A method and system for on-line measurement of 6kV insulation for power plant

Abstract

The application provides a kind of plant 6kV insulation on-line measurement method and system, related to electrical equipment insulation on-line monitoring technical field.The application is configured with micro pulse injection unit at neutral point side, with voltage zero-crossing as time window, injects polarity reversed pseudo-random micro pulse sequence and synchronously acquires response signal, through differential gate sampling and self-calibration decoupling operator, inhibits voltage mutual inductance and protection quantity coupling;Combined with primary system topology, establish reachable matrix and branch sensitivity kernel, implement joint frequency domain and time domain sparse regression to identify device level polarization time constant, leakage admittance and dielectric loss angle slice;Utilize hierarchical bayesian estimation model to calculate insulation barrier decay rate and health index, output early warning information and condition maintenance instruction, realize the safe on-line measurement and intelligent evaluation of plant 6kV system insulation state.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of on-line monitoring of electrical equipment insulation, in particular to a method and system for on-line measurement of 6kV plant insulation. BACKGROUND

[0002] The 6kV plant distribution system generally adopts neutral point small resistance grounding mode to balance the fault current control and protection selectivity in single-phase grounding. The insulation measurement in service depends on offline test during power failure or indirect estimation through operation measurement. The existing on-line monitoring device usually takes the busbar as a whole, which is difficult to distinguish the insulation changes of the feeder and the key equipment level under the condition of not stopping power, and is sensitive to weather, humidity, load fluctuation and segment switching, resulting in value fluctuation and false alarm.

[0003] State maintenance and risk pre-disposal require on-line, accurate and traceable insulation measurement capability. The trend focuses on three aspects: first, under the premise of not changing the primary wiring and operation mode, a controllable detection signal is safely injected and electromagnetically decoupled with the voltage mutual inductance loop; second, a micro-level leakage response is measured to build an integrated measurement scale for system, feeder and equipment levels; third, the working condition and environmental self-adaptive processing are introduced to form a unified evaluation result comparable across substations for early warning and maintenance decision support.

[0004] The existing scheme has common deficiencies: the detection channel is coupled with the neutral point grounding and the voltage mutual inductance loop, the on-line injection is limited; the micro-DC leakage signal has insufficient resolution in complex field background; only the equivalent index of the busbar is given, which is difficult to locate to the feeder and equipment; there is a lack of self-adaptive reduction of meteorological and working condition disturbance, and the evaluation result has insufficient stability; the threshold setting lacks a unified scale, and it is difficult to migrate across substations. SUMMARY

[0005] In order to overcome the deficiencies of the prior art, the purpose of the present application is to provide a method and system for on-line measurement of 6kV plant insulation, which realizes safe on-line measurement and intelligent evaluation of the insulation state of 6kV plant power system, and can realize accurate identification, trend warning and state maintenance decision of insulation deterioration under the condition of uninterrupted operation, significantly improving the safety and operation intelligent level of the system.

[0006] To achieve the above purpose, the present application provides the following scheme:

[0007] A method for on-line measurement of 6kV plant insulation, comprising:

[0008] A micro-impulse injection unit is configured at the neutral side of the 6kV power system, and a pseudo-random micro-impulse sequence with polarity reversal is injected into the primary circuit of the 6kV power system in a time window of voltage zero-crossing, and the injection signal and the response signal are recorded with a synchronous clock to obtain a zero-crossing synchronous reference sequence and an initial response sequence;

[0009] Difference gating sampling is performed based on three-phase phases, self-calibration decoupling operators are generated by planned closing and load transition, coupling suppression processing of voltage mutual inductance loop and protection quantity is performed on the initial response sequence to obtain a decoupling response sequence;

[0010] A reachable matrix and branch sensitivity kernel are established according to the topology of the primary system, and joint frequency domain and time domain sparse regression analysis is performed on the decoupling response sequence to identify device-level polarization time constant, leakage admittance and dielectric loss angle slice, form a device parameter vector and a position mapping of the device parameter vector;

[0011] A hierarchical Bayesian estimation model is constructed for the device parameter vector, the insulation barrier decay rate and the health index are calculated, and the credible interval is given to obtain the device-level evaluation result;

[0012] The device-level evaluation result is correspondingly associated according to the position mapping to determine the device object with abnormal insulation performance, and the advance warning information and the state maintenance instruction of the device are output.

