Intelligent monitoring method and system for high-voltage switch cabinet

Abstract

The present application belongs to the technical field of electrical equipment state monitoring, and particularly relates to a high-voltage switch cabinet intelligent monitoring method and system, comprising the following steps: collecting closing history data of each working condition of a vacuum circuit breaker in a brand-new state; calculating current mean value, power factor mean value and slope mean value of each working condition to obtain working condition reference; constructing working condition R matrix with each working condition electrical quantity variance; calculating current rising slope after each closing; comparing the current rising slope with the slope mean value of each working condition reference, taking the nearest neighbor to obtain working condition label; collecting current mean value and power factor mean value of the present steady state section; subtracting the current mean value and power factor mean value in the working condition reference corresponding to the working condition label to obtain observation deviation quantity. The present application does not need to add new sensors, and relies on existing programmable logic controller signals to realize full-cycle continuous quantitative monitoring of early deterioration of vacuum circuit breaker contacts.

Description

Technical Field

[0001] This invention belongs to the field of electrical equipment condition monitoring technology, specifically relating to an intelligent monitoring method and system for high-voltage switchgear. Background Technology

[0002] In smelting plants, vacuum circuit breakers are commonly used as the main switching equipment in the high-voltage power supply systems of electric arc furnaces and AOD refining furnaces. Vacuum circuit breakers undergo frequent opening and closing operations during electric arc furnace smelting production, accumulating 20 to 30 operations per shift. The contact resistance of the contacts slowly increases with the cumulative number of opening and closing operations. Even a slight increase in contact resistance during the early stages of contact degradation can cause abnormal deviations in the high-voltage side current and power factor. If not detected in time, the degraded contacts will experience prolonged arcing time during high-current opening and closing, which can accelerate insulation aging or even lead to contact breakdown, circuit breaker failure, and large-scale power outages, directly affecting continuous production in the electric arc furnace. Therefore, continuous online monitoring of the deterioration status of vacuum circuit breaker contacts is an urgent requirement to ensure the reliable operation of the high-voltage power supply system for electric arc furnaces.

[0003] Currently, the industry commonly uses microprocessor-based integrated relays to monitor the operating status of high-voltage switchgear. These relays collect electrical quantities such as current and voltage on the high-voltage side in real time, pre-setting fixed upper current thresholds and power factor limits. When the collected electrical quantities exceed these fixed thresholds, alarms or tripping protection are triggered. This method relies on existing electrical quantity acquisition circuits, requires no additional sensors, has low engineering implementation costs, and provides good protection in fault scenarios such as short circuits and overloads with large abnormal electrical quantity amplitudes. It is currently the mainstream method for monitoring the status of high-voltage switchgear.

[0004] However, the aforementioned microcomputer-based integrated relay monitoring method, based on fixed thresholds, cannot cope with the complex scenario of frequent switching between multiple operating conditions in the high-voltage power supply system of electric arc furnaces. During the electric arc furnace smelting process, three power supply modes—frequency conversion, bypass, and soft start—operate alternately. The normal reference values ​​of current and power factor can differ by several times between different modes. However, the abnormal deviations in current and power factor caused by early contact degradation on the high-voltage side are only on the order of magnitude, completely overlapping with the electrical fluctuations caused by normal operating condition switching. Setting the fixed threshold too high will fail to capture the weak early degradation signals, while setting it too low will result in frequent false alarms. Existing microcomputer-based integrated relays cannot identify early contact degradation under the cover of large fluctuations in normal operating conditions, resulting in contact degradation remaining in a long-term monitoring blind zone. It can only be detected during periodic shutdowns for disassembly and maintenance, failing to achieve continuous quantitative monitoring of contact degradation throughout its entire lifecycle. Summary of the Invention

[0005] This invention provides an intelligent monitoring method and system for high-voltage switchgear to solve the technical problem that the prior art cannot identify early deterioration signals of vacuum circuit breaker contacts in the context of frequent switching of multiple operating conditions.

[0006] In a first aspect, the present invention provides an intelligent monitoring method for high-voltage switchgear, comprising the following steps:

[0007] Collect historical closing data of vacuum circuit breakers under new conditions for each operating condition; calculate the mean current, mean power factor and mean slope for each operating condition to obtain the operating condition baseline; construct the operating condition R matrix based on the variance of electrical quantities for each operating condition.

[0008] After each closing, the current rise slope is calculated; the current rise slope is compared with the average slope of each operating condition reference, and the nearest neighbor is taken to obtain the operating condition label.

[0009] The mean current and mean power factor of the current during the steady-state period are collected; the mean current and mean power factor in the operating condition reference corresponding to the operating condition label are subtracted to obtain the observation deviation; using the observation deviation as the observed quantity and the operating condition R matrix corresponding to the operating condition label as the measurement noise covariance, Kalman iteration is performed to obtain the contact resistance increment and innovation, and the innovation is included in the operating condition buffer corresponding to the operating condition label; when the operating condition buffer reaches the preset window length, the operating condition R matrix is ​​corrected according to the innovation covariance of the operating condition buffer.

[0010] Divide the incremental contact resistance by the initial contact resistance of the contact to obtain the degradation index; generate and output an alarm signal based on the relationship between the degradation index and a preset threshold.

