain environmental parameter time series data and gas concentration spectra. Connect to the SCADA system via the OPC UA protocol to extract cable load curves, operation and maintenance history data, and fault tags; The above data is collected through the master control cabinet node. The master control node is equipped with GPS and Beidou dual-mode clock sources and synchronizes the clocks of all sensors through the PTP protocol. When any partial discharge channel detects a pulse >50pC, all sensors are synchronously triggered to record 1ms time window data. Establish a three-dimensional coordinate system for the cable, mark the installation location of each sensor, and automatically map the temperature data to the corresponding coordinate point; Furthermore, preprocessing includes filtering and denoising, multi-channel synchronous processing, distortion correction, and outlier screening, as follows: Get the original multi-source data matrix X according to S1 raw : X raw ={x1(t),x2(t),...,x a (t),...,x n (t)}; Among them, x a (t) represents the sensor signal of type a, including electrical signal, acoustic signal, and temperature field, and t is the time series; For different sensor signals, corresponding filtering and denoising processing is used. Elliptical filtering is used for electrical signals, EMD is used for acoustic signals, and wavelet threshold is used for optical signals to obtain the signal data X after denoising. denoised ; For the signal data X after noise reduction denoised , using multi-channel synchronous processing distortion correction, correcting the electrical signal amplitude through the transmission line model, and compensating the temperature field through heat conduction: Get the real value signal data X of the physical quantity corrected ; Finally, the local outlier factor and Mahalanobis distance are combined to eliminate abnormal data caused by hardware failure or environmental interference to obtain the standardized spatiotemporal alignment matrix X proc .

[0007] Furthermore, based on the spatiotemporally aligned standardized data matrix, feature extraction is performed to obtain multi-dimensional feature vectors and fuse them to obtain a joint multimodal feature representation, as follows: The standardized spatiotemporal alignment matrix X obtained from S2 proc , extract features from each modal data, including partial discharge signal features, acoustic signal features, and temperature field features; the partial discharge signal uses ResNet-18 to extract spatial features as partial discharge signal features; the ultrasonic signal uses a bidirectional LSTM to capture the evolution trend as the acoustic signal feature; the temperature field uses a Transformer encoder to model spatial dependencies as the temperature field feature; Based on the extracted modal features, Z-score standardization is used to calculate the f i Mutual information with the fault label y : ; in, is the joint probability density; Features f i Independent probability distribution of; p(y) is the prior probability of fault label y; Preserve the feature F of Top-K high mutual information MI , delete the ones that satisfy |p(f i ,f j )|>0.7 redundant features, where p(f i ,f j ) is f i and f j The joint probability distribution of A multimodal deep network is used to learn the intrinsic correlation between high mutual information features, obtain a highly expressive representation of the fusion features, and construct a joint multimodal feature h fusion .

[0008] Furthermore, the multimodal deep network includes a feature embedding layer, a feature interaction layer, an attention mechanism, and a feature fusion layer, as follows: The feature embedding layer maps the features of different modalities into a unified feature space: h pd =W pd ·f pd +b pd ; h sound =W sound ·f sound +b sound ; h temp =W temp ·f temp +b temp ; Where: f pd 、f sound 、f temp are the characteristic vectors of partial discharge, acoustic signal and temperature field respectively; W pd 、W sound 、W temp is a learnable weight matrix; b pd 、b sound 、b temp is the bias vector; h pd 、h sound 、h temp are the embedded eigenvectors of partial discharge, acoustic signal and temperature field respectively; The feature interaction layer models the complex associations between features of different modalities and uses a multimodal interaction network to learn the complex associations Z between features: Z=σ(W v ·[h pd ;h sound ;h temp ]+b v ); Among them, [;] represents the concatenation operation of features; W v and b v are the weights and biases of the feature interaction layer; σ is the activation function; The self-attention mechanism MultiHead is used to capture long-distance dependencies between features: Q=W Q Z, K=W K Z, V=W V ·Z; ; ; Among them, Q, K, V represent table query, key and value matrices respectively, and WQ 、W K 、W V is the parameter matrix corresponding to Q, K, and V; , , is the learnable projection matrix corresponding to Q, K, and V of the h-th attention head; head h is the hth attention head; H is the number of attention heads; Attention represents the attention mechanism calculation function; W O is the output weight matrix; Concat represents the concatenation function; Adaptively weight the importance of different modes: ; in, is the attention weight of each modality; W a1 , W a2 , W a3 and b a1 , b a2 ,b a3 is a learnable parameter; The feature fusion layer combines the results of multi-head attention and modal attention to generate the final joint multimodal feature representation h fusion : ; ; ; Among them, h attn is the output of the self-attention mechanism; h modal is the output of the modal attention mechanism; W f and b f are the weights and biases of the feature fusion layer.

