Remote diagnosis method for non-contact voltage sensor of grounding box

Abstract

The application provides a grounding box non-contact voltage sensor remote diagnosis method, comprising the following steps: obtaining the input-output response curve change trend of the voltage sensor under different voltage levels of the grounding box, confirming whether there is a preliminary abnormal signal, and obtaining an abnormal preliminary confirmation result; extracting an electrostatic force aggregation parameter from a metal debris distribution feature, mapping out a metal debris accumulation path, superimposing the metal debris accumulation path and the input-output response curve change trend of the real-time voltage sensor, and confirming an electric field abnormal source; based on the electric field abnormal source, fusing the degree of influence of the metal debris accumulation and the metal debris distribution feature to generate a grounding box non-contact voltage sensor remote diagnosis report, sending the remote diagnosis report to a diagnosis system, and obtaining an updated voltage sensor calibration reference.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to a remote diagnosis method for a non-contact voltage sensor of a grounding box. BACKGROUND

[0002] Non-contact voltage measurement technology has important significance in the safety monitoring and operation maintenance of power systems, which can effectively avoid the safety risks of traditional contact measurement and improve the reliability and efficiency of equipment operation. In particular, in the voltage monitoring of high-voltage equipment such as grounding boxes, non-contact voltage sensors achieve voltage measurement by sensing electric field changes, which can significantly reduce the risk of operating personnel in complex environments, and lay the foundation for remote diagnosis and intelligent management. Current research and application of non-contact voltage sensors have made some progress, but existing methods have obvious shortcomings in dealing with interference factors in complex operating environments, especially some special situations in long-term operation of equipment can challenge the measurement accuracy and system stability, which need to be solved to ensure the reliability of monitoring. In the closed environment of the grounding box, small metal debris generated by vibration or friction of metal components will gradually accumulate. These debris change the electric field distribution around the sensor under the action of the electric field. The surface free electrons of the metal debris redistribute to form a dipole structure. If the debris is charged (such as tribocharging), it will directly become a space charge source, interfering with the original electric field distribution. The accumulation of metal debris will directly interfere with the electric field between the sensor and the measured conductor. This interference is not only related to the amount of debris, but also closely related to the distribution position and shape of the debris, because the metal debris as a conductor forms a local charge redistribution in the electric field. The closer the debris to the sensor, the stronger the field distortion it produces. Irregularly shaped debris can produce field enhancement effects at their sharp ends, and multiple debris can form a composite interference field, causing the measured value to deviate from the actual voltage. Debris may form irregular accumulations near the sensor, changing the uniformity of the local electric field and causing directional differences in the sensor's response sensitivity to voltage changes, i.e., the sensitivity significantly decreases in some directions, while it may be abnormally amplified in other directions. Specifically, in the actual operation of the grounding box, when metal debris is generated inside the device due to long-term vibration, these debris will gradually accumulate in the key area between the sensor and the conductor under the action of electrostatic force, possibly accumulating on the insulating medium surface below the sensor, forming a structure similar to a "micro shielding layer". This structure not only weakens the electric field strength, but also can cause significant deviations in the output characteristic curves of the sensor at high and low voltages, and even in some cases, the sensor cannot correctly respond to voltage changes. SUMMARY

[0003] The present application provides a remote diagnosis method for a non-contact voltage sensor of a grounding box, mainly comprising:

[0004] The input-output response curve trend of the voltage sensor under different voltage levels of the grounding box is obtained, an abnormal signal is confirmed, and an abnormal preliminary confirmation result is obtained; based on the abnormal preliminary confirmation result, a matching degree analysis is performed on the input-output response curve trend and theoretical standard characteristic data of the voltage sensor, an influence of metal debris accumulation on the grounding box is identified, and a preliminary identification tag of the influence of metal debris accumulation is obtained; a response sensitivity directionality difference feature of the voltage sensor is extracted from the input-output response curve trend, and the degree of the influence of metal debris accumulation is determined; based on the degree of the influence of metal debris accumulation, a data point spatial distribution mode in the input-output response curve trend is analyzed, the data points in the input-output response curve trend are clustered and grouped, a distribution cluster is obtained, a metal debris distribution feature is identified by comparing a center of the distribution cluster and a position corresponding to the voltage sensor; an electrostatic force accumulation parameter is extracted from the metal debris distribution feature, a metal debris accumulation path is mapped, the metal debris accumulation path and the input-output response curve trend are superimposed, and an electric field abnormal source is confirmed; based on the electric field abnormal source, the degree of the influence of metal debris accumulation and the metal debris distribution feature are fused, diagnostic data is generated, and a voltage sensor calibration reference is updated; based on the updated voltage sensor calibration reference, the input-output response curve trend is monitored, a matching analysis is performed on the voltage sensor theoretical standard characteristic data, a monitoring matching result is obtained, and a stable state of the voltage sensor is determined by comparing the monitoring matching result with the preliminary identification tag of the influence of metal debris accumulation.

[0005] Further, the input-output response curve trend of the voltage sensor under different voltage levels of the grounding box is obtained, an abnormal signal is confirmed, and an abnormal preliminary confirmation result is obtained, including:

[0006] A test voltage is applied to the grounding box under multiple voltage levels, an output voltage value of the voltage sensor at each test point is recorded, a ratio of an output voltage difference value to an input voltage difference value of adjacent test points is calculated, a slope sequence is obtained, and the input-output response curve trend is formed; for the slope sequence, a relative deviation value of adjacent slopes is calculated, a test point corresponding to a deviation value exceeding a threshold value or a negative value in the slope sequence is marked as an abnormal signal point, a proportion of the abnormal signal points is counted, and the abnormal preliminary confirmation result is determined.

[0007] Further, the matching degree analysis is performed on the input-output response curve trend and the theoretical standard characteristic data of the voltage sensor based on the abnormal preliminary confirmation result, the influence of metal debris accumulation on the grounding box is identified, and the preliminary identification tag of the influence of metal debris accumulation is obtained, including:

[0008] The input-output response data of multiple calibration records is extracted from the voltage sensor historical calibration database, the average of the slopes of each test point is calculated, and a weighted average is formed to form a historical reference slope sequence; the corresponding test point slope value sequence in the theoretical standard characteristic data of the voltage sensor is obtained, the difference between the slope value in the input-output response curve change trend and the historical reference slope sequence is calculated, and a deviation vector is obtained; using a support vector machine algorithm, the deviation vector is used as a characteristic input, a two-classification model is trained, the distance from the deviation vector to the classification hyperplane is calculated, the influence of metal debris accumulation is determined, and a preliminary identification label of the influence of metal debris accumulation is generated.

