An intelligent monitoring method for insulation status of power cable joints based on complex environment

Abstract

The present invention proposes an intelligent monitoring method for the insulation status of power cable joints based on complex environments. The method belongs to the technical field of power system monitoring and comprises: collecting relevant data by multiple means, the relevant data including ambient temperature data, partial discharge signals, and temperature distribution images; inputting the collected environmental parameters and partial discharge signals into an intelligent analysis system for preprocessing and feature extraction, establishing an insulation status prediction model based on a machine learning algorithm, and predicting the degree of aging and potential fault points of the insulation layer of the cable joint based on historical data and real-time monitoring data. By collecting ambient temperature data, partial discharge signals, and temperature distribution images by multiple means, combined with an intelligent analysis system and a machine learning algorithm, the degree of aging and potential fault points of the insulation layer of the cable joint can be more accurately predicted.

Description

Technical Field

[0001] The invention proposes an intelligent monitoring method for the insulation status of a power cable joint based on a complex environment, and belongs to the technical field of power system monitoring. Background Art

[0002] In the complex operating environment of the power system, cold shrink cable intermediate joints are key connecting components, and the performance of their insulation materials is directly related to the safe and stable operation of the entire system. However, since cable joints are affected by various factors such as high temperature, high pressure, and mechanical damage during operation, the insulation layer is prone to aging and cracking, and may even cause partial discharge and breakdown, leading to power failures and safety accidents. Traditional cable insulation monitoring methods, such as partial discharge monitoring, temperature measurement, and insulation resistance testing, although they can detect insulation defects to a certain extent, often have problems such as low monitoring efficiency, insufficient accuracy, and the inability to monitor in real time. Therefore, it is particularly important to develop an innovative intelligent monitoring method to achieve real-time monitoring and early warning of the insulation status of power cable joints in complex environments. Summary of the Invention

[0003] The present invention provides an intelligent monitoring method for the insulation status of power cable joints based on a complex environment, which is used to solve the problems mentioned in the above background technology:

[0004] The present invention proposes an intelligent monitoring method for the insulation status of power cable joints based on a complex environment, the method comprising:

[0005] S1. Collect relevant data by multiple means, wherein the relevant data includes ambient temperature data, partial discharge signals, and temperature distribution images;

[0006] S2. Input the collected environmental parameters and partial discharge signals into the intelligent analysis system for preprocessing and feature extraction. Based on the machine learning algorithm, an insulation status prediction model is established to predict the aging degree and potential failure points of the cable joint insulation layer based on historical data and real-time monitoring data.

[0007] S3. Using infrared thermal imaging data and image processing technology, identify areas of abnormal temperature and calculate the temperature gradient. Based on the temperature gradient calculation results, evaluate the thermal stability of the insulation layer.

[0008] S4. Set a warning threshold based on the output of the insulation status prediction model. When the prediction result exceeds the threshold, the warning mechanism is triggered. Combined with partial discharge monitoring data and infrared thermal imaging results, the location of the insulation defect is accurately located and the fault type and severity are determined.

[0009] S5. Send early warning information and fault location results to operation and maintenance personnel in real time. Based on the monitoring data and early warning results, intelligently optimize the structural design of the cold shrink cable intermediate joint, regularly inspect and maintain the cable joints, and formulate targeted maintenance plans based on the monitoring data.

[0010] Furthermore, the S1 includes:

[0011] S11. Use a high-precision temperature sensor to collect real-time ambient temperature, a humidity sensor to monitor the humidity level around the connector, and an electric field strength sensor to monitor the electric field distribution around the cable connector;

[0012] S12. Capturing partial discharge signals in the cable insulation layer through a high-frequency current sensor, and performing filtering and noise reduction processing on the collected partial discharge signals;

[0013] S13. Use a high-resolution infrared thermal imaging camera to regularly capture temperature distribution images of the cable joint insulation layer, and use an image enhancement algorithm to enhance the collected temperature distribution images.

[0014] Furthermore, the S2 includes:

[0015] S21, pre-processing the collected environmental parameters and partial discharge signals;

[0016] S22. Based on statistical analysis and machine learning algorithms, feature selection is performed on the preprocessed data to identify characteristic variables that have an impact on insulation status prediction;

[0017] S23. Extract features from the pre-processed data using time domain, frequency domain, and time-frequency domain analysis techniques, and fuse the extracted feature variables to form a comprehensive feature vector;

[0018] S24. Divide the preprocessed dataset into a training set and a validation set, use the training set to train the built-in machine learning model, and use the validation set to evaluate the performance of the model;

[0019] S25. Evaluate the performance of the model through cross-validation technology. Adjust the model's hyperparameters and optimize the model structure based on the validation results.

[0020] S26. Use the trained machine learning model to predict the test set, output the insulation status prediction results, and evaluate the uncertainty of the prediction results;

[0021] S27. Design early warning strategies based on the uncertainty of the prediction results. When the uncertainty of the prediction results is high, trigger a higher level of early warning mechanism.

[0022] S28. Compare the predicted results with the actual situation to evaluate the accuracy and reliability of the model. Based on the evaluation results, iteratively optimize the model. Based on the continuous learning and updating mechanism, regularly collect new monitoring data to update and optimize the model.

[0023] Furthermore, the S23 includes:

[0024] The mean and variance of each characteristic variable after preprocessing are calculated, and the volatility and stability of the data are evaluated based on the evaluation record. The trend line of the time series data is calculated using the sliding window technology to identify the changing trends of environmental parameters and partial discharge signals over time.

[0025] Identify and record the maximum and minimum values ​​of each characteristic variable within a time period, analyze abnormal fluctuations, evaluate the correlation of time series data at different time intervals, and reveal the time dependence within the data;

[0026] Convert the time domain signal into frequency domain representation, analyze the spectral components of each characteristic variable, and identify the main frequency components and their energy distribution;

[0027] Calculate and plot power spectral density to evaluate signal energy at different frequencies and identify potential periodic or random characteristics; calculate the entropy of the spectrum to measure the complexity of the signal frequency distribution;

[0028] Perform Fourier transform on the signal within the time window to obtain the spectrum information that changes with time, capture the instantaneous frequency characteristics of the partial discharge signal, decompose the signal using wavelet functions of different scales, and extract the time-frequency characteristics at multiple scales;

[0029] Decompose the complex signal into a series of intrinsic mode functions, each IMF represents a different frequency component of the signal; normalize all extracted features;

[0030] Based on correlation analysis, the most critical feature subset for insulation status prediction is further screened, and the screened features are fused into a comprehensive feature vector using weighted averaging.

[0031] And through random forest, the contribution of each feature in the comprehensive feature vector is evaluated.

[0032] Furthermore, the S3 includes:

[0033] S31. Based on high-resolution infrared thermal imaging technology, obtain the temperature distribution image of the cable joint insulation layer and apply image processing algorithms to preliminarily identify the temperature abnormality area;

[0034] S32, performing detailed analysis on the initially identified temperature anomaly area and removing noise and artifacts using morphological processing techniques;

[0035] S33, calculating geometric features such as area, perimeter, shape factor, and statistical features of the temperature anomaly region;

[0036] S34, setting a plurality of temperature gradient calculation points around the temperature anomaly area, calculating the temperature gradient between each calculation point and its adjacent points, and forming a temperature gradient distribution map;

[0037] S35. Analyze the thermal conductivity of the insulation layer based on the temperature gradient distribution diagram to determine whether there is a thermal conductivity barrier or heat accumulation phenomenon;

[0038] S36. Combine the thermal properties of the insulation material to evaluate the thermal stability of the insulation layer and predict the risk of thermal failure;

[0039] S37. Establish a thermal stability assessment model to quantitatively assess the thermal stability of the insulation layer based on the geometric characteristics, statistical characteristics, and temperature gradient distribution of the temperature anomaly region.

[0040] S38. Based on the results of the thermal stability assessment model, combined with historical data and industry standards, set the warning threshold, and dynamically adjust the warning threshold based on the aging degree of the insulation layer and changes in the operating environment.

[0041] Furthermore, the S35 includes:

[0042] On the temperature gradient distribution map, the areas where the temperature gradient changes significantly are identified. By setting the gradient change threshold, the areas where the temperature gradient exceeds or falls below the threshold are marked as potential heat conduction barrier areas.

[0043] Using connected domain analysis in image processing technology, the points on the temperature gradient map are connected into lines to form a heat conduction path network;

[0044] According to the thermal conductivity characteristics of the insulation material, a weight is assigned to each path, and the weight is inversely proportional to the thermal conductivity of the material. The thermal conductivity efficiency of the entire insulation layer is calculated;

[0045] Evaluate the deviation of heat transfer efficiency from normal values, identify areas with significantly reduced heat transfer efficiency, analyze local high or low temperature concentration areas in the temperature gradient distribution map, and evaluate the rate and degree of heat accumulation by calculating the temperature gradient change rate of the area;

[0046] Combining the results of temperature gradient anomaly identification and heat conduction efficiency assessment, the heat conduction obstruction points in the insulation layer are located and classified based on the temperature gradient characteristics of the obstruction points, their location on the heat conduction path, and possible material defects.

[0047] Using time series analysis technology, we compare historical data of temperature gradient distribution maps, analyze the trend of thermal conductivity performance over time, identify areas where thermal conductivity performance continues to decline, and predict possible thermal failure risks in the future.

[0048] Furthermore, the S4 includes:

[0049] S41. Based on historical analysis of power system operation data, combined with industry standards and practical experience, set different levels of warning thresholds. Based on the warning trigger mechanism, when the monitoring data reaches or exceeds the set threshold, the corresponding level of warning will be automatically triggered; and a warning information transmission channel will be established.

