A power transformation equipment maintenance platform intelligent early warning signal identification method

Abstract

The application relates to an intelligent early warning signal identification method for a power transformation equipment maintenance platform, and belongs to the technical field of power system maintenance; in view of the poor contact and chain wear problems occurring when a disconnecting switch is closed, a strain sensor array and a pressure distribution monitor are deployed to collect deformation, contact surface pressure and voltage data, form an initial data set and perform pretreatment, evaluate pressure uniformity and voltage offset, group data by adopting clustering analysis, determine wear grade classification, identify potential poor contact hidden danger areas, further extract deformation and voltage abnormality features, combine distortion type identification and linear regression analysis of pressure change trend, judge abnormal contact change trend, trigger maintenance early warning if the change slope exceeds a threshold value, and update a historical database to support subsequent comparison; the application can accurately identify contact state abnormalities and give early warning through multi-dimensional data fusion and dynamic analysis, and effectively improve the operation reliability and maintenance efficiency of the disconnecting switch.

Description

Technical Field

[0001] This invention relates to an intelligent early warning signal identification method for a power equipment maintenance platform, belonging to the field of power system maintenance technology. Background Technology

[0002] Maintenance of power equipment is an important guarantee for the stable operation of the power system. Its core lies in timely detection and handling of potential faults in equipment operation to ensure the safety and efficiency of the transmission network.

[0003] With the growth of electricity demand, substation equipment operates under high loads for extended periods, leading to an increase in potential faults. This is particularly true in critical components such as disconnectors, where abnormalities in the contact state between the drive chain and the disconnector directly affect the reliability of power transmission. However, current technologies often overlook the interaction between drive chain wear and abnormal contact pressure on the disconnector, resulting in misjudgments or missed fault signals.

[0004] For example, traditional monitoring only focuses on the physical wear of the chain, failing to fully consider its dynamic impact on the contact pressure distribution, and cannot analyze the correlation between uneven pressure and sensor output signal distortion in real time. This one-sidedness results in insufficient accuracy in identifying potential faults, making it difficult to meet the real-time and accurate requirements for maintenance early warning in complex operating environments. The core technical challenge lies in the signal distortion problem caused by the interaction between drive chain wear and abnormal contact pressure of the switch. Chain wear leads to uneven force on the contact surface when the switch is closed, and uneven pressure distribution interferes with the voltage waveform output by the strain sensor, causing distortion and thus affecting the accurate judgment of the contact state.

[0005] The core of this problem lies in the technical attribute of uneven pressure distribution, as it is both a direct consequence of chain wear and a contributing factor to signal distortion. Uneven pressure distribution stems not only from the physical wear of the chain but is also influenced by a combination of factors such as dynamic load and vibration during equipment operation, making it difficult for sensors to accurately distinguish between normal and abnormal signals.

[0006] Taking a real-world business scenario as an example, during the operation of a disconnector switch, chain wear may lead to insufficient pressure at some contact points when the switch is closed, causing localized overheating or arcing risks. In this situation, strain sensors may output distorted waveforms due to uneven pressure, misjudging the contact state as normal and failing to trigger a timely warning. Similar problems are particularly prominent in substations operating under high loads, as the interaction between chain wear and abnormal pressure is amplified, increasing the concealment of potential faults.

[0007] Therefore, accurately capturing the dynamic correlation between transmission chain wear and uneven distribution of contact pressure of the disconnector, analyzing the impact of uneven pressure on the distortion of sensor output waveform in real time, and identifying potential contact defects accordingly has become a key issue in the intelligent early warning signal identification of power equipment. Summary of the Invention

[0008] Based on the problems described in the background, the problem to be solved by the present invention is to provide an intelligent early warning signal recognition method for a power equipment maintenance platform, so as to solve the problems mentioned above.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent early warning signal identification in a power equipment maintenance platform, comprising the following steps: (1) Collect deformation data, contact surface pressure data and sensor output voltage data when the switch is closed to generate an initial dataset; (2) Clean and preprocess the initial data set, calculate the ratio of the standard deviation of the pressure data of each measuring point to the average pressure value, obtain the pressure uniformity, extract the peak offset and amplitude change of the sensor output voltage data, and obtain the voltage offset. (3) Perform cluster analysis on the pressure uniformity and the voltage offset to generate a set of cluster center points, and use the combined value of pressure uniformity and voltage offset of the set of cluster center points to represent the wear level classification; (4) Select a wear data set representing the wear degree of the chain from the cluster center point set, calculate the deviation of the wear data set from the preset wear benchmark, and mark the potential contact failure area; (5) Extract deformation data and voltage data for the corresponding time for the area with potential poor contact, and identify the distortion type identifier, which indicates the abnormal contact type; (6) Statistically analyze the frequency of occurrence of the distortion type identifier and the corresponding pressure uniformity value, generate a corresponding relationship table, analyze the linear regression slope of pressure change in the corresponding relationship table, and determine the trend of abnormal contact change; (7) Based on the slope of the abnormal contact change trend compared with the preset threshold, a maintenance early warning signal is triggered.