[0009] Further, the voltage sensor response sensitivity directionality difference feature is extracted from the input-output response curve change trend, and the degree of the influence of the metal debris accumulation is determined, including:

[0010] The slope values of the test points are extracted from the input-output response curve change trend, the slope difference values of adjacent test points are calculated, and a directionality difference feature vector is constructed; principal component analysis is performed on the directionality difference feature vector, the principal components are extracted, and are projected into a low-dimensional space to obtain a reduced dimension feature representation; according to the numerical distribution of the reduced dimension feature representation, in combination with the intensity level in the preliminary identification label of the influence of the metal debris accumulation, a comprehensive influence index is calculated, and the degree of the influence of the metal debris accumulation is determined.

[0011] Further, according to the degree of the influence of the metal debris accumulation, the spatial distribution pattern of the data points in the input-output response curve change trend is analyzed, the data points in the input-output response curve change trend are clustered and grouped to obtain distribution clusters, and the metal debris distribution feature is identified by comparing the distribution cluster center and the voltage sensor position corresponding relationship, including:

[0012] The data point voltage values and output deviation values are extracted from the input-output response curve change trend to construct a two-dimensional feature space; K-means clustering is performed on the data points in the two-dimensional feature space to determine the cluster center and the data point set, and the distribution clusters are obtained; the voltage distribution interval of the distribution cluster is extracted, and is associated with the voltage sensor position to generate a cluster-sensor mapping relationship; according to the mapping relationship, the data point density, shape feature and distribution range in the cluster are calculated, and the metal debris distribution feature is identified.

[0013] Further, the electrostatic force aggregation parameters are extracted from the metal debris distribution feature, the metal debris accumulation path is mapped, the metal debris accumulation path and the input-output response curve change trend are superimposed, the source of the electric field anomaly is confirmed, and including:

[0014] Calculate the induced charge amount and electric field intensity gradient from the metal debris distribution characteristics, extract the electrostatic force gathering parameter; obtain the vibration event record, calculate the debris moving direction and distance, combine the electrostatic force gathering parameter, generate the metal debris accumulation path; superimpose the metal debris accumulation path and the input / output response curve trend in time sequence, calculate the overlapping area of the path and the response abnormal area, identify the area with the highest correlation degree, record the spatial coordinates of the area with the highest correlation degree, and confirm the source of the electric field anomaly.

[0015] Further, based on the source of the electric field anomaly, the degree of influence of the metal debris accumulation and the metal debris distribution characteristics are fused to generate diagnostic data, and the voltage sensor calibration reference is updated, including:

[0016] Based on the coordinates of the source of the electric field anomaly, the degree of influence of the metal debris accumulation and the metal debris distribution characteristic data are extracted to form diagnostic data; the diagnostic data is converted into a data stream by a remote transmission module and sent to a diagnostic system, a feedback calibration instruction is received, a calibration parameter is analyzed, a voltage sensor calibration configuration file is updated, and the updated voltage sensor calibration reference is obtained.

[0017] Further, according to the updated voltage sensor calibration reference, the input / output response curve trend is monitored, and matching analysis is performed with the voltage sensor theoretical standard characteristic data to obtain a monitoring matching result, and by comparing the monitoring matching result with the preliminary identification label of the influence of the metal debris accumulation, the stable state of the voltage sensor is determined, including:

[0018] According to the updated voltage sensor calibration reference, the input / output data of the voltage sensor is collected, the slope change rate is calculated, and the abnormal point is recorded; the input / output response curve trend is compared with the voltage sensor theoretical standard characteristic data, the difference mean value and standard deviation are counted, and the matching state is determined; the matching state is compared with the preliminary identification label of the influence of the metal debris accumulation, the consistency of the matching state and the influence level in the continuous period is determined, and the stable state of the voltage sensor is determined.

[0019] The technical scheme provided by the embodiments of the present application can include the following beneficial effects:

[0020] The application discloses a remote diagnosis method for a non-contact voltage sensor of a grounding box, and solves the problem of a service scenario in which metal debris accumulation causes abnormal response of a voltage sensor. By analyzing the input-output response curve trend of the sensor under different voltage levels, the application preliminarily identifies abnormal signals, and combines historical calibration data to evaluate the matching degree of the curve and the theoretical standard by using a support vector machine algorithm to confirm the influence of metal debris accumulation. Further, by extracting the response sensitivity directionality difference feature and performing dimensionality reduction processing, the accumulation degree of the debris is determined, and the debris distribution characteristics such as density, shape and range are identified by combining data point space distribution clustering analysis. Based on the electrostatic force aggregation parameters and vibration friction historical records, the application maps the debris accumulation path, superimposes the real-time curve trend to confirm the source of the electric field anomaly, and finally generates a remote diagnosis report and updates the calibration benchmark. The matching results and the debris influence label are compared to ensure the stable state of the sensor. By fusing multi-dimensional feature analysis and intelligent algorithms, the application realizes accurate diagnosis and adaptive calibration of the non-contact voltage sensor, and improves the operation reliability of the grounding box. BRIEF DESCRIPTION OF DRAWINGS

[0021] Fig. 1 A flowchart of a remote diagnosis method for a non-contact voltage sensor of a grounding box according to the application.

[0022] Fig. 2 A schematic diagram of a remote diagnosis method for a non-contact voltage sensor of a grounding box according to the application. DETAILED DESCRIPTION

[0023] The technical solutions of the application will be described below in connection with the embodiments, obviously, the described embodiments are only some 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 are within the protection scope of the application.

[0024] As Figs. 1-2 , the remote diagnosis method for the non-contact voltage sensor of the grounding box according to the embodiment can specifically include:

[0025] S101, the input-output response curve trend of the voltage sensor of the grounding box under different voltage levels is obtained, whether there is a preliminary abnormal signal is confirmed, and a preliminary abnormal confirmation result is obtained.

