Cable insulation aging multi-modal data fusion evaluation system

Abstract

The invention discloses a cable insulation aging multi-modal data fusion evaluation system, particularly relates to the field of power equipment state monitoring and evaluation, and comprises a multi-dimensional data acquisition module, a progressive data processing module, a hierarchical analysis module, a multi-stage output module, a data storage module and a self-diagnosis module. The multi-dimensional data acquisition module synchronously acquires electrical, thermal and chemical parameters; the progressive data processing module carries out three-level processing on data, namely preprocessing, feature extraction and fusion, the hierarchical analysis module evaluates the aging level through single-mode identification, multi-mode association verification and a fuzzy comprehensive evaluation model, and the multi-level output module outputs basic snapshots, early warning information and trend reports. The data storage module stores full-process data, and the self-diagnosis module regularly detects and pushes a fault report. The system solves the problems of one-sided evaluation, data asynchronization and the like in the prior art, improves the evaluation accuracy and the operation and maintenance convenience, and guarantees the safe operation of the cable.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring and assessment technology, and more specifically, to a multi-modal data fusion assessment system for cable insulation aging. Background Technology

[0002] As power systems develop towards higher voltage, larger capacity, and wider coverage, cables, as the core carrier of power transmission, directly determine the stability and security of power supply through their operational reliability. During long-term service, cable insulation layers must continuously withstand the coupled effects of electrical stress, thermal stress, and chemical stress, making them prone to irreversible aging phenomena such as molecular chain breakage, oxidative decomposition, and increased dielectric loss. If not assessed in a timely and accurate manner, this may lead to faults such as insulation breakdown and phase-to-phase short circuits, causing large-scale power outages and severely impacting industrial production and residential electricity consumption.

[0003] To address the challenges of assessing cable insulation aging, the industry's technological development has gone through three stages: the early stage focused on single-parameter monitoring, such as collecting partial discharge through high-frequency current sensors, monitoring cable surface temperature using thermocouples, or analyzing the content of a single gas in insulating oil using gas chromatography; the middle stage gradually attempted to superimpose multiple parameters, simply combining electrical and temperature data; and in recent years, the concept of multimodal data assessment has been proposed, which integrates electrical, thermal, and chemical data to meet the power industry's high-precision and high-reliability requirements for the entire life cycle management of cables.

[0004] However, in practical use, it still has some drawbacks. For example, most existing technologies still rely on single-dimensional parameters, ignoring the "electro-thermal-chemical" coupling characteristics of insulation aging. For instance, when judging solely by the amount of partial discharge, electromagnetic interference is easily misjudged as insulation defects, leading to false alarms. Furthermore, existing technologies have poor spatiotemporal synchronization and lack a unified clock reference, resulting in a time difference of hundreds of hours between real-time partial discharge data and infrared thermal imaging inspection data, making it impossible to establish a causal relationship between "discharge-temperature-chemical". Additionally, data processing is unsystematic, lacks targeted noise reduction, and the raw data has a high noise rate, resulting in poor performance. The extraction is incomplete, containing only time-domain features and lacking the frequency-domain dominant frequency and time-frequency domain wavelet packet energy spectrum. The simple fusion method results in high data dimensionality and computational complexity, with an evaluation delay exceeding 30 seconds, which cannot meet the needs of real-time monitoring. At the same time, the output of the existing system is mostly a single status indicator without hierarchical output logic. There are no basic data snapshots for technicians to review, no accurate early warnings, and no long-term trends. Maintenance personnel can only obtain a vague conclusion of "whether there is a fault" and cannot quickly locate the fault location, formulate maintenance priorities, or plan maintenance in advance. The engineering practicality of the output results is seriously insufficient. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a multi-modal data fusion evaluation system for cable insulation aging and its execution device, which solves the problems mentioned in the background art through the following solutions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-modal data fusion evaluation system for cable insulation aging, comprising: A multi-dimensional data acquisition module is used to simultaneously acquire the electrical, thermal, and chemical parameters of the cable; The progressive data processing module is connected to the multi-dimensional data acquisition module and performs three levels of processing on the acquired parameters: first-level preprocessing, second-level feature extraction, and third-level data fusion. The hierarchical analysis module is connected to the progressive data processing module and performs three levels of analysis on the processed data, namely, first-level single-modal state identification, second-level multimodal correlation verification, and third-level aging level assessment. A multi-level output module is connected to the hierarchical analysis module and outputs basic information, early warning information and trend prediction information based on the analysis results. The data storage module is connected to the multi-dimensional data acquisition module, the progressive data processing module, and the hierarchical analysis module, and is used to store the data from all modules. The self-diagnosis module is used to periodically check the hardware status, data transmission links, and algorithm operation of each module. The self-diagnosis cycle is 24 hours. When a hardware failure, data transmission interruption, or algorithm abnormality is detected, a diagnostic report is automatically generated and pushed to the operation and maintenance terminal.

