Intelligent monitoring and fault early warning system for electric lightning arrester

Abstract

The invention relates to the technical field of lightning arresters, in particular to an intelligent monitoring and fault early warning system for an electric power lightning arrester, which comprises an environment monitoring unit, an anti-interference unit, a fault prediction unit and a grading early warning unit. According to the invention, the environment monitoring unit utilizes a graphene sensor and a negative refractive index metamaterial to realize high-precision environment perception, and the anti-interference unit is utilized to open up a stable communication path in strong electromagnetic interference, so that the reliability of data transmission is guaranteed; a fault prediction unit is designed to realize encrypted data efficient decoding based on a quantum tunneling effect and a pseudo-random carrier, an early warning threshold is dynamically adjusted to accurately trigger fault early warning, a grading early warning unit is also adopted to construct a dynamic causal matrix to quantify environmental factor influence, an optimal maintenance route is autonomously planned, and the fault early warning efficiency is improved. And the aging data is fed back to the environment monitoring unit in a closed-loop manner to continuously optimize the system performance.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of lightning arresters, in particular to an intelligent monitoring and fault early warning system for a power lightning arrester. BACKGROUND

[0002] The intelligent monitoring and fault early warning system for a power lightning arrester is a comprehensive technical platform integrating environmental perception, anti-interference transmission, intelligent diagnosis and decision optimization. The system collects environmental data such as temperature, humidity and wind speed around the lightning arrester and equipment operation parameters in real time, guarantees data transmission stability by combining dynamic encryption anti-interference technology, performs fault prediction by using key features such as leakage current harmonic components and temperature gradient and dynamic threshold model, and ensures early warning reliability through secondary validation of confidence. When potential faults are detected, the system automatically generates maintenance tasks, plans the optimal maintenance route according to the environmental sensitivity weight and equipment causal matrix, and finally feeds back the maintenance time limit data to the monitoring unit to form a closed-loop management of perception, analysis, decision and feedback, realizes real-time monitoring of the lightning arrester state, early warning of faults, intelligent scheduling of maintenance resources and continuous optimization of system performance, and significantly improves the safety and operation efficiency of the power grid.

[0003] However, in the prior art, the traditional intelligent monitoring and fault early warning system for a power lightning arrester has the problem that the sensor is easily damaged or its precision is reduced in a strong electromagnetic field, high temperature and humidity, and extreme weather, and when wireless communication transmission is performed, data transmission is not complete due to electromagnetic interference, which affects the work efficiency of the staff.

[0004] Based on this, the application provides an intelligent monitoring and fault early warning system for a power lightning arrester to solve the above-mentioned technical problems. SUMMARY

[0005] The application aims to provide an intelligent monitoring and fault early warning system for a power lightning arrester. The application uses a graphene substrate sensor combined with a negative refractive index metamaterial filter array to significantly improve the accuracy and anti-noise ability of raw data such as temperature, humidity and wind speed by using a topological insulator material to correct environmental errors in real time through spatial focusing and fractal compression algorithms, solve the problem that the signal of the traditional sensor is easily distorted and the data reliability is low under strong electromagnetic interference, and fuse the device state parameters and real-time environmental data to generate a dynamic encryption key. The application generates a frequency spectrum heat map through short-time Fourier transform, solves the risk of data loss caused by electromagnetic interference in a high-voltage power scene by combining non-continuous carrier modulation and adaptive power allocation algorithm, and improves the fault early warning accuracy and reduces the operation and maintenance cost based on real-time correction of the early warning threshold of the leakage current harmonic component and the temperature gradient.

[0006] To achieve the above-mentioned purpose, the application provides the following technical scheme: The application provides a power lightning arrester intelligent monitoring and fault early warning system, which comprises an environment monitoring unit, an interference resisting unit, a fault prediction unit and a hierarchical early warning unit, wherein: The environment monitoring unit is used for acquiring first data information and second data information, and pre-processing the acquired first data information, wherein the first data information comprises temperature, humidity and wind speed information around the power lightning arrester, and the second data information comprises working voltage and working capacitance information of the power lightning arrester; The interference resisting unit is used for anti-interference processing of the pre-processed first data information, and real-time detection of a safe and stable communication path, and output of anti-interference first data information; The fault prediction unit is used for decoding the anti-interference first data information, and diagnosing the first data information and the second data information, and then judging whether to output a fault early warning; The hierarchical early warning unit is used for analyzing and processing the fault early warning, and then outputting fault early warnings of different levels if the fault early warning is output.

