Construction method of integrated test platform for main insulation fault detection and management of power distribution network cable

Abstract

The invention relates to the field of power distribution network cable main insulation fault detection and management, and discloses a power distribution network cable main insulation fault detection and management integrated test platform construction method comprising the following steps: a) data acquisition: installing a plurality of sensors including a partial discharge sensor, a current and voltage sensor, a temperature sensor and a vibration sensor; acquiring real-time operation data of the power distribution network cable; b) data transmission: transmitting the acquired cable monitoring data to a data processing system in a wired or wireless communication mode; and c) data preprocessing: denoising, missing value interpolation, standardization or normalization processing is carried out on the transmitted original data so as to improve the quality of the data. The real-time operation state of the power distribution network cable is monitored through various sensors, and various advanced data processing methods are combined, so that the accuracy of fault detection is effectively improved, and identification deviation caused by data noise or sensor errors in a traditional fault detection method is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of main insulation fault detection and management of cables in distribution networks, and in particular to a method for constructing an integrated test platform for main insulation fault detection and management of cables in distribution networks. Background Art

[0002] With the rapid development of the power industry, the safety and reliability of distribution networks have become the focus of power companies. As an important part of power transmission, the operating status of distribution network cables directly affects the stability of power supply. Traditional distribution network fault detection methods mainly rely on manual inspections and conventional equipment monitoring methods, but these methods often have shortcomings such as slow response speed, low detection accuracy, and inability to predict faults in a timely manner. Especially when faced with complex fault types and cable aging, the detection capabilities of traditional methods are particularly insufficient. Therefore, the development of an intelligent detection system that can monitor the status of distribution network cables in real time and has fault diagnosis and prediction capabilities has become an urgent need to improve the safety of distribution networks and reduce the occurrence of faults.

[0003] In the existing technology, distribution network cable fault detection mostly uses traditional physical sensors, such as temperature sensors, current sensors, etc., to identify faults through simple data collection and processing methods. Although these technologies can detect fault signals to a certain extent, due to the diversity and complexity of fault types, traditional methods have limitations in feature extraction, data processing and diagnostic accuracy. At the same time, traditional methods are usually unable to provide key information such as the time and location of the fault in real time, and it is also difficult to comprehensively evaluate the health status of the cable and predict fault trends, so it is difficult to take effective preventive measures before the fault occurs.

[0004] However, the fault detection methods in the existing technologies still have significant problems in terms of accuracy, real-time and intelligence. Traditional methods often rely on a single data source and are unable to perform multi-source data fusion and comprehensive analysis, resulting in the inability to accurately identify complex fault types and the lack of early warning capabilities for future faults. In addition, although some methods have adopted machine learning technology, they lack efficient multi-dimensional feature extraction and the application of deep learning models, and are unable to provide comprehensive fault diagnosis and prediction information. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for constructing an integrated test platform for the detection and management of main insulation faults in distribution network cables, which solves the problem that the fault detection methods in the prior art still have significant problems in accuracy, real-time and intelligence level. Traditional methods often rely on a single data source and cannot perform multi-source data fusion and comprehensive analysis.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for constructing an integrated test platform for main insulation fault detection and management of distribution network cables, comprising the following steps:

[0007] a) Data collection: by installing a variety of sensors including partial discharge sensors, current and voltage sensors, temperature sensors and vibration sensors, the real-time operation data of the distribution network cables are collected;

[0008] b) Data transmission: The collected cable monitoring data is transmitted to the data processing system via wired or wireless communication;

[0009] c) Data preprocessing: denoising, missing value interpolation, standardization or normalization of the transmitted raw data to improve the quality of the data;

[0010] d) Data fusion: Use a variety of data fusion technologies to fuse data from different sensors and generate a unified feature vector to ensure the effective integration of multi-source data;

[0011] e) Feature extraction: Extract time domain, frequency domain and high-order statistical features from the fused data for cable fault diagnosis and prediction;

[0012] f) Fault diagnosis and prediction: Based on machine learning and deep learning models, cable faults can be diagnosed in real time and the time and location of potential faults can be predicted;

[0013] g) Fault warning and location: By analyzing real-time data, it automatically generates warning information of cable faults and locates the specific location of the fault;

[0014] h) Visualization and report generation: Display the operating status of the cable, fault detection results and prediction information through a visual interface, and generate fault diagnosis and prediction reports.

[0015] Preferably, the sensors include partial discharge sensors, current and voltage sensors, temperature sensors and vibration sensors, and the installation locations of the sensors include key nodes and operating sections of the distribution network cables.

[0016] Preferably, in step c, the data preprocessing step includes performing wavelet denoising, missing value interpolation and standardization on the collected signal, wherein the missing value interpolation is implemented by the K-NearestNeighbor method, and the standardization adopts the Z-Score standardization method.