[0007] Preferably, the multi-dimensional data acquisition module includes: The electrical parameter acquisition unit uses a high-frequency current sensor, a 0.2-class voltage transformer, and a 0.1-class current transformer. The high-frequency current sensor has a bandwidth of 10kHz-100MHz, a sampling rate of 200MS / s, a data length of 8192 points per frame, a sampling interval of 50ms between the voltage transformer and the current transformer, and a timestamp accuracy of ±1μs. The thermal parameter acquisition unit includes a distributed fiber optic sensor and an infrared thermal imager. The distributed fiber optic sensor has a spatial resolution of 1m, a temperature measurement range of -50℃ to 150℃, and generates a temperature distribution curve every 30s. The infrared thermal imager has a resolution of 640×512, scans once a week, and synchronously records GPS location with an accuracy of ±3m. The chemical parameter acquisition unit includes a gas sensor array and a gas chromatograph. The gas sensor array has a detection limit of 0.1 ppm, a response time of <20 s, and stores one set of data per hour. The gas chromatograph has a detection accuracy of 0.01 μL / L and analyzes the content of furan compounds in insulating oil samples once per quarter.

[0008] Preferably, the primary preprocessing of the progressive data processing module includes: The discharge signal was denoised using a 3-layer decomposition with a db6 wavelet threshold of λ=0.8σ, where σ is the noise standard deviation, eliminating narrowband interference of 50Hz±2Hz. Voltage and current signals were filtered using a 50Hz notch filter, and a 24-hour effective value sequence was calculated to eliminate outliers greater than 1.1 times the rated value. Optical fiber temperature was filtered using a sliding window mean filter with a window size of 5m to eliminate abrupt changes deviating from adjacent points by ±5℃. Infrared images were matched with the cable's 3D model using the SIFT algorithm to extract the surface temperature field distribution, and the temperature values ​​in non-cable areas were set to zero. Gas concentration was denoted using a first-order difference method, with secondary sampling calibration initiated when |ΔC / Δt|>0.5ppm / h. Oil sample data underwent baseline correction and peak area calculation according to GB / T17623-2018 standard. The secondary feature extraction of the progressive data processing module includes: Extract time-domain features, including the average amplitude of the discharge and the maximum temperature gradient; extract frequency-domain features, including the dominant frequency f0 of the discharge signal after FFT transformation and the 12-hour spectral entropy of the gas concentration; extract time-frequency domain features: wavelet packet energy spectrum of the discharge signal, containing 8 frequency bands with a center frequency of 10kHz-80MHz; The three-level data fusion of the progressive data processing module includes: Time alignment: Based on the GPS clock, all data is interpolated to a 1-minute time scale; Spatial matching: Establish a three-dimensional coordinate system for the cable, with the X-axis representing length, the Y-axis representing depth, and the Z-axis representing height, and map the temperature and discharge location to the coordinate points; Feature fusion: Construct a 32-dimensional feature vector, including 12 electrical dimensions, 8 thermal dimensions, and 12 chemical dimensions, and reduce the dimensionality to a cumulative contribution rate of ≥95% through PCA.