[0026] The test voltage is applied to the grounding box in increments of 5% of the voltage in the range of 10% to 120% of the rated voltage, the output voltage value of the voltage sensor at each test point is recorded, the slope of each interval is calculated by dividing the output voltage difference between adjacent test points by the input voltage difference, the slopes are arranged in ascending order of test voltage to form a slope sequence, and the trend data of the input-output response curve is obtained. For the slope sequence in the trend data of the input-output response curve, the relative deviation value between adjacent slopes is calculated, if the relative deviation value exceeds the preset threshold or negative values appear in the slope sequence, the corresponding voltage test point is marked as an abnormal signal point, the proportion of abnormal signal points in the total test points is counted, and when the proportion exceeds the preset determination threshold, it is determined that there is an abnormality, and an abnormal preliminary confirmation result is obtained.

[0027] Specifically, in one embodiment, the test process of the voltage sensor of the grounding box starts from 10% of the rated voltage and gradually increases to 120%, forming a complete voltage response characteristic curve.

[0028] Specifically, when the rated voltage is 10kV, the test voltage starts from 1kV and increases by 0.5kV each time until 12kV, and the sensor output voltage value is read after 3 seconds of stabilization at each test point.

[0029] For example, the slope between adjacent test points is calculated by difference method, such as at 5kV and 5.5kV two test points, the sensor outputs 2.5V and 2.75V respectively, then the interval slope is 0.5V / kV.

[0030] It should be noted that the slope sequence of a normally working voltage sensor should present a monotonically increasing or constant characteristic, when the slope drops sharply or becomes negative, it indicates that the response characteristic of the sensor is distorted.

[0031] Preferably, the calculation of the relative deviation value is realized by dividing the difference of adjacent slopes by the previous slope.

[0032] In one possible implementation, when more than three consecutive test points are marked as abnormal, or the proportion of abnormal points exceeds 20%, it is determined that the sensor has preliminary signs of metal debris interference, at this time the abnormal preliminary confirmation result is positive, triggering the subsequent deep diagnosis process.

[0033] S102, according to the abnormal preliminary confirmation result, obtaining the historical calibration data of the voltage sensor as a reference benchmark, performing convergence analysis on the matching degree between the input-output response curve trend of the current voltage sensor and the theoretical standard characteristic data of the voltage sensor, identifying the current grounding box affected by potential metal debris accumulation, and obtaining a preliminary identification label of metal debris accumulation influence.

[0034] According to the preliminary identification result of the anomaly, the input-output response data in the last three calibration records in the voltage sensor historical calibration database is extracted, the slope value of each voltage test point at each calibration is calculated and the average value is obtained, the slope average value of the three calibrations is weighted and averaged according to the time weight, the closer to the current time, the higher the weight, and the historical reference slope sequence is obtained. The standard input-output response characteristic curve is obtained from the voltage sensor theoretical design document, the slope value sequence of the standard curve at the same voltage test point is extracted as the theoretical standard characteristic data, the slope value of the input-output response curve of the current sensor at the corresponding test point is calculated by difference with the historical reference slope sequence, and the deviation vector is obtained. Support vector machine algorithm is used, the deviation vector is used as the characteristic input, the normal state and the metal debris interference state are used as the two-class labels for training, the distance from the deviation vector to the classification hyperplane is calculated, if the distance is negative, it is determined that it is affected by the metal debris accumulation, and the influence degree quantization value is obtained after the distance absolute value is normalized. According to the influence degree quantization value, combined with the number and distribution position of the components exceeding the threshold in the deviation vector, the identification information including the influence intensity level, the abnormal test point range and the deviation direction feature is generated, and the identification information is converted into the preliminary identification label of the metal debris accumulation influence through the preset coding rule.

[0035] Specifically, in an embodiment, the extraction process of historical calibration data needs to consider the weight distribution of time dimension. The grounding box voltage sensor usually performs calibration once every quarter, and the calibration data contains the input-output corresponding relationship at different voltage levels.

[0036] Specifically, the weight coefficient of the last calibration is set to 0.5, the weight coefficient of the previous calibration is set to 0.3, and the weight coefficient of the second previous calibration is set to 0.2. Through this decreasing weight distribution method, the calibration data closer to the current time contributes more to the reference benchmark. Each calibration record contains data of at least 20 test points, covering a range of 10% to 120% of the rated voltage. The slope values of these data points are calculated by difference method, and the slope sequence is formed and fused by weighting.

[0037] It should be noted that the acquisition of the theoretical standard characteristic data involves the extraction of the sensor design parameters. The voltage sensor will determine the ideal linear response relationship according to the electromagnetic induction principle in the design stage. This ideal relationship is given in the form of a curve or a data table in the sensor technical manual. In actual operation, the standard curve data points extracted from the technical document need to be aligned with the actual test points, and the corresponding theoretical slope values are obtained at the same voltage level by interpolation method.

[0038] Exemplarily, the construction process of the deviation vector reflects the difference degree between the current sensor state and the historical reference. When the actual slope of the sensor at the 5kV test point is 0.48V / kV, and the historical reference slope is 0.5V / kV, the deviation component of this point is -0.02. The deviation components of all test points are combined to form a multi-dimensional vector, each dimension of the vector corresponds to a voltage test point, the length of the vector reflects the overall deviation degree, and the direction of the vector reveals the deviation pattern.

[0039] Preferably, the support vector machine algorithm plays a role of classification and identification in this application scenario. The support vector machine distinguishes between normal state and abnormal state affected by metal debris by finding the optimal hyperplane. The training process uses the historical accumulated labeled data, where the normal state samples come from the sensor data during the period without metal debris interference, and the abnormal state samples come from the historical cases that have been confirmed to be affected by metal debris. The kernel function of the support vector machine selects the radial basis function, which can handle nonlinearly separable data distribution. In the classification decision, the deviation vector is mapped to a high-dimensional feature space, the directed distance from the classification hyperplane is calculated, the sign of the distance determines the classification result, and the absolute value of the distance reflects the confidence. When the deviation vector falls on the abnormal side of the hyperplane and the absolute value of the distance exceeds the preset threshold, it is determined to be affected by the accumulation of metal debris, and the absolute value of the distance is normalized to convert to a quantitative value of the influence degree between 0 and 1.

[0040] In one possible implementation, the calculation of the influence degree quantitative value adopts sigmoid function for normalization processing, so that the distance value can be smoothly mapped to the influence degree interval.