[0010] Preferably, step (1) includes the following steps: (1.1) Collect vibration frequency data of the disconnector switch drive chain during its operating cycle, calculate the deviation rate between the vibration frequency data and the historical average vibration frequency, adjust the sampling frequency of the sensor array according to the deviation rate, record the three-dimensional coordinate position and sensor identification number of each measuring point, and generate a sensor deployment configuration table; (1.2) Data acquisition is triggered according to the sampling frequency in the sensor deployment configuration table. The original electrical signal of deformation is obtained from the sensor array and converted into deformation data. The contact surface pressure data is obtained from the monitor and the output voltage data is obtained from the signal conditioning circuit. The deformation data, contact surface pressure data and output voltage data are aligned and combined according to a unified timestamp to generate an initial acquisition dataset.

[0011] Preferably, step (2) includes the following steps: (2.1) Calculate the difference rate between the pressure data of each measuring point and the average pressure of adjacent measuring points from the initial collected dataset, mark the abnormal data points, replace the abnormal data points with linear interpolation, calculate the maximum, minimum and average values ​​of the corrected pressure data, and generate a pressure statistical feature set; (2.2) Calculate the standard deviation of the pressure data based on the average value of the pressure statistical feature set, and divide the standard deviation of the pressure data by the average value to obtain the pressure uniformity coefficient; (2.3) Extract the difference between the maximum voltage value and the initial voltage value within the sampling period from the output voltage data of the sensor to obtain the peak offset, analyze the variation law of the peak offset, determine the offset feature type, and combine the offset feature type with the pressure uniformity coefficient to generate the voltage offset.

[0012] Preferably, step (3) includes the following steps: (3.1) Construct a two-dimensional data space with the pressure uniformity as the abscissa and the voltage offset as the ordinate. Use a clustering algorithm to iteratively process the data points in the two-dimensional data space, assign them to the nearest group, calculate the average value of the data points in each group as the center point coordinates, and generate a cluster center point set. (3.2) Extract the average pressure uniformity and average voltage offset of each center point from the cluster center point set, combine them to form center point coordinates, sort them according to the average pressure uniformity, mark the wear level as mild, moderate and severe, and generate wear level classification.

[0013] Preferably, step (5) includes the following steps: (5.1) Extract the deformation time series data and voltage time series data of the potential poor contact area, calculate the amplitude value of the deformation time series data, calculate the cumulative sum of the differences between adjacent sampling points of the voltage time series data, and obtain the cumulative value of voltage fluctuation; (5.2) Based on the comparison between the deformation amplitude value and the preset threshold, mark the overload or underload characteristics, query the correspondence between the feature combination and the abnormal contact type, and generate a distortion type identifier, wherein the distortion type identifier represents point contact abnormality, line contact abnormality or surface contact abnormality.

[0014] Preferably, step (6) includes the following steps: (6.1) After identifying the distortion type identifier, count the number of times the distortion type identifier appears, extract the pressure uniformity value at the corresponding time, and generate a corresponding relationship table; (6.2) Extract the pressure uniformity sequence from the corresponding table, fit the linear regression equation, calculate the slope coefficient, and determine the trend of abnormal contact changes.

[0015] Preferably, step (7) includes the following steps: (7.1) After determining the abnormal contact change trend, compare the slope of the abnormal contact change trend with a preset threshold, generate a maintenance early warning signal containing the slope value and equipment identifier, and send it to the alarm receiving end. (7.2) Add timestamps to the corresponding relationship table and write it into the historical database.

[0016] The beneficial effects of this invention are: 1. By integrating multi-source data, innovating feature indicators, intelligent clustering analysis, and trend prediction models, a complete technical closed loop from state perception and accurate diagnosis to prediction and early warning has been constructed. This has greatly improved the accuracy of substation condition monitoring and the ability to detect early faults, realizing the transformation from passive maintenance to proactive predictive maintenance, and ultimately achieving the goals of ensuring power grid safety, optimizing resource allocation, and reducing total operation and maintenance costs.

[0017] 2. To address potential contact problems and chain wear issues during disconnection switch closure, a strain sensor array and pressure distribution monitor were deployed to collect deformation, contact surface pressure, and voltage data. This formed an initial dataset, which was then pre-processed and cleaned. Subsequently, pressure uniformity and voltage deviation were assessed. Cluster analysis was used to group the data, determine wear level classifications, and identify potential areas with poor contact.

[0018] 3. For areas with potential hazards, this invention further extracts deformation and voltage anomaly characteristics, and combines linear regression analysis of distortion type identification and pressure change trend to determine the trend of abnormal contact changes. If the slope of change exceeds the threshold, a maintenance warning is triggered, and the historical database is updated to support subsequent comparisons.