[0009] Furthermore, a fault type recognition model is constructed based on Bayesian reasoning and Monte Carlo sampling, and fault type discrimination is performed based on the joint multimodal features to obtain the fault type classification results, as follows: First, nonlinear transformation and feature enhancement are performed on the joint multimodal features: F enhanced =σ(W1·h fusion +b1)+h fusion ; F attention =Softmax(W2·F enhanced +b2)⊙F enhanced ; Among them, ⊙ represents element-by-element multiplication; F enhanced represents the output after nonlinear transformation; F attentionRepresents the output after feature enhancement; W1 and b1 are the weight and bias of nonlinear transformation respectively; W2 and b2 are the weight and bias after feature enhancement respectively; Construct a multi-scale feature extraction module and process it through maximum pooling MaxPool and average pooling AvgPool: F local =MaxPool(W3·F attention +b3); F global =AvgPool(W4·F attention +b4); F multi-scale =Concat(F local ,F global ); Among them, F local represents the output after maximum pooling; F global Represents the output after average pooling; W3 and b3 are the weight and bias of the maximum pooling MaxPool respectively; W4 and b4 are the weight and bias of the average pooling AvgPool respectively; F multi-scale Represents the extracted multi-scale features; Build a classifier based on Bayesian inference and output probability distribution : ; ; ; in, is random noise, is the probability density function of the noise; and are the mean vector and standard deviation vector of category y respectively; W σ and b σ are the weight and bias of standard deviation prediction respectively; and are the weight and bias of mean prediction respectively; Estimate classification uncertainty via Monte Carlo sampling: ; ; in, is the average probability expectation; T is the number of sampling times; is the random noise of the t-th sampling; Uncertainty is the information entropy of the model prediction distribution; c represents category c; Simultaneously optimize classification and uncertainty estimation, the objective function for: ; in, is the cross entropy loss; is the regularization loss; is the uncertainty constraint loss, using entropy regularization; 、 is the weight coefficient; Final classification decision, predicted fault type: ; in, is the probability of predicting the fault type c.

[0010] Furthermore, based on the joint multimodal features, combined with deep learning and physical models, the fault interval and location information are obtained, as follows: From the joint multimodal feature h fusion Extract spatiotemporal features from: F temp =TCN(h fusion ); F space =GCN([F temp ,A]); Among them, TCN is a temporal convolutional network that captures the temporal features F temp ; GCN is a graph convolutional network, combined with the adjacency matrix A to model the spatial relationship F space ; Fusion of cable physical parameters with spatiotemporal characteristics and embedding them into the positioning process: F phys =Embed(R,L,C,G); F joint =a'·F space +(1-a')·F phys ; Among them, R, L, C, G are the physical parameters of the cable, namely resistance, inductance, capacitance, and conductance; Embed is the physical parameter embedding function; a' is the learnable fusion weight; F phys is the physical characteristic vector of the cable; F joint It is a fusion feature; Construct a probability mapping from features to locations: ; ; Among them, p(l|F joint ) is the probability distribution of position l; l i is the predefined i-th candidate position; is the predicted fault location; W loc and b locare the weight and bias of the probability mapping respectively; N is the number of candidate positions.

[0011] Furthermore, based on the fault type classification results and the fault interval and location information, the fault level is obtained and a diagnostic report is generated for early warning push, as follows: Combining historical data with expert experience, a fault level assessment model is constructed to obtain the fault level based on the fault type classification results and the fault occurrence interval and location information. A detailed diagnostic report is generated based on the fault type classification results, the range and location information of the fault, and the fault level. The diagnostic report includes a multi-source evidence chain, displaying the original information and processing process of multi-source data such as partial discharge signals, acoustic signals, temperature fields, etc. used for diagnosis. At the same time, disposal suggestions are provided based on the fault situation and historical maintenance experience. Different early warning mechanisms are set according to the fault level. For emergency-level faults, strong reminder notifications are immediately sent to operation and maintenance management personnel through the Web system and mobile APP. For severe-level faults, detailed fault information and disposal suggestions are pushed within the preset time. For general-level faults, messages can be pushed within the preset time and included in daily inspection task reminders.

[0012] A high-voltage cable insulation fault intelligent diagnosis system based on multi-information fusion includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the high-voltage cable insulation fault intelligent diagnosis method based on multi-information fusion as described above.

[0013] A computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the steps of the intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion as described above.