[0041] Specifically, the coding rule of the identification label covers information in three dimensions. The influence intensity level is divided into three levels of mild, moderate and severe according to the quantitative value, corresponding to the intervals of 0-0.3, 0.3-0.7 and 0.7-1.0 respectively; the abnormal test point range is determined by counting the number and position distribution of components in the deviation vector that exceed the threshold, forming the area identification of the low-voltage section, the medium-voltage section or the high-voltage section; the deviation direction feature is judged by analyzing the positive and negative distribution of each component of the deviation vector, identifying the overall high, overall low or local fluctuation pattern.

[0042] For example, when the influence degree quantitative value is 0.65, there are 8 components in the deviation vector that exceed the threshold and mainly concentrate in the range of 60% to 90% rated voltage, and each component mainly presents a negative value, the coded identification label can be represented as "M-H-N", where M represents moderate influence, H represents high-voltage section abnormality, and N represents negative deviation.

[0043] It can be understood that this multi-dimensional label coding method provides a structured information basis for subsequent diagnosis and processing, so that the system can take targeted measures according to different label combinations. Further, in the actual grounding box operating environment, the accumulation process of metal debris has a gradual characteristic. A small amount of debris in the early stage may only cause slight deviation of individual test points. As the number of debris increases and the distribution range expands, the length of the deviation vector gradually increases, and more dimensions appear abnormal. By continuously monitoring the change trend of the identification label, the dynamic tracking of the metal debris accumulation process can be realized, and a warning can be issued in time before the impact reaches the critical value, avoiding serious degradation of sensor measurement accuracy.

[0044] S103, a preliminary identification label for the influence of metal debris accumulation, a response sensitivity directionality difference feature of the voltage sensor is extracted from the input-output response curve change trend of the current voltage sensor, and the degree of influence of the metal debris accumulation is determined after dimensionality reduction processing.

[0045] A preliminary identification label for the influence of metal debris accumulation is extracted from the input-output response curve change trend of the current voltage sensor. The slope values of each test point are calculated, the difference between the slope of each test point and the slope of the adjacent test point is calculated, and the directionality difference feature vector is constructed according to the uneven distribution of the difference in the low voltage section, the medium voltage section and the high voltage section. Obtain a multi-dimensional feature data set. The principal component analysis method is used for dimensionality reduction processing of the multi-dimensional feature data set, the covariance matrix of the feature vector is calculated, the first three feature vectors with the largest eigenvalues are extracted as the principal components, and when the cumulative contribution rate of the current three principal components exceeds the preset threshold, the original feature is projected into the principal component space to obtain a three-dimensional feature representation after dimensionality reduction. According to the numerical distribution of each dimension in the three-dimensional feature representation, combined with the influence intensity level in the preliminary identification label, a weight coefficient is assigned, and a comprehensive influence index is calculated by weighted summation. If the comprehensive influence index exceeds the preset mild threshold but does not exceed the severe threshold, it is determined as moderate influence, and the degree of influence of the metal debris accumulation is determined.

[0046] Specifically, in one embodiment, the extraction of the directionality difference feature is based on the response characteristic change of the voltage sensor at different voltage sections.

[0047] Specifically, the entire test range is divided into three intervals of low voltage section, medium voltage section and high voltage section, corresponding to 10%-40%, 40%-80% and 80%-120% of the rated voltage respectively. In each interval, the slope difference between adjacent test points is calculated. When the slope difference in a certain interval is significantly greater than that in other intervals, it indicates that the interval is more seriously affected by metal debris.

[0048] It should be noted that the application of principal component analysis in this scenario aims to extract key features that best reflect the impact pattern of metal debris. The construction of the covariance matrix is based on the correlation between each dimension in the multi-dimensional feature data set, and each element in the matrix represents the covariance value between two feature dimensions. By performing eigenvalue decomposition on the covariance matrix, the eigenvalues and corresponding eigenvectors are obtained, and the size of the eigenvalue reflects the degree of dispersion of the data in that direction. The eigenvectors corresponding to the largest three eigenvalues are selected as the principal components, because practice shows that the first three principal components can usually retain more than 85% of the information of the original data.

[0049] Exemplarily, when the sensor is slightly affected by metal debris, the slope difference of the high-voltage section will first appear abnormal, while the low-voltage section remains normal; as the accumulation of debris increases, the abnormality will gradually spread to the medium-voltage section and the low-voltage section.

[0050] Preferably, the calculation of the comprehensive impact index adopts a hierarchical weight system. According to the impact intensity level in the preliminary identification label, a weight of 0.3 is assigned to slight impact, a weight of 0.6 is assigned to moderate impact, and a weight of 1.0 is assigned to severe impact. Each dimension value in the three-dimensional feature representation is multiplied by the corresponding preset weight and summed to obtain a comprehensive impact index ranging from 0 to 3.

[0051] In one possible implementation, the threshold is set based on historical statistical data, with a slight threshold of 0.8 and a severe threshold of 2.0. When the comprehensive impact index is between the two thresholds, it is determined to be moderate impact, at which time the sensor can still maintain basic measurement function, but the accuracy is reduced, and preventive maintenance is needed.

[0052] It can be understood that this multi-dimensional impact degree quantification method can comprehensively reflect the differential impact of metal debris on the response characteristics of different voltage sections of the sensor, and provide a quantitative basis for subsequent accurate diagnosis and maintenance decision-making.

[0053] S104, according to the degree of metal debris accumulation, analyze the data point space distribution pattern in the input-output response curve trend of the voltage sensor, cluster and group each data point in the input-output response curve trend of the voltage sensor, obtain the grouped distribution clusters, and by comparing the corresponding relationship between the centers of each distribution cluster and the positions of each voltage sensor, identify the metal debris distribution characteristics.

