Cable defect early warning device

Abstract

The embodiment of the invention discloses a cable defect early warning device. The early warning device comprises a monitoring host, an air suction pipe network and a plurality of sensor groups, each sensor group comprises a pyroelectric particle detector, a plurality of gas sensors, a temperature and humidity sensor and a partial discharge sensor, and the monitoring host is used for acquiring data measured by each sensor group, preprocessing the data and then sending the data to the central analysis server; the central analysis server performs principal component analysis on the preprocessed data by using a cloud platform to obtain principal component vectors, and determines the defect type and the defect severity of the cable according to the principal component vector of each cable micro-region and a trained defect early warning model; and the man-machine interaction display screen is used for displaying defect types and defect severity. The early warning capability of the electrical defects of the micro-region cable is improved, the reliability and interpretability of an early warning result are enhanced, and a powerful guarantee is provided for safe and stable operation of a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of power cables, and in particular to a cable defect early warning device. Background Art

[0002] Cable electrical defect early warning technology based on multivariate parameter monitoring has attracted widespread attention due to its comprehensiveness and accuracy. When a cable has an electrical defect, multiple parameters such as pyrolytic particles, gas composition, ambient temperature and humidity will show abnormal changes. These abnormal changes can identify early signs of cable electrical defects.

[0003] At present, this technology mainly includes the following methods: 1. Traditional single parameter monitoring method: simple operation, but it is difficult to fully reflect the electrical status of the cable, and it is easy to cause misjudgment or missed judgment. Especially in complex environments, changes in a single parameter may be caused by multiple factors, and it is difficult to accurately determine whether it is caused by electrical defects. 2. Multi-parameter fusion analysis method: improves the accuracy of early warning, but faces problems such as parameter selection and weight distribution in practical applications. 3. Artificial intelligence-assisted diagnosis method: has the advantages of strong adaptability and high diagnostic accuracy, but there are problems such as large demand for model training data and poor algorithm interpretability. At the same time, when faced with new or rare defects, its early warning ability may be limited. 4. Combination of edge computing and cloud analysis: balances real-time and analysis depth, but still faces challenges in task division and data transmission security between the edge and cloud.

[0004] Although cable electrical defect early warning technology based on multivariate parameter monitoring plays an important role in the safe operation of power systems, the existing technology still has many problems and it is difficult to meet the requirements of early warning of cable electrical defects in increasingly complex power systems. Summary of the invention

[0005] The present invention provides a cable defect early warning device, which improves the early warning capability of micro-region cable electrical defects, enhances the reliability and interpretability of early warning results, and provides a strong guarantee for the safe and stable operation of the power system.

[0006] According to one aspect of the present invention, there is provided a cable defect early warning device, comprising:

[0007] Early warning device, central analysis server and human-computer interactive display screen;

[0008] The early warning device and the human-computer interaction display screen are both connected to the central analysis server in communication;

[0009] The early warning device includes a monitoring host, an air intake pipe network and a plurality of sensor groups; wherein the air intake pipe is arranged adjacent to the cable, and each cable micro-area is provided with a sensor group, each sensor group includes a pyroelectric particle detector, a gas sensor array, a temperature sensor, a humidity sensor and a partial discharge sensor, and the gas sensor array includes a plurality of different gas sensors;

[0010] Each gas sensor is used to detect the concentration of a gas in the suction pipe network; the pyroelectric particle detector is used to detect the number of particles released by the cable insulation material; the temperature sensor is used to detect the temperature near the cable; the humidity sensor is used to detect the humidity near the cable; the partial discharge sensor is used to detect the discharge signal of the cable;

[0011] The monitoring host is used to obtain the gas concentration data, particle number data, temperature data, humidity data and discharge signal data measured by each sensor group, and pre-process the gas concentration data, particle number data, temperature data, humidity data and discharge signal data and send them to the central analysis server;

[0012] The central analysis server includes a cloud platform, which is used to perform principal component analysis on the preprocessed gas concentration data, the preprocessed particle number data, the preprocessed temperature data, the preprocessed humidity data and the preprocessed discharge signal data to obtain the gas concentration principal component vector, the particle number principal component vector, the temperature principal component vector, the humidity principal component vector and the discharge signal principal component vector, and determine the defect type and defect severity of the cable according to the gas concentration principal component vector, the particle number principal component vector, the temperature principal component vector, the humidity principal component vector, the discharge signal principal component vector and the trained defect warning model of each cable micro-region;

[0013] The human-machine interactive display screen is used to display the defect type and defect severity.

[0014] Optionally, the monitoring host includes an edge computing unit, which is used to pre-process and preliminarily judge the gas concentration data, particle number data, temperature data, humidity data and discharge signal data;

[0015] Preprocessing includes data cleaning and feature extraction;

[0016] Among them, data cleaning includes removing noise values ​​and outliers; feature extraction includes calculating the statistical characteristic values ​​of each data, and the statistical characteristic values ​​include mean, variance and peak value; preliminary judgment includes comparing the statistical characteristic values ​​of each data with the corresponding threshold value to determine whether the data is abnormal;

[0017] The monitoring host is also used to issue an early warning when the preliminary judgment result is abnormal.

[0018] Optionally, the cloud platform is used to perform data standardization on each of the preprocessed gas concentration data, preprocessed particle number data, preprocessed temperature data, preprocessed humidity data and preprocessed discharge signal data, and analyze the correlation between the various data; determine the data that needs to be subjected to principal component analysis based on the correlation.

[0019] Optionally, the cloud platform is used to calculate the correlation coefficient matrix between various data, and determine the data that needs to be subjected to principal component analysis according to the correlation coefficients between various data.

[0020] Optionally, principal component analysis includes:

[0021] For each data that needs to be subjected to principal component analysis, including the standardized gas concentration data, the standardized particle number data, the standardized temperature data, the standardized humidity data, and the standardized discharge signal data: calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, select the first m principal components according to the size of the eigenvalues, and determine the principal component vector after dimensionality reduction according to the principal components; wherein m is an integer greater than or equal to 2.