[0014] The present invention has the following beneficial effects: 1. The present invention can fully tap into the complementary information between different modal features through a hierarchical and multi-scale feature fusion method, generating a joint feature representation with higher expressive power, thereby improving the accuracy and reliability of fault diagnosis; 2. This invention leverages the complementarity of multimodal features and implements probabilistic prediction and uncertainty quantification of fault types through a Bayesian reasoning framework, effectively improving the accuracy and reliability of fault classification. Combining the characterization capabilities of deep learning with the interpretability of physical models, it achieves high-precision fault location while providing uncertainty quantification, providing a reliable basis for maintenance decisions. 3. The present invention realizes the intelligence of the entire process from fault diagnosis to decision-making execution, reduces the cost of manual analysis, improves the efficiency and safety of cable operation and maintenance, and effectively reduces power outages and economic losses caused by cable faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0016] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] Example 1 refer to Figure 1 In this embodiment, a high-voltage cable insulation fault intelligent diagnosis method based on multi-information fusion is provided, which includes the following steps: S1: Obtain multi-source operating data of high-voltage cables and construct a multi-source raw data matrix; S2: Preprocessing the acquired multi-source data of the high-voltage cable operation to obtain a standardized data matrix aligned in time and space; S3: Based on the standardized data matrix aligned in time and space, feature extraction is performed to obtain a multi-dimensional feature vector, and the intrinsic correlation between features is learned using a multimodal deep network to obtain a joint multimodal feature; S4: Build a fault type recognition model based on Bayesian reasoning and Monte Carlo sampling, and perform fault type discrimination based on the joint multimodal features to obtain the fault type classification results; S5: Based on the joint multimodal features, combined with deep learning and physical models, the fault interval and location information are obtained; S6: Based on the fault type classification results and the fault interval and location information, the fault level is obtained and a diagnostic report is generated for early warning push.

[0018] In this embodiment, multi-source data on the operation of high-voltage cables is obtained, including partial discharge signals (multi-channel, such as electrical, acoustic, ultrasonic, and optical), dielectric loss factor (using a dielectric loss tester), cable temperature and humidity, environmental parameters (installed with environmental sensors), operation and maintenance history data, and fault tags (database access), as follows: High-frequency current transformers are deployed at cable joints to detect electrical pulse signals, and piezoelectric ceramic sensors are installed along the cable body to capture ultrasonic signals. Fiber Bragg grating (FBG) arrays are embedded in the cable shield to monitor optical pulse signals in real time. A digital dielectric loss tester is installed on the cable grounding wire. The test signal is injected using the low-frequency superposition method (0.1 Hz / 10 kV) to obtain the time series data of the dielectric loss factor and capacitance change rate. A distributed optical fiber temperature measurement system is spirally wound along the cable outer sheath to obtain the axial temperature distribution matrix. A three-axis MEMS vibration sensor is installed at the cable joint to monitor the X / Y / Z three-axis vibration acceleration time series signal. Environmental sensors are used to obtain environmental parameter time series data and gas concentration spectra. Connect to the SCADA system via the OPC UA protocol to extract cable load curves, operation and maintenance history data, and fault tags; The above data is collected through the master control cabinet node. The master control node is equipped with GPS and Beidou dual-mode clock sources and synchronizes the clocks of all sensors through the PTP protocol. When any partial discharge channel detects a pulse >50pC, all sensors are synchronously triggered to record 1ms time window data. Establish a three-dimensional cable coordinate system and mark the installation location of each sensor (e.g. "HFCT@X=32.5m"). Temperature data is automatically mapped to the corresponding coordinate point; In this embodiment, the preprocessing includes filtering and denoising, multi-channel synchronous processing, distortion correction, and outlier screening, as follows: Get the original multi-source data matrix X according to S1 raw : X raw ={x1(t),x2(t),...,x a (t),...,x n (t)}; Among them, x a (t) represents the sensor signal of type a, including electrical signal, acoustic signal, temperature field, etc., and t is the time series; For different sensor signals, corresponding filtering and denoising processing is used. Elliptical filtering is used for electrical signals, EMD is used for acoustic signals, and wavelet threshold is used for optical signals to obtain the signal data X after denoising. denoised ; For the signal data X after noise reduction denoised , using multi-channel synchronous processing distortion correction, correcting the electrical signal amplitude through the transmission line model, and compensating the temperature field through heat conduction: Get the real value signal data X of the physical quantity corrected ; Finally, the local outlier factor and Mahalanobis distance are combined to eliminate abnormal data caused by hardware failure or environmental interference to obtain the standardized spatiotemporal alignment matrix X proc , including partial discharge (electrical, acoustic, and optical signals), dielectric loss factor (tanδ), temperature field distribution, vibration spectrum, and environmental parameters.