[0007] Further, the operation scheme of outputting the anti-interference signal is as follows: The electromagnetic signals in the 1-10GHz frequency band are collected, the time-frequency characteristics of the signals are extracted, the frequency spectrum occupation heat map is generated, and the available frequency spectrum holes are identified; In this step, a wideband antenna is used to cover the 1-10GHz frequency band, an analog-to-digital converter is used to convert the analog signal into a digital signal, then a band-pass filter is used to suppress the out-of-band noise, and then the original electromagnetic signal matrix is output; The short-time Fourier transform value is obtained, the power spectrum density of each time-frequency point is calculated using the value, and then the time-frequency power matrix is obtained, and the time-frequency power matrix is converted into a frequency spectrum heat map through data normalization; The time sequence characteristics are extracted from the original environment data, the time sequence characteristics include mean, standard deviation and gradient characteristics, the time sequence characteristics are normalized, and the environment feature vector is output; The dynamic encryption key is parsed to obtain a binary key, the binary key is converted into a baseband waveform by Manchester coding, the bandwidth is limited by a raised cosine filter to eliminate high-frequency noise, and the key waveform is output; The environment feature vector is modulated into a low-frequency carrier signal to obtain an environment carrier, the key waveform and the environment carrier are weighted and fused, and then the modulated baseband signal is output; The baseband signal is mapped to the selected frequency spectrum hole, non-continuous carrier modulation is adopted, the power of each carrier is dynamically adjusted by an adaptive power allocation algorithm, and the anti-interference signal is synthesized.

[0008] Further, in order to decode the anti-interference first data information, the main operation process is as follows: generating a dynamic decryption key according to the identification in the encrypted communication data, and amplifying the key signal through quantum tunneling effect; fusing the pseudo-random carrier with the amplified key signal to synchronously decrypt each encrypted signal, and then obtaining the first data information and the second data information.

[0009] The specific operation of preprocessing the obtained first data information is as follows: respectively acquiring the temperature, relative humidity and wind speed vector around the power lightning arrester; separating the temperature from noise, and jointly separating the relative humidity and wind speed vector from noise to obtain denoised environmental data, and then obtaining the original environmental data.

[0010] Further, the operation steps of diagnosing the first data information and the second data information are as follows: calculating the dynamic change rate of each parameter in the first data information and the second data information, and integrating into a change rate sequence; processing each parameter in the change rate sequence to obtain a feature point set after dimension reduction; calculating the dynamic threshold of the feature point set after dimension reduction according to the feature point set after dimension reduction, to complete the diagnosis process of the first data information and the second data information.

[0011] Further, the specific operation scheme of obtaining the feature point set after dimension reduction is as follows: According to the change rate sequence, the nonlinear mutual information between each parameter is calculated, and a 5*5 mutual information matrix is generated; The data in the five-dimensional 5*5 mutual information matrix is mapped to a three-dimensional manifold space by using manifold learning algorithm, and then the feature point set after dimension reduction is obtained.

[0012] Further, the operation process of calculating the dynamic threshold of the feature point set after dimension reduction is as follows: The geodesic distance of all points in the feature point set after dimension reduction is calculated, and the adaptive bandwidth is obtained through calculation, and the local density of all points in the feature point set after dimension reduction is calculated combined with the adaptive bandwidth, and the historical density sequence is integrated; The manifold density of each point in the feature point set after dimension reduction and the historical density sequence is calculated, the abnormality score of the manifold density is calculated according to the abnormality formula, and finally the dynamic threshold of the feature point set after dimension reduction is calculated by using the dynamic threshold formula.

[0013] Further, the judgment method of whether to output fault warning is as follows: According to the dynamic threshold of the feature point set after dimension reduction, the fault probability distribution of the current equipment is calculated, and the fault warning value is calculated through the decision confidence algorithm, combined with the relevance verification, and then it is judged whether to output the fault warning. Wherein, when the output value of the fault early warning value and the relevance verification result output value are both 1, the fault early warning is output, when the fault early warning identification output value is 0, the relevance verification result does not output a value, at this time the output device is normal.