[0017] Preferably, in step d, the data fusion step includes Kalman filtering and mutual information fusion methods, Kalman filtering is used for data fusion of dynamic systems, and mutual information fusion is used to ensure that important information between multi-sensor data is retained.

[0018] Preferably, in step e, the feature extraction step includes time domain feature extraction, frequency domain feature extraction and high-order statistical feature extraction, the time domain features include mean, standard deviation, skewness and kurtosis, and the frequency domain features include Fourier transform results.

[0019] Preferably, the machine learning model includes support vector machine, random forest and XGBoost, and the deep learning model includes long short-term memory network and convolutional neural network. The models are combined for ensemble learning, and weighted voting method or weighted averaging method is used for result fusion.

[0020] Preferably, in step f, the fault diagnosis and prediction step includes inputting real-time monitoring data into trained machine learning and deep learning models, and outputting whether the cable has a fault, the fault type, location and severity.

[0021] Preferably, in step g, in the fault warning and location step, the system predicts the occurrence time and location of the cable fault based on historical data and real-time monitoring data in combination with the LSTM timing model, and displays the warning information and fault location on a visual interface.

[0022] Preferably, in step h, the report generation step includes automatically generating a cable fault diagnosis report and a prediction report, wherein the report includes the fault type, occurrence time, occurrence location, predicted fault trend and recommended operation and maintenance measures.

[0023] Preferably, the method is applicable to real-time monitoring, fault management, fault prediction and intelligent operation and maintenance of distribution network cables, and can be extended to fault detection and management of other power equipment.

[0024] The present invention provides a method for constructing an integrated test platform for main insulation fault detection and management of distribution network cables. It has the following beneficial effects:

[0025] 1. The present invention monitors the real-time operating status of the distribution network cables through a variety of sensors and combines a variety of advanced data processing methods to effectively improve the accuracy of fault detection. Multi-source data fusion and high-precision feature extraction methods enable the system to more accurately identify the type, location and severity of faults, avoiding the identification bias caused by data noise or sensor errors in traditional fault detection methods.

[0026] 2. The present invention uses machine learning and deep learning models, such as support vector machines, random forests, XGBoost, LSTM and convolutional neural networks, to comprehensively evaluate the health status of cables during fault diagnosis and fault prediction. The fault type, occurrence time, location and predicted fault trend can all be accurately predicted, so that the system can issue an early warning before the fault occurs, and notify the operation and maintenance personnel in advance to take corresponding preventive measures, thereby significantly reducing the incidence of sudden faults and reducing the downtime and economic losses of the power system.

[0027] 3. Through intelligent analysis based on real-time monitoring data and historical data, the present invention can timely discover potential fault risks in the distribution network and predict the time and location of the fault through the prediction model. This provides important decision-making support for power companies and helps operation and maintenance personnel prepare fault emergency plans or preventive maintenance in advance, thereby effectively improving the safety and reliability of the distribution network and ensuring the continuity and stability of power supply.

[0028] 4. The present invention can not only provide fault diagnosis and prediction information, but also automatically generate operation and maintenance measures suggestions based on the cable fault type, location and prediction trend. Through intelligent fault location and resource scheduling suggestions, operation and maintenance personnel can make more accurate decisions based on the report and achieve optimal scheduling of maintenance resources. The system can intelligently deploy maintenance personnel and equipment according to the fault location, impact range and risk level to ensure the rapid implementation of fault repair work, improve operation and maintenance efficiency and reduce maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] Please see attached Figure 1 The embodiment of the present invention provides a method for constructing an integrated test platform for main insulation fault detection and management of distribution network cables, comprising the following steps:

[0032] a) Data collection: by installing a variety of sensors including partial discharge sensors, current and voltage sensors, temperature sensors and vibration sensors, the real-time operation data of the distribution network cables are collected;

[0033] b) Data transmission: The collected cable monitoring data is transmitted to the data processing system via wired or wireless communication;

[0034] c) Data preprocessing: denoising, missing value interpolation, standardization or normalization of the transmitted raw data to improve the quality of the data;

[0035] d) Data fusion: Use a variety of data fusion technologies to fuse data from different sensors and generate a unified feature vector to ensure the effective integration of multi-source data;

[0036] e) Feature extraction: Extract time domain, frequency domain and high-order statistical features from the fused data for cable fault diagnosis and prediction;

[0037] f) Fault diagnosis and prediction: Based on machine learning and deep learning models, cable faults can be diagnosed in real time and the time and location of potential faults can be predicted;

[0038] g) Fault warning and location: By analyzing real-time data, it automatically generates warning information of cable faults and locates the specific location of the fault;

[0039] h) Visualization and report generation: Display the operating status of the cable, fault detection results and prediction information through a visual interface, and generate fault diagnosis and prediction reports.