[0019] 4. Through multi-dimensional data fusion and dynamic analysis, abnormal contact status can be accurately identified and early warnings can be issued, effectively improving the operational reliability and maintenance efficiency of disconnect switches. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a diagram illustrating the process of obtaining the initial dataset. Detailed Implementation

[0021] The embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0022] Implementation, for example Figure 1 As shown, the present invention provides a method for intelligent early warning signal identification in a power equipment maintenance platform, comprising the following steps: (1) Collect deformation data, contact surface pressure data and sensor output voltage data when the switch is closed to generate an initial dataset; Specifically, by deploying a strain sensor array and a pressure distribution monitor on the disconnector switch, deformation data, contact surface pressure data, and sensor output voltage data at each measuring point are collected when the disconnector switch is closed, an initial dataset is obtained. Step (1) includes the following steps: (1.1) Collect vibration frequency data of the disconnector switch drive chain during its operating cycle, calculate the deviation rate between the vibration frequency data and the historical average vibration frequency, adjust the sampling frequency of the sensor array according to the deviation rate, record the three-dimensional coordinate position and sensor identification number of each measuring point, and generate a sensor deployment configuration table; Specifically, the vibration frequency data of the disconnector switch drive chain during its operating cycle is obtained, and the deviation rate between the vibration frequency and the average vibration frequency during historical normal operation is calculated. If the deviation rate exceeds a preset deviation threshold, the sampling frequency of the strain sensor array is adjusted to a preset multiple of the current vibration frequency. At the same time, the three-dimensional coordinate position of each measuring point on the disconnector switch and the corresponding sensor identification number and its sensitivity parameters are recorded to obtain a sensor deployment configuration table.

[0023] The vibration frequency data is obtained by installing an acceleration sensor on the support bearing seat of the disconnector switch drive chain. The acceleration sensor continuously monitors the vibration signal generated by the chain operation at a basic sampling rate of 1 kHz.

[0024] The time-domain vibration signal is converted into a frequency-domain signal using a Fast Fourier Transform (FFT), and the dominant frequency component is extracted as the current vibration frequency. The dominant frequency component corresponds to the frequency component with the largest amplitude in the spectrum. The average vibration frequency during historical normal operation is obtained by statistically analyzing the vibration frequency data of the first 100 normal operations after the disconnector switch was put into operation. The deviation rate is calculated by dividing the difference between the current vibration frequency and the average value by the average value.

[0025] For example, when the chain becomes loose due to wear, the vibration frequency decreases and the deviation rate becomes negative; when the chain is too tight, the vibration frequency increases and the deviation rate becomes positive.

[0026] (1.2) Data acquisition is triggered according to the sampling frequency in the sensor deployment configuration table. The original electrical signal of deformation is obtained from the sensor array and converted into deformation data. The contact surface pressure data is obtained from the monitor and the output voltage data is obtained from the signal conditioning circuit. The deformation data, contact surface pressure data and output voltage data are aligned and combined according to a unified timestamp to generate an initial acquisition dataset.

[0027] Specifically, at the instant the switch closes, data acquisition is triggered according to the sampling frequency in the sensor deployment configuration table. The raw electrical signals of deformation are obtained from the strain sensors corresponding to each measuring point. The raw electrical signals are calibrated and converted according to the sensitivity parameters to obtain deformation data. At the same time, the pressure values ​​of the contact surfaces of each measuring point are obtained from the pressure distribution monitor, and the output voltage data is obtained from the sensor signal conditioning circuit. The deformation data, pressure values ​​and output voltage data are aligned and combined according to a unified timestamp to obtain the initial acquisition dataset.

[0028] The sensor deployment configuration table records the specific installation position of each strain sensor on the contact surface of the disconnector switch, including the coordinate values ​​of three dimensions: radial distance from the switch axis, circumferential angle, and axial height.

[0029] Sensitivity parameters are obtained through pre-calibration, that is, by measuring the sensor output voltage under known pressure conditions and establishing a linear relationship coefficient between pressure and voltage.

[0030] like Figure 2 The figure shown is a schematic diagram of obtaining the initial collection dataset in step (1).

[0031] (2) Clean and preprocess the initial data set, calculate the ratio of the standard deviation of the pressure data of each measuring point to the average pressure value, obtain the pressure uniformity, extract the peak offset and amplitude change of the sensor output voltage data, and obtain the voltage offset. Specifically, the initial dataset is cleaned and preprocessed. The standard deviation of the pressure data is obtained by identifying the dispersion of the pressure data at each measuring point. The pressure uniformity is determined by evaluating the ratio of the standard deviation of the pressure data to the average pressure value. At the same time, the peak offset and amplitude change of the sensor output voltage data are extracted to obtain the voltage offset.

[0032] Step (2) includes the following steps: (2.1) Calculate the difference rate between the pressure data of each measuring point and the average pressure of adjacent measuring points from the initial collected dataset, mark the abnormal data points, replace the abnormal data points with linear interpolation, calculate the maximum, minimum and average values ​​of the corrected pressure data, and generate a pressure statistical feature set; Specifically, the difference rate between the pressure data of each measuring point and the average pressure of adjacent measuring points is identified from the initial collected dataset. If the difference rate exceeds the preset difference threshold, it is marked as an abnormal data point. The abnormal data point is replaced by the average pressure value of two adjacent normal measuring points using a linear interpolation method. At the same time, the maximum value, minimum value and arithmetic mean of the pressure of each measuring point after correction are extracted to obtain the pressure statistical feature set.

[0033] The identification of abnormal data points is achieved using a neighborhood comparison-based method.