[0011] Its effects are as follows: using the current rise slope as the basis for operating condition identification, the operating condition tag is assigned independently within a few seconds after closing, earlier than the update of the programmable logic controller mode status signal, fundamentally eliminating the misjudgment of operating condition caused by communication delay during mode switching; the operating condition-specific operating condition reference, operating condition R matrix and operating condition buffer work together to ensure that the noise characteristics of different operating conditions evolve independently, and the contact resistance increment estimation of the Kalman iteration output is not affected by cross-interference of operating conditions.

[0012] Furthermore, the Kalman iteration includes: using the mean current, mean power factor, and rated line voltage on the high-voltage side in the operating condition reference corresponding to the operating condition label as parameters, and calculating the sensitivity vector according to the linear influence relationship between the increase in contact resistance on the current deviation and the power factor deviation in the three-phase equivalent circuit. The sensitivity vector refers to the coefficient of change of each component of the observation deviation caused by a unit increase in contact resistance; and performing Kalman gain calculation and state update using the sensitivity vector, the observation deviation, and the operating condition R matrix to obtain the contact resistance increment and information.

[0013] Its effect is as follows: the sensitivity vector is derived based on the three-phase equivalent circuit, and the average current and power factor in the operating condition reference are used as parameters for dynamic calculation, so that the observation equation has the correct physical mapping relationship under each operating condition, avoiding the sensitivity mismatch of a single fixed sensitivity vector under different operating conditions, and ensuring that the Kalman gain calculation is always based on the accurate physical model.

[0014] Furthermore, after obtaining the operating condition label, the process also includes: calculating the absolute value of the difference between the current rise slope and the average slope of each operating condition reference; when the difference between the minimum and the second minimum absolute values ​​of the difference is less than the preset tolerance, the current closing operation is marked as a transitional operation; when the current closing operation is a transitional operation, the new information is not included in the operating condition cache, and the operating condition R matrix correction is not performed.

[0015] Its effect is as follows: by calculating the difference between the absolute value of the minimum difference and the absolute value of the second smallest difference and comparing it with the preset tolerance, the transition operation in which the slope falls into the boundary region of the working condition is identified, and the writing of the information of this operation to the working condition cache is frozen, so as to prevent the data with ambiguous working condition attribution from polluting the adaptive correction process of the working condition R matrix and ensuring the statistical quality of the working condition R matrix.

[0016] Furthermore, after obtaining the operating condition tag, the process also includes: comparing the operating condition tag with the high-voltage power supply mode status signal of the programmable logic controller; when the two are inconsistent, marking the current closing operation as an abnormal operation; not including the current information in the operating condition buffer; and outputting an abnormal flag signal to the programmable logic controller.

[0017] Furthermore, the collection of the average current and power factor in the steady-state section includes: continuously collecting high-voltage side current sampling values ​​at a preset sampling interval; determining that the steady-state section has been entered when the adjacent differences of a preset number of consecutive current sampling values ​​are all less than a preset proportion of the rated current; and calculating the average of the current sampling values ​​and power factor sampling values ​​in the steady-state section to obtain the average current and power factor in the steady-state section.

[0018] Furthermore, constructing the operating condition R matrix based on the variances of electrical quantities for each operating condition includes: constructing a diagonal matrix using the variances of current and power factor for each operating condition; calculating the sample correlation coefficient between the sampled current values ​​and the sampled power factor values ​​for each operating condition; and when the absolute value of the sample correlation coefficient exceeds a preset correlation threshold, supplementing the off-diagonal elements of the diagonal matrix with the product of the sample correlation coefficient and the standard deviations of current and power factor for the corresponding operating condition, thus obtaining the operating condition R matrix.

[0019] Furthermore, when the number of initial data acquisitions for a certain operating condition is less than the preset minimum number, the method also includes: calculating the lower bound of the current variance based on the accuracy level and rated current of the current transformer; calculating the lower bound of the power factor variance based on the power factor measurement resolution; and using the lower bound of the current variance and the lower bound of the power factor variance as the current variance and power factor variance of the corresponding operating condition, respectively.

[0020] Furthermore, the correction of the working condition R matrix includes: taking the larger value of each diagonal element of the corrected working condition R matrix and the preset positive definite lower bound; setting the off-diagonal element to zero when the corrected off-diagonal element is less than zero; and constructing the corrected working condition R matrix with the updated diagonal and off-diagonal elements.

[0021] Its effect is that: the diagonal elements of the corrected working condition R matrix are taken to be larger than the preset positive definite lower bound, and the negative off-diagonal elements are set to zero, so as to ensure that the working condition R matrix always satisfies the positive semi-definite condition during the online correction process, avoid matrix singularity caused by numerical update, and prevent matrix inversion failure in Kalman gain calculation, thus ensuring the numerical stability of long-term operation.

[0022] Furthermore, generating and outputting an alarm signal includes: obtaining the estimated covariance obtained by performing Kalman iteration; when the estimated covariance is greater than a preset threshold, recording the degradation index in the historical record and keeping the alarm signal unchanged; when the estimated covariance is not greater than the preset threshold, comparing the degradation index with the preset threshold, updating and outputting the alarm signal.

[0023] Its effect is as follows: by introducing a confidence gating mechanism based on estimated covariance, when the working condition R matrix has not yet converged and the estimation uncertainty is high at the beginning of the working condition switch, the alarm status update is frozen and only the degradation index is recorded in the historical record; after the estimated covariance drops below the preset confidence threshold, normal judgment is restored, which effectively suppresses false alarms caused by estimation bias during the convergence transition period.