[0022] Optionally, the defect warning model includes a neural network model;

[0023] The neural network model structure includes input layer, hidden layer and output layer;

[0024] The input layer is used to receive the principal component vector;

[0025] The hidden layer includes multiple hidden sublayers, each hidden sublayer contains multiple neurons, and the hidden sublayer is used to perform nonlinear transformation through activation functions;

[0026] The output layer is used to output defect type and defect severity.

[0027] Optionally, the central analysis server is used to construct a defect warning model using the model knowledge of an existing cable neural network model; and update the defect warning model according to the actual defect type and severity of the cable and the data measured by the sensor group.

[0028] Optionally, the human-machine interactive display screen is also used to display an overview of the defect status of the entire cable, the detection data and abnormal data of each micro-area, and recommended measures;

[0029] The overview includes the defect type and defect severity of each micro-area of ​​the entire cable; the detection data includes gas concentration data, particle number data, temperature data, humidity data and discharge signal data.

[0030] Optionally, the early warning device and the human-computer interaction display screen are connected to the central analysis server via a 4G network or a fiber optic network.

[0031] Optionally, the gas sensor array includes SF6 gas sensors, H2 gas sensors, CH4 gas sensors and C2H2 gas sensors.

[0032] The cable defect early warning device provided by the embodiment of the present invention comprises: an early warning device, a central analysis server and a human-computer interaction display screen; the early warning device and the human-computer interaction display screen are both communicatively connected to the central analysis server; the early warning device comprises a monitoring host, an air intake pipe network and a plurality of sensor groups; wherein the air intake pipe is arranged adjacent to the cable, and each cable micro-area is provided with a sensor group, each sensor group comprises a pyrolytic particle detector, a gas sensor array, a temperature sensor, a humidity sensor and a partial discharge sensor, and the gas sensor array comprises a plurality of different gas sensors; each gas sensor is used to detect the concentration of a gas in the air intake pipe network; the pyrolytic particle detector is used to detect the number of particles released by the cable insulation material; the temperature sensor is used to detect the temperature near the cable; the humidity sensor is used to detect the humidity near the cable; the partial discharge sensor is used to detect the discharge signal of the cable; the monitoring host is used to obtain the gas concentration measured by each sensor group The data, particle number data, temperature data, humidity data and discharge signal data are pre-processed and sent to the central analysis server respectively; the central analysis server includes a cloud platform, which is used to perform principal component analysis on the pre-processed gas concentration data, pre-processed particle number data, pre-processed temperature data, pre-processed humidity data and pre-processed discharge signal data to obtain the gas concentration principal component vector, particle number principal component vector, temperature principal component vector, humidity principal component vector and discharge signal principal component vector, and determine the defect type and defect severity of the cable according to the gas concentration principal component vector, particle number principal component vector, temperature principal component vector, humidity principal component vector, discharge signal principal component vector and the trained defect warning model of each cable micro-region; the human-computer interactive display screen is used to display the defect type and defect severity. By real-time monitoring of multiple parameters such as pyrolytic particles, multiple gas components, ambient temperature and humidity, and partial discharge, a comprehensive perception of cable electrical defects can be achieved. Combining advanced technologies such as cloud analysis and neural network models, a multi-level data analysis system is constructed, which effectively solves the problems of insufficient comprehensiveness of parameter selection, low real-time and accuracy of data analysis, insufficient adaptive ability of early warning models, insufficient coordination between edge computing and cloud analysis, noise interference and complexity of multi-parameter fusion analysis. It improves the early warning capability of micro-regional cable electrical defects, enhances the reliability and interpretability of early warning results, and provides a strong guarantee for the safe and stable operation of the power system.

[0033] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 It is a module schematic diagram of a cable defect early warning device provided by an embodiment of the present invention;

[0036] Figure 2 It is a schematic diagram of a multivariate parameter monitoring system of a cable defect early warning device provided by an embodiment of the present invention;

[0037] Figure 3 It is a module schematic diagram of another cable defect warning device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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 should fall within the scope of protection of the present invention.

[0039] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0040] The embodiment of the present invention provides a cable defect early warning device, Figure 1is a module schematic diagram of a cable defect early warning device provided by an embodiment of the present invention, Figure 2 Schematic diagram of a multivariate parameter monitoring system of a cable defect early warning device provided by an embodiment of the present invention. Figure 1-2 , cable defect early warning device, including:

[0041] Early warning device 10, central analysis server 20 and human-computer interaction display screen 30;

[0042] The early warning device 10 and the human-machine interactive display screen 30 are both in communication connection with the central analysis server 20;

[0043] The early warning device 10 includes a monitoring host 11, an air intake pipe network 12 and a plurality of sensor groups 40; wherein the air intake pipe is arranged adjacent to the cable, each cable micro-area 13 is provided with a sensor group 40, each sensor group 40 includes a pyroelectric particle detector 41, a gas sensor array 42, a temperature sensor 43, a humidity sensor 44 and a partial discharge sensor 45, and the gas sensor array 42 includes a plurality of different gas sensors;

[0044] Each gas sensor is used to detect the concentration of a gas in the suction pipe network 12; the pyroelectric particle detector 41 is used to detect the number of particles released by the cable insulation material; the temperature sensor 43 is used to detect the temperature near the cable; the humidity sensor 44 is used to detect the humidity near the cable; the partial discharge sensor 45 is used to detect the discharge signal of the cable;

[0045] The monitoring host 11 is used to obtain the gas concentration data, particle number data, temperature data, humidity data and discharge signal data measured by each sensor group 40, and pre-process the gas concentration data, particle number data, temperature data, humidity data and discharge signal data and send them to the central analysis server 20;

[0046] The central analysis server 20 includes a cloud platform, which is used to perform principal component analysis on the preprocessed gas concentration data, the preprocessed particle number data, the preprocessed temperature data, the preprocessed humidity data and the preprocessed discharge signal data to obtain a gas concentration principal component vector, a particle number principal component vector, a temperature principal component vector, a humidity principal component vector and a discharge signal principal component vector, and determine the defect type and defect severity of the cable according to the gas concentration principal component vector, the particle number principal component vector, the temperature principal component vector, the humidity principal component vector, the discharge signal principal component vector and the trained defect warning model of each cable micro-area 13;

[0047] The human-machine interactive display screen 30 is used to display the defect type and defect severity.