[0034] For each measuring point on the contact surface of the disconnector switch, the pressure values ​​of its two adjacent measuring points are obtained and the average value is calculated. When the difference between the pressure value of the target measuring point and this average value exceeds 30% of the average value, it is determined to be an abnormal data point. During the linear interpolation correction process, the arithmetic mean of the pressure values ​​of adjacent normal measuring points is directly used to replace the abnormal value, avoiding interference from abnormal data on subsequent pressure uniformity calculations. The construction of the pressure statistical feature set involves statistical processing of the pressure data of all measuring points after correction. In the actual substation operating environment, 8 to 12 pressure monitoring points are usually arranged on the contact surface of the disconnector switch, and these measuring points are distributed in a ring or rectangular array. By extracting the maximum, minimum, and arithmetic mean of the pressure at each measuring point, the overall characteristics of the pressure distribution on the contact surface can be reflected. The arithmetic mean serves as the center value of the pressure distribution, providing a benchmark reference for subsequent standard deviation calculations. When wear occurs in the disconnector switch drive chain, it will cause the pressure in some areas to be higher and the pressure in other areas to be lower during the disconnector closing process. This non-uniformity is directly reflected in the numerical differences of the pressure statistical feature set. The pressure uniformity coefficient is calculated using the coefficient of variation method, that is, the ratio of the standard deviation to the mean. This coefficient can eliminate the influence of the absolute value of pressure, making the pressure uniformity under different working conditions comparable.

[0035] (2.2) Calculate the standard deviation of the pressure data based on the average value of the pressure statistical feature set, and divide the standard deviation of the pressure data by the average value to obtain the pressure uniformity coefficient; Specifically, based on the arithmetic mean of the pressure statistical feature set, the sum of squares of the deviations between the pressure value at each measuring point and the average value is calculated. The standard deviation of the pressure data is obtained by taking the square root of the sum of squares of the deviations divided by the total number of measuring points. The pressure uniformity coefficient is obtained by dividing the standard deviation by the arithmetic mean. At the same time, the difference between the maximum voltage value within the sampling period and the voltage value at the initial moment is extracted from the sensor output voltage data as the peak offset.

[0036] In both rated load and light load conditions, although the absolute values ​​of the contact surface pressure differ significantly, the pressure uniformity coefficient can uniformly evaluate the uniformity of the pressure distribution. The determination of the offset characteristic type is based on the offset change rate data at five consecutive sampling times. Gradual offset characteristics are characterized by a continuously positive offset change rate that gradually increases, typically corresponding to a slow decrease in contact pressure caused by the chain gradually loosening. Oscillating offset characteristics are characterized by a periodic change in the offset change rate between positive and negative values, reflecting the impact of mechanical vibration during equipment operation.

[0037] (2.3) Extract the difference between the maximum voltage value and the initial voltage value within the sampling period from the output voltage data of the sensor to obtain the peak offset, analyze the variation law of the peak offset, determine the offset feature type, and combine the offset feature type with the pressure uniformity coefficient to generate the voltage offset.

[0038] Specifically, the variation pattern of the peak offset at continuous sampling times is analyzed, and the difference between the peak offsets at adjacent times is calculated to obtain the offset change rate. If the offset change rate increases monotonically, it is determined to be a gradual offset feature. If the offset change rate alternates periodically with positive and negative values, it is determined to be an oscillating offset feature. The voltage offset is obtained by multiplying the offset feature type code value with the pressure uniformity coefficient.

[0039] The voltage offset is determined by multiplying the offset feature type code value and the pressure uniformity coefficient. The gradual offset is assigned a code value of 1.2, and the oscillating offset is assigned a code value of 0.8. This differentiated processing makes the voltage offset under different fault modes distinguishable.

[0040] (3) Perform cluster analysis on the pressure uniformity and the voltage offset to generate a set of cluster center points, and use the combined value of pressure uniformity and voltage offset of the set of cluster center points to represent the wear level classification; Specifically, cluster analysis is performed on pressure uniformity and voltage offset. The data are grouped according to the range of pressure uniformity values ​​and the range of voltage offset. The set of cluster center points is obtained by identifying the center values ​​of each group. The combined values ​​of pressure uniformity and voltage offset of the cluster center points are used to represent different wear levels.

[0041] Step (3) includes the following steps: (3.1) Construct a two-dimensional data space with the pressure uniformity as the abscissa and the voltage offset as the ordinate. Use a clustering algorithm to iteratively process the data points in the two-dimensional data space, assign them to the nearest group, calculate the average value of the data points in each group as the center point coordinates, and generate a cluster center point set. Specifically, a two-dimensional data space is constructed using pressure uniformity as the x-axis and voltage offset as the y-axis. The K-means clustering algorithm is used to iteratively process the data points in this two-dimensional data space. The Euclidean distance from each data point to each candidate center is calculated, and the data point is assigned to the nearest group. The arithmetic mean of the coordinates of all data points within each group is recalculated as the new center position. Iteration stops when the change in center position between two consecutive iterations is less than a preset convergence threshold, resulting in stable data groups. The arithmetic mean of all pressure uniformity values ​​and all voltage offset values ​​within each group is calculated. The two averages are combined to form the center point coordinates of that group. The center point coordinates of all groups are collected to obtain a set of cluster center points.