[0009] Preferably, the first-level single-modal state identification of the hierarchical analysis module includes: Electrical condition: Calculate the skewness S of the partial discharge phase distribution spectrum; if S > 1.2, it is determined to be a tip discharge. Thermal condition: Calculate the steady-state temperature rise ΔT = I using a thermal circuit model. 2 Rθ, where I is the cable load current, R is the conductor resistance, and θ is the total thermal resistance. When ΔT > 10K, it is marked as an overheated area; Chemical status: According to IEC60599 standard, a chemical warning is triggered when the total hydrocarbon content > 150μL / L; The second-level multimodal correlation verification of the hierarchical analysis module includes: Establish the correlation matrix M pqWhen |Mpq| > 0.7, a strong correlation is determined; when |Mpq| < 0.3, a historical database comparison is initiated. The historical data consists of multimodal data of similar cables under the same voltage level and laying environment over the past 3 years. The hierarchical analysis module employs a fuzzy comprehensive evaluation model for its three-level aging assessment, including a factor set U = {u1, u2, u3}, where u1 is the discharge characteristic d. pd u2 represents the temperature characteristic t cv u3 is a chemical characteristic c hc The weight vector A = [0.4, 0.3, 0.3] is determined based on the analytic hierarchy process. It is obtained by constructing a judgment matrix, calculating the eigenvector corresponding to the largest eigenvalue, and normalizing it. The comment set V = {V1, V2, V3, V4} corresponds to healthy, mildly aged, moderately aged, and severely aged, respectively, through membership degree calculation.

[0010] Preferably, the multi-level output module includes: Basic output: A state snapshot is generated every 15 minutes, containing k. pca =8-dimensional principal component values, single-mode state indicators including tip discharge / no tip discharge, overheating / no overheating, chemical warning / no chemical warning, snapshot format is both PDF and Excel; Warning output: When the level reaches V3, an audible and visual alarm is triggered. The alarm sound pressure level is ≥85dB, the alarm light flashes red at a frequency of 2Hz, and a maintenance work order containing three-dimensional positioning is pushed. The work order includes the aging level, the characteristic parameters that exceed the standard, and the recommended maintenance measures. Trend Output: Monthly aging rate reports are generated, and the remaining lifespan is predicted using the grey model GM(1,1). The model construction steps are as follows: S1.1, Historical aging level data Accumulation generation , where n gm =12 represents selecting historical data from the most recent 12 months; The original observed value of the cable insulation aging level for month α; S1.2 Establishing the differential equation , where a is the development coefficient, reflecting the development trend and rate of change of the data sequence; b is the gray action quantity, reflecting the influence of external factors during the data change process; S1.3 Solving for parameters [a, b] using the least squares method T =(B T B) −1 B T y gm,N ; S1.4. Based on the solved parameters a and b, predict the cumulative generated sequence to obtain the cumulative predicted value at future time k+1, and then subtract and restore it to obtain the original data sequence. The predicted value at future time k+1.

[0011] The technical effects and advantages of this invention are as follows: This invention uses a multi-dimensional data acquisition module to simultaneously collect electrical, thermal, and chemical parameters, avoiding misjudgment based on a single parameter. For example, by combining partial discharge and temperature data, it can distinguish between electromagnetic interference and actual insulation defects, thus improving the accuracy of the assessment. This invention uses GPS clock as a unified reference, interpolates data to a 1-minute time scale to reduce synchronization errors, establishes a causal relationship between "discharge-temperature-chemistry", and reduces the noise rate of the original data through a progressive data processing module, using targeted denoising such as db6 wavelet denoising and 50Hz notch filtering, fully extracting time and frequency domain features, and then using PCA to reduce the dimensionality to 8-dimensional principal components, reducing data dimensionality and computational complexity, significantly shortening the evaluation delay, and meeting the needs of real-time monitoring. The multi-level output module of this invention includes basic snapshots, precise early warnings, and monthly generation of long-term trend reports (using the GM(1,1) model to predict remaining lifespan). Maintenance personnel can quickly locate faults, formulate repair priorities and maintenance plans, and the output results are highly practical. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the overall structure of the present invention. Detailed Implementation