[0050] S42. Using monitoring equipment to collect real-time data, and combining historical data and industry standards, extract fault characteristics, including abnormal fluctuations, mutation points, and frequency components; and preprocessing the extracted fault characteristics;

[0051] S43. Locate the fault location based on the fault characteristics and in combination with the fault location algorithm; and perform simulation testing and optimization on the fault location algorithm;

[0052] S44. Based on the results obtained from the fault location algorithm, combined with on-site inspection and data analysis, diagnose the fault and determine the fault type and cause; establish a fault diagnosis knowledge base to record the fault characteristics, location methods and treatment measures of common faults;

[0053] S45. Based on the results of fault location and diagnosis, combined with changes in the operating status of the power system, the early warning threshold is dynamically adjusted, and the performance of the early warning mechanism is regularly tested and evaluated. The early warning threshold and trigger mechanism are optimized and improved based on the test results.

[0054] Furthermore, the S43 includes:

[0055] S431. Preliminarily apply the fault location algorithm to the fault feature dataset based on the fault features, preliminarily locate the fault location, and evaluate the basic performance and accuracy of the algorithm.

[0056] S432. Fault features extracted from multi-source monitoring data are integrated to form a comprehensive feature vector, a power system fault simulation model is constructed, different types of fault scenarios are simulated, and corresponding fault feature data sets are generated.

[0057] S433. Apply the optimized fault location algorithm to the simulation data set to verify the location accuracy and stability of the algorithm under different fault types and different operating conditions.

[0058] Furthermore, the step S432 includes:

[0059] Extract fault features from multi-source monitoring data through signal processing technology and data analysis methods, and pre-process the extracted fault features;

[0060] The extracted features are fused based on the preset feature fusion strategy, and a fault simulation model is constructed according to the structure and operation characteristics of the power system;

[0061] Based on the fault simulation model, we generate fault feature datasets for different fault types and operating conditions. We then apply the fused feature vectors to the fault location algorithm to evaluate the effect of feature fusion on improving the algorithm's performance.

[0062] According to the evaluation results of the feature fusion effect, the simulation model and feature fusion strategy are iteratively optimized.

[0063] Furthermore, the S5 includes:

[0064] S51. Based on the early warning information push mechanism, the early warning information and fault location results are sent to the mobile device or workstation of the operation and maintenance personnel in real time;

[0065] S52. Analyze the structural design and insulation material performance of cold-shrink cable intermediate joints based on monitoring data and early warning results, and propose improvement suggestions based on the analysis results.

[0066] S53. Develop a targeted maintenance plan based on the monitoring data and early warning results, the maintenance plan including inspection frequency, inspection methods, and maintenance measures;

[0067] S54. Carry out regular inspection and maintenance, record the inspection results, compare and analyze with the monitoring data, evaluate the maintenance effect, and continuously optimize the maintenance plan.

[0068] The beneficial effects of the present invention are as follows: by collecting ambient temperature data, partial discharge signals and temperature distribution images through multiple means, and combining intelligent analysis systems and machine learning algorithms, the aging degree and potential fault points of the cable joint insulation layer can be predicted more accurately; by using infrared thermal imaging data and image processing technology, temperature abnormality areas can be identified and the thermal stability of the insulation layer can be evaluated, so as to set a reasonable early warning threshold and improve the timeliness and accuracy of fault early warning; by combining partial discharge monitoring data and infrared thermal imaging results, the position of the insulation defect can be accurately located, and the type and degree of the fault can be given, which is conducive to rapid response and handling of the fault; according to the monitoring data and early warning results, the structural design of the cable joint can be intelligently optimized, regular inspection and maintenance can be carried out, and targeted maintenance plans can be formulated to improve maintenance efficiency and effectiveness; by sending early warning information and fault location results to operation and maintenance personnel in real time, timely measures can be taken Measures can reduce power outage time caused by faults and improve the reliability and stability of the power system; through intelligent monitoring and early warning systems, unnecessary maintenance work can be reduced, excessive maintenance can be avoided, and thus maintenance costs can be reduced; this method can adapt to complex environments and has good adaptability and robustness for cable joint insulation status monitoring under different environmental conditions; through time domain, frequency domain and time-frequency domain analysis techniques, as well as feature extraction and fusion, it can more comprehensively understand and process monitoring data, and improve the depth and breadth of data processing; the entire monitoring method realizes intelligent management of data collection, analysis, early warning and maintenance, reduces interference from human factors, and improves the scientific and systematic nature of management; through continuous learning and updating mechanisms, new monitoring data are regularly collected, and the model is updated and optimized, so that the monitoring method can continuously adapt to new operating conditions and environmental changes, and maintain its advanced nature and effectiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a step diagram of the method of the present invention. DETAILED DESCRIPTION

[0070] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0071] One embodiment of the present invention, as Figure 1 As shown, a method for intelligently monitoring the insulation status of a power cable joint based on a complex environment comprises:

[0072] S1. Use high-precision sensor equipment to collect real-time environmental parameters around the intermediate joint of the cold-shrink cable, including temperature, humidity, and electric field strength; use partial discharge monitoring technology to collect partial discharge signals in the cable insulation layer, including discharge amplitude, discharge frequency, etc.; use infrared thermal imaging technology to obtain temperature distribution images of the cable joint insulation layer, intuitively displaying potential temperature anomaly areas.

[0073] S2. Input the collected environmental parameters and partial discharge signals into the intelligent analysis system for preprocessing and feature extraction. Based on the machine learning algorithm, an insulation status prediction model is established to predict the aging degree and potential failure points of the cable joint insulation layer based on historical data and real-time monitoring data.

[0074] S3. Using infrared thermal imaging data and image processing technology, identify areas of abnormal temperature and calculate the temperature gradient. Based on the temperature gradient calculation results, evaluate the thermal stability of the insulation layer.

[0075] S4. Set a warning threshold based on the output of the insulation status prediction model. When the prediction result exceeds the threshold, the warning mechanism is triggered. Combined with partial discharge monitoring data and infrared thermal imaging results, the location of the insulation defect is accurately located and the fault type and severity are determined.

[0076] S5. Send early warning information and fault location results to operation and maintenance personnel in real time. Based on the monitoring data and early warning results, intelligently optimize the structural design of the cold-shrink cable intermediate joint to improve the high-temperature and high-voltage resistance of the insulation material; regularly inspect and maintain the cable joints, and formulate targeted maintenance plans based on the monitoring data.

[0077] The working principle of the above technical solution is: using high-precision sensors to measure the temperature, humidity and electric field strength around the intermediate joint of the cold shrink cable in real time; using partial discharge monitoring technology to capture the partial discharge phenomenon occurring inside the cable insulation layer, including the amplitude and frequency of the discharge, which can reflect the damage inside the insulation layer; using infrared thermal imaging technology to obtain the temperature distribution image of the cable joint insulation layer, and intuitively display the possible temperature anomaly areas; inputting the collected environmental parameters and partial discharge signals into the intelligent analysis system, first performing preprocessing such as denoising and standardization to improve data quality; then performing feature extraction to extract feature information useful for insulation status prediction from the original data; based on machine learning algorithms (such as support vector machines, neural networks, etc.), combined with historical data and real-time monitoring data, training and establishing an insulation status prediction model to predict the aging degree and potential fault points of the cable joint insulation layer; using infrared thermal imaging data, identifying temperature anomalies through image processing technology area, and calculate the temperature gradient, that is, the rate of change of temperature with spatial position; based on the temperature gradient calculation results, evaluate the thermal stability of the insulation layer, and determine whether there is an overheating risk or a thermal stress concentration area; according to the output results of the insulation status prediction model, set a reasonable warning threshold, and when the prediction result exceeds the threshold, automatically trigger the warning mechanism and notify relevant personnel; combine partial discharge monitoring data and infrared thermal imaging results, use multi-source information fusion technology to accurately locate the location of insulation defects, and determine the type and severity of the fault based on data analysis; send warning information and fault location results to operation and maintenance personnel in real time for rapid response and processing; based on monitoring data and warning results, intelligently optimize the structural design of cold shrink cable intermediate joints, such as improving the formulation of insulation materials to improve their high temperature and high voltage resistance; regularly inspect and maintain cable joints, and formulate targeted maintenance plans based on monitoring data, such as increasing inspection frequency, replacing aging components, etc., to ensure the long-term safe operation of cable joints.

[0078] The effects of the above technical solutions are as follows: through high-precision sensors and partial discharge monitoring technology, key environmental parameters around the intermediate joints of cold-shrink cables and partial discharge signals inside the insulation layer can be collected in real time to realize continuous monitoring of the insulation status of the cable joints; the insulation status prediction model established in combination with the machine learning algorithm can accurately predict the aging degree and potential fault points of the cable joint insulation layer, and timely set and trigger the early warning mechanism to effectively avoid the occurrence or expansion of faults; using infrared thermal imaging technology and image processing technology, the temperature distribution image of the cable joint insulation layer can be intuitively displayed, and the temperature abnormality area can be identified. Combined with the temperature gradient calculation results, the thermal stability of the insulation layer can be evaluated; combined with partial discharge monitoring data and infrared thermal imaging results, the location of the insulation defect can be accurately located, and the fault type and degree can be given, providing detailed fault diagnosis information for operation and maintenance personnel; according to the monitoring data and early warning results, intelligent optimization The structural design of the cold-shrink cable intermediate joint improves the high-temperature and high-voltage resistance of the insulation material, and improves the insulation performance of the cable joint from the source; the cable joints are regularly inspected and maintained, and targeted maintenance plans are formulated based on the monitoring data to ensure the long-term stable operation of the cable joints and reduce power outages and maintenance costs caused by faults; this technical solution realizes intelligent monitoring and early warning of the insulation status of the cable joints, reduces the frequency and intensity of manual inspections, and improves operation and maintenance efficiency; at the same time, through precise fault location and intelligent optimization, it reduces safety risks in the operation and maintenance process and ensures the life safety of operation and maintenance personnel; the application of this technical solution not only improves the insulation monitoring level of power cable joints, but also provides strong support for the intelligent and digital transformation of the power industry; with the continuous development and improvement of technology, this technical solution is expected to be more widely used and promoted in the power industry, promoting the progress and development of the entire industry.