[0040] In one embodiment, the data collection and sensor network module

[0041] 2.1 Sensor selection and function

[0042] Partial discharge (PD) sensor: Partial discharge is an early signal of cable main insulation defects. This embodiment uses UHF (ultra-high frequency) partial discharge sensors installed at key nodes of the cable to capture high-frequency discharge signals.

[0043] Current and voltage sensors: used to monitor the load condition of the cable. Current and voltage sensors are installed at the starting and end points of each cable section to ensure full control of the cable working status.

[0044] Temperature sensor: monitors the cable surface temperature and the surrounding environment temperature, and monitors temperature changes in real time by installing it outside the insulation layer of the cable.

[0045] Vibration sensor: used to detect physical shock in the cable operating environment. Vibration sensors are installed at cable joints and nodes to monitor the occurrence of mechanical damage to the cable in real time.

[0046] 2.2 Sensor installation method and data collection frequency

[0047] Each sensor is placed at a key position on the cable, and the sampling frequency of each type of sensor is set according to the fault characteristics:

[0048] The PD sensor sampling frequency is set to 500MHz to capture instantaneous discharge events.

[0049] The sampling frequency of the current and voltage sensors is set to 5kHz to ensure that electrical characteristic fluctuations within a short period of time can be detected.

[0050] The temperature sensor samples once per second to detect temperature change trends.

[0051] The vibration sensor samples 10 times per second to monitor external physical shock.

[0052] Each sensor transmits the collected analog signal to the data acquisition unit (DAQ), which converts it into a digital signal and transmits it to the data transmission and preprocessing module in real time.

[0053] Data transmission and preprocessing module

[0054] 3.1 Data Transmission Method

[0055] In order to adapt to different power grid environments, this platform transmits the collected data to the central server through wired communication (such as optical fiber) and wireless communication (such as NB-IoT, LoRa). Remote sensors use low-power wireless communication to ensure stable and efficient operation of sensors.

[0056] 3.2 Data Preprocessing Process

[0057] Denoising

[0058] Method selection: Use wavelet transform to perform signal denoising. Wavelet transform can decompose the signal into different frequency levels, retain effective fault features and remove noise.

[0059] Denoising steps:

[0060] Decompose the original signal by wavelet and select the appropriate wavelet basis function (such as Daubechies wavelet).

[0061] Threshold processing is performed on each layer of decomposed signal to retain the frequency band of the fault characteristic signal and remove high-frequency noise.

[0062] The processed signal is reconstructed by wavelet to restore the denoised signal.

[0063] Missing Value Interpolation

[0064] Interpolation method: KNN (K-NearestNeighbor) method is used to interpolate missing data. The specific process is as follows:

[0065] Perform KNN search on missing data to find the k nearest data points.

[0066] The mean of k neighboring data is used to fill missing values ​​to ensure data continuity.

[0067] Standardization and normalization

[0068] Standardization method: Z-Score standardization is used to convert sensor data from different sources to the same scale for subsequent analysis.

[0069] Normalization formula:

[0070]

[0071] Among them, X is the original data, μ is the mean, and σ is the standard deviation.

[0072] Data storage and fusion module

[0073] 4.1 Data Storage System

[0074] The platform uses a distributed big data storage system (such as Hadoop or Spark) to manage and store data. The system has horizontal expansion capabilities, can efficiently process massive monitoring data, achieve rapid data access and processing, and provide a basis for subsequent analysis.

[0075] 4.2 Multi-source data fusion method

[0076] Through multi-source data fusion technology, the data of various sensors are uniformly processed and integrated to generate a comprehensive feature vector. The fusion process is as follows:

[0077] Kalman filter fusion

[0078] Principle: Kalman filtering is a recursive algorithm used to estimate the state of a system. The update formula of the Kalman gain is as follows:

[0079]

[0080] in, is the fused state estimate, yk is the measured value, P k∣k is the covariance matrix.

[0081] Application: Dynamically integrate various sensor data to improve the real-time and accuracy of data.

[0082] Mutual Information Fusion

[0083] Principle: Mutual information measures the amount of shared information between different sensor data, ensuring that important information is retained during the fusion process.

[0084] Application: Improve the information utilization between different signals and ensure the effectiveness of data fusion.

[0085] Feature extraction and selection module

[0086] After data fusion, feature extraction is performed on the fused signal to form an input feature vector suitable for the machine learning model. Feature extraction includes time domain features, frequency domain features, and high-order statistical features, which are described in detail as follows:

[0087] Time domain feature extraction

[0088] Mean and standard deviation: describe the central trend and fluctuation of the signal.