[0019] In this embodiment, feature extraction is performed based on the spatiotemporally aligned standardized data matrix to obtain multi-dimensional feature vectors and fuse them to obtain a joint multimodal feature representation, as follows: The standardized spatiotemporal alignment matrix X obtained from S2 proc, extract features from each modal data, including partial discharge signal features, acoustic signal features, and temperature field features; the partial discharge signal uses ResNet-18 to extract spatial features as partial discharge signal features; the ultrasonic signal uses a bidirectional LSTM to capture the evolution trend as the acoustic signal feature; the temperature field uses a Transformer encoder to model spatial dependencies as the temperature field feature; Based on the extracted modal features, Z-score standardization is used to calculate the f i Mutual information with the fault label y: ; in, is the joint probability density; Features f i Independent probability distribution of; p(y) is the prior probability of fault label y; Retain the top-K (e.g. K=50) features with high mutual information F MI , delete the redundant features that satisfy |p(fi,fj)|>0.7; p(f i ,f j ) is f i and f j The joint probability distribution of A multimodal deep network is used to learn the intrinsic correlation between high mutual information features, obtain a highly expressive representation of the fusion features, and construct a joint multimodal feature h fusion .

[0020] In this embodiment, the multimodal deep network includes a feature embedding layer, a feature interaction layer, an attention mechanism, and a feature fusion layer, as follows: The feature embedding layer maps the features of different modalities into a unified feature space: h pd =W pd ·f pd +b pd ; h sound =W sound ·f sound +b sound ; h temp =W temp ·f temp +b temp ; Where: f pd 、f sound 、f temp are the characteristic vectors of partial discharge, acoustic signal and temperature field respectively; W pd 、W sound 、Wtemp is a learnable weight matrix; b pd 、b sound 、b temp is the bias vector; h pd 、h sound 、h temp are the embedded eigenvectors of partial discharge, acoustic signal and temperature field respectively; The feature interaction layer models the complex associations between features of different modalities and uses a multimodal interaction network to learn the complex associations Z between features: Z=σ(W v ·[h pd ;h sound ;h temp ]+b v ); Among them, [;] represents the concatenation operation of features; W v and b v are the weights and biases of the feature interaction layer; σ is the activation function; The self-attention mechanism MultiHead is used to capture long-distance dependencies between features: Q=W Q Z, K=W K Z, V=W V ·Z; ; ; Among them, Q, K, V represent table query, key and value matrices respectively, and W Q 、W K 、W V is the parameter matrix corresponding to Q, K, and V; , , is the learnable projection matrix corresponding to Q, K, and V of the h-th attention head; head h is the hth attention head; H is the number of attention heads; Attention represents the attention mechanism calculation function; W O is the output weight matrix; Concat represents the concatenation function; Adaptively weight the importance of different modes: ; in, is the attention weight of each modality; W a1 , W a2 , W a3 and b a1 , b a2 ,b a3 is a learnable parameter; The feature fusion layer combines the results of multi-head attention and modal attention to generate the final joint multimodal feature representation h fusion : ; ; ; Among them, h attn is the output of the self-attention mechanism; h modal is the output of the modal attention mechanism; W f and b f are the weights and biases of the feature fusion layer.

[0021] In this embodiment, a fault type recognition model is constructed based on Bayesian reasoning and Monte Carlo sampling, and fault type discrimination is performed based on the combined multimodal features to obtain the fault type classification results, as follows: First, nonlinear transformation and feature enhancement are performed on the joint multimodal features: F enhanced =σ(W1·h fusion +b1)+h fusion ; F attention =Softmax(W2·F enhanced +b2)⊙F enhanced ; Among them, ⊙ represents element-by-element multiplication; F enhanced represents the output after nonlinear transformation; F attention Represents the output after feature enhancement; W1 and b1 are the weight and bias of nonlinear transformation respectively; W2 and b2 are the weight and bias after feature enhancement respectively; Construct a multi-scale feature extraction module and process it through maximum pooling MaxPool and average pooling AvgPool: F local =MaxPool(W3·F attention +b3); F global =AvgPool(W4·F attention +b4); F multi-scale =Concat(F local ,F global ); Among them, F local represents the output after maximum pooling; F global Represents the output after average pooling; W3 and b3 are the weight and bias of the maximum pooling MaxPool respectively; W4 and b4 are the weight and bias of the average pooling AvgPool respectively; Fmulti-scale Represents the extracted multi-scale features; Build a classifier based on Bayesian inference and output probability distribution : ; ; ; in, is random noise, is the probability density function of the noise; and are the mean vector and standard deviation vector of category y respectively; W σ and b σ are the weight and bias of standard deviation prediction respectively; and are the weight and bias of mean prediction respectively; ; ; in, is the average probability expectation; T is the number of sampling times; is the random noise of the t-th sampling; Uncertainty is the information entropy of the model prediction distribution; c represents category c; Simultaneously optimize classification and uncertainty estimation, the objective function for: ; in, is the cross entropy loss; is the regularization loss; is the uncertainty constraint loss, using entropy regularization; 、 is the weight coefficient; Final classification decision, predicted fault type: ; in, is the probability of predicting the fault type c.