[0024] Secondly, the present invention provides an intelligent monitoring system for high-voltage switchgear, including a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned intelligent monitoring method for high-voltage switchgear is implemented.

[0025] The beneficial effects are as follows: This invention eliminates the differences in electrical quantity benchmarks between different high-voltage power supply modes by using operating condition benchmarks, quantifies the inherent noise level of each operating condition using the operating condition R matrix, and continuously accumulates weak contact resistance increment signals across the number of operations using Kalman iteration. This enables continuous and quantitative monitoring of the early deterioration process of vacuum circuit breaker contacts throughout the entire cycle, under the complex background of frequent switching between three power supply modes and large fluctuations in electrical quantities in electric arc furnaces. All monitoring functions are implemented based on existing programmable logic controller signals, without the need for additional sensors or hardware modifications. The monitoring results are output as a deterioration index and graded alarm signals, directly compatible with existing alarm interfaces, transforming contact deterioration from a monitoring blind spot into a quantifiable basis for operation and maintenance decisions. Attached Figure Description

[0026] Figure 1 This is a flowchart of an intelligent monitoring method for high-voltage switchgear.

[0027] Figure 2 A comparison chart showing the evolution of contact resistance increments across multiple operation cycles.

[0028] Figure 3 Comparison curves of robust degradation under frequent switching of harsh operating conditions. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] An embodiment of the intelligent monitoring method for high-voltage switchgear provided by this invention:

[0031] like Figure 1 As shown, the intelligent monitoring method for high-voltage switchgear includes the following steps:

[0032] S101: Collect historical closing data of the vacuum circuit breaker under new conditions for each operating condition; calculate the mean current, mean power factor and mean slope for each operating condition to obtain the operating condition baseline; construct the operating condition R matrix based on the variance of electrical quantities for each operating condition.

[0033] In one embodiment, this method is applied to the online monitoring scenario of vacuum circuit breaker contact deterioration in a high-voltage power supply system of a smelting plant. A typical scenario is a combined production line of a 70-ton electric arc furnace and an AOD refining furnace, equipped with a KYN61-40.5 type metal-clad enclosed high-voltage switchgear. The vacuum circuit breaker adopts a Siemens 3AH4 type spring operating mechanism. The high-voltage power supply system has three operating modes: frequency conversion, bypass, and soft start. The programmable logic controller has collected signals such as high-voltage side current, power factor, and opening / closing status in real time, with 20 to 30 opening / closing operations per shift. Due to the frequent switching between the three modes during the electric arc furnace smelting process, the magnitude difference in current and power factor between different modes can reach several times. If a single fixed benchmark is used to measure the contact condition, the fluctuations under normal operating conditions will completely mask the slight deviations caused by early deterioration. Therefore, in the initial stage of operation when the vacuum circuit breaker contacts are in a brand-new state, it is necessary to collect historical closing data according to the high-voltage power supply mode, establish operating condition benchmarks and operating condition R matrices for each operating condition, and provide an independent reference system for subsequent online estimation.

[0034] There are three high-voltage power supply modes: frequency conversion mode, bypass mode, and soft start mode. It is recommended to collect no less than 15 operation data for each mode, with a total of no less than 45 initializations, corresponding to approximately 3 to 5 production shifts. During initialization, the operating mode of each operation is confirmed by relying on the existing high-voltage power supply mode status signals in the programmable logic controller.

[0035] For each closing operation, high-voltage side current samples are continuously collected at 20ms intervals from the closing moment until the current enters the steady-state phase. The method for determining the steady-state phase is detailed in S103. After entering the steady-state phase, the average values ​​of the current and power factor samples within the steady-state phase are calculated to obtain the average steady-state current and power factor for this operation. Simultaneously, the current rise slope is calculated using the least squares linear regression slope formula for the sampling sequence after closing, obtaining the recorded slope value for this operation.

[0036] For each operating condition, the average values ​​of the steady-state current, power factor, and slope recorded for each of the above operations are calculated using the standard sample mean formula. This yields the average current, power factor, and slope for that operating condition, which together constitute the operating condition baseline. The variance of the current and power factor for each operating condition is calculated using the unbiased sample variance formula, serving as input for constructing the operating condition R matrix.

[0037] To obtain the initial value of the operating condition R matrix, the operating condition R matrix is ​​constructed using the variances of electrical quantities for each operating condition. This construction includes: forming a diagonal matrix using the variances of current and power factor for each operating condition; calculating the sample correlation coefficient between the sampled current values ​​and the sampled power factor values ​​for each operating condition; and when the absolute value of the sample correlation coefficient exceeds a preset correlation threshold, supplementing the off-diagonal elements of the diagonal matrix with the product of the sample correlation coefficient and the standard deviations of current and power factor for the corresponding operating condition, thus obtaining the operating condition R matrix.

[0038] The initial working condition R matrix for working condition m satisfies the following relationship:

[0039]

[0040] In the formula, Let R be the initial working condition matrix for working condition m. The current variance under operating condition m, in units of , Let be the power factor variance under operating condition m, which is dimensionless. and These are the corresponding standard deviations of current and power factor, respectively. The correlation coefficient between the sampled current value and the sampled power factor value under operating condition m is given; the preset correlation threshold is 0.3. When the off-diagonal elements are zero, the matrix degenerates into a diagonal matrix.