[0079] In one embodiment of the present invention, the S1 includes:

[0080] S11. Use a high-precision temperature sensor to collect ambient temperature in real time, and monitor changes in insulation material performance caused by temperature changes based on the collected ambient temperature; use a humidity sensor to monitor the humidity level around the connector, and use an electric field strength sensor to monitor the electric field distribution around the cable connector;

[0081] S12. Capturing partial discharge signals in the cable insulation layer using a high-frequency current sensor, including discharge amplitude and discharge frequency, and performing filtering and noise reduction on the collected partial discharge signals;

[0082] S13. Use a high-resolution infrared thermal imaging camera to regularly capture temperature distribution images of the cable joint insulation layer, and use an image enhancement algorithm to enhance the collected temperature distribution images.

[0083] The working principle of the above technical solution is as follows: High-precision temperature sensors are deployed in the environment surrounding the cold-shrink cable intermediate joint to collect real-time ambient temperature data. This data is input into an intelligent analysis system to monitor changes in insulation material performance caused by temperature fluctuations. Insulation material performance varies with temperature. Excessively high or low temperatures can accelerate insulation aging, making real-time temperature monitoring crucial for assessing insulation condition. Humidity sensors are also installed around the cable joint to monitor humidity levels. High humidity can cause moisture in the insulation material, which, when absorbed, accelerates aging and degrades insulation performance. Real-time humidity monitoring allows for timely detection and prevention of moisture damage. Electric field strength sensors monitor the electric field distribution around the cable joint. Areas of concentrated electric field strength are often the starting point of insulation failure because these areas are subject to higher electric field strength, leading to localized damage to the insulation material. Monitoring electric field strength can identify potential areas of concentrated electric field strength, providing important information for subsequent insulation condition assessment and fault warning. High-frequency current sensors are installed near the cable insulation layer to capture signals of partial discharge within the insulation layer. Partial discharge (PD) is an electrical discharge phenomenon that occurs at tiny defects or damage within insulating materials. The discharge signal includes information such as discharge amplitude and frequency. The collected PD signal is first filtered to remove background noise and interference signals to ensure signal accuracy. Noise reduction is then used to further purify the signal, providing high-quality data input for subsequent intelligent analysis and insulation condition prediction. A high-resolution infrared thermal imaging camera is used to regularly capture temperature distribution images of the insulation layer of cable connectors. Infrared thermal imaging technology captures infrared radiation emitted from an object's surface and converts it into a visible temperature distribution image. The collected temperature distribution image may be affected by factors such as ambient lighting and camera performance, resulting in image quality degradation. Therefore, an image enhancement algorithm is used to process the collected image to improve image clarity and contrast, making areas of temperature anomalies more intuitive and easy to identify.

[0084] The above technical solution achieves the following: Through the integrated application of high-precision temperature sensors, humidity sensors, and electric field strength sensors, comprehensive and real-time monitoring of environmental parameters around cold-shrink cable intermediate joints, including temperature, humidity, and electric field strength, is possible. High-frequency current sensors accurately capture partial discharge signals in the cable insulation layer, including discharge amplitude and frequency. Partial discharge occurs at tiny defects or damage within the insulation material and serves as an early warning sign of insulation degradation. Filtering and noise reduction processes further improve the accuracy and reliability of partial discharge signals, providing strong support for subsequent insulation condition assessment and fault warning. A high-resolution infrared thermal imaging camera regularly captures temperature distribution images of the cable joint insulation layer and enhances these images using an image enhancement algorithm. This makes temperature anomalies more easily visible, helping maintenance personnel promptly identify and address potential insulation defects. Furthermore, infrared thermal imaging technology provides the overall temperature distribution of the cable joint insulation layer, providing an important basis for assessing the insulation layer's thermal stability and developing targeted maintenance plans. This technical solution, through automated and intelligent monitoring, enables real-time monitoring and early warning of the insulation condition of cold-shrink cable intermediate joints. Compared to traditional manual inspection methods, this technical solution significantly improves monitoring efficiency and accuracy, reducing operation and maintenance costs. Furthermore, through real-time monitoring and early warning, potential insulation defects can be promptly detected and addressed, preventing the occurrence or escalation of faults and ensuring the safe operation of cable joints. The application of this technical solution not only enhances insulation monitoring of power cable joints but also provides strong support for the intelligent and digital transformation of the power industry. With the continuous development and improvement of the technology, this technical solution is expected to be more widely applied and promoted in the power industry, driving the intelligent development of the entire industry.

[0085] In one embodiment of the present invention, the S2 includes:

[0086] S21, pre-processing the collected environmental parameters (temperature, humidity, electric field strength) and partial discharge signals;

[0087] S22. Based on statistical analysis and machine learning algorithms, feature selection is performed on the preprocessed data to identify characteristic variables that have an impact on insulation status prediction;

[0088] S23. Extract features from the pre-processed data using time domain, frequency domain, and time-frequency domain analysis techniques, and fuse the extracted feature variables to form a comprehensive feature vector;

[0089] S24. Divide the preprocessed dataset into a training set and a validation set, use the training set to train the built-in machine learning model, and use the validation set to evaluate the performance of the model;

[0090] S25. Evaluate the model performance through cross-validation techniques, such as K-fold cross-validation. Based on the validation results, adjust the model's hyperparameters and optimize the model structure. The model performance is obtained using the following formula:

[0091]

[0092] Among them, M represents the model, D represents the dataset, and K represents the number of cross-validation folds. represents the accuracy of the k-fold verification;

[0093] S26. Use the trained machine learning model to predict the test set (or real-time monitoring data), output the insulation status prediction result, and evaluate the uncertainty of the prediction result, that is, the confidence or reliability of the prediction result. The uncertainty of the prediction result is calculated by the following method:

[0094]

[0095] in, Represents the model's predicted value for sample i, N represents the number of test set samples, represents the average of all predicted values;

[0096] S27. Design early warning strategies based on the uncertainty of the prediction results. When the uncertainty of the prediction results is high, trigger a higher level of early warning mechanism.

[0097] S28. Compare the predicted results with the actual situation to evaluate the accuracy and reliability of the model. Based on the evaluation results, iteratively optimize the model, including adjusting feature selection, improving algorithm design, and increasing training data. Based on the continuous learning and updating mechanism, regularly collect new monitoring data to update and optimize the model.

[0098] The working principle of the above technical solution is as follows: the collected environmental parameters (temperature, humidity, electric field strength) and partial discharge signals (discharge amplitude, discharge frequency) are first preprocessed. The main task of the preprocessing stage is to remove outliers and missing values ​​caused by sensor failure, data transmission errors, etc. Based on statistical analysis and machine learning algorithms, feature selection is performed on the preprocessed data. Feature selection is the process of identifying characteristic variables that have a significant impact on the prediction of insulation status. It involves calculating the correlation between features and target variables, using feature importance evaluation methods (such as feature importance scores in random forests, coefficients in Lasso regression, etc.), and model-based feature selection techniques (such as recursive feature elimination); through time domain, frequency domain and time-frequency domain analysis techniques, feature extraction is performed on the preprocessed data. Time domain analysis focuses on the change of the signal over time, frequency domain analysis focuses on the distribution of the signal in frequency, and time-frequency domain analysis combines the two to reveal the joint characteristics of the signal in time and frequency. The extracted feature variables are combined using fusion techniques (such as principal component analysis and linear discriminant analysis) to form a comprehensive feature vector to improve the performance of the prediction model. The preprocessed dataset is divided into a training set and a validation set. The training set is used to train the built-in machine learning model, while the validation set is used to evaluate the model's performance. The model training process involves selecting an appropriate machine learning algorithm (such as a support vector machine, decision tree, neural network, etc.) and adjusting its parameters to minimize prediction error. Cross-validation techniques (such as K-fold cross-validation) are used to conduct a more comprehensive performance evaluation of the model. Cross-validation divides the dataset into K parts, alternately using K-1 parts as the training set and the remaining part as the validation set. The experiment is repeated K times, and the average performance metric is calculated. Based on the validation results, the model's hyperparameters (such as the learning rate and regularization parameter) are adjusted to optimize the model structure until the model achieves optimal performance on the validation set. The trained machine learning model is then used to predict the test set (or real-time monitoring data) and output the insulation status prediction results. To assess the uncertainty of the prediction results—that is, the confidence or reliability of the prediction results—the variance, confidence interval, or Bayesian methods can be calculated. Early warning strategies can be designed based on the uncertainty of the prediction results. When the uncertainty of the prediction results is high, a higher-level early warning mechanism is triggered to alert operations and maintenance personnel to take more timely maintenance measures. The design of the early warning strategy should comprehensively consider the accuracy, uncertainty, and actual operation and maintenance needs of the prediction results. The prediction results should be compared with the actual situation to evaluate the accuracy and reliability of the model. Based on the evaluation results, the model is iteratively optimized. Optimization measures include adjusting feature selection, improving algorithm design, and increasing training data. Based on a continuous learning and updating mechanism, new monitoring data is regularly collected to update and optimize the model to adapt to the ever-changing working environment and insulation conditions.

[0099] The above technical solution improves data accuracy and completeness by removing outliers and missing values ​​caused by sensor failures, data transmission errors, and other factors. This provides a high-quality data foundation for subsequent feature selection, feature extraction, and model training, helping to improve the accuracy and reliability of prediction results. Feature selection based on statistical analysis and machine learning algorithms can identify characteristic variables that significantly influence insulation condition prediction, reducing the interference of redundant information on model performance. Furthermore, feature extraction using time-domain, frequency-domain, and time-frequency-domain analysis techniques, and fusion of the extracted characteristic variables to form a comprehensive feature vector, further enhances the model's sensitivity to insulation condition and predictive power. Model training using preprocessed datasets and performance evaluation using validation sets ensures model accuracy in practical applications. Furthermore, cross-validation is used to evaluate model performance, and hyperparameters and model structure are adjusted based on the validation results, further improving the model's predictive power and generalization performance. Evaluating the uncertainty of prediction results—that is, the confidence or reliability of the prediction results—helps design more appropriate early warning strategies. When the uncertainty of the prediction results is high, a higher-level early warning mechanism is triggered, which can remind operation and maintenance personnel to take timely measures to avoid potential safety hazards; the prediction results are compared with the actual situation to evaluate the accuracy and reliability of the model, and the model is iteratively optimized based on the evaluation results. This includes adjusting feature selection, improving algorithm design, increasing training data, etc. to ensure that the model can adapt to the ever-changing working environment and insulation status. Based on the continuous learning and updating mechanism, new monitoring data is regularly collected to update and optimize the model, which can maintain the effectiveness and accuracy of the model in long-term operation; this technical solution realizes real-time monitoring and prediction of the insulation status of cold-shrink cable intermediate joints in an intelligent and automated manner, thereby improving operation and maintenance efficiency. At the same time, through the design and implementation of early warning strategies, potential safety hazards can be discovered and dealt with in a timely manner, ensuring the safe operation of cable joints and improving the overall safety of the power system.