[0022] In this embodiment, based on the joint multimodal features, combined with deep learning and physical models, the fault interval and location information are obtained, as follows: From the joint multimodal feature h fusion Extract spatiotemporal features from: F temp =TCN(h fusion ); F space =GCN([F temp ,A]); Among them, TCN is a temporal convolutional network that captures the temporal features F temp ; GCN is a graph convolutional network, combined with the adjacency matrix A to model the spatial relationship F space ; Fusion of cable physical parameters with spatiotemporal characteristics and embedding them into the positioning process: F phys =Embed(R,L,C,G); F joint =a'·F space +(1-a')·F phys ; Among them, R, L, C, G are the physical parameters of the cable, namely resistance, inductance, capacitance, and conductance; Embed is the physical parameter embedding function; a' is the learnable fusion weight; F phys is the physical characteristic vector of the cable; F joint It is a fusion feature; Construct a probability mapping from features to locations: ; ; Among them, p(l|F joint ) is the probability distribution of position l; l i is the predefined i-th candidate position; is the predicted fault location; W loc and b loc are the weight and bias of the probability mapping respectively; N is the number of candidate positions.

[0023] In this embodiment, based on the fault type classification result and the fault occurrence interval and location information, the fault level is obtained and a diagnostic report is generated for early warning push, as follows: Combining historical data with expert experience, a fault level assessment model is constructed to obtain the fault level based on the fault type classification results and the interval and location information of the fault occurrence.

[0024] Different fault types are weighted accordingly. For example, partial discharge faults are given a higher weight due to their potential damage to cable insulation, while minor temperature anomalies are weighted relatively lower. The fault's location is also considered. If a fault occurs at a critical cable node or on a power line in a densely populated area, the impact and potential damage are greater, increasing the fault level score. By calculating a weighted composite score based on fault type and location, faults are classified into different levels, such as urgent (requiring immediate attention), severe (requiring attention within 24 hours), and general (routine maintenance can be scheduled).

[0025] Specifically, the fault type is scored according to its severity (e.g., 1-10 points): major fault 10 points, general fault 5 points, minor fault 2 points; the fault location is scored according to its importance (e.g., 1-10 points): core components 10 points, key components 6 points, minor components 3 points. The comprehensive score = (fault type score × fault type weight) + (fault location score × fault location weight).

[0026] For example, when a serious partial discharge fault is detected at a cable joint, combined with its location in the power supply line in the core area of ​​the city, the assessment model calculates that it is an emergency level fault.

[0027] A detailed diagnostic report is generated based on the fault type classification, fault location, and fault severity. This report includes a multi-source evidence chain, showcasing the raw information and processing of multiple data sources, including partial discharge signals, acoustic signals, and temperature fields, used for diagnosis. The report clearly presents the diagnostic conclusion, such as "Analysis indicates a type XX fault has occurred in the cable at location XX, and the fault severity is urgent." Furthermore, based on the fault situation and historical maintenance experience, recommendations are provided, such as "Immediately cut off power to the area, dispatch a professional maintenance team to replace the faulty cable connector, and conduct a comprehensive inspection of surrounding cables." Furthermore, the report includes information such as the probability of the fault occurring and the predicted impact range, presented in intuitive charts to facilitate quick understanding of the fault's full scope.

[0028] Different early warning mechanisms are set according to the fault level. For emergency-level faults, strong reminder notifications such as pop-ups, vibrations, and voice broadcasts are immediately sent to operation and maintenance management personnel through the Web system and mobile APP to ensure timely communication of information; for serious-level faults, detailed fault information and handling suggestions are pushed within the preset time (10 minutes); for general-level faults, messages can be pushed within the preset time (1 hour) and included in daily inspection task reminders.

[0029] For example, a 10kV high-voltage cable line in a certain city experienced a sudden fault during operation. The operation and maintenance center collected multi-source data, including partial discharge signals (time series electrical data), temperature distribution data (infrared thermal imaging), geographic location information (GIS coordinates), historical operation and maintenance records, and current / voltage waveforms. The intelligent diagnosis method for high-voltage cable insulation faults based on this solution, which is based on multi-information fusion, is as follows: Obtain multi-source data of high-voltage cables (as shown in Table 1) and construct a multi-source original data matrix.