[0054] According to the degree of influence of metal debris accumulation, the voltage value and the corresponding output value of each data point are extracted from the trend of the input-output response curve of the voltage sensor, the deviation of the output value of each data point from the theoretical value is calculated, the voltage value is taken as the abscissa, and the deviation value is taken as the ordinate to construct a two-dimensional feature space, and the spatial distribution representation of the data points is obtained. The K-means clustering algorithm is used to group the data points in the spatial distribution representation, the number of clusters is determined according to the distribution range of the deviation value, and the position of each cluster center is updated through iterative calculation. The iteration is stopped when the sum of the squares of the distances of the data points in the cluster to the cluster center no longer changes, and the center coordinates of each cluster and the data point set contained are recorded to obtain the grouped distribution clusters. The distribution interval of each cluster in the voltage axis is extracted, the voltage value corresponding to the midpoint of the interval is associated with the position of the sensor working at that voltage level in the grounding box, and if the measurement range of a certain sensor contains the voltage value, the corresponding relationship between the cluster and the sensor is established, and a mapping relationship table of the cluster and the sensor is obtained. According to the mapping relationship table, the number of data points in each distribution cluster is divided by the voltage range covered by the cluster to obtain the debris density, the shape feature is judged by the standard deviation ratio of the data points in the cluster in the horizontal and vertical directions, and the distribution range is determined by measuring the distance from the farthest data point in the cluster to the cluster center. The three indicators are combined to identify the metal debris distribution characteristics.

[0055] Specifically, in one embodiment, the construction of the two-dimensional feature space is based on the deviation relationship between the output value of the voltage sensor and the theoretical standard value.

[0056] Specifically, for each test voltage point, the theoretical output value is obtained through the calibration curve of the sensor, and the difference between the actual output value and the theoretical value is the deviation value. When there is metal debris in the grounding box, the deviation values at different voltage levels will show a specific distribution pattern. The deviation at the low voltage segment may be small, while at the high voltage segment, the influence of metal debris is more obvious due to the increase of electric field strength, and the deviation value increases accordingly. By mapping the voltage value to the horizontal axis and the deviation value to the vertical axis, the two-dimensional space can intuitively display the abnormal distribution of the sensor response characteristics.

[0057] It should be noted that the key of the K-means clustering algorithm in this application is to reasonably determine the number of clusters. The determination of the number of clusters is based on the distribution characteristics of the deviation value, and the clustering effect is evaluated by calculating the silhouette coefficient under different numbers of clusters. The silhouette coefficient reflects the tightness of the data points in the cluster and the separation degree between clusters, and the closer the coefficient value is to 1, the better the clustering effect is. In actual operation, the number of clusters corresponding to the peak value of the silhouette coefficient is usually selected. The iteration process of the algorithm starts from randomly selected initial center points, and in each iteration, each data point is assigned to the nearest cluster center, and then the center position of each cluster is recalculated as the mean coordinate of all points in the cluster. This process continues until the change of the cluster center position is less than a preset convergence threshold or the maximum number of iterations is reached.

[0058] Exemplarily, the mapping relationship establishment process between the cluster and the sensor needs to consider the correspondence of voltage levels. There are usually multiple voltage sensors configured in the grounding box, and each sensor is responsible for the measurement of a specific voltage range. When the distribution interval of a certain distribution cluster on the voltage axis is 30kV to 40kV, the sensor responsible for the measurement of this voltage range is found. If the sensor of the 35kV level happens to cover this interval, the corresponding relationship between the cluster and this sensor is established. This mapping relationship reveals the approximate location of the metal debris in the grounding box, because the debris tends to accumulate in the area close to the sensor affected by it.

[0059] Preferably, the calculation of the debris density uses the number of data points in a unit voltage range to represent. The specific calculation method is to divide the total number of data points in the cluster by the span range of the cluster on the voltage axis. The larger the density value, the more densely the test points in the voltage range are affected by the metal debris, and the more serious the accumulation of the debris.

[0060] In one possible implementation, the shape feature is determined by calculating the dispersion of the data points in the cluster in the horizontal and vertical directions. The horizontal standard deviation reflects the distribution width of the data points on the voltage axis, and the vertical standard deviation reflects the variation amplitude of the deviation value. The ratio of the two can describe the shape feature of the cluster: when the ratio is close to 1, the cluster presents a circular distribution, indicating that the metal debris has a relatively uniform effect on each voltage point; when the ratio is greater than 2, the cluster presents an oval shape that is horizontally elongated, indicating that the debris affects a wide voltage range but has a relatively consistent deviation amplitude; when the ratio is less than 0.5, the cluster presents a vertically elongated shape, indicating that the debris causes severe deviation fluctuations at a specific voltage point. This shape feature analysis helps to infer the accumulation shape of the metal debris, and flaky debris tends to form a horizontal distribution, while granular debris may cause local point-like effects.

[0061] Specifically, the determination of the distribution range is achieved by calculating the maximum distance from the boundary points in the cluster to the center. This distance value directly reflects the spatial span of the metal debris affected area. Further, the comprehensive analysis of the three indicators can fully characterize the distribution characteristics of the metal debris. The combination of high density, circular distribution, and small range indicates that the debris is highly concentrated in the local area; the combination of low density, oval distribution, and large range indicates that the debris is in a dispersed state.

[0062] It can be understood that by converting the abstract electrical parameter change into specific spatial distribution characteristics, the maintenance personnel can more accurately locate the debris accumulation position and develop a targeted cleaning plan, thereby improving the operation reliability of the grounding box.

[0063] S105, extract the electrostatic force gathering parameter from the metal debris distribution characteristics, map out the metal debris accumulation path, superimpose the metal debris accumulation path and the input-output response curve trend of the real-time voltage sensor, and confirm the source of the electric field anomaly.

[0064] From the density value and shape characteristics in the metal debris distribution characteristics, the surface induced charge of the debris is calculated according to the polarization principle of the metal debris in the electric field, the local electric field intensity gradient is calculated combining the voltage level in the grounding box and the debris distribution position, the charge density spatial distribution is obtained through the distribution relationship between the electric field intensity gradient and the debris density, and the charge density, field strength distribution, and polarization intensity are extracted as the electrostatic force gathering parameter. The vibration event record in the grounding box equipment operation log is obtained, the vibration occurrence time, intensity level, and duration are extracted, the displacement direction and distance of the debris in each time period are calculated according to the empirical relationship between the vibration intensity and the metal debris displacement speed, the displacement vector caused by the vibration is superimposed on the charge distribution position in the electrostatic force gathering parameter, and the metal debris accumulation path is obtained. The input-output response curve trend data of the real-time voltage sensor is read, the metal debris accumulation path is unfolded in time sequence, the sensor response deviation value of each point on the path at the corresponding time is extracted, the spatial overlap area of the path passing area and the response abnormal area is calculated, and the path and anomaly correlation degree distribution map is obtained. According to the path and anomaly correlation degree distribution map, the area position with the highest correlation degree is identified, the source and accumulation process of the debris at the position are traced back, and if the correlation degree exceeds a preset threshold, the area is determined as the main source of the electric field anomaly, the spatial coordinates of the abnormal source are recorded, and the source of the electric field anomaly is confirmed.