[0013] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] As attached Figure 1 The cable insulation aging multimodal data fusion evaluation system shown includes: A multi-dimensional data acquisition module is used to simultaneously acquire the electrical, thermal, and chemical parameters of the cable; The multi-dimensional data acquisition module includes: The electrical parameter acquisition unit uses a high-frequency current sensor, a 0.2-class voltage transformer, and a 0.1-class current transformer. The high-frequency current sensor has a bandwidth of 10kHz-100MHz, a sampling rate of 200MS / s, a data length of 8192 points per frame, a sampling interval of 50ms between the voltage transformer and the current transformer, and a timestamp accuracy of ±1μs. The thermal parameter acquisition unit includes a distributed fiber optic sensor and an infrared thermal imager. The distributed fiber optic sensor has a spatial resolution of 1m, a temperature measurement range of -50℃ to 150℃, and generates a temperature distribution curve every 30s. The infrared thermal imager has a resolution of 640×512, scans once a week, and synchronously records GPS location with an accuracy of ±3m. The chemical parameter acquisition unit includes a gas sensor array and a gas chromatograph. The gas sensor array has a detection limit of 0.1 ppm, a response time of <20 s, and stores one set of data per hour. The gas chromatograph has a detection accuracy of 0.01 μL / L and analyzes the content of furan compounds in insulating oil samples once per quarter. It should be further explained that the installation method of the electrical parameter acquisition unit is as follows: One set of high-frequency current sensors (RC-100) is installed at each of the J1, J2, and J3 connector wells, sleeved on the cable grounding wire, with the sensor axis parallel to the cable axis and a spacing ≥10cm; voltage transformers (JDZ10-10) and current transformers (LMK-0.66) are installed at the cable terminals of the East Substation and the First Line Substation, with the secondary side of the transformers connected to a data acquisition card (NIPCIe-6363), which is connected to the local server via a network cable; all electrical acquisition equipment is grounded (grounding resistance ≤4Ω), and shielded cables (single-end grounding of the shielding layer) are used for signal transmission lines to reduce electromagnetic interference; The installation method of the thermal parameter acquisition unit is as follows: a distributed optical fiber sensor (G.652D) is laid parallel to the cable laying path, with a distance of ≤30cm between the optical fiber and the cable. Bending should be avoided during laying, and the bending radius should be ≥30mm. The two ends are respectively connected to the OTDR (AnritsuMS9740A) of the East Substation and the First Line Substation; a track-type inspection robot (CR-100) is installed above the cable path with a track length of 5km. The robot is equipped with an infrared thermal imager (FLIRA655sc), and a GPS module (UBloxNEO-7M) is fixed on the top of the robot to ensure positioning accuracy. The chemical parameter acquisition unit is installed as follows: one gas sensor array is installed in each of the J1, J2, and J3 joint wells. The sensor box is fixed to the well wall, 30cm away from the joint surface. The sampling tube extends into the space around the joint through a Φ5mm hole. Insulating oil samples are collected quarterly at the cable terminals of the East Substation and the First Line Substation (using clean sampling bottles, rinsing three times before sampling, according to GB / T7597-2007 standard) and sent to the laboratory for analysis by gas chromatograph.