[0041] Understandably, the initialization phase is completed in a brand-new state for the contacts, with zero increase in contact resistance. Therefore, the deviations in the electrical quantities collected under various operating conditions originate entirely from the measurement noise itself. and It accurately reflects the noise level under no-deterioration conditions and is the physical source of the initial value of the R matrix for the operating condition. The diagonal structure reflects the physical fact that current noise and power factor noise are approximately independent within the same operating condition; off-diagonal elements are only added when the absolute value of the sample correlation coefficient exceeds 0.3 to avoid introducing spurious covariance structures due to a small number of accidental fluctuations.

[0042] In the initial stage of actual operation, if the number of closing operations for a certain operating condition is insufficient, the lower bound of variance is obtained according to the following method to replace the actual sampled variance. When the number of initial sampling operations for a certain operating condition is less than the preset minimum number, the lower bound of the current variance is calculated based on the accuracy class and rated current of the current transformer; the lower bound of the power factor variance is calculated based on the power factor measurement resolution; the lower bounds of the current variance and power factor variance are used as the current variance and power factor variance of the corresponding operating condition, respectively. This lower bound of variance is directly derived from the accuracy parameters on the instrument nameplate, which can ensure that the R matrix of the operating condition still has reasonable positive definiteness when the data accumulation is insufficient, and avoid the failure of Kalman gain calculation due to the singularity of the R matrix.

[0043] It should be noted that the minimum number of cycles is preset to 15, with a KYN61-40.5 type switchgear equipped with a 0.2-class current transformer and a rated current of 1200A as a reference. The lower bound of the current variance is... The lower bound of the power factor variance is calculated to be 0.0001 with a power factor measurement resolution of 0.01.

[0044] After the initialization of each operating condition is completed, The current value of the operating condition R matrix during the operation phase is denoted as... It is continuously updated in S103. Simultaneously, the Kalman filter is initialized: the initial estimate of the contact resistance increment is zero, and the initial estimate covariance is set to... The corresponding estimated standard deviation is 10 μΩ, reflecting the uncertainty estimate of the absolute value of the initial contact resistance of the contact.

[0045] Three independent operating condition references and operating condition R matrices record the inherent electrical quantity reference levels and noise characteristics of the three modes: frequency conversion, bypass, and soft start. This eliminates the cross-interference of magnitude differences between different operating conditions on subsequent deviation estimation, and enables the corresponding operating condition reference to be loaded immediately after any mode switch, providing a reliable data foundation for the accurate extraction of early contact degradation signals.

[0046] S102, after each closing, calculate the current rise slope; compare the current rise slope with the average slope of each operating condition reference, take the nearest neighbor, and obtain the operating condition label.

[0047] In one embodiment, the three high-voltage power supply modes frequently alternate during the electric arc furnace production process. Each time the circuit is closed, it is necessary to independently determine which operating condition the operation belongs to at the moment of closure. Relying on the mode status signal of the programmable logic controller (PLC) is not feasible because this signal has a communication delay and may lag behind the actual electrical state during rapid mode switching. The current rise rate differs significantly by order of magnitude among the three operating conditions: in frequency conversion mode, the frequency converter actively limits the current, and the current rises slowly, typically ranging from 50 to 200 A / s; in bypass mode, the mains frequency grid is directly connected, and the current rises rapidly, typically ranging from 500 to 2000 A / s; the soft-start mode falls between the two, typically ranging from 200 to 500 A / s. This difference stems from the different main circuit control mechanisms of the three modes, which is a stable physical law unaffected by contact wear. Therefore, the current rise rate can be used as an electrical physical fingerprint for operating condition identification.

[0048] After each closing, high-voltage side current samples are continuously acquired at 20ms intervals from the closing moment until the steady-state phase is reached. The current rise slope is calculated using the least-squares linear regression slope formula for the sampling sequence after closing. The unit is A / s, where the subscript k indicates the k-th closing operation.

[0049] To assign operating condition labels, the current rise slope was calculated separately. The absolute value of the difference between the slope and the mean slope of each working condition baseline is used as the working condition label for this operation.

[0050] Understandably, the magnitude difference between the average slopes of the three operating conditions is approximately one order of magnitude, and the nearest neighbor judgment results have high confidence in most closing operations. However, during the transition period of mode switching, the current rise slope may fall into the middle region of the average slopes of two adjacent operating conditions. If the information of the transition operation is included in the operating condition buffer at this time, it will introduce inaccurate measurement noise samples, polluting the adaptive correction result of the operating condition R matrix. This method introduces a transition operation identification mechanism: calculate the absolute value of the difference between the current rise slope and the average slope of each operating condition reference; when the difference between the minimum and second minimum absolute values ​​of the difference is less than the preset tolerance, the current closing operation is marked as a transition operation; when the current closing operation is a transition operation, the information is not included in the operating condition buffer, and the operating condition R matrix correction is not performed.

[0051] Among them, the preset tolerance The maximum standard deviation of the slope among the three operating conditions during the S101 initialization phase is taken as 0.5 times. This method automatically calibrates based on the actual dispersion of the slope for each operating condition, eliminating the need for manual setting of a fixed value. For example, if the standard deviations of the slopes for the inverter mode, bypass mode, and soft start mode during the initialization phase are 40 A / s, 200 A / s, and 80 A / s, respectively, then the preset tolerance is... The slope of the current rise during a certain closing operation Mean slope of soft start mode Mean slope of bypass mode The minimum absolute value of the difference is 30A / s, and the second minimum absolute value of the difference is 830A / s. The difference between the two, 800A / s, is much greater than 100A / s. This is not a transitional operation. The operating condition label is assigned to soft start mode, and subsequent steps are executed normally.