[0065] Specifically, in one embodiment, the polarization phenomenon of the metal debris in the electric field is the physical basis for extracting the electrostatic force gathering parameter. When the metal debris is in the electric field environment of the grounding box, the free electrons inside the debris will redistribute under the action of the electric field, one end of the debris will gather negative charges, and the other end will present positive charges, forming an electric dipole structure. The induced charge on the surface of the debris depends on the size, shape, and electric field strength of the position of the debris. For spherical debris, the induced charge is proportional to the cube of the radius of the debris; for sheet-shaped debris, the induced charge is mainly distributed in the edge region. By analyzing the debris density value and shape characteristics, combining the boundary conditions in the Maxwell equations, the charge density distribution on the surface of the debris can be calculated. The calculation of the electric field intensity gradient is based on the voltage distribution in the grounding box. Near the high-voltage conductor, the electric field intensity can reach several thousand volts per meter, while in the area far from the conductor, the electric field intensity decays sharply.

[0066] It is noted that the mechanism of the vibration event affecting the movement of the metal debris involves the interaction of mechanical vibration and electrostatic force. The grounding box will vibrate due to the vibration of the transformer core, the operation of the cooling fan or external mechanical impact during operation. These vibrations are conducted through the box, causing the originally stationary metal debris to gain kinetic energy. There is an empirical relationship between the vibration intensity and the displacement speed of the debris: for every 1g increase in vibration acceleration, the average displacement speed of the debris increases by about 0.1 meters per second. The directionality of the vibration also affects the movement path of the debris, and vertical vibration mainly causes jumping motion of the debris, while horizontal vibration leads to sliding of the debris. When the debris moves under the action of vibration, the electrostatic force acting on the debris changes with the position, and the vector superposition of the electrostatic force and the vibration force determines the actual motion trajectory of the debris.

[0067] Illustratively, the formation process of the metal debris accumulation path can be described as a time sequence of position changes. At the initial moment, the debris can be located in a corner at the bottom of the grounding box, and at the first vibration event, the debris gains an initial speed and starts to move. As the debris approaches the high electric field region, the electrostatic force begins to dominate its motion direction, causing the debris to tend to accumulate in the position where the electric field gradient is the largest. After many vibration events, the movement path of the debris presents a certain pattern: gradually migrating from the low electric field region to the high electric field region, and finally forming an accumulation near the sensor.

[0068] Preferably, the superposition processing of the path and the response curve adopts a space-time matching method. At each time point, the position of the metal debris in space corresponds to a coordinate, and the response deviation value of the sensor at the same time reflects the influence degree of the debris on the measurement. By projecting the spatial coordinates of each point on the path to the sensing area of the sensor, the distance between the projection point and the center of the sensor is calculated, and the closer the distance, the greater the influence weight.

[0069] In one possible implementation, the spatial overlap area calculation between the path passing area and the response abnormal area is based on a geometric intersection algorithm. The path is regarded as a series of continuous line segments, and the response abnormal area is regarded as a circular or elliptical area centered on the sensor. By calculating the intersection length of the line segment and the area, a quantitative indicator of the overlap degree is obtained.

[0070] Specifically, the calculation of the correlation degree takes into account both the spatial overlap area and the time consistency. The larger the spatial overlap area, the stronger the spatial correlation between the debris path and the abnormal area; the time consistency is evaluated by comparing the time difference between the time when the debris reaches a certain position and the time when the response abnormality occurs at that position, and the smaller the time difference, the more explicit the causal relationship. Further, the confirmation of the source of the electric field anomaly requires tracing back the entire accumulation process. Starting from the position with the highest correlation degree, the source path of the debris is deduced in reverse, and the initial position of the debris and the vibration event that triggered the movement of the debris are identified.

[0071] It can be understood that by establishing a coupling model of metal debris movement and electric field distribution, accurate positioning from macroscopic abnormal phenomena to microscopic physical processes is realized.

[0072] For example, when the abnormal source is identified to be located 5 cm directly below the sensor, the maintenance personnel can accurately locate the area that needs to be cleaned, avoiding blind disassembly of the entire grounding box, improving maintenance efficiency and reducing equipment downtime.

[0073] S106, based on the electric field abnormal source, the degree of influence of metal debris accumulation and the metal debris distribution characteristics are fused to generate a grounding box non-contact voltage sensor remote diagnosis report, the remote diagnosis report is sent to the diagnosis system, and an updated voltage sensor calibration reference is obtained.

[0074] Based on the spatial coordinates of the electric field abnormal source, the degree of influence of metal debris accumulation and the density, shape, and range data in the metal debris distribution characteristics are extracted, the abnormal position coordinates, influence degree level, debris density value, shape coefficient, and distribution radius are organized according to the diagnosis report data structure, and a grounding box non-contact voltage sensor remote diagnosis report containing fault location, influence evaluation, and debris characteristics is generated. Through the network interface of the remote transmission module, the remote diagnosis report is converted into a JSON format data stream using the TCP / IP protocol, sent to the diagnosis system server port, and the feedback calibration instruction returned by the server is received. The calibration parameter name and value are parsed from the instruction data to obtain the parameter list to be updated. It is judged whether the parameter list contains the sensor sensitivity coefficient or the zero point offset value. If it contains, the calibration configuration file of the current voltage sensor is read, the corresponding parameter value in the configuration file is replaced with the new value in the parameter list, and the updated configuration file is saved to the sensor storage unit to obtain the updated voltage sensor calibration reference.

[0075] Specifically, in one embodiment, the data structure design of the diagnosis report adopts a hierarchical organization method, and the fault information, influence evaluation, and debris characteristics are independently encapsulated. The fault location part contains the three-dimensional spatial coordinates of the abnormal source, and the coordinate system is established with the geometric center of the grounding box as the origin; the influence evaluation part maps the degree quantization value to five levels, corresponding to 0.2, 0.4, 0.6, 0.8, and 1.0 quantization values from slight to severe; and the debris characteristics part records the specific values of density, shape coefficient, and distribution radius.