[0014] Further, the operation process of analyzing and processing the fault early warning is as follows: According to the geodesic distance of all points in the reduced feature point set, the fault feature vector of the system is calculated, and then each parameter in the first data information and the second data information is processed twice to obtain relevance verification data; The risk index is calculated to obtain the fault definition index.

[0015] Further, the operation steps of outputting different levels of fault early warning are as follows: When the fault definition index is greater than or equal to a, the output level is first; When b is less than or equal to the fault definition index and less than a, the output level is second; When c is less than or equal to the fault definition index and less than b, the output level is third; Wherein, a is the first warning limit value, b is the second warning limit value, and c is the third warning limit value.

[0016] Further, the specific operation of processing each parameter in the first data information and the second data information twice is as follows: The contribution of each parameter in the first data information and the second data information to the fault is calculated, and then integrated into a contribution degree vector; The parameter with the highest contribution degree is selected, and the coupling strength value is calculated according to the mutual information coupling formula, and the coupling strength value is verified for relevance to obtain relevance verification data.

[0017] Further, the specific calculation method of calculating the fault feature vector of the system is: According to the reduced feature point set, the dynamic sensitivity coefficient is obtained; Then, the dynamic sensitivity coefficient is weighted to obtain a weighted distance vector, and the weighted distance vector is fused with the dynamic threshold to obtain the fault feature vector.

[0018] Further, the specific operation of obtaining the dynamic sensitivity coefficient is: The geodesic distance is used as the basic distance measure to calculate the geodesic distance matrix between all points in the reduced feature point set; Then, the sensitivity of each point in the geodesic distance matrix to the fault is calculated, and the average value of multiple sensitivity degrees is taken to obtain the dynamic sensitivity coefficient.

[0019] Compared with the prior art, the beneficial effects of the present application are: The application realizes high-precision collection and signal enhancement of environmental parameters such as temperature, humidity and wind speed by adopting a graphene sensor and a negative refractive index metamaterial filter array in the environmental monitoring unit, preserves the purity of original data by fractal compression and topological insulator error correction, dynamically generates an encryption key based on equipment state and environmental data in the interference resistance unit, opens a stable communication path in a strong electromagnetic interference environment by using spectrum hole identification and non-continuous carrier modulation technology, guarantees data transmission reliability, and the fault prediction unit amplifies the key signal through quantum tunneling effect, realizes efficient decoding of encrypted data by combining pseudo-random carrier fusion, and then extracts key features such as leakage current harmonic components and temperature gradient, dynamically adjusts the warning threshold according to environmental sensitivity, accurately triggers the fault warning after secondary verification of confidence, and the hierarchical warning unit constructs a dynamic causal matrix with fault features as nodes, combines environmental sensitivity weight and action space benefit function, and autonomously plans the optimal maintenance path, and feeds back the maintenance time limit data to the environmental monitoring unit to form a closed-loop optimization, finally realizes the whole-process intelligentization from environmental perception, anti-interference transmission, intelligent diagnosis to autonomous decision-making, significantly improves the operation reliability of the lightning arrester and reduces the operation and maintenance cost. BRIEF DESCRIPTION OF DRAWINGS

[0020] Fig. 1 A system diagram of the intelligent monitoring and fault warning system of the power lightning arrester according to the application; Fig. 2 A flowchart of the fault prediction unit in the intelligent monitoring and fault warning system of the power lightning arrester according to the application; Fig. 3 A framework diagram for analyzing and processing the fault warning in the intelligent monitoring and fault warning system of the power lightning arrester according to the application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0022] Embodiment: As shown in the figure, the embodiment provides an intelligent monitoring and fault warning system of a power lightning arrester, which comprises an environmental monitoring unit, an interference resistance unit, a fault prediction unit and a hierarchical warning unit, wherein: Figs. 1-3 ​Environmental monitoring unit: used to acquire first data information and second data information, and to preprocess the acquired first data information, wherein the first data information includes temperature, humidity and wind speed information around the power surge arrester, and the second data information includes the operating voltage and operating capacitance information of the power surge arrester; Anti-interference unit: performs anti-interference processing on the preprocessed first data information, detects a safe and stable communication path in real time, and outputs the anti-interference first data information; Fault prediction unit: used to decode the first data information to combat interference, and to perform diagnostic processing on the first data information and the second data information, thereby determining whether to output a fault warning; Tiered early warning unit: If a fault warning is output, the fault warning is analyzed and processed, and then different levels of fault warnings are output.