[0100] In one embodiment of the present invention, the step S23 includes:

[0101] S231. Calculate the mean and variance of each characteristic variable (such as temperature, humidity, electric field strength, and discharge amplitude) after preprocessing, and evaluate the volatility and stability of the data based on the evaluation score; calculate the trend line of the time series data using a sliding window technique to identify the changing trends of environmental parameters and partial discharge signals over time; the mean is obtained by:

[0102]

[0103] Where X={x1…x N} represents the dataset of feature variables, where N represents the number of data points;

[0104] The variance is obtained by the following formula:

[0105]

[0106] S232. Identify and record the maximum and minimum values ​​of each characteristic variable within a time period, analyze abnormal fluctuations, evaluate the correlation of time series data at different time intervals, and reveal the time dependence within the data; the correlation is obtained using the following formula:

[0107]

[0108] S233, converting the time domain signal into a frequency domain representation, analyzing the spectral components of each characteristic variable, and identifying the main frequency components and their energy distribution;

[0109] S234. Calculate and plot a power spectrum density diagram to evaluate signal energy at different frequencies and identify potential periodic or random characteristics; calculate the entropy value of the spectrum to measure the complexity of the signal frequency distribution; the entropy value of the spectrum is obtained using the following formula:

[0110]

[0111] Where M represents the number of spectral fractions, and P(i) represents the ratio of the power of the i-th spectral component to the total power, which is obtained by the following formula:

[0112]

[0113] Where S(i) represents the power of the i-th spectral component;

[0114] S235. Performing Fourier transform on the signal within the time window to obtain spectrum information that varies with time, capturing instantaneous frequency characteristics of the partial discharge signal, decomposing the signal using wavelet functions of different scales, and extracting time-frequency features at multiple scales;

[0115] S236, decomposing the complex signal into a series of intrinsic mode functions (IMFs), each IMF representing a different frequency component of the signal; and normalizing all extracted features;

[0116] S237. Based on the correlation analysis, further screen the feature subset that is most critical for insulation state prediction, and fuse the screened features into a comprehensive feature vector using weighted averaging;

[0117] S238, and use random forest to evaluate the contribution of each feature in the comprehensive feature vector.

[0118] The working principle of the above technical solution is as follows: the mean and variance of each characteristic variable (such as temperature, humidity, electric field strength, and discharge amplitude) after preprocessing are calculated. The mean reflects the average level of the data, while the variance reveals the degree of dispersion or volatility of the data. These statistics can be used to assess the stability and volatility of the data, providing a basis for subsequent analysis. The sliding window technique is applied to time series data to calculate the trend line within each window. The sliding window technique can capture the changing trends of the data over time, helping to identify long-term or short-term changes in environmental parameters and partial discharge signals. The maximum and minimum values ​​of each characteristic variable are identified and recorded within each time period to analyze abnormal fluctuations. In addition, the correlation of time series data at different time intervals is evaluated to reveal the time dependence within the data, which is crucial for understanding the interaction between characteristic variables and predicting insulation status. The time domain signal is converted to the frequency domain representation, and spectral analysis is used to identify the main frequency components of each characteristic variable and their energy distribution. The power spectral density plot is calculated to evaluate the signal energy at different frequencies and identify potential periodic or random characteristics. In addition, the entropy of the spectrum is calculated to measure the complexity of the signal's frequency distribution, serving as an indicator for assessing the complexity of insulation state changes. A Fourier transform is performed on the signal within a time window to obtain time-varying spectral information, capturing the instantaneous frequency characteristics of the partial discharge signal. The signal is decomposed using wavelet functions of different scales to extract multi-scale time-frequency features. These features more comprehensively reflect the signal's joint characteristics in time and frequency. The complex signal is decomposed into a series of intrinsic mode functions (IMFs), each representing a different frequency component of the signal. This decomposition helps extract the signal's intrinsic characteristics and improves the accuracy of subsequent analysis. All extracted features are normalized to eliminate dimensional differences between features and ensure comparability. Correlation analysis is used to further select the most critical feature subset for insulation state prediction. Correlation analysis reveals the degree of correlation between features and helps select the most representative features. The selected features are then fused into a comprehensive feature vector using methods such as weighted averaging. The comprehensive feature vector more comprehensively and accurately reflects the insulation state information. Machine learning algorithms such as random forests are used to evaluate the contribution of each feature to the comprehensive feature vector. Random forests can assess the importance of each feature to the model's predictive performance, helping to understand which features are most critical for insulation status prediction.

[0119] The effects of the above technical solutions are as follows: by calculating the mean and variance, the volatility and stability of the data are evaluated, providing basic information for subsequent insulation state prediction; using the sliding window technology to identify the trend line of the time series data, it can clearly show the changing trend of environmental parameters and partial discharge signals over time, which is helpful to capture potential insulation state changes; analyzing the maximum and minimum values ​​of each characteristic variable within the time period, as well as the correlation of time series data, reveals the time dependence within the data, and further enhances the accuracy and comprehensiveness of feature extraction; converting the time domain signal into a frequency domain representation, and identifying the main frequency components and their energy distribution through spectrum analysis, which helps to identify potential periodic or random characteristics and provides more dimensional information for insulation state prediction; performing Fourier transform on the signal within the time window to capture the instantaneous frequency characteristics of the partial discharge signal, and combining wavelet function decomposition to extract time-frequency characteristics at multiple scales, which can more accurately reflect Changes in insulation status; Based on correlation analysis, the most critical feature subset for insulation status prediction is selected, avoiding the interference of redundant information on the prediction results and improving the prediction efficiency and accuracy; Using weighted averaging to fuse the selected features into a comprehensive feature vector simplifies the input of subsequent models while retaining key information, which helps to improve the prediction performance of the model; Using machine learning algorithms such as random forest to evaluate the contribution of each feature in the comprehensive feature vector, it provides a basis for further optimization and selection of features, which helps to build a more accurate and efficient insulation status prediction model; The combined use of the above technical solutions can more comprehensively extract and fuse feature information related to insulation status, improving the accuracy and reliability of the prediction model; By optimizing the feature selection and extraction process, the uncertainty of the prediction results is reduced, providing a strong guarantee for the safe operation of the power system; This technical solution provides strong support for the intelligent operation and maintenance and preventive maintenance of the power system. By real-time monitoring and prediction of the changing trends of insulation status, potential safety hazards can be discovered in a timely manner, and corresponding maintenance measures can be taken to avoid accidents and improve the reliability and stability of the power system.

[0120] In one embodiment of the present invention, S3 includes:

[0121] S31. Based on high-resolution infrared thermal imaging technology, obtain the temperature distribution image of the cable joint insulation layer and apply image processing algorithms (such as threshold segmentation and edge detection) to preliminarily identify the temperature abnormality area;

[0122] S32. Perform detailed analysis on the initially identified temperature anomaly area and remove noise and artifacts using morphological processing techniques (such as dilation, erosion, opening operation, closing operation, etc.);

[0123] S33. Calculate the geometric characteristics of the temperature anomaly region, such as the area, perimeter, and shape factor, as well as the statistical characteristics such as the maximum temperature value, minimum temperature value, average temperature value, and standard deviation. The area is obtained using the following formula:

[0124]

[0125] Where I(i, j) represents the pixel value of the binary image at position (i, j) (1 represents the pixel within the abnormal area, and 0 represents the pixel outside the area), N and M represent the number of rows and columns of the image respectively;

[0126] The perimeter is obtained using the following formula:

[0127]

[0128] Among them, (x k ,y k ) represents the continuous points on the boundary, and L represents the number of boundary points;

[0129] The shape factor is obtained by the following formula:

[0130]

[0131] The maximum temperature is obtained by the following formula:

[0132] T max =max(I(i,j))

[0133] The minimum temperature is obtained by the following formula:

[0134] T min =min(I(i,j))

[0135] The average temperature is obtained by the following formula:

[0136]

[0137] Where T(i,j) represents the temperature value at position (i,j);

[0138] The temperature standard deviation is obtained by the following formula:

[0139]

[0140] S34. Set multiple temperature gradient calculation points around the temperature anomaly area, calculate the temperature gradient between each calculation point and its adjacent points, and form a temperature gradient distribution map; the temperature gradient is obtained by the following formula:

[0141]

[0142] in, represents the temperature gradient at position (i, j); as well as Represent the partial derivatives of temperature in x and y, respectively, and are approximately calculated using the following difference formula:

[0143]

[0144] Among them, Δ x and Δ y Represents the pixel spacing in the x and y directions of the image respectively;

[0145] S35. Analyze the thermal conductivity of the insulation layer based on the temperature gradient distribution diagram to determine whether there is a thermal conductivity barrier or heat accumulation phenomenon;

[0146] S36. Combine the thermal properties of the insulation material (such as thermal conductivity, thermal expansion coefficient, etc.) to evaluate the thermal stability of the insulation layer and predict the possible thermal failure risk;

[0147] S37. Establish a thermal stability assessment model to quantitatively assess the thermal stability of the insulation layer based on the geometric characteristics, statistical characteristics, and temperature gradient distribution of the temperature anomaly region.

[0148] S38. Based on the results of the thermal stability assessment model, combined with historical data and industry standards, set warning thresholds, including area thresholds for temperature anomaly areas, temperature gradient thresholds, etc., and dynamically adjust the warning thresholds based on the aging degree of the insulation layer and changes in the operating environment.