[0048] Among them, the pyroelectric particle detector 41 detects tiny particles released by the cable insulation material at the beginning of electrical defects. The tiny particles are usually caused by partial discharge or overheating inside the cable, and they will be captured by the detector as the air flows. The detector usually contains a highly sensitive particle detection module that can identify and count these tiny particles to determine whether the cable has potential electrical defects. The gas sensor array 42 is used to monitor characteristic gases that may be produced inside the cable. Including sulfur hexafluoride (SF6), hydrogen (H2), methane (CH4), acetylene (C2H2), etc., which are released when the cable fails. Each gas sensor is optimized for a specific gas component and can detect the concentration of the target gas with high accuracy. For example, SF6 sensors usually use electrochemical or optical detection technology, which can accurately measure the presence of SF6 at extremely low concentrations. H2 sensors use semiconductor or catalytic combustion technology and can work stably in high temperature environments. Temperature sensor 43 and humidity sensor 44 are used to monitor temperature and humidity changes in the environment around the cable. The temperature sensor 43 and the humidity sensor 44 can measure the temperature and humidity of the environment in real time. For example, the capacitive humidity sensor reflects the change of the ambient humidity by measuring the change of capacitance, while the resistive temperature sensor reflects the change of temperature by measuring the change of resistance. The partial discharge sensor 45 is used to detect the partial discharge phenomenon that may occur in the cable. Partial discharge usually occurs inside the insulation layer of the cable and is an early sign of cable aging. The partial discharge sensor 45 usually adopts ultrasonic or electromagnetic wave detection technology, which can detect these weak discharge signals outside the cable. For example, the ultrasonic partial discharge sensor receives the ultrasonic signal generated by the discharge inside the cable and then converts it into an electrical signal for analysis. The electromagnetic wave partial discharge sensor analyzes by detecting the electromagnetic wave signal generated by the discharge inside the cable.

[0049] For example, suppose that in an electric power system, the length of the cable line is 1 km, and a monitoring point needs to be set every 100 meters, and each monitoring point is a cable micro-area. Each monitoring point is equipped with a pyroelectric particle detector 41, a gas sensor array 42, a temperature sensor 43, a humidity sensor 44 and a partial discharge sensor 45. First, the pyroelectric particle detector 41 collects air samples through the air sampling system, and then detects the number of particles in the sample through a particle counter. Assume that in a certain detection, the particle counter detects 100 particles per cubic centimeter of air, while the number of particles per cubic centimeter of air should be less than 10 under normal circumstances, so it can be judged that the cable near the monitoring point may have electrical defects. Secondly, the gas sensor array 42 analyzes the air sample. Assume that the reading of the SF6 sensor is 10ppm, while the concentration of SF6 should be less than 1ppm under normal circumstances; the reading of the H2 sensor is 50ppm, while the concentration of H2 should be less than 10ppm under normal circumstances. These abnormal readings indicate that there may be a fault inside the cable, resulting in the release of these characteristic gases. At the same time, the temperature sensor 43 and the humidity sensor 44 recorded an ambient temperature of 35 degrees Celsius and a humidity of 80%. These data indicate that the cable is in a harsh environment, and high temperature and high humidity may accelerate the aging process of the cable. Finally, the partial discharge sensor 45 detected a weak ultrasonic signal, indicating that there may be partial discharge inside the cable. The frequency and intensity of these signals are significantly different from the background noise under normal conditions, further confirming that the cable may have electrical defects. Through this multi-parameter collaborative monitoring method, the electrical status of the cable can be comprehensively and accurately evaluated, and potential fault hazards can be discovered in a timely manner.

[0050] Specifically, the pyroelectric particle detector 41 detects the number of particles released by the cable insulation material, and transmits the particle number data to the monitoring host 11 in real time. The gas sensor array 42 includes a variety of different gas sensors, each of which is responsible for detecting the concentration of the corresponding gas in the suction pipe network 12, and transmitting the concentration data to the monitoring host 11 in real time. The temperature sensor 43 detects the temperature near the cable, and transmits the temperature data to the monitoring host 11 in real time. The humidity sensor 44 detects the humidity near the cable, and transmits the humidity data to the monitoring host 11 in real time. The partial discharge sensor 45 detects the discharge signal of the cable, and transmits the discharge signal data to the monitoring host 11 in real time. The suction pipe network 12 is arranged adjacent to the cable, and guides the gas near the cable to the gas sensor array 42 for detection. The monitoring host 11 receives all the raw data (gas concentration data, particle number data, temperature data, humidity data and discharge signal data) from the sensor group 40, and performs preprocessing to ensure the quality and consistency of the data. The preprocessed data is sent from the monitoring host 11 to the central analysis server 20 via 4G or optical fiber communication. The cloud platform of the central analysis server 20 performs principal component analysis on the received preprocessed data, and extracts the principal component vector of gas concentration, the principal component vector of particle number, the principal component vector of temperature, the principal component vector of humidity, and the principal component vector of discharge signal. Principal component analysis helps to reduce the dimension of data while retaining the main features of the data. The cloud platform uses the trained neural network model to evaluate the defect type and defect severity of the cable according to the extracted principal component vector. The central analysis server 20 sends the defect type and defect severity information obtained by the analysis to the human-computer interaction display screen 30 for display. The human-computer interaction display screen 30 can intuitively display the defect status of the cable, making it convenient for the operation and maintenance personnel to take corresponding measures in time. The operation and maintenance personnel understand the defect status of the cable by viewing the information on the human-computer interaction display screen 30, and take corresponding maintenance or repair measures according to the defect type and severity to improve the accuracy and reliability of the early warning.