[0023] In this embodiment, in order to obtain the first data information and the second data information, a temperature sensor, a relative humidity sensor and a wind speed sensor are deployed around the power surge arrester, and a voltage sensor and a capacitance sensor are deployed at the electrical connection point of the surge arrester. By using multiple sensors working simultaneously, the first and second data information of the power surge arrester are collected.

[0024] Meanwhile, the operation scheme for outputting anti-interference signals is as follows: Electromagnetic signals in the 1-10GHz frequency band are collected, their time-frequency characteristics are extracted, a spectrum occupancy heatmap is generated, and available spectrum holes are identified. In this step, a broadband antenna is used to cover the 1-10 GHz frequency band, and an analog-to-digital converter is used to convert the analog signal into a digital signal. Then, a bandpass filter is used to suppress out-of-band noise, and the original electromagnetic signal matrix is ​​output. Obtain the short-time Fourier transform value, use the value to calculate the power spectral density at each time frequency point, and then obtain the time-frequency power matrix. The time-frequency power matrix is ​​then converted into a spectral heatmap by data normalization. The formula used to obtain the short-time Fourier transform value is as follows:

[0025] In this formula, These are the original signal sample values. For the Hanning window function, For frequency resolution, This represents the STFT window length.

[0026] The dynamic encryption key is combined with the original environmental data to generate the baseband signal to be modulated; In this step, time-series features are extracted from the first data information. These time-series features include mean, standard deviation, and gradient features. The time-series features are then normalized to output an environmental feature vector. The dynamic encryption key is parsed to obtain a binary key. The binary key is converted into a baseband waveform using Manchester encoding. The bandwidth is limited by a raised cosine filter to eliminate high-frequency noise and output the key waveform. The environmental feature vector is modulated into a low-frequency carrier signal to obtain the environmental carrier. The key waveform and the environmental carrier are weighted and fused, and then the baseband signal to be modulated is output. The formula for calculating the environmental carrier is as follows:

[0027] In this formula, For environmental carrier, For characteristic frequencies, For random phase, For environmental feature vector components, For time, The sampling period; Furthermore, the formula used for weighted fusion of the key waveform and the ambient carrier is:

[0028] In this formula, and These are the key and the environment weight, respectively. =0.7, =0.3, The key waveform, Environmental carrier.

[0029] The baseband signal is mapped to a selected spectral hole, non-continuous carrier modulation is used, the power of each carrier is dynamically adjusted through an adaptive power allocation algorithm, and an anti-interference signal is synthesized.

[0030] The specific process for synthesizing the anti-interference signal in this step is as follows: Multiple carrier center frequencies are extracted from available spectrum holes, and the channel gain vector is obtained through the path loss algorithm. The optimized power allocation vector is calculated using a gradient iterative update formula. Then, a carrier modulation algorithm can be used to map the spectral components of the baseband signal to each carrier and synthesize an anti-interference signal. The formula for the path loss algorithm is as follows:

[0031] In this formula, For reference gain, For transmission distance, At the speed of light, For shadow decay; Meanwhile, the gradient iterative update formula used in the above steps is:

[0032] In this formula, For the total rate, For power update rules, For carrier bandwidth, Let m be the noise power of the m-th carrier. and, The calculation formula is:

[0033] Furthermore, the main operational procedures for decoding encrypted communication data are as follows: A dynamic decryption key is generated based on the identifier in the encrypted communication data, and the key signal is amplified through the quantum tunneling effect. By fusing a pseudo-random carrier wave with an amplified key signal, each encrypted signal is synchronously decrypted to obtain the first data information and the second data information.

[0034] In this step, to form a pseudo-random carrier, the dynamic decryption key needs to be initialized to obtain initial data, which is then output in binary sequence form, and the substitution formula for each element in the sequence is... for:

[0035] in, For time step index, The length of the shift register. For feedback coefficients, Regarding the register state, it should be noted that... and The values ​​are all 0 or 1; The binary sequence is converted into a carrier signal to generate a pseudo-random carrier. This pseudo-random carrier is then multiplied with the amplified key signal to generate a fused signal. The encrypted signal is delayed, and the fused signal is used as the decryption key for decryption.