[0015] The progressive data processing module is connected to the multi-dimensional data acquisition module and performs three levels of processing on the acquired parameters: first-level preprocessing, second-level feature extraction, and third-level data fusion. The primary preprocessing of the progressive data processing module includes: The discharge signal was denoised using a 3-layer decomposition with a db6 wavelet threshold of λ=0.8σ, where σ is the noise standard deviation, eliminating narrowband interference of 50Hz±2Hz. Voltage and current signals were filtered using a 50Hz notch filter, and a 24-hour effective value sequence was calculated to eliminate outliers greater than 1.1 times the rated value. Optical fiber temperature was filtered using a sliding window mean filter with a window size of 5m to eliminate abrupt changes deviating from adjacent points by ±5℃. Infrared images were matched with the cable's 3D model using the SIFT algorithm to extract the surface temperature field distribution, and the temperature values ​​in non-cable areas were set to zero. Gas concentration was denoted using a first-order difference method, with secondary sampling calibration initiated when |ΔC / Δt|>0.5ppm / h. Oil sample data underwent baseline correction and peak area calculation according to GB / T17623-2018 standard. The secondary feature extraction of the progressive data processing module includes: Extract time-domain features, including the average amplitude of the discharge. Where N is the number of discharges counted, and A i The amplitude and maximum temperature gradient of the i-th discharge. , and The values ​​represent the temperature values ​​of two adjacent monitoring points, with 1m representing the distance between the two points; frequency domain features are extracted, including the dominant frequency f0 of the discharge signal after FFT transformation and the 12-hour spectral entropy of the gas concentration. p i It is the probability of the i-th frequency component in the gas concentration spectrum; extract time-frequency domain features: wavelet packet energy spectrum of the discharge signal, containing 8 frequency bands, with a center frequency of 10kHz-80MHz; The three-level data fusion of the progressive data processing module includes: Time alignment: Based on the GPS clock, all data is interpolated to a 1-minute time scale; Spatial matching: Establish a three-dimensional coordinate system for the cable, with the X-axis representing length, the Y-axis representing depth, and the Z-axis representing height, and map the temperature and discharge location to the coordinate points; Feature fusion: A 32-dimensional feature vector is constructed, consisting of 12 electrical features, 8 thermal features, and 12 chemical features. Dimensionality is reduced using PCA to achieve a cumulative contribution rate of ≥95%. The electrical characteristics include: average amplitude of partial discharge, maximum amplitude, number of discharge pulses within 1 minute, main frequency of partial discharge signal after FFT transformation, and energy proportion of 8 wavelet packet frequency bands (E1 / Etotal to E8 / Etotal). Thermal characteristics include: the maximum temperature, average temperature, maximum temperature gradient, temperature at four key locations: cable joint, intermediate section, bend section, and terminal, and the highest temperature in the infrared thermal image. Chemical characteristics include: real-time concentrations of H2, CO, and CH4; 1-hour change rates of H2, CO, and CH4; total hydrocarbon content; 1-hour change rate of total hydrocarbons; and the content of four furan compounds: 2-furan carbaldehyde, 2-methylfuran, furan, and furfuryl alcohol.

[0016] The hierarchical analysis module is connected to the progressive data processing module and performs three levels of analysis on the processed data, namely, first-level single-modal state identification, second-level multimodal correlation verification, and third-level aging level assessment. The first-level single-modal state identification of the hierarchical analysis module includes: Electrical condition: The skewness S of the partial discharge phase distribution spectrum is calculated using the formula: , where n ph Let n be the number of phase windows, where n is the number of phase windows. ph This refers to the number of phase windows. The average number of pulses in the phase window, s ph The standard deviation of the phase window pulse number is used. When S > 1.2, it is determined to be a tip discharge. Thermal state: The steady-state temperature rise ΔT = I is calculated using a thermal circuit model. 2 Rθ, where I is the cable load current, R is the conductor resistance, and θ is the total thermal resistance. When ΔT > 10K, it is marked as an overheated area; Chemical status: According to IEC60599 standard, a chemical warning is triggered when the total hydrocarbon content > 150μL / L; According to IEC 60599 "Guideline for the analysis of dissolved and free gases in oil-immersed power transformers, reactors and bushings", a chemical warning is triggered when the total hydrocarbon content is >150 μL / L. The total hydrocarbon content is calculated as the sum of the concentrations of CH4, C2H6, C2H4 and C2H2. The concentration of each gas is obtained by using the calibration curve of the gas chromatograph peak area and the standard gas concentration. For example, the calibration curve of CH4 is y=0.02x+0.05, where x is the peak area and y is the concentration in μL / L.

[0017] The second-level multimodal correlation verification of the hierarchical analysis module includes: Establish the correlation matrix M 3×3 M pq The Pearson correlation coefficient between the p-th and q-th modes is expressed by the formula: , where n sam The total number of samples, Let r be the feature value of the r-th sample in the p-th modality. The mean of the i-th modal sample is used; when |Mpq| > 0.7, a strong correlation is determined; when |Mpq| < 0.3, a historical database comparison is initiated. The historical data consists of multimodal data of the same type of cable under the same voltage level and the same laying environment in the past 3 years. For example, the "electrical-thermal" correlation: if the "electrical-thermal" correlation is greater than 0.7, it indicates a strong correlation between "high partial discharge" and "high temperature," which can be identified as a synergistic phenomenon caused by insulation aging. The historical database stores multimodal data of similar cables (same voltage level, same insulation type, same laying environment) for the past 3 years. Each data entry includes a timestamp, electrical / thermal / chemical characteristic values, and correlation values. When the current |Mpq| < 0.3 (e.g., high partial discharge but normal temperature), the database is searched for historical records of "same characteristic combinations." If more than 80% of the historical records are identified as "false signals caused by interference," the current data is marked as "to be verified" and needs to be re-analyzed in conjunction with the data from the next hour.