[0051] The cable defect early warning device provided by the embodiment of the present invention comprises: an early warning device 10, a central analysis server 20 and a human-computer interaction display screen 30; the early warning device 10 and the human-computer interaction display screen 30 are both connected to the central analysis server 20 for communication; the early warning device 10 comprises a monitoring host 11, an air intake pipe network 12 and a plurality of sensor groups 40; wherein the air intake pipe is arranged adjacent to the cable, and each cable micro-area 13 is provided with a sensor group 40, each sensor group 40 comprises a pyroelectric particle detector 41, a gas sensor array 42, a temperature sensor 43, a humidity sensor 44 and a partial discharge sensor 45, and the gas sensor array 42 comprises a plurality of different gas sensors; each gas sensor is used to detect the concentration of a gas in the air intake pipe network 12; the pyroelectric particle detector 41 is used to detect the number of particles released by the cable insulation material; the temperature sensor 43 is used to detect the temperature near the cable; the humidity sensor 44 is used to detect the humidity near the cable; the partial discharge sensor 45 is used to detect the discharge signal of the cable; the monitoring The host 11 is used to obtain the gas concentration data, particle number data, temperature data, humidity data and discharge signal data measured by each sensor group 40, and pre-process the gas concentration data, particle number data, temperature data, humidity data and discharge signal data and send them to the central analysis server 20; the central analysis server 20 includes a cloud platform, which is used to perform principal component analysis on the pre-processed gas concentration data, pre-processed particle number data, pre-processed temperature data, pre-processed humidity data and pre-processed discharge signal data to obtain the gas concentration principal component vector, particle number principal component vector, temperature principal component vector, humidity principal component vector and discharge signal principal component vector, and determine the defect type and defect severity of the cable according to the gas concentration principal component vector, particle number principal component vector, temperature principal component vector, humidity principal component vector, discharge signal principal component vector and the trained defect warning model of each cable micro-area 13; the human-computer interaction display screen 30 is used to display the defect type and defect severity. By real-time monitoring of multiple parameters such as pyrolytic particles, multiple gas components, ambient temperature and humidity, and partial discharge, a comprehensive perception of cable electrical defects is achieved. Combining advanced technologies such as cloud analysis and neural network models, building a multi-level data analysis system can increase the comprehensiveness of parameter selection, improve the real-time and accuracy of data analysis, enhance the adaptive ability of early warning models, reduce noise interference, and simplify multi-parameter fusion analysis. It improves the early warning ability of micro-regional cable electrical defects, enhances the reliability and interpretability of early warning results, and provides a strong guarantee for the safe and stable operation of the power system.

[0052] Optionally, the monitoring host 11 includes an edge computing unit, which is used to pre-process and preliminarily judge the gas concentration data, particle number data, temperature data, humidity data and discharge signal data;

[0053] Preprocessing includes data cleaning and feature extraction;

[0054] Among them, data cleaning includes removing noise values ​​and outliers; feature extraction includes calculating the statistical characteristic values ​​of each data, and the statistical characteristic values ​​include mean, variance and peak value; preliminary judgment includes comparing the statistical characteristic values ​​of each data with the corresponding threshold value to determine whether the data is abnormal;

[0055] The monitoring host 11 is also used to issue an early warning when the preliminary judgment result is abnormal.

[0056] Among them, the purpose of data cleaning is to remove obvious noise and outliers to ensure the data quality of subsequent analysis. Noise is usually caused by errors in the sensor itself, environmental interference, or signal distortion during transmission. Outliers may be caused by sensor failure, external interference, or real abnormal events. Data cleaning methods include filtering, smoothing, and outlier detection. For example, a low-pass filter can be used to remove high-frequency noise, a sliding average filter can be used to smooth data, and statistical methods (such as Z-score) can be used to detect and remove outliers. Suppose that during a data acquisition process, the temperature data recorded by a sensor fluctuates violently in a short period of time. These fluctuations can be smoothed out by a sliding average filter to obtain more stable temperature data. The purpose of feature extraction is to extract useful statistical features from the raw data for subsequent analysis and judgment. Common statistical features include mean, variance, peak, peak-to-peak value, slope, etc. For example, the mean can reflect the overall level of the data, the variance can reflect the degree of fluctuation of the data, the peak and peak-to-peak value can reflect the maximum value and fluctuation range of the data, and the slope can reflect the trend of data changes. The purpose of preliminary judgment is to perform preliminary anomaly detection based on a preset threshold so as to quickly identify potential abnormal situations locally. The preset thresholds are set based on historical data to distinguish between normal and abnormal situations.

[0057] For example, taking pyrolysis particle data as an example, the edge computing unit uses a sliding window method to calculate the moving average and standard deviation of particle counts:

[0058]

[0059] where μ t is the moving average at time t, σ t is the standard deviation, ω is the window size, x i Count the particles.

[0060] Formula 1 means: at time t, calculate the average of the most recent ω data points. For example, assuming the window size is 10, then at time t, the moving average is the average of the most recent 10 particle counts. By calculating the moving average, the overall level of particle counts can be reflected, thereby identifying long-term trend changes.

[0061] Formula 2 means: at time t, calculate the average of the sum of the squares of the deviations of the most recent ω data points from the moving average, and then take the square root. By calculating the standard deviation, the degree of fluctuation of the particle count can be reflected, thereby identifying short-term abnormal fluctuations.

[0062] If x t >μ t +3σ t , it is considered that there may be an abnormality, triggering an alarm and transmitting the data to the cloud.