[0042] The construction of the two-dimensional data space uses pressure uniformity and voltage offset as two independent dimensions to form a visualized scatter plot. During the long-term operation of substation equipment, disconnectors with different wear levels exhibit obvious clustering characteristics in this two-dimensional space. Data points from normal equipment are concentrated in areas with high pressure uniformity and small voltage offset, while data points from severely worn equipment are distributed in areas with low pressure uniformity and large voltage offset. In this embodiment, the K-means clustering algorithm sets the initial number of clusters to 3, corresponding to three wear levels: mild, moderate, and severe. During algorithm execution, three data points are randomly selected as initial centers. Then, the Euclidean distance from all data points to these three centers is calculated. The Euclidean distance is calculated by taking the square root of the sum of the squares of the pressure uniformity difference and the squares of the voltage offset difference. After each data point is assigned to the nearest group, the center position of each group is recalculated, which is the arithmetic mean of the coordinates of all data points within the group. When the change in center position between two consecutive iterations is less than 0.01, the algorithm is considered to have converged, and the group division tends to be stable. This iterative optimization process typically converges within 5 to 10 iterations, and the resulting clustering effectively distinguishes equipment data under different wear conditions. The formation of the cluster centroid set involves statistical processing of the data within each group. The average pressure uniformity and voltage offset are calculated for each group, and these two averages are combined to form the coordinates of the representative centroid of that group, reflecting the typical characteristics of that wear level. The wear level is classified using a single-dimensional determination method based on pressure uniformity. A first preset threshold is set to 0.3, and a second preset threshold is set to 0.6. When the average pressure uniformity of the cluster centroids is less than 0.3, it indicates a relatively uniform pressure distribution and mild chain wear; when the average pressure uniformity is greater than 0.6, it indicates a severely uneven pressure distribution and chain wear that requires immediate repair.

[0043] (3.2) Extract the average pressure uniformity and average voltage offset of each center point from the cluster center point set, combine them to form center point coordinates, sort them according to the average pressure uniformity, mark the wear level as mild, moderate and severe, and generate wear level classification.

[0044] Specifically, based on the average pressure uniformity of each center point in the cluster center point set, sorting them from smallest to largest, the group corresponding to the center point with an average pressure uniformity less than a first preset threshold is marked as a mild wear level, the group corresponding to the center point with an average pressure uniformity greater than a second preset threshold is marked as a severe wear level, and the remaining groups are marked as a moderate wear level. The different wear levels are represented by the combination of pressure uniformity and voltage offset of each center point.

[0045] The combination of pressure uniformity and voltage offset provides a more accurate characterization of wear level. Typical combinations for mild wear are: pressure uniformity 0.2 with a voltage offset of 10mV; moderate wear is represented by pressure uniformity 0.45 with a voltage offset of 25mV; and severe wear is represented by pressure uniformity 0.75 with a voltage offset of 40mV or higher. This two-dimensional combination improves the accuracy of wear level determination.

[0046] (4) Select a wear data set representing the wear degree of the chain from the cluster center point set, calculate the deviation of the wear data set from the preset wear benchmark, and mark the potential contact failure area; Specifically, center points with pressure uniformity values ​​greater than a preset threshold are selected from the cluster center point set as wear data groups. The pressure uniformity value and voltage offset value of the wear data group are read. The difference between these two values ​​is calculated by subtracting them from the standard pressure uniformity value and standard voltage offset value in the preset standard wear benchmark. The square root of the sum of the squares of these two differences is then used to obtain the wear deviation value. This wear deviation value is compared with a preset wear threshold. If the wear deviation value exceeds the preset wear threshold, the contact surface area of ​​the disconnector corresponding to the original data point of that wear data group is marked as a potential contact failure area.

[0047] The selection of wear data groups is based on the magnitude of pressure uniformity values. A preset selection threshold is set to 0.5. When the pressure uniformity value at the cluster center is greater than 0.5, it indicates that the disconnector corresponding to that data group exhibits significant pressure unevenness, and chain wear has affected the normal contact state of the disconnector. The preset standard wear benchmark is obtained through statistical analysis of historical data from newly installed or normally operating equipment. The standard pressure uniformity value is typically set to 0.15, representing an ideal state of relatively uniform pressure distribution; the standard voltage deviation value is set to 5mV, representing the reference output of the sensor under normal contact pressure. The wear deviation value is calculated using the Euclidean distance method, by adding the squares of the pressure uniformity difference and the voltage deviation difference, and then taking the square root to obtain a single numerical index comprehensively reflecting the degree of wear.

[0048] The preset wear threshold is dynamically adjusted based on the equipment's operating environment and maintenance requirements. In high-load substations, the threshold setting is relatively strict, typically 0.35; under normal load conditions, the threshold can be relaxed to 0.45. When the wear deviation exceeds the threshold, the corresponding contact surface area is automatically marked as a potential fault point requiring close monitoring.

[0049] (5) Extract deformation data and voltage data for the corresponding time for the area with potential poor contact, and identify the distortion type identifier, which indicates the abnormal contact type; Specifically, deformation data and sensor output voltage data for the corresponding time are extracted for the marked potential contact failure areas. The data are then identified based on the deformation exceeding the limit and the abnormal voltage fluctuation characteristics to obtain a distortion type identifier, which indicates the abnormal contact type.