[0018] The hierarchical analysis module employs a fuzzy comprehensive evaluation model for its three-level aging assessment, including a factor set U = {u1, u2, u3}, where u1 is the discharge characteristic d. pd u2 represents the temperature characteristic t cv u3 is a chemical characteristic c hc The weight vector A = [0.4, 0.3, 0.3] is determined based on the analytic hierarchy process (AHP) and a judgment matrix is ​​constructed. The eigenvector corresponding to the largest eigenvalue is calculated and normalized; the evaluation set V={V1,V2,V3,V4} corresponds to healthy, mildly aged, moderately aged, and severely aged, respectively, through membership degree calculation. The membership degree of the comment set is calculated as follows: V1 Health: ; ; ; V2 Mild Aging: ; ; ; V3 moderate aging: ; ; ; V4 is severely aged: ; ; ; Membership function application: Taking a 110kV XLPE cable as an example, if the average amplitude of partial discharge is 400pC, the operating temperature is 70℃, and the total hydrocarbon content is 200μL / L, then the membership degree of each feature is calculated as follows—the membership degree μ of the discharge feature to V2. V2 (400)==0, membership degree μ of V3 V3 (400)==1.5 (corrected to 1); Membership degree μ of temperature feature to V2 V2 (70)==0, membership degree μ of V3 V3 (70) = ≈0.667; Membership degree μ of chemical characteristic to V2V2 (200)==0, membership degree μ of V3 V3 (200) = ≈ 0.333; Fuzzy synthesis operation: Using the "weighted average" synthesis operator, calculated in the above example, V1 = 0.4 × 0 + 0.3 × 0 + 0.3 × 0 = 0, V2 = 0.4 × 0 + 0.3 × 0 + 0.3 × 0 = 0, V3 = 0.4 × 1 + 0.3 × 0.667 + 0.3 × 0.333 ≈ 0.4 + 0.2 + 0.1 = 0.7, V4 = 0.4 × 0 + 0.3 × 0 + 0.3 × 0 = 0. According to the principle of maximum membership degree, V3 = 0.7 is the largest, and it is determined to be V3.

[0019] A multi-level output module is connected to the hierarchical analysis module and outputs basic information, early warning information and trend prediction information based on the analysis results. The multi-level output module includes: Basic output: A state snapshot is generated every 15 minutes, containing k. pca =8-dimensional principal component values, single-mode state indicators including tip discharge / no tip discharge, overheating / no overheating, chemical warning / no chemical warning, snapshot format is both PDF and Excel; Warning output: When the level reaches V3, an audible and visual alarm is triggered. The alarm sound pressure level is ≥85dB, the alarm light is red and flashing at a frequency of 2Hz, and a maintenance work order containing three-dimensional positioning is pushed with a latitude and longitude error ≤±3m and a depth error ≤±0.5m. The work order includes the aging level, the characteristic parameters that exceed the standard, and the recommended maintenance measures. Trend Output: Monthly aging rate reports are generated, and the remaining lifespan is predicted using the grey model GM(1,1). The model construction steps are as follows: S1.1, Historical aging level data Accumulation generation , where n gm =12 represents selecting historical data from the most recent 12 months; (α=1,2,⋯,n) gm () represents the original observed value of the cable insulation aging level in month α; S1.2 Establishing the differential equation , where a is the development coefficient, reflecting the development trend and rate of change of the data sequence; b is the gray action quantity, reflecting the influence of external factors during the data change process; S1.3 Solving for parameters [a, b] using the least squares method T =(B T B) −1 B T y gm,N ,in , ; S1.4. Based on the solved parameters a and b, predict the cumulative generated sequence to obtain the cumulative predicted value at future time k+1. and cumulatively reduce and restore The original data sequence is obtained. The predicted value at future time k+1.