[0063] The sliding window method is a commonly used data processing technique that reflects the dynamic changes of data by performing statistical analysis on the data within a window of fixed size. First of all, the core idea of ​​the sliding window method is to perform statistical analysis on the data within a window of fixed size. The window size (ω) determines the time range of the statistical analysis and is usually set according to the specific application scenario and data characteristics. For example, if the window size is 10, then the most recent 10 data points will be considered in each calculation. The calculation process of the sliding window method includes two main steps: calculating the moving average and calculating the standard deviation (refer to Formula 1 and Formula 2).

[0064] If the particle count x at a certain moment t Exceeds the moving average μ t plus three times the standard deviation σ t , it is considered that there may be an anomaly, triggering an alarm and transmitting the data to the cloud. This judgment standard is based on the "three standard deviation principle" in statistics, that is, in the case of normal distribution, about 99.7% of the data points will fall within the range of the mean plus or minus three standard deviations. Therefore, if a data point is outside this range, it is likely to be an outlier.

[0065] Gas data, temperature and humidity data are similar to partial discharge data. For gas data, the concentration data of gases such as SF6, H2, CH4, C2H2, etc. are first collected from the gas sensor array 42. A low-pass filter is used to remove high-frequency noise, a sliding average filter is used to smooth the data, and a statistical method (such as Z-score) is used to detect and remove outliers. The variance of the concentration of each gas collected at different times in a day is calculated, and the average of all variances is calculated. Finally, preliminary anomaly detection is performed based on a preset threshold. For example, if the average value of the variance exceeds the preset threshold, it is considered that there may be an anomaly.

[0066] For temperature and humidity data, first collect temperature and humidity data from the temperature and humidity sensor. Use a low-pass filter to remove high-frequency noise, use a sliding average filter to smooth the data, and use statistical methods (such as Z-score) to detect and remove outliers. Calculate the mean of temperature and humidity, and finally perform preliminary anomaly detection based on the preset threshold. For example, if the temperature mean exceeds the preset threshold, it is considered that there may be an anomaly.

[0067] For partial discharge data, partial discharge usually occurs inside the cable insulation layer and is an early sign of cable aging. The ultrasonic signal generated by the discharge inside the cable is received and then converted into an electrical signal for analysis. In order to capture weak partial discharge signals, high-frequency sampling technology is usually used, and the sampling frequency can reach tens of kHz to hundreds of kHz. According to the frequency range of the partial discharge signal, only the signal within a specific frequency range is retained through a bandpass filter, and statistical methods (such as Z-score) are used to detect and remove outliers. The peak value of the partial discharge signal is determined, and finally preliminary anomaly detection is performed based on the preset threshold. For example, if the peak value of the partial discharge signal exceeds the preset threshold, it is considered that there may be an anomaly.

[0068] Specifically, the pyroelectric particle detector 41, the gas sensor array 42, the temperature sensor 43, the humidity sensor 44 and the partial discharge sensor 45 respectively collect the particle number data, gas concentration data, temperature data, humidity data and discharge signal data related to the cable. The collected data is transmitted to the edge computing unit of the monitoring host 11 via 4G or optical fiber communication. The edge computing unit identifies and removes the noise value in the data, and the edge computing unit identifies and removes the abnormal value in the data through the set threshold or algorithm. The edge computing unit calculates the statistical characteristic values ​​such as the mean, variance and peak value of the particle number data, gas concentration data, temperature data, humidity data and discharge signal data. The edge computing unit compares the calculated statistical characteristic value with the preset threshold value to determine whether the data is abnormal. If the statistical characteristic value of a certain data exceeds the corresponding threshold value, the edge computing unit will determine that the data is abnormal data. When the edge computing unit determines that a certain type of data or multiple data is abnormal, the monitoring host 11 will trigger the early warning mechanism. The monitoring host 11 generates early warning information and sends it to the central analysis server 20 or the human-computer interaction display screen 30 via 4G or optical fiber communication so that the operation and maintenance personnel can obtain and process it in time. Based on the early warning information, the operation and maintenance personnel conduct further inspection and maintenance of the cables to ensure the safe operation of the cables.

[0069] Optionally, the cloud platform is used to perform data standardization on each of the preprocessed gas concentration data, preprocessed particle number data, preprocessed temperature data, preprocessed humidity data and preprocessed discharge signal data, and analyze the correlation between the various data; determine the data that needs to be subjected to principal component analysis based on the correlation.

[0070] The data standardization process uses the following formula:

[0071] Where x is the original data, x′ is the standardized data, min(x) and max(x) are the minimum and maximum values ​​of the data, respectively. Standardization converts the data into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0072] Among them, μ is the mean of the data, and σ is the standard deviation of the data. For example, the minimum value of gas concentration data is 0ppm and the maximum value is 100ppm; the minimum value of temperature data is 0 degrees Celsius and the maximum value is 50 degrees Celsius; the minimum value of humidity data is 0% and the maximum value is 100%; the minimum value of particle count data is 0 and the maximum value is 1000. Through standardization, the data is unified to the [0,1] interval.

[0073] The purpose of correlation analysis is to calculate the correlation coefficient matrix between various parameters in order to understand the relationship between various parameters. The correlation coefficient matrix can help identify which parameters have strong correlations and which parameters have weak correlations, providing a basis for subsequent dimensionality reduction processing. The correlation analysis formula is as follows:

[0074]

[0075] Among them, r xy is the correlation coefficient between parameters x and y, x k and k are the kth data points of parameters x and y respectively, and are the means of the parameters x and y respectively.

[0076] Specifically, at the monitoring host 11, gas concentration data, particle number data, temperature data, humidity data and discharge signal data are pre-processed and uploaded to the cloud platform. The pre-processed data is uploaded to the cloud platform via 4G or fiber optic communication. The cloud platform determines the value range of each data to understand the distribution and variation of the data. According to the characteristics of the data and analysis requirements, select a suitable standardization method, such as Z-score standardization, Min-Max standardization, etc. Each data is standardized so that it has the same scale and distribution. The cloud platform calculates the correlation coefficient between each data, organizes the calculated correlation coefficient into a correlation matrix, and intuitively displays the correlation between each data. According to the correlation matrix, the strength of the correlation and the positive and negative relationship between the data are analyzed to provide a basis for the subsequent principal component analysis.