[0050] Step (5) includes the following steps: (5.1) Extract the deformation time series data and voltage time series data of the potential poor contact area, calculate the amplitude value of the deformation time series data, calculate the cumulative sum of the differences between adjacent sampling points of the voltage time series data, and obtain the cumulative value of voltage fluctuation; Specifically, the deformation time series data and sensor output voltage time series data corresponding to the marked potential poor contact areas are extracted from the historical database. The difference between the maximum and minimum values ​​of the deformation data within the sampling period is calculated to obtain the deformation amplitude value. At the same time, the absolute values ​​of the differences between adjacent sampling points in the voltage data are accumulated to obtain the voltage fluctuation cumulative value.

[0051] The historical database stores monitoring data from various measuring points of the disconnector switch at different times, with a data acquisition frequency of 100 times per second. The deformation time-series data reflects the degree of mechanical deformation of the disconnector contact surface during the closing process. Under normal circumstances, the deformation fluctuates within a small range. When chain wear causes abnormal contact pressure, the deformation will show obvious periodic changes or continuous deviations. The cumulative voltage fluctuation value is obtained by accumulating the absolute value of the voltage difference between adjacent sampling points. This value can quantitatively reflect the stability of the sensor output signal. When there is poor contact, the voltage signal will frequently jump, leading to a significant increase in the cumulative value. The deformation amplitude value is calculated using the difference between the maximum and minimum values, directly reflecting the dynamic range of deformation. The preset upper limit threshold for deformation is determined based on the allowable deformation under the rated load conditions of the equipment, usually set to 1.5 times the normal deformation value; the preset lower limit threshold for deformation is set to 0.5 times the normal deformation value. When the deformation amplitude value exceeds the upper limit threshold, it indicates that the contact pressure is too high, causing excessive material deformation; when it is below the lower limit threshold, it indicates insufficient contact pressure, which may pose a risk of contact surface separation.

[0052] (5.2) Based on the comparison between the deformation amplitude value and the preset threshold, mark the overload or underload characteristics, query the correspondence between the feature combination and the abnormal contact type, and generate a distortion type identifier, wherein the distortion type identifier represents point contact abnormality, line contact abnormality or surface contact abnormality.

[0053] Specifically, the deformation amplitude value is compared with a preset upper deformation threshold. If it exceeds the upper threshold, it is marked as an overload characteristic; if it is below the preset lower deformation threshold, it is marked as an underload characteristic. The fluctuation level is determined based on the ratio of the cumulative voltage fluctuation value to a preset fluctuation reference value, where a ratio less than a first preset ratio indicates a low fluctuation level, and a ratio greater than a second preset ratio indicates a high fluctuation level. Based on the combination of the overload characteristic and the fluctuation level, or the combination of the underload characteristic and the fluctuation level, a pre-established correspondence between feature combinations and contact anomaly types is queried to determine the distortion type identifier. The distortion type identifier uses different numerical codes to represent three abnormal contact types: point contact anomaly, line contact anomaly, or surface contact anomaly.

[0054] The fluctuation level is determined based on the ratio of the cumulative voltage fluctuation value to a preset fluctuation benchmark value. The preset fluctuation benchmark value is obtained by statistically analyzing voltage fluctuation characteristics under normal operating conditions. The first preset ratio is set to 1.2, and the second preset ratio is set to 2.0. A low fluctuation level corresponds to a ratio less than 1.2, indicating a relatively stable voltage signal; a high fluctuation level corresponds to a ratio greater than 2.0, indicating severe signal distortion; and levels in between are considered medium fluctuation levels.

[0055] Meanwhile, the correspondence between characteristic combinations and contact anomaly types was established through statistical analysis of a large number of fault cases. Point contact anomalies are usually characterized by overload characteristics combined with high fluctuation levels. In this case, the reduced contact area leads to local pressure concentration, causing severe voltage fluctuations. Line contact anomalies are characterized by normal load combined with medium fluctuation levels, with the contact surface showing a linear distribution. Surface contact anomalies are characterized by underload characteristics combined with low fluctuation levels. Although the contact area is large, the pressure distribution is sparse.

[0056] The distortion type identifier uses a three-digit code: the hundreds digit indicates the type of deformation characteristic, the tens digit indicates the fluctuation level, and the units digit indicates the specific abnormal pattern. This coding method facilitates the rapid location and classification of faults.

[0057] (6) Statistically analyze the frequency of occurrence of the distortion type identifier and the corresponding pressure uniformity value, generate a corresponding relationship table, analyze the linear regression slope of pressure change in the corresponding relationship table, and determine the trend of abnormal contact change; Specifically, a correlation table is obtained by statistically analyzing the frequency of occurrence of distortion type identifiers and corresponding pressure uniformity values. By analyzing the linear regression slope of the pressure change trend in the correlation table, the abnormal contact change trend is determined.

[0058] Step (6) includes the following steps: (6.1) After identifying the distortion type identifier, count the number of times the distortion type identifier appears, extract the pressure uniformity value at the corresponding time, and generate a corresponding relationship table; Specifically, the frequency of occurrence of each distortion type identifier within a preset time period is counted, and the pressure uniformity value corresponding to each identifier at that time is extracted. The distortion type identifier is then paired with its corresponding average pressure uniformity value and arranged in chronological order to form a corresponding relationship table.

[0059] The construction of the correspondence table is based on statistical analysis of long-term monitoring data.