[0089] Skewness and kurtosis: measure the symmetry and sharpness of the signal, which facilitates the identification of abnormal signal characteristics.

[0090] Frequency domain feature extraction

[0091] Fourier transform: Convert the signal to the frequency domain and extract features such as spectrum energy and main frequency. The spectrum features reveal the distribution of the signal in different frequency bands and help identify specific fault modes.

[0092] High-order statistical features

[0093] Kurtosis and skewness: These higher-order statistics are used to identify nonlinear features in a signal, especially those that are characteristic of sudden faults.

[0094] The sensors include partial discharge sensors, current and voltage sensors, temperature sensors and vibration sensors. The installation locations of the sensors include key nodes and operating sections of the distribution network cables.

[0095] In one embodiment, partial discharge (PD) is an early signal of insulation aging, damage or defects in the cable insulation system. The partial discharge sensor mainly captures the high-frequency discharge signal in the cable insulation layer through ultra-high frequency (UHF) sensor and high-frequency current transformer (HFCT) technology.

[0096] UHF sensor: UHF sensor monitors by detecting high-frequency electromagnetic waves generated by partial discharge in cable insulation. When cracks or holes appear in the insulation material of the cable, partial discharge will occur, and the high-frequency electromagnetic waves generated can be received by the UHF sensor and converted into digital signals.

[0097] HFCT sensor: HFCT sensor is mainly used to monitor current changes. Partial discharge will cause current fluctuations in the cable. These fluctuation signals can be effectively captured by the current transformer installed outside the cable.

[0098] Furthermore, the current and voltage sensors are important sensors used to monitor the operation status of cables, and can collect the current and voltage data in the cables in real time. The current and voltage signals are important bases for judging the load status of cables and identifying cable faults.

[0099] Current sensor: By installing a current transformer (CT), the current flowing through the cable is measured in real time. The current sensor can reflect the change of cable load and whether it is overloaded. When the current exceeds the designed carrying range of the cable, the current sensor can promptly feedback the current abnormality, indicating possible overload or short circuit fault.

[0100] Voltage sensor: Voltage sensors are installed at both ends of the cable to monitor the voltage value of the cable in real time. When the cable voltage is abnormal (such as overvoltage or undervoltage), the sensor will generate an alarm signal. Voltage fluctuations are usually related to cable faults, equipment problems, or voltage instability in the power system.

[0101] Temperature sensors are used to detect the operating temperature of cables. Temperature changes are usually early signs of cable insulation damage, overload or electrical failure. Overheating inside the cable can cause insulation material aging, melting, and even fire accidents. Therefore, by monitoring the cable temperature in real time, abnormal conditions of the cable can be discovered in time.

[0102] Infrared temperature sensor: Non-contact temperature monitoring can be achieved through infrared temperature sensor. Infrared sensor can determine the surface temperature of cable by detecting thermal radiation on the cable surface.

[0103] Thermocouple temperature sensor: For situations where accurate temperature monitoring is required, thermocouples can be installed on the cable surface or at key locations to measure temperature changes through thermoelectric potential.

[0104] Vibration sensors are used to monitor the mechanical vibration of cables. Mechanical vibration may be caused by external physical shock, equipment failure or damage to the cable itself. When the cable is subjected to external mechanical shock (such as wind, vehicle passing, etc.) or internal mechanical damage (such as cracks in the cable insulation layer, wire core breakage, etc.), the vibration sensor can capture these tiny vibration signals in time.

[0105] Accelerometer: An accelerometer is used to detect tiny vibrations of the cable, especially those related to changes in the external environment or the internal structure of the cable, to indicate whether the cable has structural damage.

[0106] Piezoelectric sensor: Piezoelectric sensor is used to detect vibration caused by changes inside the cable, converting mechanical vibration into electrical signals through the piezoelectric effect.

[0107] In step c, the data preprocessing step includes wavelet denoising, missing value interpolation and standardization processing on the collected signal. The missing value interpolation is achieved by the K-NearestNeighbor method, and the standardization processing adopts the Z-Score standardization method.

[0108] In one embodiment, wavelet denoising is a signal processing technique, which is particularly suitable for processing noisy signals. It decomposes the signal at multiple scales, converts the signal from the time domain to the wavelet domain, and removes noise by thresholding the wavelet coefficients.

[0109] The core idea of ​​wavelet transform is to decompose the signal into frequency bands of different scales and process these frequency bands. Specifically, through wavelet transform, we can decompose the signal into low-frequency part (i.e. the smooth component of the signal) and high-frequency part (i.e. the detail part of the signal). In signal processing, noise is usually mainly concentrated in the high-frequency part, so the noise can be removed by thresholding the coefficients of the high-frequency part in the wavelet domain.