[0149] The working principle of the above technical solution is as follows: using this technology to obtain a temperature distribution image of the cable joint insulation layer, which can intuitively show the temperature differences on the insulation layer surface; applying image processing algorithms (such as threshold segmentation and edge detection) to process the temperature distribution image to preliminarily identify temperature anomaly areas. These areas usually appear as hot or cold spots with higher or lower temperatures than the surrounding normal areas, indicating potential thermal failure risks; refining the preliminarily identified temperature anomaly areas, and using morphological processing techniques (such as expansion, erosion, opening and closing operations) to remove noise and artifacts to improve the accuracy and reliability of identification; calculating the geometric characteristics of the temperature anomaly area, such as the area, perimeter, and shape factor, which can reflect the scale and shape of the anomaly area; calculating the statistical characteristics of the temperature anomaly area, such as the maximum, minimum, average, and standard deviation, which can reflect the temperature distribution and dispersion of the anomaly area; setting multiple temperature gradient calculation points around the temperature anomaly area, and these points should be evenly distributed to cover the entire anomaly area; calculating the temperature gradient between each calculation point and its adjacent points to form a temperature gradient distribution map. This graph reflects the temperature variations on the insulation layer's surface and is an important basis for assessing the insulation layer's thermal conductivity and thermal stability. The thermal conductivity of the insulation layer is analyzed based on the temperature gradient distribution graph. An abnormal increase or decrease in the temperature gradient may indicate a thermal conductivity barrier or heat accumulation, a key indicator of thermal failure risk. The thermal stability of the insulation layer is assessed by combining the thermal properties of the insulation material (such as thermal conductivity and thermal expansion coefficient). These properties reflect the insulation layer's ability to withstand thermal stress. Based on the thermal stability assessment results, the potential risk of thermal failure is predicted. If the insulation layer has poor thermal stability and exhibits thermal conductivity barriers or heat accumulation, the risk of thermal failure is high. A thermal stability assessment model is developed to quantitatively assess the thermal stability of the insulation layer based on the geometric and statistical characteristics of the temperature anomaly region and the temperature gradient distribution. This model provides specific assessment results and scores, facilitating an objective evaluation of the insulation layer's thermal stability. Warning thresholds are set based on the results of the thermal stability assessment model, combined with historical data and industry standards. These thresholds, including the area threshold of the temperature anomaly region and the temperature gradient threshold, are used to determine whether the insulation layer is in a hazardous state. The warning threshold is dynamically adjusted based on the aging of the insulation layer and changes in the operating environment. This ensures the accuracy and reliability of the warning system, allowing for the timely detection and resolution of potential thermal failure risks.

[0150] The effects of the above technical solutions are as follows: through high-resolution infrared thermal imaging technology, the temperature distribution of the insulation layer of the cable joint can be accurately captured, which has higher monitoring accuracy and efficiency than traditional methods; the application of image processing algorithms, such as threshold segmentation and edge detection, can accurately identify temperature abnormal areas, providing a reliable basis for subsequent detailed analysis; the application of morphological processing technology effectively removes noise and artifacts, and improves the accuracy and reliability of identifying temperature abnormal areas; the calculation of the geometric and statistical characteristics of the temperature abnormal area provides rich data support, which helps to more comprehensively understand the thermal state of the insulation layer; the construction of the temperature gradient distribution map intuitively displays the temperature changes around the temperature abnormal area, and provides a powerful tool for evaluating the thermal conductivity of the insulation layer; by analyzing the temperature gradient distribution map, heat conduction obstacles or heat accumulation phenomena can be discovered in time, providing a basis for preventing thermal failure Important basis; combining the thermal properties of the insulation material to evaluate the thermal stability of the insulation layer can more accurately predict possible thermal failure risks; the establishment of a thermal stability evaluation model realizes the quantitative evaluation of the thermal stability of the insulation layer, providing a scientific basis for formulating targeted maintenance strategies; based on the results of the thermal stability evaluation model, combined with historical data and industry standards, reasonable early warning thresholds are set, including area thresholds and temperature gradient thresholds for temperature anomaly areas; the dynamic adjustment of the early warning threshold takes into account the aging degree of the insulation layer and changes in the operating environment, ensuring the accuracy and reliability of the early warning system; this technical solution monitors the temperature distribution and changes of the insulation layer of the cable joint in real time, and promptly discovers and handles potential thermal failure risks, which helps to ensure the safe operation of the power system; by optimizing the early warning system, the ability to prevent thermal failure risks is improved, and power accidents and losses caused by thermal failure are reduced.

[0151] In one embodiment of the present invention, the step S35 includes:

[0152] On the temperature gradient distribution map, the areas where the temperature gradient changes significantly are identified. By setting the gradient change threshold, the areas where the temperature gradient exceeds or falls below the threshold are marked as potential heat conduction barrier areas.

[0153] Using connected domain analysis in image processing technology, the points on the temperature gradient map are connected into lines to form a heat conduction path network;

[0154] Based on the thermal conductivity characteristics of the insulation material, each path is weighted, and the weight is inversely proportional to the thermal conductivity of the material. That is, the lower the thermal conductivity, the larger the path weight. Based on the reconstructed heat conduction path network, the thermal conduction efficiency of the entire insulation layer is calculated, measured by the average value, maximum value, or cumulative weight of the path weight on a specific path. The thermal conduction efficiency is calculated using the following formula:

[0155]

[0156] Where N represents the total number of paths in the heat conduction path network, W i represents the weight of the i-th path, L i represents the length of the i-th path, k i represents the thermal conductivity of the material in the i-th path, Indicates the average thermal conductivity of the entire insulation material; represents the weight of a unit length path, and

[0157] Evaluate deviations from normal thermal conductivity to identify areas with significantly reduced thermal conductivity, which may indicate thermal conductivity barriers. Analyze localized high or low temperature concentrations in the temperature gradient distribution map, which may indicate heat accumulation or poor heat dissipation in the insulation layer. Calculate the rate of change of the temperature gradient in these areas to assess the rate and extent of heat accumulation.

[0158] Combining the results of temperature gradient anomaly identification and heat conduction efficiency assessment, the heat conduction obstruction points in the insulation layer are located. Based on the temperature gradient characteristics of the obstruction point, its location on the heat conduction path, and possible material defects, the heat conduction obstruction is classified into poor contact, material aging, and embedded foreign matter.

[0159] Using time series analysis technology, we compare historical data of temperature gradient distribution maps, analyze the trend of thermal conductivity performance over time, identify areas where thermal conductivity performance continues to decline, and predict possible thermal failure risks in the future.

[0160] The working principle of the above technical solution is as follows: First, on the temperature gradient distribution map, by comparing the temperature differences between adjacent points, areas with significant temperature gradient changes are identified. These areas may indicate anomalies or thermal resistance points in the heat conduction path. To accurately identify potential heat conduction obstructions, a gradient change threshold is set. When the temperature gradient exceeds or falls below the threshold, the corresponding area is marked as a potential heat conduction obstruction area. Using connected domain analysis, image processing techniques, the points on the temperature gradient map are connected into lines to form a heat conduction path network. This step helps to visually display the heat transfer paths in the insulation layer. Based on the thermal conductivity properties of the insulation material, each path is assigned a weight. The weight is inversely proportional to the thermal conductivity of the material; that is, the lower the thermal conductivity, the greater the path weight. This step reflects the ease with which heat transfer occurs along different paths. Based on the reconstructed heat conduction path network, the thermal conduction efficiency of the entire insulation layer is calculated. This can be measured as the average, maximum, or cumulative weight of the path weights for a specific path. The calculated thermal conduction efficiency is compared with the normal value to identify areas with significantly reduced thermal conduction efficiency. These areas may contain thermal conductivity barriers, hindering heat transfer. Analyze localized areas of high or low temperature concentration in the temperature gradient distribution map. These areas may indicate heat accumulation or poor heat dissipation within the insulation layer. By calculating the rate of change of the temperature gradient in these areas, the rate and extent of heat accumulation can be assessed. This step helps understand the dynamic changes in heat within the insulation layer. Combining the results of temperature gradient anomaly identification and thermal conductivity efficiency assessment, thermal conductivity barriers within the insulation layer can be located. These barriers may be key factors causing heat transfer obstruction. Thermal conductivity barriers are classified based on their temperature gradient characteristics, location along the heat transfer path, and possible material defects. Common classifications include poor contact, material aging, and embedded foreign matter. Time series analysis techniques are used to compare historical data on the temperature gradient distribution map. This step helps understand the changing trends of thermal conductivity performance over time. By identifying areas of consistently declining thermal conductivity, the risk of future thermal failure can be predicted. This step provides a scientific basis for developing targeted maintenance strategies.

[0161] The effect of the above technical solution is as follows: by identifying areas with significant temperature gradient changes on the temperature gradient distribution map and setting a gradient change threshold, the technology can accurately mark potential areas of heat conduction obstruction. This method is more intuitive and accurate than traditional methods, helping to detect and address heat conduction problems early. Using connected domain analysis in image processing technology, the points on the temperature gradient map are connected into lines to form a heat conduction path network. This step not only helps to intuitively display the heat propagation path in the insulation layer, but also provides a basis for subsequent calculations of heat conduction efficiency. Based on the thermal conductivity characteristics of the insulation material, each path is weighted, and the thermal conduction efficiency of the entire insulation layer is calculated based on the reconstructed heat conduction path network. This method can comprehensively and scientifically evaluate the thermal conductivity performance of the insulation layer and provide strong support for identifying heat conduction obstacles. It analyzes the local high-temperature or low-temperature cluster areas in the temperature gradient distribution map and calculates the temperature gradient change rate in these areas to assess the rate and extent of heat accumulation. This step provides a deeper understanding of the dynamics of heat within the insulation layer, providing an important basis for preventing thermal failure. Combining the results of temperature gradient anomaly identification and thermal conductivity efficiency assessment, thermal conductivity obstructions within the insulation layer are located and classified based on their temperature gradient characteristics, location along the heat conduction path, and possible material defects. This method accurately identifies different types of thermal conductivity obstructions, providing guidance for implementing appropriate maintenance measures. Using time series analysis techniques, the temperature gradient distribution map is compared with historical data to analyze the trends in thermal conductivity performance over time. By identifying areas of consistently declining thermal conductivity, this technology can predict the risk of future thermal failures, enabling proactive preventative measures. This comprehensive and in-depth analysis of the thermal conductivity and thermal conductivity obstructions within the cable joint insulation layer helps promptly identify and address potential thermal failure risks, thereby improving the safety and reliability of the power system. This is crucial for ensuring the stable operation of the power system and reducing failures caused by thermal failure.