[0060] The preset time period is typically set to a continuous monitoring cycle of 30 days, during which the system records hundreds of operation data points of the disconnect switches daily. Whenever a distortion type indicator appears, the system simultaneously records the pressure uniformity value at that moment. By calculating the arithmetic mean of all pressure uniformity values ​​corresponding to the same distortion type indicator, the typical pressure characteristic value for that type of distortion is obtained. The distortion type indicators at different time points and their corresponding average pressure uniformity values ​​are arranged chronologically to form a time-series correspondence table, which visually reflects the evolution of equipment wear status over time.

[0061] (6.2) Extract the pressure uniformity sequence from the corresponding table, fit the linear regression equation, calculate the slope coefficient, and determine the trend of abnormal contact changes.

[0062] Specifically, the average pressure uniformity sequence arranged in chronological order is extracted from the correspondence table. Using the time sequence number as the independent variable and the average pressure uniformity as the dependent variable, a linear regression equation is fitted using the least squares method. The slope coefficient of the linear regression equation is calculated as the pressure change rate. The abnormal contact change trend is determined based on the sign and absolute value of the pressure change rate. If the pressure change rate is positive and the absolute value is greater than a preset deterioration threshold, it is determined to be a deterioration trend; if the pressure change rate is negative and the absolute value is greater than the preset deterioration threshold, it is determined to be an improvement trend; otherwise, it is determined to be a stabilization trend.

[0063] The fitting process of the linear regression equation is achieved using the least squares method.

[0064] Using the sampling time sequence as the independent variable x and the average pressure uniformity at the corresponding time as the dependent variable y, the slope coefficient a and intercept b in the linear equation y=ax+b are obtained by minimizing the sum of squares error between the actual and fitted values. The slope coefficient a represents the pressure change rate, which physically indicates the change in pressure uniformity per unit time. A positive slope indicates that the pressure distribution is less uniform, and this non-uniformity worsens with time; a negative slope indicates that the pressure distribution tends to become more uniform.

[0065] The determination of the preset deterioration threshold is based on statistical analysis of equipment maintenance cycles and fault development rates. In actual substation operation, when the absolute value of the pressure change rate exceeds 0.01 / day, it means that the equipment condition will change significantly within a month, requiring timely maintenance measures. A deterioration trend indicates that chain wear is accelerating and poor contact problems are gradually becoming more serious; an improvement trend usually appears in the initial stage after equipment maintenance, indicating that the maintenance measures have produced positive effects; a stable trend indicates that the equipment is in the normal wear stage.

[0066] The determination of abnormal contact trends directly affects the formulation of maintenance strategies. A deteriorating trend triggers an emergency maintenance warning, an improving trend extends the maintenance cycle, and a stable trend maintains the regular inspection frequency.

[0067] Therefore, by analyzing trends over multiple consecutive monitoring cycles, it is possible to identify the accelerated and stable periods of equipment degradation, providing data support for predictive maintenance.

[0068] (7) Based on the slope of the abnormal contact change trend compared with the preset threshold, a maintenance early warning signal is triggered.

[0069] Specifically, if the slope of the abnormal contact change trend is greater than the preset threshold, a maintenance early warning signal will be triggered, and the corresponding relationship table will be updated to the historical database for benchmark comparison of subsequent data sets.

[0070] Step (7) includes the following steps: (7.1) After determining the abnormal contact change trend, compare the slope of the abnormal contact change trend with a preset threshold, generate a maintenance early warning signal containing the slope value and equipment identifier, and send it to the alarm receiving end. Specifically, the slope of the abnormal contact change trend is compared with a preset threshold. If the slope is greater than the preset threshold, a maintenance warning signal containing the slope value, trigger time, and device identifier is generated, and the maintenance warning signal is sent to a preset alarm receiver.

[0071] The preset threshold is determined based on the historical fault data of the disconnecting switch, and is usually set to 0.02 / day.

[0072] When the slope of the abnormal contact change trend exceeds this threshold, it indicates that the equipment deterioration rate has exceeded the normal range, requiring immediate maintenance. The maintenance early warning signal contains several key information fields, including a slope value used to assess the severity of deterioration, a trigger time recording the time node of fault development, and equipment identification pinpointing the specific disconnector switch number and installation location. Preset alarm receivers include the main system at the substation monitoring center and mobile terminals of maintenance personnel. Upon receiving the early warning signal, the main system automatically records and displays it on the monitoring screen, while the mobile terminals promptly alert on-site personnel via push notifications.

[0073] (7.2) Add timestamps to the corresponding relationship table and write it into the historical database.

[0074] Specifically, the corresponding relationship table is written to the historical database after the current timestamp is added. The timestamp is used to identify the data version. At the same time, the maintenance early warning signal is stored as an associated record for subsequent comparison and analysis of newly collected datasets with historical benchmarks.

[0075] The historical database employs a time-series structure, specifically designed to store timestamped monitoring data. When the corresponding data is written to the database, the timestamp is accurate to the millisecond, serving as a unique identifier for the data version. When new collected data requires benchmark comparison, the most recent historical records are retrieved based on the timestamp. By comparing the changes in pressure uniformity and voltage deviation, the evolution trend of the equipment's condition is determined.