[0030] Table 1 Original data:

[0031] Output joint multimodal features, h fusion=[0.72,0.10,...,0.94], representing the multidimensional comprehensive health status characteristics of this time and space point.

[0032] The fault type identification network output (fault classification probability) is performed through Bayesian inference Monte Carlo sampling, as shown in Table 2.

[0033] Table 2 Fault classification probability

[0034]

[0035] The high probability of an insulation breakdown event is inferred. Using a deep neural network based on the characteristics, the high-probability fault interval is inferred: "1.4km to 1.8km from substation A." Finally, the fault is automatically determined to be an emergency level, requiring urgent repair. A diagnostic report is automatically generated and pushed to the operation and maintenance app and the main screen, along with recommendations: Diagnostic report summary: Fault type: insulation breakdown; Fault range: 1.4km~1.8km; Fault level: emergency level fault; Speculated cause: Partial discharge and high temperature superimposed, resulting in continuous insulation degradation; Recommended measures: Immediately isolate the faulty section, send personnel to conduct on-site inspection and confirmation, and prepare spare cables.

[0036] This embodiment also provides an intelligent diagnosis system for high-voltage cable insulation faults based on multi-information fusion, including a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the intelligent diagnosis method for high-voltage cable insulation faults based on multi-information fusion as described above.

[0037] This embodiment further provides a computer storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the steps of the intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion as described above.

[0038] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0039] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0040] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0041] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other manner. Any person skilled in the art may utilize the above-disclosed technical content to modify or modify the present invention into equivalent embodiments. However, any simple modifications, equivalent variations, and modifications to the above embodiments that do not depart from the technical content of the present invention and are based on the technical essence of the present invention remain within the scope of protection of the present invention.

Claims

1. An intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion, characterized in that: The following steps are involved: S1: Obtain multi-source operating data of high-voltage cables and construct a multi-source raw data matrix; S2: Preprocessing the acquired multi-source data of the high-voltage cable operation to obtain a standardized data matrix aligned in time and space; S3: Based on the standardized data matrix aligned in time and space, feature extraction is performed to obtain a multi-dimensional feature vector, and the intrinsic correlation between features is learned using a multimodal deep network to obtain a joint multimodal feature; S4: Build a fault type recognition model based on Bayesian reasoning and Monte Carlo sampling, and perform fault type discrimination based on the joint multimodal features to obtain the fault type classification results; S5: Based on the joint multimodal features, combined with deep learning and physical models, the fault interval and location information are obtained; S6: Based on the fault type classification results and the fault interval and location information, the fault level is obtained and a diagnostic report is generated for early warning push.

2. The intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion according to claim 1 is characterized in that: The acquisition of multi-source data on the operation of high-voltage cables, including partial discharge signals, dielectric loss factor, cable temperature and humidity, environmental parameters, operation and maintenance history data, and fault tags, is as follows: High-frequency current transformers are deployed at cable joints to detect electrical pulse signals, and piezoelectric ceramic sensors are installed along the cable body to capture ultrasonic signals. Fiber Bragg grating arrays are embedded in the cable shield to monitor optical pulse signals in real time. Install a digital dielectric loss tester on the cable grounding wire, inject the test signal using the low-frequency superposition method, and obtain the time series data of the dielectric loss factor and capacitance change rate; A distributed optical fiber temperature measurement system is spirally wound along the cable outer sheath to obtain the axial temperature distribution matrix. A three-axis MEMS vibration sensor is installed at the cable joint to monitor the X / Y / Z three-axis vibration acceleration time series signal. Environmental sensors are used to obtain environmental parameter time series data and gas concentration spectra. Connect to the SCADA system via the OPC UA protocol to extract cable load curves, operation and maintenance history data, and fault tags; The above data is collected through the master control cabinet node. The master control node is equipped with GPS and Beidou dual-mode clock sources and synchronizes the clocks of all sensors through the PTP protocol. When any partial discharge channel detects a pulse >50pC, all sensors are synchronously triggered to record 1ms time window data. Establish a three-dimensional coordinate system for the cable, mark the installation location of each sensor, and automatically map the temperature data to the corresponding coordinate points.