[0020] The data storage module is connected to the multi-dimensional data acquisition module, the progressive data processing module, and the hierarchical analysis module, and is used to store the data from all modules. The self-diagnosis module is used to periodically check the hardware status, data transmission links, and algorithm operation of each module. The self-diagnosis cycle is 24 hours. When a hardware failure, data transmission interruption, or algorithm abnormality is detected, a diagnostic report is automatically generated and pushed to the operation and maintenance terminal.

[0021] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-modal data fusion evaluation system for cable insulation aging, characterized in that, include: A multi-dimensional data acquisition module is used to simultaneously acquire the electrical, thermal, and chemical parameters of the cable; The progressive data processing module is connected to the multi-dimensional data acquisition module and performs three levels of processing on the acquired parameters: first-level preprocessing, second-level feature extraction, and third-level data fusion. The hierarchical analysis module is connected to the progressive data processing module and performs three levels of analysis on the processed data, namely, first-level single-modal state identification, second-level multimodal correlation verification, and third-level aging level assessment. A multi-level output module is connected to the hierarchical analysis module and outputs basic information, early warning information and trend prediction information based on the analysis results. The data storage module is connected to the multi-dimensional data acquisition module, the progressive data processing module, and the hierarchical analysis module, and is used to store the data from all modules. The self-diagnosis module is used to periodically check the hardware status, data transmission links, and algorithm operation of each module. The self-diagnosis cycle is 24 hours. When a hardware failure, data transmission interruption, or algorithm abnormality is detected, a diagnostic report is automatically generated and pushed to the operation and maintenance terminal.

2. The cable insulation aging multimodal data fusion evaluation system according to claim 1, characterized in that: The multi-dimensional data acquisition module includes: The electrical parameter acquisition unit uses a high-frequency current sensor, a 0.2-class voltage transformer, and a 0.1-class current transformer. The high-frequency current sensor has a bandwidth of 10kHz-100MHz, a sampling rate of 200MS / s, a data length of 8192 points per frame, a sampling interval of 50ms between the voltage transformer and the current transformer, and a timestamp accuracy of ±1μs. The thermal parameter acquisition unit includes a distributed fiber optic sensor and an infrared thermal imager. The distributed fiber optic sensor has a spatial resolution of 1m, a temperature measurement range of -50℃ to 150℃, and generates a temperature distribution curve every 30s. The infrared thermal imager has a resolution of 640×512, scans once a week, and synchronously records GPS location with an accuracy of ±3m. The chemical parameter acquisition unit includes a gas sensor array and a gas chromatograph. The gas sensor array has a detection limit of 0.1 ppm, a response time of <20 s, and stores one set of data per hour. The gas chromatograph has a detection accuracy of 0.01 μL / L and analyzes the content of furan compounds in insulating oil samples once per quarter.

3. The cable insulation aging multimodal data fusion evaluation system according to claim 1, characterized in that: The primary preprocessing of the progressive data processing module includes: The discharge signal was denoised using a 3-layer decomposition with a db6 wavelet threshold of λ=0.8σ, where σ is the noise standard deviation, eliminating narrowband interference of 50Hz±2Hz. Voltage and current signals were filtered using a 50Hz notch filter, and a 24-hour effective value sequence was calculated to eliminate outliers greater than 1.1 times the rated value. Optical fiber temperature was filtered using a sliding window mean filter with a window size of 5m to eliminate abrupt changes deviating from adjacent points by ±5℃. Infrared images were matched with the cable's 3D model using the SIFT algorithm to extract the surface temperature field distribution, and the temperature values ​​in non-cable areas were set to zero. Gas concentration was denoted using a first-order difference method, with secondary sampling calibration initiated when |ΔC / Δt|>0.5ppm / h. Oil sample data underwent baseline correction and peak area calculation according to GB / T17623-2018 standard.