[0110] Wavelet denoising steps:

[0111] Signal decomposition: Perform wavelet transform on the collected signal x(t) to obtain low-frequency components and high-frequency components of different scales.

[0112]

[0113] Among them, c j is the wavelet transform coefficient, ψ j (t) is the wavelet basis function, and N is the number of decomposition levels.

[0114] Denoising thresholding: Set a threshold λ to process the wavelet coefficients. When the absolute value of the coefficient is less than the threshold, it is set to zero; when the absolute value of the coefficient is greater than the threshold, the coefficient is kept unchanged or soft thresholding is performed.

[0115]

[0116] in, is the denoised wavelet coefficient, and λ is the selected threshold.

[0117] Reconstruct the signal: The denoised coefficients are reconstructed into the denoised signal through inverse wavelet transform:

[0118]

[0119] Missing value interpolation (K-NearestNeighbor method);

[0120] In the process of data collection, missing data may occur due to various reasons, and these missing values ​​may affect the accuracy of subsequent analysis results. The K-NearestNeighbor (KNN) interpolation method is a similarity-based interpolation method that fills missing values ​​by using the mean or weighted mean of the k nearest neighboring samples to the missing value.

[0121] Select neighboring samples: For the sample with missing values, calculate the similarity between it and other samples (such as Euclidean distance):

[0122]

[0123] Among them, x i and x j are the feature vectors of sample i and sample j respectively, and n is the number of feature dimensions.

[0124] Select k nearest neighbors: Select the k nearest neighboring samples to the missing value based on the distance value.

[0125] Interpolation of missing values: Estimate missing values ​​by the mean or weighted mean of neighboring samples. For example, take the mean of k neighboring samples for interpolation:

[0126]

[0127] Among them, x missing is the interpolated value, x i is the characteristic value of the neighboring samples.

[0128] Standardization (Z-Score Standardization)

[0129] Working principle:

[0130] When performing machine learning and data analysis, the numerical ranges of different features may vary greatly, which may affect the performance of the algorithm and the accuracy of the results. Standardization is to transform the distribution of data so that the mean of each feature is 0 and the variance is 1, thereby eliminating the dimensional differences between the data.

[0131] Z-Score standardization is a common standardization method, and its formula is:

[0132]

[0133] Where: z i is the standardized data value; x i is the original data; μ is the mean of the sample data; σ is the standard deviation of the sample data.

[0134] Calculate the mean and standard deviation: Calculate the mean μ and standard deviation σ for all feature data x:

[0135]

[0136]

[0137] Where, is the sample size.

[0138] Transform data: Use the Z-Score standardization formula to standardize each feature data.

[0139] In step d, the data fusion step includes Kalman filtering and mutual information fusion methods. Kalman filtering is used for data fusion of dynamic systems, and mutual information fusion is used to ensure that important information between multi-sensor data is retained.

[0140] Kalman filtering is a recursive optimal estimation method widely used in dynamic systems. Its core idea is to update the system state based on the current observation value and the previous state estimate combined with the dynamic model to obtain the optimal state estimate at the current moment. In the distribution network cable fault detection, Kalman filtering can combine dynamic data from different sensors (such as current, voltage, temperature, vibration, etc.) to estimate the health status of the cable in real time, filter out noise and improve the accuracy of the data.

[0141] Kalman filtering achieves data fusion through two main steps: prediction step and update step. In the prediction step, the Kalman filter uses the system state estimate at the previous moment and combines it with the control input (such as grid load) to predict the current state. In the update step, the system corrects the predicted state through the current sensor observation value, and finally obtains the optimal state estimate.

[0142] Mutual Information (MI) is a metric that measures the correlation and shared information between two random variables. In multi-sensor data fusion, mutual information is used to quantify the correlation between different sensor data and ensure that the fused data retains the unique information provided by each sensor to the greatest extent. Unlike traditional weighted averaging or direct data concatenation methods, mutual information fusion focuses on effective fusion through the amount of shared information, thereby minimizing redundant information between multi-sensor data.

[0143] The core idea of ​​mutual information fusion is to determine which sensors provide more unique information by calculating the mutual information between different sensor data, and weight or selectively fuse the sensor data accordingly. By maximizing the mutual information between sensor data, the quality of the fusion result can be improved and important fault features can be ensured not to be missed.