3. The intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion according to claim 2 is characterized in that: The preprocessing includes filtering and denoising, multi-channel synchronous processing, distortion correction and outlier screening, as follows: Get the original multi-source data matrix X according to S1 raw :X raw ={x1(t),x2(t),...,x a (t),...,x n (t)}; Among them, x a (t) represents the sensor signal of type a, including electrical signal, acoustic signal, and temperature field, and t is the time series; For different sensor signals, corresponding filtering and denoising processing is used. Elliptical filtering is used for electrical signals, EMD is used for acoustic signals, and wavelet threshold is used for optical signals to obtain the signal data X after denoising. denoised ; For the signal data X after noise reduction denoised , using multi-channel synchronous processing distortion correction, correcting the electrical signal amplitude through the transmission line model, and compensating the temperature field through heat conduction: Get the real value signal data X of the physical quantity corrected ; Finally, the local outlier factor and Mahalanobis distance are combined to eliminate abnormal data caused by hardware failure or environmental interference to obtain the standardized spatiotemporal alignment matrix X proc .

4. The intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion according to claim 1 is characterized in that: The feature extraction is performed based on the standardized data matrix aligned in time and space, and multi-dimensional feature vectors are obtained and fused to obtain a joint multimodal feature representation, as follows: The standardized spatiotemporal alignment matrix X obtained from S2 proc , extract features from each modal data, including partial discharge signal features, acoustic signal features, and temperature field features; the partial discharge signal uses ResNet-18 to extract spatial features as partial discharge signal features; the ultrasonic signal uses a bidirectional LSTM to capture the evolution trend as the acoustic signal feature; the temperature field uses a Transformer encoder to model spatial dependencies as the temperature field feature; Based on the extracted modal features, Z-score standardization is used to calculate the f i Mutual information with the fault label y : ； in, is the joint probability density; is the feature f i Independent probability distribution of; p(y) is the prior probability of fault label y; Preserve the feature F of Top-K high mutual information MI , delete the ones that satisfy |p(f i ,f j )|>0.7 redundant features, where p(f i ,f j ) is f i and f j The joint probability distribution of A multimodal deep network is used to learn the intrinsic correlation between high mutual information features, obtain a highly expressive representation of the fusion features, and construct a joint multimodal feature h fusion .

5. The intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion according to claim 4 is characterized in that: The multimodal deep network includes a feature embedding layer, a feature interaction layer, an attention mechanism, and a feature fusion layer, as follows: The feature embedding layer maps the features of different modalities into a unified feature space: h pd =W pd ·f pd +b pd ; h sound =W sound ·f sound +b sound ; h temp =W temp ·f temp +b temp ; Where: f pd 、f sound 、f temp are the eigenvectors of partial discharge, acoustic signal and temperature field respectively; W pd 、W sound 、W temp is a learnable weight matrix; b pd 、b sound 、b temp is the bias vector; h pd 、h sound 、h temp are the embedded eigenvectors of partial discharge, acoustic signal and temperature field respectively; The feature interaction layer models the complex associations between features of different modalities and uses a multimodal interaction network to learn the complex associations Z between features: Z=σ(W v ·[h pd ;h sound ;h temp ]+b v ); Among them, [;] represents the concatenation operation of features; W v and b v are the weights and biases of the feature interaction layer; σ is the activation function; The self-attention mechanism MultiHead is used to capture long-distance dependencies between features: Q=W Q ·Z,K=W K ·Z,V=W V ·Z; ； ； Among them, Q, K, V represent table query, key and value matrices respectively, and W Q 、W K 、W V is the parameter matrix corresponding to Q, K, and V; , , is the learnable projection matrix corresponding to Q, K, and V of the h-th attention head; head h is the hth attention head; H is the number of attention heads; Attention represents the attention mechanism calculation function; W O is the output weight matrix; Concat represents the concatenation function; Adaptively weight the importance of different modes: ； in, is the attention weight of each modality; W a1 、W a2 、W a3 and b a1 、b a2 、b a3 is a learnable parameter; The feature fusion layer combines the results of multi-head attention and modal attention to generate the final joint multimodal feature representation h fusion : ； ； ； Among them, h attn is the output of the self-attention mechanism; h modal is the output of the modal attention mechanism; W f and b f are the weights and biases of the feature fusion layer.