[0077] Optionally, the cloud platform is used to calculate the correlation coefficient matrix between various data, and determine the data that needs to be subjected to principal component analysis according to the correlation coefficients between various data.

[0078] For example, the correlation coefficient between gas concentration and temperature is 0.8, indicating that there is a strong positive correlation between the two; the correlation coefficient between gas concentration and humidity is -0.5, indicating that there is a weak negative correlation between the two. Then, the gas concentration and humidity are subsequently subjected to principal component analysis and dimensionality reduction processing.

[0079] Specifically, the correlation coefficients between the data are calculated using statistical methods. The calculated correlation coefficients are organized into a matrix form, where the rows and columns of the matrix represent different data variables, and the elements in the matrix represent the correlation coefficients between the corresponding variables. By analyzing the correlation coefficient matrix, it is possible to identify which data variables have strong correlations (it is generally believed that the absolute value of the correlation coefficient is greater than a certain threshold, such as 0.7 or 0.8, indicating a strong correlation).

[0080] Optionally, principal component analysis includes:

[0081] For each data that needs to be subjected to principal component analysis, including the standardized gas concentration data, the standardized particle number data, the standardized temperature data, the standardized humidity data, and the standardized discharge signal data: calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, select the first m principal components according to the size of the eigenvalues, and determine the principal component vector after dimensionality reduction according to the principal components; wherein m is an integer greater than or equal to 2.

[0082] For example, assume that the standardized data received by the cloud includes gas concentration, temperature, humidity and particle count. The covariance matrix is ​​calculated for each of these data, and the eigenvalue decomposition is performed to obtain the eigenvalue and the corresponding eigenvector. For example, assuming that n types of gas concentration, temperature, humidity, particle count and partial discharge characteristics are monitored, the covariance matrix can be calculated and the eigenvector can be determined. Taking gas concentration as an example, the constructed gas eigenvector is: G = [g 1 , g 2 , ..., g n ], assuming that the first two principal components are selected, the original data is projected onto these two principal components to obtain the data after dimensionality reduction. For example, the principal component vector of gas concentration data is: P = [p 1 , p 2 ]

[0083] Among them, p 1 and p 2 are the first and second principal components respectively.

[0084] After dimensionality reduction through principal component analysis, the principal component vector of each eigenvector is obtained. Let’s take the gas principal component vector as an example: P g =[p g1 , p g2 , ..., p gm ] (m<n)

[0085] Among them, p gi Represents the i-th gas principal component. Similarly, perform principal component analysis and dimensionality reduction on the temperature feature vector T, humidity feature vector H, particle count feature vector Pc and partial discharge feature vector PD to obtain their respective principal component vectors. Finally, all principal component vectors are input into the trained neural network model. It is assumed that all principal component vectors include the gas principal component vector Pg, the temperature principal component vector Pt, the humidity principal component vector Ph, the particle count principal component vector Pcp and the partial discharge principal component vector Ppd. Input these principal component vectors into the trained neural network model: Y = f(W·[P g , P t , P h , P cp , P pd ]+b)

[0086] Among them, Y is the output result, f is the activation function, W is the weight matrix, and b is the bias vector.

[0087] Specifically, the gas concentration data, particle number data, temperature data, humidity data and discharge signal data monitored by the cable defect early warning device are collected. Since these data may have different dimensions and ranges, in order to ensure the accuracy of principal component analysis, these data need to be standardized. The standardized data are constructed into a data matrix according to the variables (i.e., gas concentration, particle number, temperature, humidity and discharge signal) and samples (i.e., monitoring data at different time points). According to the data matrix, the covariance matrix of each variable is calculated respectively. The covariance matrix is ​​decomposed by eigenvalue decomposition to obtain the eigenvalue and the corresponding eigenvector. According to the size of the eigenvalue, the eigenvector corresponding to the first m largest eigenvalues ​​is selected as the principal component. The selection of m is usually determined according to actual needs and the characteristics of the data set, and can be an integer greater than or equal to 2. The eigenvector corresponding to the selected principal component is used as a column vector to construct the principal component matrix. The original data matrix is ​​projected onto the principal component matrix to obtain the principal component vector after dimensionality reduction. The principal component vector after dimensionality reduction is used in combination with the algorithm model of the cable defect early warning device to monitor and warn the operation status of the cable in real time. When the neural network model outputs defect types and severity levels that are high, an early warning signal is triggered, prompting the operation and maintenance personnel to conduct inspections and maintenance.

[0088] Optionally, the defect warning model includes a neural network model;

[0089] The neural network model structure includes input layer, hidden layer and output layer;

[0090] The input layer is used to receive the principal component vector;

[0091] The hidden layer includes multiple hidden sublayers, each hidden sublayer contains multiple neurons, and the hidden sublayer is used to perform nonlinear transformation through activation functions;

[0092] The output layer is used to output defect type and defect severity.

[0093] Among them, the neural network model is a commonly used machine learning model, which aims to identify defect types and evaluate severity through a multi-layer perceptron (MLP). The neural network model can learn the characteristics of data through training, thereby realizing the prediction and classification of unknown data.

[0094] For example, assume that the principal component vector after dimensionality reduction received by the cloud is P, and input it into the trained neural network model. The input layer of the neural network model receives P, the hidden layer performs nonlinear transformation through the activation function, and the output layer outputs the defect type and severity assessment results. For example, the output result Y of the neural network model is: Y = f(W·P+b)

[0095] Among them, Y is the output result, f is the activation function, W is the weight matrix, and b is the bias vector. Assuming that the output results of the neural network model are "partial discharge" and "high severity", it is believed that the cable may have partial discharge defects and the severity is high.