[0076] It should be noted that the remote transmission module uses the TCP / IP protocol stack to realize reliable data transmission.

[0077] Specifically, the diagnostic report is first converted into JSON format, whose data structure contains four fields of timestamp, device identification, report type and report content. During transmission, a TCP connection with the diagnostic server is established, the connection state is confirmed through three-way handshake, and then the JSON data stream is sent in accordance with the maximum transmission unit fragmentation. After the server receives the complete data, it performs parsing processing and generates a feedback calibration instruction.

[0078] Exemplarily, the parameter types contained in the feedback calibration instruction mainly have two types, the sensor sensitivity coefficient is used to adjust the response amplitude of the sensor to the electric field change, and the zero point offset value is used to correct the reference output of the sensor in the absence of electric field. When the sensitivity coefficient is adjusted from 1.0 to 0.95, the output amplitude of the sensor will decrease by 5%, thereby compensating for the signal amplification effect caused by metal debris.

[0079] Preferably, the update of the configuration file adopts an atomic operation mode, the new parameters are written into a temporary file first, and the original configuration file is replaced after verifying the file integrity, thereby avoiding configuration damage caused by interruption during the update process.

[0080] In a possible implementation manner, the updated calibration reference takes effect immediately, and the sensor recalculates the output value according to the new parameters. The history of each parameter update is recorded, including the update time, the original parameter value, the new parameter value and the update reason, to form a traceable calibration log. Especially for the sensor installed in a high-voltage environment that is difficult to access, the parameter adjustment is completed through a remote mode, thereby improving the maintenance efficiency and safety.

[0081] S107, according to the updated voltage sensor calibration reference, real-time monitoring the input-output response curve change trend of the subsequent voltage sensor, reanalyzing the matching degree between the input-output response curve change trend of the current voltage sensor and the theoretical standard characteristic data of the voltage sensor, obtaining a monitoring matching result, and through continuous comparison between the monitoring matching result and the preliminary identification label of the influence of metal debris accumulation, determining the stable state of the diagnostic grounding box non-contact voltage sensor.

[0082] According to the sensitivity coefficient and zero point offset value in the updated voltage sensor calibration reference, the input and output data pairs of the voltage sensor are collected in real time at a preset sampling period, the slope change rate between adjacent sampling points is calculated, the number and position of abnormal points whose slope change rate exceeds a threshold value are recorded, and a real-time response curve change trend is obtained. The actual output value of each test point in the real-time response curve change trend is subtracted from the standard output value of the corresponding point in the theoretical standard characteristic data of the voltage sensor, and the mean and standard deviation of the difference value are calculated. When the mean is less than a preset threshold value and the standard deviation is within an allowed range, it is determined that the matching is good, otherwise it is determined that the matching is abnormal, and a monitoring matching result is obtained. The matching state identifier of the monitoring matching result is extracted and compared with a preliminary identification tag affected by metal debris accumulation. If the matching state is good and consistent with the impact level displayed by the preliminary identification tag within three consecutive monitoring periods and remains in the mild range, it is determined that the sensor is running stably, and a stable state of the non-contact voltage sensor of the grounding box is determined.

[0083] Specifically, in an embodiment, the sampling period of real-time monitoring needs to consider the running characteristics of the grounding box and the time scale of metal debris accumulation.

[0084] Specifically, the sampling period is usually set to 5 minutes, and complete input and output data pairs are collected within this interval. The slope change rate is calculated using the difference method. For two adjacent sampling points, the slope change rate is equal to the slope of the latter minus the slope of the former divided by the time interval. When the change rate exceeds the threshold value of 0.01 V / kV per minute, it is marked as an abnormal point.

[0085] It should be noted that the determination of the monitoring matching result adopts a double standard. The mean value reflects systematic deviation, and the standard deviation reflects the degree of random fluctuation. In actual application, the mean threshold value is set to 2% of the theoretical value, and the standard deviation is allowed to be 0.05 V. Only when both conditions are met, it is determined that the matching is good. This double determination mechanism can distinguish between systematic failure and random interference.

[0086] By way of example, the setting of the continuous monitoring period is based on the physical process of metal debris accumulation. The movement and aggregation of metal debris under the action of the electric field usually takes several hours to reach stability, so three monitoring periods of 15 minutes are set, which is sufficient to observe the dynamic change process of the debris distribution.

[0087] Preferably, the comparison between the preliminary identification tag and the monitoring result adopts a level matching method. The impact level in the preliminary identification tag is divided into mild, moderate and severe, corresponding to different impact degree intervals. When the impact degree displayed by the monitoring result is consistent with the tag level, and remains at the same level within three consecutive periods, it is considered that the system state is stable.

[0088] In one possible implementation, the determination of the stable state further comprises trend analysis. By calculating the trend of the matching degree in a plurality of continuous monitoring periods, if the trend line slope is close to zero, it indicates that the system has reached dynamic balance.

[0089] It can be understood that this real-time monitoring and stability determination mechanism provides a quantitative basis for preventive maintenance of the grounding box. When it is determined to be a stable state, it indicates that the current metal debris accumulation has reached a balance and will not further affect the measurement accuracy of the sensor in the short term, and cleaning can be arranged at the next planned maintenance.

[0090] It can be clearly understood by the person skilled in the art that, for the convenience and brevity of description, the specific working processes of the above-described system, modules and units can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein again. It should be understood that the protection scope of the present specification is not limited to this, and any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed in the present specification, and these modifications or replacements should be covered within the protection scope of the present specification.

Claims

1. A remote diagnostic method for a non-contact voltage sensor of a grounding box, characterized by, The method comprises the following steps: Obtaining the input-output response curve trend of the voltage sensor under different voltage levels of the grounding box, confirming the abnormal signal, and obtaining the preliminary confirmation result of the abnormality; According to the preliminary confirmation result of the abnormality, the matching degree analysis is performed on the input-output response curve trend and the theoretical standard characteristic data of the voltage sensor, the influence of the metal debris accumulation on the grounding box is identified, and the preliminary identification label of the influence of the metal debris accumulation is obtained; For the preliminary identification label of the influence of the metal debris accumulation, the response sensitivity directionality difference feature of the voltage sensor is extracted from the input-output response curve trend, the degree of the influence of the metal debris accumulation is determined, the data point spatial distribution mode in the input-output response curve trend is analyzed, the data points in the input-output response curve trend are clustered and grouped, the distribution cluster is obtained, the metal debris distribution feature is identified by comparing the center of the distribution cluster and the position corresponding to the voltage sensor, the electrostatic force accumulation parameter is extracted from the metal debris distribution feature, the metal debris accumulation path is mapped, and the metal debris accumulation path and the input-output response curve trend are superimposed to confirm the source of the electric field anomaly; Based on the source of the electric field anomaly, the degree of the influence of the metal debris accumulation and the metal debris distribution feature are fused to generate the diagnosis data and update the calibration reference of the voltage sensor; according to the updated calibration reference of the voltage sensor, the input-output response curve trend is monitored, the matching analysis is performed on the voltage sensor theoretical standard characteristic data, the monitoring matching result is obtained, and the stable state of the voltage sensor is determined by comparing the monitoring matching result with the preliminary identification label of the influence of the metal debris accumulation.