[0144] By combining Kalman filtering and mutual information fusion, the data fusion step can achieve efficient integration of multiple sensor data. Kalman filtering estimates data in real time in a dynamic system, which can effectively filter noise and update the health status of the cable; while mutual information fusion maximizes information sharing between different sensors to ensure that the data of each sensor is optimally fused, reduce redundant information and retain key fault characteristics. This combination can make the cable fault detection and management system more accurate, real-time and reliable.

[0145] In step e, the feature extraction step includes time domain feature extraction, frequency domain feature extraction and high-order statistical feature extraction. The time domain features include mean, standard deviation, skewness and kurtosis, and the frequency domain features include Fourier transform results.

[0146] In one embodiment, time domain feature extraction extracts statistics reflecting signal fluctuation characteristics and distribution patterns by analyzing the change patterns of signals in the time domain. Time domain features can intuitively reflect the basic characteristics of signals, especially when the signal fluctuates greatly or changes dramatically, time domain features can effectively provide early information of faults.

[0147] In the process of monitoring distribution network cables, common time domain features include:

[0148] Mean: The mean of a signal describes the center position of the signal and reflects the overall deviation of the signal. For cable health monitoring, the mean helps understand the baseline level of the signal.

[0149] Standard Deviation: Standard Deviation reflects the amplitude of signal fluctuation, that is, the dispersion of signal value relative to the mean. A signal with a large standard deviation may mean that there is strong external interference or cable failure.

[0150] Skewness: Skewness measures the asymmetry of the signal distribution. If the skewness of the signal distribution is large, it means that there may be anomalies in the peak or tail of the signal.

[0151] Kurtosis: Kurtosis measures the sharpness of the signal distribution, reflecting whether the signal has prominent peaks or heavy tails. Higher kurtosis usually indicates the presence of sudden events or abnormal fluctuations in the signal.

[0152] Frequency domain feature extraction converts time domain signals into frequency domain, analyzes the frequency components in the signal, and reveals the frequency distribution characteristics of the signal. Frequency domain analysis can help identify periodic fluctuations or noise characteristics that are not easily detected in the time domain. In cable fault monitoring, frequency domain features can reveal periodic fault signals inside the cable or at the connection, such as the frequency response generated by cable insulation faults, partial discharges, etc.

[0153] High-order statistical feature extraction provides a deeper statistical characteristic analysis of the signal by calculating the high-order moments of the signal (such as skewness, kurtosis, etc.). High-order statistical features can capture the signal structure characteristics that cannot be revealed by time domain and frequency domain analysis, and are particularly suitable for analyzing complex phenomena such as nonlinearity and non-Gaussian noise in the signal.

[0154] Machine learning models include support vector machines, random forests, and XGBoost. Deep learning models include long short-term memory networks and convolutional neural networks. The models are combined for ensemble learning, and weighted voting or weighted averaging methods are used to fuse the results.

[0155] In one embodiment, by combining traditional machine learning models such as support vector machines, random forests, and XGBoost with deep learning models such as long short-term memory networks and convolutional neural networks, and using ensemble learning methods, the advantages of each model can be fully utilized, and multiple features can be comprehensively considered to greatly improve the accuracy and stability of cable fault detection. Ensemble learning not only improves the accuracy of diagnosis by fusing the results of multiple models, but also enhances the robustness of the system in complex environments, ensuring the safety and reliability of the power system.

[0156] In step f, the fault diagnosis and prediction step includes inputting real-time monitoring data into the trained machine learning and deep learning models to output whether the cable is faulty, the fault type, location and severity.

[0157] In one embodiment, by inputting real-time monitoring data into trained machine learning and deep learning models, the system can comprehensively diagnose cable faults, accurately identify fault types, locations, and severity, and provide accurate fault predictions. Whether it is the diagnosis of current faults or the prediction of future faults, the integrated test platform can provide strong technical support for the operation and maintenance personnel of the distribution network, improve the fault response speed, operation reliability, and maintenance efficiency of the power system, and ultimately realize the intelligent and automated management of the power system.

[0158] In step g, the fault warning and location step, the system predicts the occurrence time and location of the cable fault based on historical data and real-time monitoring data, combined with the LSTM time series model, and displays the warning information and fault location on the visual interface.

[0159] In one embodiment, by combining historical data with real-time monitoring data, the system uses the LSTM time series model to predict faults, accurately determine the time and location of cable faults, and provide early warning information through a visual interface. This method can greatly improve the intelligent management level of the power system, identify potential fault risks in advance, optimize the allocation and use of operation and maintenance resources, and ensure the efficient, stable and safe operation of the distribution network.

[0160] In step h, the report generation step includes automatically generating a cable fault diagnosis report and a prediction report, the report including the fault type, occurrence time, occurrence location, predicted fault trend and recommended operation and maintenance measures.