[0036] Furthermore, the diagnostic processing steps for the first data information and the second data information are as follows: Calculate the dynamic rate of change of each parameter in the first data information and the second data information, and integrate them into a rate of change sequence; In this step, the formula for the change in the dynamic rate of change is:

[0037] in, These are parameters, including: temperature T, humidity H, wind speed W, voltage U, and capacitance C. It represents the current moment. Represents the time difference between two different moments; Furthermore, the expression for the rate of change sequence is: ; Manifold processing is performed on each parameter within the rate of change sequence to obtain the dimensionality-reduced feature point set; The dynamic threshold of the dimensionality-reduced feature point set is calculated based on the dimensionality-reduced feature point set to complete the diagnostic processing of the first data information and the second data information.

[0038] Furthermore, the specific operation scheme for obtaining the dimensionality-reduced feature point set is as follows: Based on the rate of change sequence, calculate the nonlinear mutual information between each parameter and generate a 5×5 mutual information matrix; In this step, the formula for generating the 5×5 mutual information matrix is:

[0039] in, and These are any two parameters from the first data information and the second data information, respectively. For joint probability density; The manifold learning algorithm is used to map the data in the five-dimensional 5×5 mutual information matrix to the three-dimensional manifold space, thereby obtaining the dimensionality-reduced feature point set.

[0040] In this embodiment, the dynamic change rate is calculated by reflecting the evolution of the equipment state through the real-time change trend of parameters, making the calculated value more sensitive. Furthermore, the operation of reducing the 5-dimensional parameter sequence to a 3-dimensional feature point set can preserve the nonlinear relationship between parameters and avoid misjudgment under changing operating conditions with a fixed threshold.

[0041] Furthermore, the procedure for calculating the dynamic threshold of the feature point set after dimensionality reduction is as follows: Calculate the geodesic distance of all points in the dimensionality-reduced feature point set, obtain the adaptive bandwidth, calculate the local density of all points in the dimensionality-reduced feature point set using the adaptive bandwidth, and integrate the historical density sequence. In this step, the formula for calculating the geodetic distance is:

[0042] in, and Let each be any two feature points within the dimensionality-reduced feature point set. For geodetic distance, is a point. Time The set of paths This represents the straight-line distance between adjacent points on the path. This represents the number of points in the path. The above steps can be used to calculate The average value of the adaptive bandwidth is... The average value × 1.5; Furthermore, the formula for calculating the local density of all points within the dimensionality-reduced feature point set is:

[0043] in, For any feature point in the feature point set after dimensionality reduction, This is the upper limit value; Therefore, the expression for the historical density sequence is: ; The manifold density of each point in the dimensionality-reduced feature point set and the historical density sequence is calculated. The anomaly score of the manifold density is calculated according to the anomaly formula. Finally, the dynamic threshold of the dimensionality-reduced feature point set is calculated using the dynamic threshold formula.

[0044] It should be noted that the formula for calculating the anomaly score is:

[0045] in, The standard deviation of the historical density sequence. This is the mean of the historical density sequence.

[0046] Furthermore, the method for determining whether to output a fault warning is as follows: The fault probability distribution of the current device is calculated based on the dynamic threshold of the feature point set after dimensionality reduction, and the fault warning value is calculated through the decision confidence algorithm. Combined with correlation verification, it is then determined whether to output a fault warning. Specifically, when both the output value of the fault warning value and the output value of the correlation verification result are 1, a fault warning is output. When the output value of the fault warning indicator is 0, the correlation verification result does not output a value, and the device is output as normal.

[0047] It should be noted that the formula used for the failure probability distribution is:

[0048] in, For dynamic thresholds, The steepness coefficient.

[0049] In this embodiment, the fault probability distribution transforms the dynamic threshold into the probability of fault occurrence, thereby quantifying the risk level. At the same time, the error can be reduced by calculating the warning value through multi-parameter fusion, and the correlation verification eliminates interference signals through parameter coupling relationships.

[0050] Furthermore, the operational procedure for analyzing and processing fault warnings is as follows: The fault feature vector of the system is calculated based on the geodesic distance of all points in the feature point set after dimensionality reduction. Then, each parameter in the first data information and the second data information is processed twice to obtain the correlation verification data. Calculate the risk index to obtain the fault identification index.