[0096] Optionally, the central analysis server 20 is used to construct a defect warning model using the model knowledge of an existing cable neural network model; and update the defect warning model according to the actual defect type and defect severity of the cable and the data measured by the sensor group.

[0097] Specifically, in order to improve the adaptability and accuracy of the model, a combination of online learning and transfer learning is adopted: Online learning: Continuously update model parameters based on actual operation data. Transfer learning: Use the model knowledge of existing cable types to quickly build an early warning model for new cables.

[0098] Specifically, online learning uses the gradient descent method to update the model parameters based on the new data set D. Assume that the parameters of the original model are θ 0 , the new data set is D, then the model parameter update formula is:

[0099] Among them, η is the learning rate, L is the loss function, The learning rate η determines the step size of the parameter update, and the loss function L measures the difference between the model prediction result and the actual result. Represents the partial derivative of the loss function with respect to the model parameters.

[0100] The purpose of online learning is to continuously update model parameters based on actual operation data and expert feedback to adapt to the ever-changing data environment and actual needs. The core idea of ​​online learning is to adjust model parameters in real time by continuously receiving new data so that the model can better fit the current data distribution. The purpose of transfer learning is to use the model knowledge of existing cable types to quickly build an early warning model for new cables. The core idea of ​​transfer learning is to reduce the training time and data requirements of the new model by transferring the knowledge of the existing model to the new model, thereby improving the accuracy and adaptability of the new model.

[0101] The steps of transfer learning are as follows: Model selection: select an existing model similar to the new cable type as the basic model. Knowledge transfer: transfer some or all parameters of the basic model to the early warning model of the new cable. Fine-tuning: use the data of the new cable to fine-tune the migrated model and update the model parameters so that the model can better fit the data distribution of the new cable.

[0102] For example, assume that the parameters of the existing model are θ 0, the data set of the new cable is D. Select an existing model similar to the new cable type as the basic model and change its parameter θ 0 Migrate to the early warning model of the new cable. Use the data of the new cable to fine-tune the migrated model and update the model parameters. For example, assuming the learning rate η = 0.01, the model parameter update formula is: In this way, an early warning model for new cables can be quickly constructed, reducing the training time and data requirements of the new model, and improving the accuracy and adaptability of the new model.

[0103] Through the combination of online learning and transfer learning, the model can continuously update parameters, adapt to the ever-changing data environment and actual needs, and improve the adaptability of the model. By continuously receiving new data, the model can adjust parameters in real time, improve the accuracy of the prediction results, avoid misjudgment and omission of single parameter monitoring, and through online learning, the model can receive new data in real time, adjust parameters in time, improve the real-time nature of the early warning system, and avoid further development of abnormal situations.

[0104] Optionally, the human-machine interactive display screen 30 is also used to display an overview of the defect status of the entire cable, the detection data and abnormal data of each micro-area, and recommended measures;

[0105] The overview includes the defect type and defect severity of each micro-area of ​​the entire cable; the detection data includes gas concentration data, particle number data, temperature data, humidity data and discharge signal data.

[0106] Among them, the overview map: shows the health status of the entire cable line. Micro-area details: displays multi-parameter data and abnormal indicators of a single micro-area. Defect diagnosis report: gives the suspected defect type, location, severity and recommended measures.

[0107] For example, for a micro-area suspected of insulation aging, the following diagnosis is given: "Potential insulation aging problems were detected in micro-area A (milepost 15+200). The pyroelectric particle count increased by 150% compared to the normal value, the SF6 concentration increased by 30%, and the partial discharge amplitude increased by 5dB. It is recommended to conduct a detailed inspection at the next planned maintenance and consider replacing this section of cable in advance."

[0108] Specifically, the central analysis server 20 uses the existing cable neural network model knowledge and combines the preprocessed data to build an initial defect warning model. The model can identify the defect type and severity of the cable and provide warning information based on the detection data. According to the actual defect type, defect severity and data measured by the sensor group of the cable, the central analysis server continuously optimizes and updates the defect warning model. The human-computer interaction display screen displays an overview of the defect status of the entire cable, which presents the defect type and defect severity of each micro-area of ​​the cable in an intuitive way. Operation and maintenance personnel can quickly understand the overall health status of the cable and locate potential problem areas through the overview map. The display screen displays the detection data of each micro-area in detail, including gas concentration, particle number, temperature, humidity and discharge signal. These data are presented in the form of charts, curves, etc., which is convenient for operation and maintenance personnel to analyze and compare data changes in different areas. When the detection data exceeds the preset threshold, the display screen will issue an abnormal data alarm to remind the operation and maintenance personnel. The alarm information includes the type, location, severity, etc. of the abnormal data, so that the operation and maintenance personnel can take timely measures. According to the output results of the defect warning model and the analysis of the detection data, the human-computer interaction display screen will automatically propose corresponding recommended measures. Recommended measures may include regular inspection, repair and replacement, and adjustment of operating parameters, etc., aimed at preventing or mitigating the impact of cable defects on grid operation.

[0109] Figure 3 is a schematic diagram of another module of a cable defect warning device provided by an embodiment of the present invention, referring to Figure 3 Optionally, the early warning device 10 and the human-computer interaction display screen 30 are connected to the central analysis server 20 via a 4G network or an optical fiber network.

[0110] Specifically, the 4G network and the fiber optic network have the characteristics of high bandwidth and low latency, which can ensure real-time data transmission between the early warning device 10 and the human-computer interaction display screen 30 and the central analysis server 20.

[0111] Optionally, the gas sensor array 42 includes a SF6 gas sensor, a H2 gas sensor, a CH4 gas sensor, and a C2H2 gas sensor.