[0162] In one embodiment of the present invention, the S4 includes:

[0163] S41. Based on historical analysis of power system operating data, combined with industry standards and practical experience, different levels of warning thresholds are set, including minor warning, moderate warning, and severe warning. Based on the warning trigger mechanism, when the monitoring data reaches or exceeds the set threshold, the corresponding level of warning is automatically triggered. Warning information includes key information such as warning level, warning time, warning location, possible impact range, and recommended response measures. Warning information delivery channels are established, including SMS, email, app push, and other methods to ensure that warning information can be quickly and accurately delivered to relevant personnel.

[0164] Differentiated early warning response strategies are formulated for different levels of early warning, including on-site inspections, remote monitoring, equipment shutdown and other measures to control the development of faults and reduce losses.

[0165] Establish an early warning response team, clarify the responsibilities and tasks of team members, and ensure that early warning response measures can be implemented quickly and effectively.

[0166] S42. Using real-time data such as current, voltage, and temperature collected by monitoring equipment, combined with historical data and industry standards, extract fault characteristics, including abnormal fluctuations, mutation points, and frequency components; and preprocessing the extracted fault characteristics;

[0167] S43. Based on the fault characteristics, combined with fault location algorithms, including advanced algorithms such as wavelet transform, neural network, and support vector machine, the fault location is located; and the fault location algorithm is simulated, tested, and optimized;

[0168] S44. Based on the results obtained from the fault location algorithm, combined with on-site inspection and data analysis, diagnose the fault and determine the fault type and cause; establish a fault diagnosis knowledge base to record the fault characteristics, location methods and treatment measures of common faults;

[0169] S45. Based on the results of fault location and diagnosis, combined with changes in the operating status of the power system, the early warning threshold is dynamically adjusted, and the performance of the early warning mechanism is regularly tested and evaluated. The early warning threshold and trigger mechanism are optimized and improved based on the test results.

[0170] The working principle of the above technical solution is as follows: based on the historical analysis of power system operation data, combined with industry standards and practical experience, different levels of warning thresholds (minor warning, medium warning, severe warning) are set; these thresholds are used to determine whether the current state of the power system deviates from the normal range, thereby triggering the corresponding warning; when the data collected by the monitoring equipment (such as current, voltage, temperature, etc.) reaches or exceeds the set warning threshold, the corresponding level of warning is automatically triggered; the warning information includes key information such as warning level, warning time, warning location, possible impact range and recommended response measures; establish multiple warning information transmission channels (such as SMS, email, APP push, etc.) to ensure that the warning information can be quickly and accurately delivered to relevant personnel; formulate differentiated warning response strategies for different levels of warnings (such as on-site inspection, remote monitoring, equipment shutdown, etc.); establish a warning response team, clarify the responsibilities and tasks of team members, and ensure that warning response measures can be implemented quickly and effectively; use monitoring equipment to collect current, voltage, temperature and other data of the power system in real time; combine historical data and industry standards to extract fault features from the collected data, including abnormal fluctuations, mutation points and frequency components; pre-process the extracted fault features, such as denoising and filtering , normalization, etc. to improve the accuracy of subsequent fault location and diagnosis; select appropriate fault location algorithms (such as wavelet transform, neural network, support vector machine and other advanced algorithms); locate the fault location based on the extracted fault features using the fault location algorithm; conduct simulation tests on the fault location algorithm to evaluate its accuracy and robustness under different fault types and operating environments; optimize and improve the algorithm based on the test results to improve the accuracy and reliability of fault location; diagnose the fault based on the results obtained by the fault location algorithm, combined with on-site inspection and data analysis, to determine the fault type and cause; establish a fault diagnosis knowledge base to record the fault characteristics, location methods and treatment measures of common faults; the knowledge base can be used to assist fault diagnosis and provide treatment suggestions to improve the efficiency and accuracy of fault diagnosis; dynamically adjust the warning threshold based on the results of fault location and diagnosis, combined with changes in the operating status of the power system; this helps ensure that the warning mechanism maintains sensitivity and accuracy under different operating conditions; regularly test and evaluate the performance of the warning mechanism, including warning trigger accuracy, information transmission speed, warning response effect, etc.; optimize and improve the warning threshold and trigger mechanism based on the test results to improve the overall performance and reliability of the warning mechanism.

[0171] The effects of the above technical solutions are: by setting warning thresholds based on historical data and industry standards, combined with real-time monitoring data, it is possible to accurately trigger warnings of corresponding levels, ensuring the accuracy and timeliness of warning information; the establishment of multiple warning information transmission channels (such as text messages, emails, APP push, etc.) ensures that warning information can be quickly and accurately transmitted to relevant personnel, shortening response time; using advanced algorithms (such as wavelet transform, neural networks, support vector machines, etc.) to locate faults, which improves the accuracy and efficiency of fault location; combining on-site inspections and data analysis to diagnose faults, it is possible to accurately determine the type and cause of faults, and provide strong support for fault handling; for different levels of warnings, formulate differentiated warning response strategies, including on-site inspections, remote monitoring, equipment shutdown and other measures, to effectively control the development of faults and reduce losses; establish a warning response team and clarify the team members' responsibilities and tasks to ensure that early warning response measures can be implemented quickly and effectively; establish a fault diagnosis knowledge base to record the fault characteristics, location methods and treatment measures of common faults, provide a convenient reference for fault handling, and improve the efficiency and quality of fault handling; according to the results of fault location and diagnosis, combined with the changes in the operating status of the power system, dynamically adjust the early warning threshold to ensure the sensitivity and accuracy of the early warning mechanism; regularly perform performance testing and evaluation of the early warning mechanism, optimize and improve the early warning threshold and trigger mechanism according to the test results, and realize continuous optimization and improvement of the early warning mechanism; through the implementation of the above technical solutions, potential faults in the power system can be discovered and handled in a timely manner, the occurrence and development of faults can be effectively prevented, and the safety and reliability of the power system can be enhanced; this is of great significance to ensure the stable operation of the power system, reduce power outages and ensure the safety of users' electricity use.

[0172] In one embodiment of the present invention, the step S43 includes:

[0173] S431. Preliminarily apply the fault location algorithm to the fault feature dataset based on the fault features, preliminarily locate the fault location, and evaluate the basic performance and accuracy of the algorithm.

[0174] S432. Fault features extracted from multi-source monitoring data such as current, voltage, and temperature are integrated to form a comprehensive feature vector. A power system fault simulation model is constructed to simulate different types of fault scenarios, including short circuit, open circuit, overload, etc., and generate corresponding fault feature data sets.

[0175] S433. Apply the optimized fault location algorithm to the simulation data set to verify the location accuracy and stability of the algorithm under different fault types and different operating conditions.

[0176] The working principle of the above technical solution is as follows: based on the real-time data of current, voltage, temperature, etc. collected from the monitoring equipment, combined with historical data and industry standards, fault characteristics are extracted, including abnormal fluctuations, mutation points, and frequency components; the extracted fault characteristic data set is input into the preliminarily selected fault location algorithm to preliminarily locate the fault location; the basic performance and accuracy of the algorithm are evaluated, including indicators such as positioning accuracy, calculation speed, and stability; the evaluation results will be used to guide the optimization and improvement of subsequent algorithms; the fault characteristics extracted from multi-source monitoring data such as current, voltage, and temperature are fused to form a comprehensive feature vector; the fused feature vector can more comprehensively reflect the characteristic information of the fault and improve the accuracy of fault location; according to the structure and characteristics of the power system, a power system fault simulation model is constructed. ; The simulation model can simulate different types of fault scenarios, including short circuit, open circuit, overload, etc., and generate corresponding fault feature data sets; by running the simulation model, fault feature data sets under different fault types and different operating conditions are generated; these data sets will be used for subsequent algorithm optimization and verification; based on the results of preliminary application and performance evaluation, the fault location algorithm is optimized and improved; the optimization may include adjustment of algorithm parameters, improvement of algorithm structure, etc.; the optimized fault location algorithm is applied to the simulation data set to verify the algorithm's positioning accuracy and stability under different fault types and different operating conditions; the verification results will be used to evaluate the optimization effect of the algorithm and guide further improvement of the subsequent algorithm; based on the verification results, the algorithm is iteratively optimized until satisfactory positioning accuracy and stability are achieved.

[0177] The above technical solution achieves the following: By initially applying the fault location algorithm to an actual fault signature dataset, the fault location can be preliminarily located. This step not only provides preliminary guidance for subsequent fault handling, but also, by evaluating the algorithm's basic performance and accuracy, promptly identifies potential algorithm issues, laying the foundation for subsequent optimization. By integrating fault signatures extracted from multiple monitoring data sources, such as current, voltage, and temperature, a comprehensive feature vector is formed. This step not only improves the comprehensiveness of the fault signature, enabling the algorithm to more accurately capture the essential characteristics of the fault, but also generates a rich fault signature dataset by constructing a power system fault simulation model and simulating different fault scenarios. This provides strong data support for subsequent algorithm optimization and validation. By applying the optimized fault location algorithm to the simulation dataset, the algorithm's accuracy and stability under different fault types and operating conditions are verified. This step not only ensures the algorithm's reliability and effectiveness in practical applications, but also, through continuous iterative optimization, further enhances its performance, making it more adaptable to the complexity and diversity of power systems. Through the implementation of the above technical solution, the algorithm's performance has been significantly improved. This not only makes fault location more accurate and rapid, but also reduces the time and labor costs of troubleshooting, improving the operational efficiency and reliability of the power system. Accurate fault location and timely troubleshooting can effectively prevent the development and expansion of faults, reducing power outages and safety incidents in the power system. This not only ensures the safe and stable operation of the power system, but also improves user satisfaction and the economic benefits of power companies.