4. The cable insulation aging multimodal data fusion evaluation system according to claim 1, characterized in that: The secondary feature extraction of the progressive data processing module includes: Extract time-domain features, including the average amplitude of the discharge and the maximum temperature gradient; extract frequency-domain features, including the main frequency f0 of the discharge signal after FFT transformation and the 12-hour spectral entropy of the gas concentration; extract time-frequency domain features: wavelet packet energy spectrum of the discharge signal, containing 8 frequency bands with a center frequency of 10kHz-80MHz.

5. The cable insulation aging multimodal data fusion evaluation system according to claim 1, characterized in that: The three-level data fusion of the progressive data processing module includes: Time alignment: Based on the GPS clock, all data is interpolated to a 1-minute time scale; Spatial matching: Establish a three-dimensional coordinate system for the cable, with the X-axis representing length, the Y-axis representing depth, and the Z-axis representing height, and map the temperature and discharge location to the coordinate points; Feature fusion: Construct a 32-dimensional feature vector, including 12 electrical dimensions, 8 thermal dimensions, and 12 chemical dimensions, and reduce the dimensionality to a cumulative contribution rate of ≥95% through PCA.

6. The cable insulation aging multimodal data fusion evaluation system according to claim 1, characterized in that: The first-level single-modal state identification of the hierarchical analysis module includes: Electrical condition: Calculate the skewness S of the partial discharge phase distribution spectrum; if S > 1.2, it is determined to be a tip discharge. Thermal condition: Calculate the steady-state temperature rise ΔT = I using a thermal circuit model. 2 Rθ, where I is the cable load current, R is the conductor resistance, and θ is the total thermal resistance. When ΔT > 10K, it is marked as an overheated area; Chemical status: According to IEC60599 standard, a chemical warning is triggered when the total hydrocarbon content > 150μL / L.

7. The cable insulation aging multimodal data fusion evaluation system according to claim 1, characterized in that: The second-level multimodal correlation verification of the hierarchical analysis module includes: Establish the correlation matrix M pq When |Mpq| > 0.7, a strong correlation is determined; when |Mpq| < 0.3, a historical database comparison is initiated. The historical data consists of multimodal data of similar cables under the same voltage level and laying environment over the past 3 years. The hierarchical analysis module employs a fuzzy comprehensive evaluation model for its three-level aging assessment, including a factor set U = {u1, u2, u3}, where u1 is the discharge characteristic d. pd u2 represents the temperature characteristic t cv u3 is a chemical characteristic c hc The weight vector A = [0.4, 0.3, 0.3] is determined based on the analytic hierarchy process. It is obtained by constructing a judgment matrix, calculating the eigenvector corresponding to the largest eigenvalue, and normalizing it. The comment set V = {V1, V2, V3, V4} corresponds to healthy, mildly aged, moderately aged, and severely aged, respectively, through membership degree calculation.

8. The cable insulation aging multimodal data fusion evaluation system according to claim 1, characterized in that: The multi-level output module includes: Basic output: A state snapshot is generated every 15 minutes, containing k. pca =8-dimensional principal component values, single-mode state indicators including tip discharge / no tip discharge, overheating / no overheating, chemical warning / no chemical warning, snapshot format is both PDF and Excel; Warning output: When the level reaches V3, an audible and visual alarm is triggered. The alarm sound pressure level is ≥85dB, the alarm light flashes red at a frequency of 2Hz, and a maintenance work order containing three-dimensional positioning is pushed. The work order includes the aging level, the characteristic parameters that exceed the standard, and the recommended maintenance measures. Trend Output: Monthly aging rate reports are generated, and a grey model is used to predict remaining lifespan. The model construction steps are as follows: S1.1, Historical aging level data Accumulation generation , where n gm =12 represents selecting historical data from the most recent 12 months; The original observed value of the cable insulation aging level for month α; S1.2 Establishing the differential equation , where a is the development coefficient, reflecting the development trend and rate of change of the data sequence; b is the gray action quantity, reflecting the influence of external factors during the data change process; S1.3 Solving for parameters [a, b] using the least squares method T =(B T B) −1 B T y gm,N ; S1.

4. Based on the solved parameters a and b, predict the cumulative generated sequence to obtain the cumulative predicted value at future time k+1, and then subtract and restore it to obtain the original data sequence. The predicted value at future time k+1.

Images