6. The intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion according to claim 5 is characterized in that: The fault type recognition model is constructed based on Bayesian reasoning and Monte Carlo sampling, and the fault type is distinguished according to the joint multimodal features to obtain the fault type classification results, as follows: First, nonlinear transformation and feature enhancement are performed on the joint multimodal features: F enhanced =σ(W1 h fusion +b1)+h fusion ; <h2 style=";text-align:left;direction:ltr">F<h2 style=";text-align:left;direction:ltr"> attention <h2 style=";text-align:left;direction:ltr"> =Softmax(W2·F<h2 style=";text-align:left;direction:ltr"> enhanced <h2 style=";text-align:left;direction:ltr"> +b2)⊙F<h2 style=";text-align:left;direction:ltr"> enhanced <h2 style=";text-align:left;direction:ltr"> ; Among them, ⊙ represents element-by-element multiplication; F enhanced represents the output after nonlinear transformation; F attention Represents the output after feature enhancement; W1 and b1 are the weight and bias of nonlinear transformation respectively; W2 and b2 are the weight and bias after feature enhancement respectively; Construct a multi-scale feature extraction module and process it through maximum pooling MaxPool and average pooling AvgPool: <h2 style=";text-align:left;direction:ltr">F<h2 style=";text-align:left;direction:ltr"> local <h2 style=";text-align:left;direction:ltr"> =MaxPool(W3·F<h2 style=";text-align:left;direction:ltr"> attention <h2 style=";text-align:left;direction:ltr"> +b3); F global =AvgPool(W4·F attention +b4); F multi-scale =Concat(F local ,F global ); Among them, F local represents the output after maximum pooling; F global Represents the output after average pooling; W3 and b3 are the weight and bias of the maximum pooling MaxPool respectively; W4 and b4 are the weight and bias of the average pooling AvgPool respectively; F multi-scale Represents the extracted multi-scale features; Build a classifier based on Bayesian inference and output probability distribution : ； ； ； in, is random noise, is the probability density function of the noise; and are the mean vector and standard deviation vector of category y respectively; W σ and b σ are the weight and bias of standard deviation prediction respectively; and are the weight and bias of mean prediction respectively; Estimate classification uncertainty via Monte Carlo sampling: ； ； in, is the average probability expectation; T is the number of sampling times; is the random noise of the t-th sampling; Uncertainty is the information entropy of the model prediction distribution; c represents category c; Simultaneously optimize classification and uncertainty estimation, the objective function for: ； in, is the cross entropy loss; is the regularization loss; is the uncertainty constraint loss, using entropy regularization; 、 is the weight coefficient; Final classification decision, predicted fault type: ； in, is the probability of predicting the fault type c.

7. The intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion according to claim 6 is characterized in that: The above method combines the multimodal features, deep learning and physical models to obtain the fault interval and location information, as follows: From the joint multimodal feature h fusion Extract spatiotemporal features from: F temp =TCN(h fusion ); F space =GCN([F temp ,A]); Among them, TCN is a temporal convolutional network that captures the temporal features F temp ; GCN is a graph convolutional network, combined with the adjacency matrix A to model the spatial relationship F space ; Fusion of cable physical parameters with spatiotemporal characteristics and embedding them into the positioning process: F phys =Embed(R,L,C,G); F joint =a’·F space +(1-a’)·F phys ; Among them, R, L, C, G are the physical parameters of the cable, namely resistance, inductance, capacitance, and conductance respectively; Embed is the physical parameter embedding function; a' is the learnable fusion weight; F phys is the physical characteristic vector of the cable; F joint It is a fusion feature; Construct a probability mapping from features to locations: ； ； Among them, p(l|F joint ) is the probability distribution of position l; l i is the predefined i-th candidate position; is the predicted fault location; W loc and b loc are the weight and bias of the probability mapping respectively; N is the number of candidate positions.

8. The intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion according to claim 1 is characterized in that: Based on the fault type classification results and the fault occurrence interval and location information, the fault level is obtained and a diagnostic report is generated for early warning push, as follows: Combining historical data with expert experience, a fault level assessment model is constructed to obtain the fault level based on the fault type classification results and the fault occurrence interval and location information. A detailed diagnostic report is generated based on the fault type classification, fault location, and fault severity. This report includes a multi-source evidence chain, showcasing the original information and processing of the multi-source data used for diagnosis, including partial discharge signals, acoustic signals, and temperature fields. Furthermore, treatment recommendations are provided based on the fault situation and historical maintenance experience. Different early warning mechanisms are set based on fault severity. For emergency faults, strong alerts are immediately sent to maintenance personnel via the web system and mobile app. For serious faults, detailed fault information and handling suggestions will be pushed within the preset time; for general faults, messages can be pushed within the preset time and included in the daily inspection task reminders.

9. An intelligent diagnosis system for high-voltage cable insulation faults based on multi-information fusion, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory. When the processor executes the computer program, the method specifically performs the steps in the intelligent diagnosis method of high-voltage cable insulation fault based on multi-information fusion as described in any one of claims 1 to 8.

10. A computer storage medium, characterized in that The computer storage medium stores a plurality of instructions, which are suitable for being loaded by a processor and executing the steps in the intelligent diagnosis method for high-voltage cable insulation fault based on multi-information fusion as described in any one of claims 1 to 8.

Images