[0052] After assigning a value to the operating condition label, the label needs to be cross-verified to detect abnormalities in the programmable logic controller (PLC) mode signal link. After obtaining the operating condition label, the process also includes: comparing the operating condition label with the PLC's high-voltage power supply mode status signal; if they do not match, marking the current closing operation as an abnormal operation; not including this information in the operating condition buffer; and outputting an abnormal flag signal to the PLC.

[0053] It should be noted that the determination of abnormal operation does not affect the normal execution of this Kalman iteration. The estimation of contact resistance increment continues as usual. Only the writing of new information to the operating condition buffer is frozen to avoid erroneous data from polluting the operating condition R matrix during programmable logic controller (PLC) signal failures. The abnormal flag signal output to the PLC is a switching signal, prompting maintenance personnel to check the high-voltage power supply mode status signal link, thus achieving redundant verification of the PLC's own signal reliability.

[0054] By assigning operating condition labels based on the current rise slope after closing, operating condition identification can be completed independently within approximately 1 to 3 seconds after each closing operation, earlier than the communication update of the programmable logic controller mode status signal. This fundamentally eliminates misjudgments of operating conditions caused by signal delays during rapid mode switching. The transition operation identification and abnormal operation verification mechanism further ensures the purity of the operating condition cache data, ensuring that the adaptive correction of the operating condition R matrix is ​​always based on valid operating condition attribution data.

[0055] S103: Collect the mean current and mean power factor of the current steady-state period; subtract the mean current and mean power factor in the operating condition reference corresponding to the operating condition label to obtain the observation deviation; use the observation deviation as the observed quantity and the operating condition R matrix corresponding to the operating condition label as the measurement noise covariance, perform Kalman iteration to obtain the contact resistance increment and innovation, and incorporate the innovation into the operating condition buffer corresponding to the operating condition label; when the operating condition buffer reaches the preset window length, correct the operating condition R matrix according to the innovation covariance of the operating condition buffer.

[0056] In one embodiment, after assigning a working condition label in S102, the average current and average power factor are collected from the steady-state segment after the current closing, and the working condition reference established in S101 is subtracted to construct an observation deviation reflecting the change in contact resistance. Then, a Kalman iteration is driven by a working condition-specific working condition R matrix to continuously extract weak contact degradation signals from the noise background. Taking a 70-ton electric arc furnace production line as an example, the high-voltage side operating current is approximately 1000 to 1200A, while the current deviation caused by early degradation is only in the milliampere range. If the original current measurement value is directly compared, the degradation signal is completely submerged in the normal operating condition fluctuations. Only by subtracting the working condition reference and accumulating the weak deviation through Kalman iteration across the number of operations can the invisible early degradation process be transformed into a quantifiable estimate of the contact resistance increment.

[0057] The method for determining the steady-state section and collecting the average current and power factor of the steady-state section is as follows: continuously collect the current sampling value of the high-voltage side at a preset sampling interval; when the adjacent difference of a preset number of consecutive current sampling values ​​is less than the preset ratio of the rated current, it is determined that the steady-state section has been entered; calculate the average of the current sampling value and the power factor sampling value in the steady-state section to obtain the average current and power factor of the steady-state section.

[0058] The preset sampling interval is 20ms, the preset number of samples is 5, and the preset ratio is 0.5% of the rated current. Using a rated current of 1200A as a reference, the steady-state criterion threshold is... When the difference between five consecutive sampling points is less than 6A, the system is considered to have entered the steady-state phase. The duration of the steady-state phase varies depending on the operating conditions, typically ranging from several seconds to tens of seconds. The arithmetic mean of the current and power factor sampling values ​​at all sampling points within the steady-state phase is calculated to obtain the mean steady-state current for this operation. and power factor mean .

[0059] The observation deviation is obtained by subtracting the average current of the operating condition reference corresponding to the operating condition label from the average current of the steady-state section, and by subtracting the average power factor of the operating condition reference from the average power factor.

[0060] Understandably, when the contacts are in good condition, the expected value of the observed deviation is a zero vector. Subtracting the operating condition reference eliminates the magnitude difference in electrical quantity reference levels between different operating conditions, ensuring that the systematic non-zero offset of the observed deviation originates solely from changes in contact resistance. When the increase in contact resistance is positive, the current deviation component is systematically negative, and the power factor deviation component is systematically positive. The two signals are in opposite directions but point to the same physical cause. This inverse symmetry characteristic helps the Kalman iteration distinguish between systematic degradation offset and random noise.

[0061] Before performing Kalman iteration, the sensitivity vector needs to be determined, which is the coefficient of change of each component of the observation deviation caused by a unit increase in contact resistance. Performing Kalman iteration includes: using the mean current, mean power factor, and rated line voltage on the high-voltage side in the operating condition reference corresponding to the operating condition label as parameters, and calculating the sensitivity vector according to the linear influence relationship of the increase in contact resistance on the current deviation and power factor deviation in the three-phase equivalent circuit. The sensitivity vector refers to the coefficient of change of each component of the observation deviation caused by a unit increase in contact resistance; using the sensitivity vector, the observation deviation, and the operating condition R matrix, Kalman gain calculation and state update are performed to obtain the contact resistance increment and information.