[0161] In one embodiment, the report generation step generates cable fault diagnosis reports and prediction reports through an automated process. The system can combine real-time monitoring data and historical data to intelligently generate detailed fault analysis and operation and maintenance suggestions, greatly improving fault response speed and operation and maintenance efficiency. Through these reports, operation and maintenance personnel can obtain key fault information, warn of fault trends, and take corresponding measures in the shortest time to ensure the safe operation of the power system and reduce downtime and economic losses caused by faults. At the same time, the automated generation of reports and intelligent suggestions help improve the overall level of distribution network operation and maintenance management, making it more efficient, accurate, and intelligent.

[0162] The method is suitable for real-time monitoring, fault management, fault prediction and intelligent operation and maintenance of distribution network cables, and can be extended to fault detection and management of other power equipment.

[0163] The operation and maintenance measures in the report are based on the fault diagnosis results and fault prediction information, providing targeted emergency response and preventive maintenance suggestions to the operation and maintenance personnel. The suggestions include handling measures after the fault occurs and preventive measures after the fault is predicted.

[0164] Operation and maintenance measures for fault types: Based on the fault type diagnosed in the report, the system will provide specific treatment measures for the operation and maintenance personnel. For example, if the report shows an insulation fault, the system may recommend shutting down and checking the insulation layer; if it is an overload fault, it may recommend load distribution or adding a cooling system.

[0165] Fault severity and response strategy: The report will give different emergency response strategies for different fault severities. For high-severity faults, the report may recommend immediate shutdown and repair; for minor faults, the report may recommend further monitoring and regular inspections.

[0166] Preventive maintenance measures for predicted failures: For prediction reports, the system will provide a series of preventive measures to help operation and maintenance personnel avoid potential risks in advance. If it is predicted that a certain section of cable may be overloaded in the future, the system will recommend adjusting the load, increasing the monitoring frequency or rearranging the cable.

[0167] Resource scheduling suggestions: The report also provides reasonable resource scheduling suggestions based on the location and nature of the fault, ensuring that maintenance personnel and equipment can arrive at the fault site in time and complete the repair task. The system will provide the best path and work arrangement for the maintenance team based on geographic information or cable topology maps.

[0168] Comparative Example 1:

[0169] The existing distribution network cable main insulation fault detection and management integrated test platform is usually implemented through the following key methods:

[0170] Fault Detection Technology Integration

[0171] The platform integrates a variety of fault detection technologies, such as insulation resistance testing, cable partial discharge detection, vibration and sound wave monitoring, infrared thermal imaging, etc. These technologies can monitor the health status of the cable in real time and detect insulation faults in a timely manner.

[0172] Data management and remote monitoring: The platform uploads data to the centralized management system through the data acquisition unit, supports real-time monitoring and remote alarm, and displays fault information through the human-machine interface (HMI) to facilitate operators to respond in a timely manner.

[0173] Integration testing and optimization: The platform conducts comprehensive integration testing to verify the collaborative working effect of each detection module, continuously optimizes the fault diagnosis algorithm and hardware configuration through real-time data feedback, and improves the stability and accuracy of the system.

[0174] Test comparison content:

[0175] Fault detection accuracy:

[0176] Experimental method: Different types of faults (such as partial discharge, overload, mechanical damage, temperature anomaly, etc.) are simulated on multiple distribution network cable sections, and the detection results are recorded.

[0177] Evaluation criteria: True Positive Rate (TPR), False Positive Rate (FPR), Accuracy.

[0178] Expected Results: Example 1 should demonstrate higher accuracy and lower false positive rate because it uses the integration of multi-sensor data fusion technology, machine learning and deep learning.

[0179] Response time:

[0180] Experimental method: Measure the time from the occurrence of a fault to the system issuing an early warning, including signal acquisition, data transmission, data processing, fault diagnosis and prediction, and report generation.

[0181] Evaluation criteria: Average response time (seconds).

[0182] Expected results: Since Example 1 introduces efficient data transmission and processing mechanisms (such as NB-IoT, LoRa communication and real-time machine learning models), its response time should be lower than that of existing platforms.

[0183] System stability:

[0184] Experimental method: Run the system under different environmental conditions (such as temperature fluctuations, current changes, etc.) and record the system stability and failure rate.

[0185] Evaluation criteria: System stability score (based on failure frequency and system recovery time).

[0186] Expected results: Since the platform uses distributed big data storage and multi-source data fusion, the system stability of embodiment 1 should be higher than that of the existing platform.

[0187] Data processing efficiency:

[0188] Experimental method: Measure the time required from data acquisition to fusion, feature extraction, and fault diagnosis.

[0189] Evaluation criteria: Data processing time (seconds).

[0190] Expected results: The optimized data processing flow of Example 1 (such as wavelet denoising, KNN interpolation, etc.) should improve processing efficiency.