[0112] Specifically, SF6 is an important insulating medium in electrical equipment. Its leakage will not only reduce the insulation performance of the equipment, but also may increase greenhouse gas emissions. Through the SF6 gas sensor, the concentration of SF6 gas in the cable can be monitored in real time, leakage can be detected in time, and potential safety risks can be prevented. If H2 leaks, it may accumulate and form an explosive mixture. Especially in cases such as cable faults or arc discharges, H2 may be produced in large quantities. Through the H2 gas sensor, the change of H2 concentration can be continuously monitored. Once a leak is detected, the alarm system is immediately triggered to effectively curb the risk of fire or explosion. CH4 is a colorless and odorless gas that is widely present in the natural environment and is also a by-product of many industrial processes. In the cable operating environment, CH4 may come from the decomposition of cable insulation materials or the penetration of the external environment. Through the CH4 gas sensor, the CH4 concentration can be monitored in real time to ensure the safety of the cable operating environment. C2H2 is a flammable and explosive gas that may be produced during cable faults or discharges. Through the C2H2 gas sensor, the accumulation of C2H2 can be detected in time to prevent the occurrence of fire or explosion accidents. The gas sensor array can monitor the concentration changes of multiple gases at the same time, improving the accuracy and reliability of the early warning system. By comprehensively analyzing the concentration data of different gases, the operating status and potential risks of the cable can be judged more accurately.

[0113] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0114] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A cable defect early warning device, characterized in that: include: Early warning device, central analysis server and human-computer interactive display screen; The early warning device and the human-computer interaction display screen are both communicatively connected to the central analysis server; The early warning device includes a monitoring host, an air intake pipe network and a plurality of sensor groups; wherein the air intake pipe is arranged adjacent to the cable, each cable micro-area is provided with a sensor group, each of the sensor groups includes a pyroelectric particle detector, a gas sensor array, a temperature sensor, a humidity sensor and a partial discharge sensor, and the gas sensor array includes a plurality of different gas sensors; Each of the gas sensors is used to detect the concentration of a gas in the suction pipe network; the pyroelectric particle detector is used to detect the number of particles released by the cable insulation material; the temperature sensor is used to detect the temperature near the cable; the humidity sensor is used to detect the humidity near the cable; the partial discharge sensor is used to detect the discharge signal of the cable; The monitoring host is used to obtain the gas concentration data, particle number data, temperature data, humidity data and discharge signal data measured by each sensor group, and pre-process the gas concentration data, particle number data, temperature data, humidity data and discharge signal data and send them to the central analysis server; The central analysis server includes a cloud platform, which is used to perform principal component analysis on the preprocessed gas concentration data, the preprocessed particle number data, the preprocessed temperature data, the preprocessed humidity data and the preprocessed discharge signal data to obtain the gas concentration principal component vector, the particle number principal component vector, the temperature principal component vector, the humidity principal component vector and the discharge signal principal component vector, and determine the defect type and defect severity of the cable according to the gas concentration principal component vector, the particle number principal component vector, the temperature principal component vector, the humidity principal component vector, the discharge signal principal component vector and the trained defect warning model of each cable micro-region; The human-machine interactive display screen is used to display the defect type and defect severity.

2. The cable defect warning device according to claim 1, characterized in that: The monitoring host includes an edge computing unit, which is used to pre-process and preliminarily judge the gas concentration data, particle number data, temperature data, humidity data and discharge signal data; The preprocessing includes data cleaning and feature extraction; The data cleaning includes removing noise values ​​and outliers; the feature extraction includes calculating the statistical characteristic values ​​of each data, and the statistical characteristic values ​​include mean, variance and peak value; the preliminary judgment includes comparing the statistical characteristic values ​​of each data with the corresponding threshold value to judge whether the data is abnormal; The monitoring host is also used to issue an early warning when the preliminary judgment result is abnormal.

3. The cable defect warning device according to claim 1, characterized in that: The cloud platform is used to perform data standardization processing on each of the preprocessed gas concentration data, preprocessed particle number data, preprocessed temperature data, preprocessed humidity data and preprocessed discharge signal data, and analyze the correlation between the various data; determine the data that needs to be subjected to principal component analysis based on the correlation.

4. The cable defect warning device according to claim 3 is characterized in that: The cloud platform is used to calculate the correlation coefficient matrix between various data, and determine the data that need to be subjected to principal component analysis according to the correlation coefficients between various data.

5. The cable defect warning device according to claim 3 is characterized in that: The principal component analysis includes: For each data that needs to be subjected to principal component analysis, including the standardized gas concentration data, the standardized particle number data, the standardized temperature data, the standardized humidity data, and the standardized discharge signal data: calculate the covariance matrix, perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and corresponding eigenvectors, select the first m principal components according to the size of the eigenvalues, and determine the principal component vector after dimensionality reduction according to the principal components; wherein m is an integer greater than or equal to 2.

6. The cable defect early warning device according to claim 1, characterized in that: The defect early warning model includes a neural network model; The neural network model structure includes input layer, hidden layer and output layer; The input layer is used to receive the principal component vector; The hidden layer includes multiple hidden sublayers, each hidden sublayer contains multiple neurons, and the hidden sublayer is used to perform nonlinear transformation through activation functions; The output layer is used to output defect type and defect severity.

7. The cable defect early warning device according to claim 1, characterized in that: The central analysis server is used to construct the defect warning model by using the model knowledge of the existing cable neural network model; and update the defect warning model according to the actual defect type and defect severity of the cable and the data measured by the sensor group.

8. The cable defect warning device according to claim 1, characterized in that: The human-machine interactive display screen is also used to display an overview of the defect status of the entire cable, the detection data and abnormal data of each micro-area, and recommended measures; The overview includes the defect type and defect severity of each micro-area of ​​the entire cable; the detection data includes gas concentration data, particle number data, temperature data, humidity data and discharge signal data.

9. The cable defect early warning device according to claim 1, characterized in that: The early warning device and the human-computer interaction display screen are connected to the central analysis server via a 4G network or an optical fiber network.

10. The cable defect warning device according to claim 1, characterized in that: The gas sensor array includes SF6 gas sensor, H2 gas sensor, CH4 gas sensor and C2H2 gas sensor.

Images