[0051] Furthermore, the operation steps for outputting different levels of fault warnings are as follows: When the fault determination index is ≥ a, the output level is level one; When b ≤ fault determination index < a, the output level is level two; When c ≤ fault determination index < b, the output level is level three; Where a is the threshold for Level 1 warning, b is the threshold for Level 2 warning, and c is the threshold for Level 3 warning.

[0052] In this step, the possible values ​​for a, b, and c are as follows: The risk index is obtained by fusing the fault feature vector and the dynamic threshold, and the fusion formula is as follows:

[0053] in, The i-th element of the fault feature vector The i-th element of the dynamic threshold vector As a dynamic adjustment factor; It should be noted that the formula for calculating the dynamic adjustment factor is:

[0054] in, and These are the mean and standard deviation of the fault feature vector.

[0055] By correlating the risk index with the corresponding local density, the values ​​of a, b, and c are obtained. The formula required for this correlation is:

[0056]

[0057]

[0058] in, The function is a weighted quantile function, with quantiles of 0.95, 0.75, and 0.55 corresponding to the critical probabilities of Level 1, Level 2, and Level 3 early warnings, respectively.

[0059] Therefore, in this embodiment, the values ​​of a, b, and c are 0.8, 0.5, and 0.2, respectively.

[0060] In this step, the fault conditions are classified, so that different measures can be taken to deal with different fault conditions, which saves more manpower and material resources in operation and maintenance.

[0061] Furthermore, the specific operations for secondary processing of each parameter in the first and second data information are as follows: Calculate the contribution of each parameter in the first and second data information to the fault, and then integrate them into a contribution vector; The formula for calculating contribution is as follows:

[0062] In this formula, .

[0063] The parameter with the highest contribution is selected, and the coupling strength value is calculated according to the mutual information coupling formula. The correlation verification formula is then used to obtain the correlation verification data.

[0064] Furthermore, the specific calculation method for the fault feature vector of this system is as follows: The dynamic sensitivity coefficient is obtained by initializing the feature point set after dimensionality reduction. Subsequently, the dynamic sensitivity coefficients are weighted to obtain a weighted distance vector, which is then fused with the dynamic threshold to obtain the fault feature vector. Furthermore, the specific steps to obtain the dynamic sensitivity coefficient are as follows: Using geodesic distance as the basic distance metric, the geodesic distance matrix between all points in the dimensionality-reduced feature point set is calculated. Then, the sensitivity of each point in the geodesic distance matrix to the fault is calculated, and the average of multiple sensitivity values ​​is taken to obtain the dynamic sensitivity coefficient.

[0065] In this step, the formula used to calculate the dynamic sensitivity coefficient is:

[0066] in, This represents the total number of feature points; Furthermore, to calculate the weighted distance vector, the following formula is used:

[0067] in,

[0068] It should be noted that in this formula, For weighted distance components, This is the bandwidth parameter.

[0069] The weighted distance vector is fused with the dynamic threshold to obtain the fault feature vector; In this step, it is necessary to compare the geodesic distance with the dynamic threshold. When the geodesic distance is greater than or equal to the dynamic threshold, the attenuation factor is calculated and the fusion result is obtained; When the geodetic distance is less than the dynamic threshold, the weighted distance is directly retained; It should be noted that the fusion formula used is:

[0070] in, As the attenuation factor, For the fusion result; Ultimately, Expanded into vector form, it serves as the most faulty feature vector.

[0071] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0072] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A smart monitoring and fault early warning system for power surge arresters, characterized in that, It includes an environmental monitoring unit, an interference resistance unit, a fault prediction unit, and a graded early warning unit, among which: Environmental monitoring unit: used to acquire first data information and second data information, and to preprocess the acquired first data information, wherein the first data information includes temperature, humidity and wind speed information around the power surge arrester, and the second data information includes the operating voltage and operating capacitance information of the power surge arrester; Anti-interference unit: performs anti-interference processing on the preprocessed first data information, detects a safe and stable communication path in real time, and outputs the anti-interference first data information; Fault prediction unit: used to decode the first data information to combat interference, and to perform diagnostic processing on the first data information and the second data information, thereby determining whether to output a fault warning; Tiered early warning unit: If a fault warning is output, the fault warning is analyzed and processed, and then different levels of fault warnings are output.