[0178] In one embodiment of the present invention, the step S432 includes:

[0179] Extract fault features such as abnormal fluctuations, mutation points, and frequency components from multi-source monitoring data such as current, voltage, and temperature through signal processing technology and data analysis methods, and pre-process the extracted fault features.

[0180] The extracted features are fused based on a preset feature fusion strategy, and a fault simulation model is constructed based on the structure and operating characteristics of the power system, including the grid topology, equipment parameters, fault type, etc.

[0181] Based on the fault simulation model, we generate fault feature datasets for different fault types and operating conditions. We then apply the fused feature vectors to the fault location algorithm to evaluate the effect of feature fusion on improving the algorithm's performance.

[0182] According to the evaluation results of the feature fusion effect, the simulation model and feature fusion strategy are iteratively optimized.

[0183] The working principle of the above technical solution is as follows: real-time data is collected from multi-source monitoring equipment such as current, voltage, and temperature of the power system; signal processing technology and data analysis methods such as filtering, transformation, and statistical analysis are used to process the collected data and extract fault characteristics; fault characteristics may include abnormal fluctuations, mutation points, frequency components, etc., which can reflect the abnormal state of the power system; the extracted fault characteristics are pre-processed, such as denoising, normalization, and standardization, to improve the quality of the characteristics and the robustness of the algorithm; according to a preset feature fusion strategy, the extracted fault characteristics are fused to form a comprehensive feature vector; feature fusion strategies may include weighted averaging, principal component analysis (PCA), neural network fusion, etc.; the structure and operating characteristics of the power system, such as the grid topology and equipment parameters, are analyzed to provide a basis for building a fault simulation model; Design a fault simulation model based on the structure and operating characteristics of the power system; the simulation model should be able to simulate different types of fault scenarios, such as short circuit, open circuit, overload, etc.; based on the fault simulation model, generate fault feature data sets under different fault types and different operating conditions; these data sets will be used for subsequent algorithm optimization and verification; apply the fused comprehensive feature vector to the fault location algorithm to evaluate the algorithm's positioning accuracy and stability under different fault types and different operating conditions; verify the effectiveness of the feature fusion strategy by comparing the positioning accuracy, stability and other indicators of the algorithm before and after fusion; iteratively optimize the simulation model parameters and fault feature extraction method based on the feature fusion effect evaluation results; continuously optimize the feature fusion strategy to improve the effectiveness of feature fusion and the accuracy of the algorithm; repeat the above steps until satisfactory algorithm performance and feature fusion effect are achieved.

[0184] The effects of the above technical solution are as follows: through signal processing technology and data analysis methods, fault features such as abnormal fluctuations, mutation points, frequency components, etc. are extracted from multi-source monitoring data such as current, voltage, and temperature, which can more accurately capture the abnormal state of the power system; the extracted fault features are pre-processed, such as denoising and normalization, to improve the quality and reliability of the features, providing a solid foundation for subsequent feature fusion and algorithm application; based on a preset feature fusion strategy, the extracted fault features are fused to form a comprehensive feature vector. This fusion strategy can comprehensively consider the correlation and complementarity between different fault features, and improve the effectiveness and robustness of the features; by comparing the positioning accuracy, stability and other indicators of the algorithm before and after fusion, the effectiveness of the feature fusion strategy can be verified, providing guidance for subsequent optimization; according to the structure and operating characteristics of the power system, a fault simulation model is constructed, including the grid topology, equipment parameters, fault type, etc. This simulation model can simulate different types of fault scenarios and generate realistic fault feature data sets. By continuously optimizing the simulation model parameters and fault feature extraction methods, the accuracy and practicality of the simulation model can be improved, providing strong support for subsequent algorithm verification and optimization. Applying the fused comprehensive feature vector to the fault location algorithm can significantly improve the algorithm's positioning accuracy and stability. By evaluating the improvement effect of feature fusion on algorithm performance, the algorithm's structure and parameters can be continuously optimized to better adapt to the complexity and diversity of the power system. Accurate fault location and timely fault handling can effectively avoid the development and expansion of faults, and reduce power outages and safety accidents in the power system. The application of this technical solution can improve the safe and stable operation level of the power system, ensure users' electricity needs and the economic benefits of power companies. Automated fault feature extraction, fusion and algorithm application can reduce manual intervention and operation and maintenance costs. At the same time, the application of this technical solution can improve the efficiency of fault handling, shorten fault recovery time, and improve the reliability and availability of the power system.

[0185] In one embodiment of the present invention, the S5 includes:

[0186] S51. Based on the early warning information push mechanism, the early warning information and fault location results are sent to the mobile device or workstation of the operation and maintenance personnel in real time;

[0187] S52. Analyze the structural design and insulation material performance of cold-shrink cable intermediate joints based on monitoring data and early warning results, and propose improvement suggestions based on the analysis results.

[0188] S53. Develop a targeted maintenance plan based on the monitoring data and early warning results, the maintenance plan including inspection frequency, inspection methods, and maintenance measures;

[0189] S54. Carry out regular inspection and maintenance, record the inspection results, compare and analyze with the monitoring data, evaluate the maintenance effect, and continuously optimize the maintenance plan.

[0190] The working principle of the above technical solution is as follows: the monitoring equipment in the power system collects key data such as current, voltage, temperature, etc. of the intermediate joints of the cold shrink cable in real time; through data analysis and algorithm processing, potential faults or abnormal conditions are identified and early warning information is generated; the early warning information push mechanism sends the early warning information and fault location results to the mobile devices (such as mobile phones, tablets) or workstations of the operation and maintenance personnel in real time; the push method may include SMS, email, APP push, etc., to ensure that the operation and maintenance personnel can obtain the early warning information in the first time; the early warning information contains the fault location results, indicating the specific location where the problem may occur in the intermediate joints of the cold shrink cable; this helps the operation and maintenance personnel to quickly locate the problem and take corresponding measures; collect monitoring data of the intermediate joints of the cold shrink cable , including historical and real-time data of parameters such as current, voltage, and temperature; conduct in-depth analysis of the early warning results to understand the abnormal conditions that occur in the operation of the cold shrink cable intermediate joint; evaluate the structural design and insulation material performance of the cold shrink cable intermediate joint based on the monitoring data and early warning results; analyze whether there are defects in the structural design and whether the insulation material is aging or its performance has deteriorated; based on the evaluation results, put forward improvement suggestions for the structural design and insulation material of the cold shrink cable intermediate joint; improvement suggestions may include optimizing the design, replacing high-performance materials, strengthening maintenance, etc.; determine the maintenance needs of the cold shrink cable intermediate joint based on the monitoring data and early warning results; formulate a targeted maintenance plan, including detection frequency, detection methods and maintenance measures. The inspection frequency should be determined based on the operating status and historical fault records of the cold shrink cable intermediate joints; inspection methods may include appearance inspection, electrical performance testing, insulation resistance measurement, etc.; maintenance measures may include cleaning, tightening, replacement of parts, etc.; regular inspection and maintenance should be carried out according to the maintenance plan to ensure the normal operation of the cold shrink cable intermediate joints; the details of each inspection and maintenance should be recorded, including the inspection results, maintenance measures and problems encountered; the results of regular inspections should be compared and analyzed with the monitoring data to evaluate the effectiveness of the maintenance measures; comparative analysis can reveal the changing trends and potential problems of the cold shrink cable intermediate joints during operation; based on the results of comparative analysis, the maintenance plan should be continuously optimized; optimization may include adjusting the inspection frequency, improving the inspection methods, increasing maintenance measures, etc.; the goal is to ensure the long-term safe and stable operation of the cold shrink cable intermediate joints and reduce the failure rate.

[0191] The effects of the above technical solutions are: through the early warning information push mechanism, operation and maintenance personnel can receive early warning information and fault location results of cold shrink cable intermediate joints in real time; this enables operation and maintenance personnel to respond quickly and take timely measures to avoid the expansion of faults, reduce power outage time and economic losses; analysis of the structural design and insulation material properties of cold shrink cable intermediate joints can help to discover potential design defects and material aging problems; improvement suggestions based on the analysis results can guide manufacturers to optimize product design and improve the reliability and durability of cold shrink cable intermediate joints; the formulated maintenance plan is tailored according to monitoring data and early warning results to ensure the pertinence and efficiency of maintenance work; through reasonable detection frequency, detection methods and maintenance measures, operation and maintenance personnel can more effectively manage the maintenance work of cold shrink cable intermediate joints and reduce maintenance costs. Cost, improve maintenance efficiency; regular inspection and maintenance, as well as comparative analysis of inspection results and monitoring data, provide a strong basis for evaluating maintenance effects; continuous optimization of maintenance plans based on evaluation results ensures the long-term stability and reliability of cold shrink cable intermediate joints; this continuous improvement mechanism helps to improve the professional skills of operation and maintenance personnel and improve the overall management level of the power system; through the above series of measures, the safety and stability of cold shrink cable intermediate joints can be significantly improved; this helps to reduce power outages and safety accidents caused by cold shrink cable intermediate joint failures, and ensure the safe and stable operation of the power system; through real-time early warning, targeted maintenance, effect evaluation and continuous optimization and other measures, the operation and maintenance costs can be significantly reduced; at the same time, the power outage time and economic losses caused by failures are reduced, and the economic benefits of the power system are improved.