Claims

1. A method for intelligent early warning signal recognition in a power equipment maintenance platform, characterized in that, Includes the following steps: (1) Collect deformation data, contact surface pressure data and sensor output voltage data when the switch is closed to generate an initial dataset; (2) Clean and preprocess the initial data set, calculate the ratio of the standard deviation of the pressure data of each measuring point to the average pressure value, obtain the pressure uniformity, extract the peak offset and amplitude change of the sensor output voltage data, and obtain the voltage offset. (3) Perform cluster analysis on the pressure uniformity and the voltage offset to generate a set of cluster center points, and use the combined value of pressure uniformity and voltage offset of the set of cluster center points to represent the wear level classification; (4) Select a wear data set representing the wear degree of the chain from the cluster center point set, calculate the deviation of the wear data set from the preset wear benchmark, and mark the potential contact failure area; (5) Extract deformation data and voltage data for the corresponding time for the potential poor contact area, identify the distortion type identifier, and the distortion type identifier indicates the abnormal contact type; (6) Statistically analyze the frequency of occurrence of the distortion type identifier and the corresponding pressure uniformity value, generate a corresponding relationship table, analyze the linear regression slope of pressure change in the corresponding relationship table, and determine the trend of abnormal contact change; (7) Based on the slope of the abnormal contact change trend compared with the preset threshold, a maintenance early warning signal is triggered.

2. The intelligent early warning signal identification method for a substation equipment maintenance platform according to claim 1, characterized in that, Step (1) includes the following steps: (1.1) Collect vibration frequency data of the disconnector switch drive chain during its operating cycle, calculate the deviation rate between the vibration frequency data and the historical average vibration frequency, adjust the sampling frequency of the sensor array according to the deviation rate, record the three-dimensional coordinate position and sensor identification number of each measuring point, and generate a sensor deployment configuration table; (1.2) Data acquisition is triggered according to the sampling frequency in the sensor deployment configuration table. The original electrical signal of deformation is obtained from the sensor array and converted into deformation data. The contact surface pressure data is obtained from the monitor and the output voltage data is obtained from the signal conditioning circuit. The deformation data, contact surface pressure data and output voltage data are aligned and combined according to a unified timestamp to generate an initial acquisition dataset.

3. The intelligent early warning signal identification method for a substation equipment maintenance platform according to claim 1, characterized in that, Step (2) includes the following steps: (2.1) Calculate the difference rate between the pressure data of each measuring point and the average pressure of adjacent measuring points from the initial collected dataset, mark the abnormal data points, replace the abnormal data points with linear interpolation, calculate the maximum, minimum and average values ​​of the corrected pressure data, and generate a pressure statistical feature set; (2.2) Calculate the standard deviation of the pressure data based on the average value of the pressure statistical feature set, and divide the standard deviation of the pressure data by the average value to obtain the pressure uniformity coefficient; (2.3) Extract the difference between the maximum voltage value and the initial voltage value within the sampling period from the output voltage data of the sensor to obtain the peak offset, analyze the variation law of the peak offset, determine the offset feature type, and combine the offset feature type with the pressure uniformity coefficient to generate the voltage offset.

4. The intelligent early warning signal identification method for a substation equipment maintenance platform according to claim 1, characterized in that, Step (3) includes the following steps: (3.1) Construct a two-dimensional data space with the pressure uniformity as the abscissa and the voltage offset as the ordinate. Use a clustering algorithm to iteratively process the data points in the two-dimensional data space, assign them to the nearest group, calculate the average value of the data points in each group as the center point coordinates, and generate a cluster center point set. (3.2) Extract the average pressure uniformity and average voltage offset of each center point from the cluster center point set, combine them to form center point coordinates, sort them according to the average pressure uniformity, mark the wear level as mild, moderate and severe, and generate wear level classification.

5. The intelligent early warning signal identification method for a substation equipment maintenance platform according to claim 1, characterized in that, Step (5) includes the following steps: (5.1) Extract the deformation time series data and voltage time series data of the potential poor contact area, calculate the amplitude value of the deformation time series data, calculate the cumulative sum of the differences between adjacent sampling points of the voltage time series data, and obtain the cumulative value of voltage fluctuation; (5.2) Based on the comparison between the deformation amplitude value and the preset threshold, mark the overload or underload characteristics, query the correspondence between the feature combination and the abnormal contact type, and generate a distortion type identifier, wherein the distortion type identifier represents point contact abnormality, line contact abnormality or surface contact abnormality.

6. The intelligent early warning signal identification method for a substation equipment maintenance platform according to claim 1, characterized in that, Step (6) includes the following steps: (6.1) After identifying the distortion type identifier, count the number of times the distortion type identifier appears, extract the pressure uniformity value at the corresponding time, and generate a corresponding relationship table; (6.2) Extract the pressure uniformity sequence from the corresponding table, fit the linear regression equation, calculate the slope coefficient, and determine the trend of abnormal contact changes.

7. The intelligent early warning signal identification method for a substation equipment maintenance platform according to claim 1, characterized in that, Step (7) includes the following steps: (7.1) After determining the abnormal contact change trend, compare the slope of the abnormal contact change trend with a preset threshold, generate a maintenance early warning signal containing the slope value and equipment identifier, and send it to the alarm receiving end. (7.2) Add timestamps to the corresponding relationship table and write it into the historical database.

Images