2. The intelligent monitoring and fault early warning system for power surge arresters according to claim 1, characterized in that, The diagnostic processing steps for the first data information and the second data information are as follows: Calculate the dynamic rate of change of each parameter in the first data information and the second data information, and integrate them into a rate of change sequence; Manifold processing is performed on each parameter within the rate of change sequence to obtain the dimensionality-reduced feature point set; The dynamic threshold of the dimensionality-reduced feature point set is calculated based on the dimensionality-reduced feature point set to complete the diagnostic processing of the first data information and the second data information.

3. The intelligent monitoring and fault early warning system for power surge arresters according to claim 2, characterized in that, The specific steps to obtain the dimensionality-reduced feature point set are as follows: Based on the rate of change sequence, calculate the nonlinear mutual information between each parameter and generate a 5×5 mutual information matrix; The manifold learning algorithm is used to map the data in the five-dimensional 5×5 mutual information matrix to the three-dimensional manifold space, thereby obtaining the dimensionality-reduced feature point set.

4. The intelligent monitoring and fault early warning system for power surge arresters according to claim 2, characterized in that, The procedure for calculating the dynamic threshold of the feature point set after dimensionality reduction is as follows: Calculate the geodesic distance of all points in the dimensionality-reduced feature point set, obtain the adaptive bandwidth, calculate the local density of all points in the dimensionality-reduced feature point set using the adaptive bandwidth, and integrate the historical density sequence. The manifold density of each point in the dimensionality-reduced feature point set and the historical density sequence is calculated. The anomaly score of the manifold density is calculated according to the anomaly formula. Finally, the dynamic threshold of the dimensionality-reduced feature point set is calculated using the dynamic threshold formula.

5. The intelligent monitoring and fault early warning system for power surge arresters according to claim 1, characterized in that, The method for determining whether to output a fault warning is as follows: The fault probability distribution of the current device is calculated based on the dynamic threshold of the feature point set after dimensionality reduction, and the fault warning value is calculated through the decision confidence algorithm. Combined with correlation verification, it is then determined whether to output a fault warning. Specifically, when both the output value of the fault warning value and the output value of the correlation verification result are 1, a fault warning is output. When the output value of the fault warning indicator is 0, the correlation verification result does not output a value, and the device is output as normal.

6. The intelligent monitoring and fault early warning system for power surge arresters according to claim 1, characterized in that, The operational procedure for analyzing and processing fault warnings is as follows: The fault feature vector of the system is calculated based on the geodesic distance of all points in the feature point set after dimensionality reduction. Then, each parameter in the first data information and the second data information is processed twice to obtain the correlation verification data. Calculate the risk index to obtain the fault identification index.

7. The intelligent monitoring and fault early warning system for power surge arresters according to claim 6, characterized in that, The operation steps for outputting different levels of fault warnings are as follows: When the fault determination index is ≥ a, the output level is level one; When b ≤ fault determination index < a, the output level is level two; When c ≤ fault determination index < b, the output level is level three; Where a is the threshold for Level 1 warning, b is the threshold for Level 2 warning, and c is the threshold for Level 3 warning.

8. The intelligent monitoring and fault early warning system for power surge arresters according to claim 1, characterized in that, The specific steps for performing secondary processing on each parameter in the first and second data information are as follows: Calculate the contribution of each parameter in the first and second data information to the fault, and then integrate them into a contribution vector; The parameter with the highest contribution is selected, and the coupling strength value is calculated according to the mutual information coupling formula. The correlation of the coupling strength value is verified to obtain the correlation verification data.

9. The intelligent monitoring and fault early warning system for power surge arresters according to claim 6, characterized in that, The specific calculation method for the fault feature vector of this system is as follows: The dynamic sensitivity coefficient is obtained by initializing the feature point set after dimensionality reduction. Subsequently, the dynamic sensitivity coefficient is weighted to obtain a weighted distance vector, which is then fused with the dynamic threshold to obtain the fault feature vector.

10. The intelligent monitoring and fault early warning system for power surge arresters according to claim 9, characterized in that, The specific steps to obtain the dynamic sensitivity coefficient are as follows: Using geodesic distance as the basic distance metric, the geodesic distance matrix between all points in the dimensionality-reduced feature point set is calculated. Then, the sensitivity of each point in the geodesic distance matrix to the fault is calculated, and the average of multiple sensitivity values ​​is taken to obtain the dynamic sensitivity coefficient.

Images