[0062] Under the first-order small-quantity approximation, the contact resistance increment is much smaller than the load impedance modulus. This condition is satisfied throughout the entire service life of the contact in 10kV to 35kV high-voltage systems. (Sensitivity vector) Satisfying the relation:

[0063]

[0064] In the formula, For working conditions The corresponding sensitivity vector, first component The unit is A / Ω, second component The unit is Ω - ¹; For working conditions The average current, expressed in amperes (A). For working conditions The average power factor; ; This is the rated line voltage on the high-voltage side, a fixed value on the switchgear nameplate, in volts (V). The AOD furnace system uses... electric arc furnace system Before implementation, each switch cabinet nameplate must be checked.

[0065] Understandably, According to working condition label Automatic switching, exemplarily, in bypass mode , , ,but , If the contact resistance increment is The predicted current deviation is approximately The predicted power factor deviation is approximately The magnitude is extremely small, completely masked by normal operating condition fluctuations, which is precisely the weak signal that Kalman iteration needs to continuously accumulate and extract across multiple operations.

[0066] Based on the observation bias, sensitivity vector, and operating condition R matrix corresponding to the operating condition label, Kalman iteration is performed to obtain the contact resistance increment and innovation. The Kalman iteration includes a prediction step, innovation calculation, gain calculation, and state update. The prediction step includes: adding the estimated contact resistance increment after the previous operation to the average drift of each operation to obtain the predicted value before the current operation; simultaneously, the prediction covariance is obtained by adding the estimated covariance of the previous operation to the process noise covariance. The state transition is incremented by the number of closing operations, with each operation advancing one step, consistent with the physical nature of contact arc erosion accumulation. The average drift of the contact resistance in each operation... Based on the rated operating life parameters in the Siemens 3AH4 vacuum circuit breaker technical manual, and using the rated contact resistance... The rated mechanical life is 10,000 cycles, which is taken as a reference. Typical value is approximately Second; process noise covariance The innovation calculation includes subtracting the theoretical deviation calculated based on the predicted value from the current observation deviation to obtain the innovation, which reflects the amount of new information in the current observation that exceeds the interpretable range of the prediction model. Gain calculation and state update include calculating based on the prediction covariance, sensitivity vector, and condition-specific measurement noise covariance. The Kalman gain is calculated; then, the gain-weighted innovation is used to correct the predicted value, obtaining the optimal estimate of the contact resistance increment after this operation; at the same time, the estimated covariance is updated to prepare for the next iteration. The optimal estimate of the contact resistance increment will be passed to S104, and the estimated covariance will also be passed to S104.

[0067] After the new information calculation is completed, the new information will be included in the operating condition label only if the current operation is neither a transient operation nor an abnormal operation. Corresponding working condition cache Each operating condition has its own independent operating condition cache, and the three sets of operating condition caches are independent of each other and do not interfere with each other.

[0068] When the number of valid samples in the operating condition cache reaches the preset window length At that time, based on the most recent cached data... The R matrix for the next new information correction condition. Preset window length. The optimal value is 10, which balances statistical stability and convergence speed. Based on the reference of about 8 to 10 operations per shift in bypass mode, the first correction can be completed in about 1 to 2 shifts.

[0069] Specifically, the correction process includes: calculating the sample covariance of the innovation within the window as the measured innovation covariance; subtracting the contribution of state prediction uncertainty to the innovation covariance from the measured innovation covariance to obtain the R-bias correction amount; and multiplying this correction amount by the learning rate. This is then accumulated into the R matrix for the current operating condition. This correction method is based on Sage. Husa's adaptive Kalman filtering theory enables the operating condition R matrix to automatically adapt to changes in noise levels caused by frequency converter replacement, line modification, etc. during operation.

[0070] To ensure the positive definiteness of the corrected working condition R matrix, each diagonal element is assigned the larger value between itself and a preset positive definite lower bound. When a corrected off-diagonal element is less than zero, it is set to zero. The corrected working condition R matrix is ​​then constructed using the updated diagonal and off-diagonal elements. The preset positive definite lower bound is... The lower bound of the diagonal element is determined by the accuracy of the current transformer and the resolution of the power factor measurement, respectively, and is consistent with the lower bound of the variance in S101.

[0071] The Kalman iteration, which uses the observation deviation and sensitivity vector as inputs and the operating condition R matrix as a noise characterization tool, continuously outputs the optimal estimate of the contact resistance increment after each closing operation. The operating condition-specific buffer mechanism ensures that the noise characteristics of the three operating conditions evolve independently. The operating condition R matrix automatically adapts to changes during operation without manual recalibration, continuously and accurately extracting the nanoohmic-level early degradation signals buried under large operating condition fluctuations during the electric arc furnace smelting process.

[0072] S104: Divide the incremental contact resistance by the initial contact resistance of the contact to obtain the degradation index; generate an alarm signal and output it based on the relationship between the degradation index and the preset threshold.