2. The ground box non-contact voltage sensor remote diagnosis method according to claim 1, characterized by, The method comprises the following steps: The test voltage is applied to the grounding box under multiple voltage levels, the output voltage value of the voltage sensor at each test point is recorded, the ratio of the output voltage difference value and the input voltage difference value of adjacent test points is calculated to obtain the slope sequence, and the input-output response curve trend is formed; for the slope sequence, the relative deviation value of adjacent slopes is calculated, the test points corresponding to the deviation value exceeding the threshold value or the negative value in the slope sequence are marked as abnormal signal points, the proportion of the abnormal signal points is counted, and the preliminary confirmation result of the abnormality is determined.

3. The ground box non-contact voltage sensor remote diagnostic method according to claim 1, characterized by, The method comprises the following steps: The input-output response data of multiple calibration records is extracted from the voltage sensor historical calibration database, the average of the slopes of each test point is calculated, and a weighted average is formed to form a historical reference slope sequence; the corresponding test point slope value sequence in the voltage sensor theoretical standard characteristic data is obtained, the difference between the slope value in the input-output response curve change trend and the historical reference slope sequence is calculated, and a deviation vector is obtained; using a support vector machine algorithm, the deviation vector is used as a characteristic input, a two-classification model is trained, the distance of the deviation vector to the classification hyperplane is calculated, the influence of metal debris accumulation is determined, and a preliminary identification label of the influence of metal debris accumulation is generated.

4. The ground box non-contact voltage sensor remote diagnostic method according to claim 1, characterized by, The preliminary identification label of the influence of metal debris accumulation extracts the response sensitivity directionality difference features of the voltage sensor from the input-output response curve change trend, and determines the degree of the influence of the metal debris accumulation, including: The slope values of the test points are extracted from the input-output response curve change trend, the slope difference values of adjacent test points are calculated, the directionality difference feature vector is constructed, principal component analysis is performed on the directionality difference feature vector, the principal components are extracted, and the dimensionality reduction feature representation is obtained by projecting to a low-dimensional space; according to the numerical distribution of the dimensionality reduction feature representation, in combination with the intensity level in the preliminary identification label of the influence of metal debris accumulation, a comprehensive influence index is calculated, and the degree of the influence of the metal debris accumulation is determined.

5. The ground box non-contact voltage sensor remote diagnostic method of claim 1, wherein, The data point space distribution mode in the input-output response curve change trend is analyzed, the data points in the input-output response curve change trend are clustered and grouped to obtain distribution clusters, and the metal debris distribution features are identified by comparing the distribution cluster centers and the corresponding relationship between the voltage sensor positions, including: The data point voltage values and output deviation values are extracted from the input-output response curve change trend to construct a two-dimensional feature space; the data points in the two-dimensional feature space are K-means clustered to determine the cluster centers and data point sets, and the distribution clusters are obtained; the voltage distribution interval of the distribution cluster is extracted, associated with the voltage sensor position, and a cluster-sensor mapping relationship is generated; according to the mapping relationship, the data point density, shape feature and distribution range in the cluster are calculated, and the metal debris distribution features are identified.

6. The ground box non-contact voltage sensor remote diagnostic method of claim 1, wherein, The electrostatic force aggregation parameters are extracted from the metal debris distribution features, the metal debris accumulation path is mapped, the metal debris accumulation path and the input-output response curve change trend are superimposed, and the source of the electric field anomaly is confirmed, including: The induced charge amount and the electric field intensity gradient are calculated from the metal debris distribution features to extract the electrostatic force aggregation parameters; vibration event records are obtained, the debris movement direction and distance are calculated, the electrostatic force aggregation parameters are combined to generate the metal debris accumulation path; the metal debris accumulation path and the input-output response curve change trend are superimposed in time sequence, the overlapping area of the path and the response abnormal area is calculated, the area with the highest correlation degree is identified, the spatial coordinates of the area with the highest correlation degree are recorded, and the source of the electric field anomaly is confirmed.

7. The ground box non-contact voltage sensor remote diagnostic method of claim 1, wherein, The diagnostic data is generated based on the electric field anomaly source, the degree of influence of the metal debris accumulation and the metal debris distribution characteristics, and the voltage sensor calibration benchmark is updated, including: Based on the electric field anomaly source coordinates, the degree of influence of the metal debris accumulation and the metal debris distribution characteristic data are extracted and organized into diagnostic data; the diagnostic data is converted into a data stream by a remote transmission module and sent to a diagnostic system, receives feedback calibration instructions, analyzes calibration parameters, updates voltage sensor calibration configuration files, and obtains the updated voltage sensor calibration benchmark.

8. The ground box non-contact voltage sensor remote diagnostic method of claim 1, wherein, According to the updated voltage sensor calibration benchmark, the input-output response curve change trend is monitored, matched with the voltage sensor theoretical standard characteristic data, and the monitoring matching result is obtained, and the voltage sensor stable state is determined by comparing the monitoring matching result with the preliminary identification label of the metal debris accumulation influence, including: According to the updated voltage sensor calibration benchmark, the input-output data of the voltage sensor is collected, the slope change rate is calculated, and the abnormal points are recorded; the input-output response curve change trend is compared with the voltage sensor theoretical standard characteristic data, the difference mean and standard deviation are counted, and the matching state is determined; the matching state is compared with the preliminary identification label of the metal debris accumulation influence, the consistency of the matching state and the influence level in the continuous period is determined, and the voltage sensor stable state is determined.

Images