[0192] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An intelligent monitoring method for the insulation status of power cable joints based on complex environments, characterized in that: The method comprises: S1. Collect relevant data by multiple means, wherein the relevant data includes ambient temperature data, partial discharge signals, and temperature distribution images; S2. Input the collected environmental parameters and partial discharge signals into the intelligent analysis system for preprocessing and feature extraction. Based on the machine learning algorithm, an insulation status prediction model is established to predict the aging degree and potential failure points of the cable joint insulation layer based on historical data and real-time monitoring data. S3. Using infrared thermal imaging data and image processing technology, identify areas of abnormal temperature and calculate the temperature gradient. Based on the temperature gradient calculation results, evaluate the thermal stability of the insulation layer. S4. Set a warning threshold based on the output of the insulation status prediction model. When the prediction result exceeds the threshold, the warning mechanism is triggered. Combined with partial discharge monitoring data and infrared thermal imaging results, the location of the insulation defect is accurately located and the fault type and severity are determined. S5. Send the warning information and fault location results to the operation and maintenance personnel in real time. Based on the monitoring data and warning results, optimize the structural design of the cold shrink cable intermediate joint; Said S3 comprises: S31. Based on high-resolution infrared thermal imaging technology, obtain the temperature distribution image of the cable joint insulation layer and apply image processing algorithms to preliminarily identify the temperature abnormality area; S32, performing detailed analysis on the initially identified temperature anomaly area and removing noise and artifacts using morphological processing techniques; S33, calculating the geometric characteristics and statistical characteristics of the temperature anomaly area; S34, setting a plurality of temperature gradient calculation points around the temperature anomaly area, calculating the temperature gradient between each calculation point and its adjacent points, and forming a temperature gradient distribution map; S35. Analyze the thermal conductivity of the insulation layer based on the temperature gradient distribution diagram to determine whether there is a thermal conductivity barrier or heat accumulation phenomenon; S36. Combine the thermal properties of the insulation material to evaluate the thermal stability of the insulation layer and predict the risk of thermal failure; S37. Establish a thermal stability assessment model to quantitatively assess the thermal stability of the insulation layer based on the geometric characteristics, statistical characteristics, and temperature gradient distribution of the temperature anomaly region. S38. Based on the results of the thermal stability assessment model, combined with historical data and industry standards, set the warning threshold, and dynamically adjust the warning threshold based on the aging degree of the insulation layer and changes in the operating environment.

2. The intelligent monitoring method for the insulation status of power cable joints in a complex environment according to claim 1, characterized in that: Said S1 comprises: S11. Use a high-precision temperature sensor to collect real-time ambient temperature, a humidity sensor to monitor the humidity level around the connector, and an electric field strength sensor to monitor the electric field distribution around the cable connector; S12. Capturing partial discharge signals in the cable insulation layer through a high-frequency current sensor, and performing filtering and noise reduction processing on the collected partial discharge signals; S13. Use a high-resolution infrared thermal imaging camera to regularly capture temperature distribution images of the cable joint insulation layer, and use an image enhancement algorithm to enhance the collected temperature distribution images.

3. The intelligent monitoring method for the insulation status of power cable joints in a complex environment according to claim 1, characterized in that: Said S2 comprises: S21, pre-processing the collected environmental parameters and partial discharge signals; S22. Based on statistical analysis and machine learning algorithms, feature selection is performed on the preprocessed data to identify characteristic variables that have an impact on insulation status prediction; S23. Extract features from the pre-processed data using time domain, frequency domain, and time-frequency domain analysis techniques, and fuse the extracted feature variables to form a comprehensive feature vector; S24. Divide the preprocessed dataset into a training set and a validation set, use the training set to train the built-in machine learning model, and use the validation set to evaluate the performance of the model; S25. Evaluate the performance of the model through cross-validation technology. Adjust the model's hyperparameters and optimize the model structure based on the validation results. S26. Use the trained machine learning model to predict the test set, output the insulation status prediction results, and evaluate the uncertainty of the prediction results; S27. Design early warning strategies based on the uncertainty of the prediction results. When the uncertainty of the prediction results is high, trigger a higher level of early warning mechanism. S28. Compare the predicted results with the actual situation to evaluate the accuracy and reliability of the model. Based on the evaluation results, iteratively optimize the model. Based on the continuous learning and updating mechanism, regularly collect new monitoring data to update and optimize the model.

4. The intelligent monitoring method for insulation status of power cable joints in a complex environment according to claim 3, characterized in that: Said S23 comprises: The mean and variance of each characteristic variable after preprocessing are calculated, and the volatility and stability of the data are evaluated based on the evaluation record. The trend line of the time series data is calculated using the sliding window technology to identify the changing trends of environmental parameters and partial discharge signals over time. Identify and record the maximum and minimum values ​​of each characteristic variable within a time period, analyze abnormal fluctuations, evaluate the correlation of time series data at different time intervals, and reveal the time dependence within the data; Convert the time domain signal into frequency domain representation, analyze the spectral components of each characteristic variable, and identify the main frequency components and their energy distribution; Calculate and plot power spectral density to evaluate signal energy at different frequencies and identify potential periodic or random characteristics; calculate the entropy of the spectrum to measure the complexity of the signal frequency distribution; Perform Fourier transform on the signal within the time window to obtain the spectrum information that changes with time, capture the instantaneous frequency characteristics of the partial discharge signal, decompose the signal using wavelet functions of different scales, and extract the time-frequency characteristics at multiple scales; Decompose the complex signal into a series of intrinsic mode functions, each IMF represents a different frequency component of the signal; normalize all extracted features; Based on correlation analysis, the most critical feature subset for insulation status prediction is further screened, and the screened features are fused into a comprehensive feature vector using weighted averaging. And through random forest, the contribution of each feature in the comprehensive feature vector is evaluated.

5. The intelligent monitoring method for the insulation status of power cable joints in a complex environment according to claim 1, characterized in that: The S35 includes: On the temperature gradient distribution map, the areas where the temperature gradient changes significantly are identified. By setting the gradient change threshold, the areas where the temperature gradient exceeds or falls below the threshold are marked as potential heat conduction barrier areas. Using connected domain analysis in image processing technology, the points on the temperature gradient map are connected into lines to form a heat conduction path network; According to the thermal conductivity characteristics of the insulation material, a weight is assigned to each path, and the weight is inversely proportional to the thermal conductivity of the material. The thermal conductivity efficiency of the entire insulation layer is calculated; Evaluate the deviation of heat transfer efficiency from normal values, identify areas with significantly reduced heat transfer efficiency, analyze local high or low temperature concentration areas in the temperature gradient distribution map, and evaluate the rate and degree of heat accumulation by calculating the temperature gradient change rate of the area; Combining the results of temperature gradient anomaly identification and heat conduction efficiency assessment, the heat conduction obstruction points in the insulation layer are located and classified based on the temperature gradient characteristics of the obstruction points, their location on the heat conduction path, and possible material defects. Using time series analysis technology, we compare historical data of temperature gradient distribution maps, analyze the trend of thermal conductivity performance over time, identify areas where thermal conductivity performance continues to decline, and predict possible thermal failure risks in the future.

6. The intelligent monitoring method for the insulation status of power cable joints in a complex environment according to claim 1, characterized in that: Said S4 comprises: S41. Based on historical analysis of power system operation data, combined with industry standards and practical experience, set different levels of warning thresholds. Based on the warning trigger mechanism, when the monitoring data reaches or exceeds the set threshold, the corresponding level of warning will be automatically triggered; and a warning information transmission channel will be established. S42. Using monitoring equipment to collect real-time data, and combining historical data and industry standards, extract fault characteristics, including abnormal fluctuations, mutation points, and frequency components; and preprocessing the extracted fault characteristics; S43. Locate the fault location based on the fault characteristics and in combination with the fault location algorithm; and perform simulation testing and optimization on the fault location algorithm; S44. Based on the results obtained from the fault location algorithm, combined with on-site inspection and data analysis, diagnose the fault and determine the fault type and cause; establish a fault diagnosis knowledge base to record the fault characteristics, location methods and treatment measures of common faults; S45. Based on the results of fault location and diagnosis, combined with changes in the operating status of the power system, the early warning threshold is dynamically adjusted, and the performance of the early warning mechanism is regularly tested and evaluated. The early warning threshold and trigger mechanism are optimized and improved based on the test results.

7. The intelligent monitoring method for insulation status of power cable joints in a complex environment according to claim 6, characterized in that: The S43 includes: S431. Preliminarily apply the fault location algorithm to the fault feature dataset based on the fault features, preliminarily locate the fault location, and evaluate the basic performance and accuracy of the algorithm. S432. Fusing fault features extracted from multi-source monitoring data to form a comprehensive feature vector, constructing a power system fault simulation model, simulating different types of fault scenarios, and generating corresponding fault feature datasets; S433. Apply the optimized fault location algorithm to the simulation data set to verify the location accuracy and stability of the algorithm under different fault types and different operating conditions.

8. The intelligent monitoring method for insulation status of power cable joints in a complex environment according to claim 7, characterized in that: The S432 includes: Extract fault features from multi-source monitoring data through signal processing technology and data analysis methods, and pre-process the extracted fault features; The extracted features are fused based on the preset feature fusion strategy, and a fault simulation model is constructed according to the structure and operation characteristics of the power system; Based on the fault simulation model, we generate fault feature datasets for different fault types and operating conditions. We then apply the fused feature vectors to the fault location algorithm to evaluate the effect of feature fusion on improving the algorithm's performance. According to the evaluation results of the feature fusion effect, the simulation model and feature fusion strategy are iteratively optimized.

9. The intelligent monitoring method for insulation status of power cable joints in a complex environment according to claim 1, characterized in that: Said S5 comprises: S51. Based on the early warning information push mechanism, the early warning information and fault location results are sent to the mobile device or workstation of the operation and maintenance personnel in real time; S52. Analyze the structural design and insulation material performance of cold-shrink cable intermediate joints based on monitoring data and early warning results, and propose improvement suggestions based on the analysis results. S53. Develop a targeted maintenance plan based on the monitoring data and early warning results, the maintenance plan including inspection frequency, inspection methods, and maintenance measures; S54. Carry out regular inspection and maintenance, record the inspection results, compare and analyze with the monitoring data, evaluate the maintenance effect, and continuously optimize the maintenance plan.

Images