[0073] In one embodiment, S103 outputs the contact resistance increment after each Kalman iteration. and estimate covariance , Expressed in absolute ohms, the physical meaning is clear but not intuitive, making it difficult for maintenance personnel to directly determine whether the contacts require attention. Normalizing the increase in contact resistance to a relative increase ratio yields a degradation index, which facilitates direct comparison with international and national standard circuit breaker contact resistance degradation criteria, forming a standardized output compatible with existing programmable logic controller alarm systems.

[0074] Deterioration Index Satisfying the relation:

[0075]

[0076] In the formula, For the first The degradation index after the operation is dimensionless and reflects the percentage increase in contact resistance relative to the initial value. The optimal estimate of the contact resistance increment output by the S103 Kalman iteration is given in Ω. The initial contact resistance of the contacts is expressed in Ω. The measured value from a milliohm meter in the commissioning record during switchgear operation is preferred, and the typical value should not exceed [value missing]. If no actual measurement records are available, use the upper limit of the factory technical specifications of the Siemens 3AH4 circuit breaker for the KYN61-40.5 type switchgear. As a conservative estimate.

[0077] Understandably, the degradation index As the number of closing operations increases monotonically, combined with the existing closing count counter of the programmable logic controller, a historical curve of the number of operations versus the deterioration index can be output with the number of operations as the horizontal axis and the deterioration index as the vertical axis. The slope of the curve represents the average loss rate of a single operation. If the slope increases abnormally in a short period of time, it indicates that there are frequent high current impacts recently, which have led to the aggravation of single burn-out. The remaining life of the contacts can be predicted in advance, transforming passive post-event maintenance into proactive predictive maintenance.

[0078] Before generating an alarm signal based on the degradation index, the confidence level of the current estimate must be determined to prevent estimation deviations during the convergence transition period of the operating condition R matrix from triggering false alarms. The alarm signal generation and output process includes: obtaining the estimated covariance obtained from the Kalman iteration; when the estimated covariance is greater than a preset confidence threshold, adding the degradation index to the historical record and keeping the alarm signal unchanged; when the estimated covariance is not greater than the preset confidence threshold, comparing the degradation index with the preset threshold, updating and outputting the alarm signal.

[0079] Among them, the preset threshold value is taken The corresponding standard deviation of the contact resistance increment estimate is When the estimated covariance is greater than the threshold, it indicates that the current estimation confidence is insufficient. The degradation index will only be recorded in the historical record and will not trigger an alarm status update. When the estimated covariance is not greater than the preset confidence threshold, the alarm signal will be updated and output according to the relationship between the degradation index and the preset threshold.

[0080] Alarm classification satisfies the following relation:

[0081]

[0082] In the formula, the warning threshold of 0.20 and the alarm threshold of 0.40 refer to the relevant provisions of IEC62271-100 and GB / T11022 regarding the criteria for deterioration of circuit breaker contact resistance, respectively, corresponding to an increase in contact resistance to 1.20 times and 1.40 times the initial value. In the warning state, a set signal is output to the programmable logic controller (PLC) digital output channel DI1, prompting maintenance personnel to pay attention to the contacts and suggesting that the switchgear be inspected during the next planned shutdown. In the alarm state, a set signal is output to DI2, suggesting that a shutdown for maintenance be arranged during the gap between tapping and charging after the current furnace smelting is completed to avoid continued operation with defective contacts. Both signals are normally open dry contact outputs, consistent with the existing high-voltage switchgear alarm signal interface, requiring no modification to the PLC program.

[0083] By normalizing the deterioration index and using a confidence gating mechanism based on estimated covariance, the contact deterioration state is transformed from an invisible micro-ohmic continuous physical process into a graded alarm signal that can be directly read by maintenance personnel. Without adding any hardware or modifying the existing programmable logic controller program, full-cycle continuous intelligent monitoring of early deterioration of vacuum circuit breaker contacts is achieved, minimizing the risk of unplanned downtime caused by contact deterioration in electric arc furnace production lines.

[0084] Figure 2 This is a comparative diagram of the evolution of contact resistance increment across multiple operation cycles between existing technologies and the method of this invention. As shown in the figure, under the background of frequent switching of multiple operating conditions in the high-voltage power supply system of electric arc furnace, the existing technology suffers from severe oscillations in the extraction result of contact resistance increment due to the masking effect of large fluctuations in electrical quantities. In contrast, this invention effectively isolates normal operating condition fluctuations by constructing an independent operating condition benchmark and Kalman iteration accumulation across multiple cycles, accurately fitting the real physical process of slow deterioration of contact resistance, and estimating stability to be about 20% higher than that of existing technologies.

[0085] Figure 3 The curves show the robustness degradation of the system under frequent switching of harsh operating conditions. The horizontal axis represents the fluctuation range of electrical quantities caused by rapid switching of operating conditions. As can be seen from the curves, when the fluctuation range is small and in a stable period, the two methods perform similarly. However, once the system enters the harsh fluctuation range of frequent switching, the degradation index estimation error of the existing technology increases sharply, which can easily trigger false alarms. In contrast, thanks to the transition operation identification mechanism and the confidence gating mechanism based on the estimated covariance, the error of this invention only increases slightly and then quickly stabilizes. Under extreme operating conditions, it maintains an accuracy advantage of about 20% over the existing technology.

[0086] An embodiment of the intelligent monitoring system for high-voltage switchgear provided by this invention:

[0087] The intelligent monitoring system for high-voltage switchgear includes a processor and a memory. The memory stores computer program instructions, which are executed by the processor to implement the aforementioned intelligent monitoring method for high-voltage switchgear.