[0191] index Embodiment 1 Comparative Example 1 Fault detection accuracy 98％ 92％ True Positive Rate 97％ 90％ False Positive Rate 1％ 6％ Average response time (seconds) 3 6 Data processing time (seconds) 1.2 2.5

[0192] Table 1

[0193] Compared with comparative example 1, embodiment 1 shows better performance, especially in terms of fault detection accuracy, response time and system stability. This advantage is due to its comprehensive use of multiple advanced technologies, including sensor data fusion, machine learning, deep learning and real-time data processing, which ensures the efficiency and accuracy of cable fault detection and prediction.

[0194] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing an integrated test platform for main insulation fault detection and management of distribution network cables, characterized in that: The following steps are involved: a) Data collection: by installing a variety of sensors including partial discharge sensors, current and voltage sensors, temperature sensors and vibration sensors, the real-time operation data of the distribution network cables are collected; b) Data transmission: The collected cable monitoring data is transmitted to the data processing system via wired or wireless communication; c) Data preprocessing: denoising, missing value interpolation, standardization or normalization of the transmitted raw data to improve the quality of the data; d) Data fusion: Use a variety of data fusion technologies to fuse data from different sensors and generate a unified feature vector to ensure the effective integration of multi-source data; e) Feature extraction: Extract time domain, frequency domain and high-order statistical features from the fused data for cable fault diagnosis and prediction; f) Fault diagnosis and prediction: Based on machine learning and deep learning models, cable faults can be diagnosed in real time and the time and location of potential faults can be predicted; g) Fault warning and location: By analyzing real-time data, it automatically generates warning information of cable faults and locates the specific location of the fault; h) Visualization and report generation: Display the operating status of the cable, fault detection results and prediction information through a visual interface, and generate fault diagnosis and prediction reports.

2. The method for constructing an integrated test platform for power distribution network cable main insulation fault detection and management according to claim 1, characterized in that: The sensors include partial discharge sensors, current and voltage sensors, temperature sensors and vibration sensors, and the installation locations of the sensors include key nodes and operating sections of the distribution network cables.

3. The method for constructing an integrated test platform for power distribution network cable main insulation fault detection and management according to claim 1, characterized in that: In the step c, the data preprocessing step includes performing wavelet denoising, missing value interpolation and standardization on the collected signal, wherein the missing value interpolation is implemented by the K-NearestNeighbor method, and the standardization adopts the Z-Score standardization method.

4. The method for constructing an integrated test platform for power distribution network cable main insulation fault detection and management according to claim 1, characterized in that: In the step d, the data fusion step includes Kalman filtering and mutual information fusion methods. Kalman filtering is used for data fusion of dynamic systems, and mutual information fusion is used to ensure that important information between multi-sensor data is retained.

5. The method for constructing an integrated test platform for power distribution network cable main insulation fault detection and management according to claim 1, characterized in that: In the step e, the feature extraction step includes time domain feature extraction, frequency domain feature extraction and high-order statistical feature extraction, the time domain features include mean, standard deviation, skewness and kurtosis, and the frequency domain features include Fourier transform results.

6. The method for constructing an integrated test platform for power distribution network cable main insulation fault detection and management according to claim 1, characterized in that: The machine learning models include support vector machines, random forests and XGBoost, and the deep learning models include long short-term memory networks and convolutional neural networks. The models are combined for integrated learning, and weighted voting or weighted averaging is used to fuse the results.

7. The method for constructing an integrated test platform for distribution network cable main insulation fault detection and management according to claim 1, characterized in that: In step f, the fault diagnosis and prediction step includes inputting real-time monitoring data into trained machine learning and deep learning models, and outputting whether the cable has a fault, the fault type, location and severity.

8. The method for constructing an integrated test platform for distribution network cable main insulation fault detection and management according to claim 1, characterized in that: In step g, in the fault warning and location step, the system predicts the occurrence time and location of the cable fault based on historical data and real-time monitoring data in combination with the LSTM time series model, and displays the warning information and fault location on a visual interface.

9. The method for constructing an integrated test platform for main insulation fault detection and management of distribution network cables according to claim 1, characterized in that: In the step h, the report generation step includes automatically generating a cable fault diagnosis report and a prediction report, wherein the report includes the fault type, occurrence time, occurrence location, predicted fault trend and recommended operation and maintenance measures.

10. The method for constructing an integrated test platform for power distribution network cable main insulation fault detection and management according to claim 1, characterized in that: The method is applicable to real-time monitoring, fault management, fault prediction and intelligent operation and maintenance of distribution network cables, and can be extended to fault detection and management of other power equipment.

Images