[0088] The intelligent monitoring system for high-voltage switchgear also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0089] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0090] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent monitoring of high-voltage switchgear, characterized in that, Includes the following steps: Collect historical closing data of vacuum circuit breakers under new conditions for each operating condition; calculate the mean current, mean power factor and mean current rise slope for each operating condition to obtain the operating condition baseline; construct the operating condition R matrix based on the current variance and power factor variance for each operating condition. Calculate the current rise slope after each closing. The current rise slope is compared with the average current rise slope of each operating condition reference, and the nearest neighbor is taken to obtain the operating condition label. Collect the average current and average power factor during the steady-state period; subtract the average current and average power factor from the operating condition reference corresponding to the operating condition label to obtain the observation deviation. Using the observed deviation as the observed quantity and the working condition R matrix corresponding to the working condition label as the measurement noise covariance, Kalman iteration is performed to obtain the contact resistance increment and the new information, and the new information is included in the working condition cache corresponding to the working condition label. When the operating condition buffer reaches the preset window length, the operating condition R matrix is ​​corrected based on the information covariance of the operating condition buffer. The Kalman iteration includes: using the mean current, mean power factor, and rated line voltage on the high-voltage side in the operating condition reference corresponding to the operating condition label as parameters, and calculating the sensitivity vector according to the linear influence relationship between the increase in contact resistance and the current deviation and power factor deviation in the three-phase equivalent circuit. The sensitivity vector refers to the coefficient of change of each component of the observation deviation caused by the unit increase in contact resistance. The Kalman gain calculation and state update are performed using the sensitivity vector, the observation deviation, and the operating condition R matrix to obtain the contact resistance increment and information. Divide the incremental contact resistance by the initial contact resistance of the contact to obtain the degradation index; generate and output an alarm signal based on the relationship between the degradation index and a preset threshold.

2. The intelligent monitoring method for high-voltage switchgear according to claim 1, characterized in that, After obtaining the operating condition label, the following steps are also included: calculating the absolute value of the difference between the current rise slope and the average current rise slope of each operating condition reference; when the difference between the minimum and the second minimum absolute values ​​of the difference is less than the preset tolerance, the current closing operation is marked as a transitional operation; when the current closing operation is a transitional operation, the new information is not included in the operating condition cache and the operating condition R matrix correction is not performed.

3. The intelligent monitoring method for high-voltage switchgear according to claim 1, characterized in that, After obtaining the operating condition tag, the process also includes: comparing the operating condition tag with the high-voltage power supply mode status signal of the programmable logic controller; if the two do not match, marking the current closing operation as an abnormal operation; not including the current information in the operating condition buffer; and outputting an abnormal flag signal to the programmable logic controller.

4. The intelligent monitoring method for high-voltage switchgear according to claim 1, characterized in that, The collection of the mean current and mean power factor in the steady-state section includes: continuously collecting the current sample value on the high-voltage side at a preset sampling interval; when the difference between adjacent current sample values ​​of a preset number of consecutive current sample values ​​is less than a preset proportion of the rated current, it is determined that the steady-state section has been entered; the mean values ​​of the current sample value and the mean power factor sample value in the steady-state section are calculated respectively to obtain the mean current and mean power factor in the steady-state section.

5. The intelligent monitoring method for high-voltage switchgear according to claim 1, characterized in that, Constructing the operating condition R matrix based on the variances of electrical quantities for each operating condition includes: forming a diagonal matrix using the variances of current and power factor for each operating condition; calculating the sample correlation coefficient between the sampled current values ​​and the sampled power factor values ​​for each operating condition; and when the absolute value of the sample correlation coefficient exceeds a preset correlation threshold, supplementing the off-diagonal elements of the diagonal matrix with the product of the sample correlation coefficient and the standard deviations of current and power factor for the corresponding operating condition, thus obtaining the operating condition R matrix.

6. The intelligent monitoring method for high-voltage switchgear according to claim 5, characterized in that, When the number of initial data acquisitions for a certain operating condition is less than the preset minimum number, the method also includes: calculating the lower bound of the current variance based on the accuracy class and rated current of the current transformer; calculating the lower bound of the power factor variance based on the power factor measurement resolution; and using the lower bound of the current variance and the lower bound of the power factor variance as the current variance and power factor variance of the corresponding operating condition, respectively.

7. The intelligent monitoring method for high-voltage switchgear according to claim 1, characterized in that, The modified operating condition R matrix includes: taking the larger value of each diagonal element of the modified operating condition R matrix and the preset positive definite lower bound; setting the off-diagonal element to zero when the modified off-diagonal element is less than zero; and constructing the modified operating condition R matrix with the updated diagonal and off-diagonal elements.

8. The intelligent monitoring method for high-voltage switchgear according to claim 1, characterized in that, The process of generating and outputting an alarm signal includes: obtaining the estimated covariance obtained by performing Kalman iteration; when the estimated covariance is greater than a preset threshold, adding the degradation index to the historical record and keeping the alarm signal unchanged; when the estimated covariance is not greater than the preset threshold, comparing the degradation index with the preset threshold, updating and outputting the alarm signal.

9. A high-voltage switchgear intelligent monitoring system, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent monitoring method for high-voltage switchgear as described in any one of claims 1-8 is implemented.

Images