[0013] Preferably, a micro-impulse injection unit is configured at the neutral side of the 6kV power system, and a pseudo-random micro-impulse sequence with polarity reversal is injected into the primary circuit of the 6kV power system in a time window of voltage zero-crossing, and the injection signal and the response signal are recorded with a synchronous clock to obtain a zero-crossing synchronous reference sequence and an initial response sequence, comprising:

[0014] A zero-point detection reference is established by the micro-impulse injection unit based on the power frequency voltage of the primary circuit to determine the time mark and phase relationship of the zero-crossing transient to obtain a zero-crossing synchronous reference;

[0015] A pseudo-random micro-impulse sequence with polarity reversal is generated by the micro-impulse injection unit under the constraint of safe injection parameters, and is aligned with the zero-crossing synchronous reference within the window to form a to-be-injected sequence;

[0016] The to-be-injected sequence is injected into the primary circuit through the parallel isolation channel of the micro-impulse injection unit, and the injection signal is recorded using the synchronous clock to obtain an injection signal record;

[0017] The voltage response and the current response generated by the pseudo-random micro-pulse sequence to the primary circuit at predetermined measurement points are collected under a sampling reference consistent with the synchronous clock to form a response signal, and a component outside a gated sampling suppression window is obtained to obtain a response signal record;

[0018] The injection signal record and the response signal record are time-aligned according to the zero-crossing synchronization reference to extract a window-in sequence to obtain the zero-crossing synchronization reference sequence and the initial response sequence.

[0019] Preferably, differential gated sampling is performed based on a three-phase phase, a self-calibration decoupling operator is generated by planned closing and load transition, and a voltage mutual inductance circuit and a protection quantity coupling suppression process is performed on the initial response sequence to obtain a decoupling response sequence, including:

[0020] A phase alignment reference is established based on the three-phase phase, a symmetric gated time window around zero-crossing and a three-phase pairing relationship are determined;

[0021] The initial response sequence is extracted before and after the transition and a differential observation pair is formed based on the planned closing and load transition;

[0022] The transfer characteristics of the voltage mutual inductance circuit coupling and the protection quantity coupling are estimated according to the differential observation pair, and a parameter set of the self-calibration decoupling operator is determined;

[0023] The initial response sequence is decoupled according to the self-calibration decoupling operator to obtain the decoupling response sequence.

[0024] Preferably, a reachability matrix and a branch sensitivity kernel are established according to the primary system topology, joint frequency domain and time domain sparse regression analysis is performed on the decoupling response sequence, device-level polarization time constant, leakage admittance and dielectric loss angle slice are identified, a device parameter vector and a position mapping of the device parameter vector are formed, including:

[0025] A reachability matrix is generated according to the switching position and the connectivity of the primary circuit, and the reachability matrix is updated when the operating mode is changed;

[0026] A branch sensitivity kernel is established according to the branch-to-ground admittance and the dielectric loss characteristics, and the dimensions of different sections are unified in scale;

[0027] Under the joint constraints of the reachability matrix and the branch sensitivity kernel, joint frequency domain and time domain sparse regression is performed on the decoupling response sequence to obtain the estimated value of the device-level polarization time constant, the estimated value of the leakage admittance and the estimated value of the dielectric loss angle slice;

[0028] The validity of each of the estimated values is confirmed by a stability test across time windows and a residual consistency test, and the estimated values are aggregated into the device parameter vectors according to the device numbers and installation positions, and a position mapping of the device parameter vectors is established.

[0029] Preferably, under the joint constraints of the reachability matrix and the branch sensitivity kernel, joint frequency-domain and time-domain sparse regression is performed on the decoupling response sequence to obtain estimated values of device-level polarization time constants, estimated values of leakage admittances, and estimated values of medium loss angle slices, including:

[0030] A joint dictionary comprising frequency-domain bases and time-domain bases is constructed according to the branch sensitivity kernel, and a basis set that can participate in regression is determined under the reachability relationship given by the reachability matrix;

[0031] Under the constraints of the reachability relationship and the joint dictionary, sparse regression is performed on an observation vector formed by the decoupling response sequence, where a coefficient sub-vector divided according to devices and physical quantities is taken as the minimum unit of sparse constraints, and the formula is: ; wherein, is an observation vector composed of the decoupling response sequence; is a joint dictionary matrix constructed according to the branch sensitivity kernel; is a basis coefficient vector to be solved; is an optimal solution of the basis coefficient vector; is an index set of coefficient sub-vectors divided according to devices and physical quantities; is a coefficient sub-vector with index ; and is a diagonal weight matrix weighted by observation sample confidence; is a sparse regularization coefficient;

[0032] The obtained coefficient sub-vectors are merged according to device numbers, and mapped into each of the estimated values of device-level parameters, and the formula is: ; wherein, is an estimated value of a device-level polarization time constant; is an estimated value of a device-level leakage admittance; is an estimated value of a device-level medium loss angle slice; are index sets of coefficient sub-vectors corresponding to polarization time constants, leakage admittances, and medium loss angle slices, respectively; is a discrete candidate value of a polarization time constant basis; is a discrete candidate value of a medium loss angle slice basis; are basis coefficients corresponding to the index sets, respectively.

[0033] Preferably, a hierarchical Bayesian estimation model is constructed for the device parameter vector to calculate the insulation barrier decay rate and health index, and a confidence interval is given to obtain the device-level evaluation results, including:

[0034] Using the device parameter vector, which consists of the estimated values ​​of the device-level polarization time constant, the estimated value of the leakage admittance, and the estimated value of the dielectric loss angle slice, as the observation, the prior of the group of devices of the same type is converged under the hierarchical Bayesian framework to obtain the posterior mean and posterior covariance of the device type layer, which are used to characterize the degree of deviation of the device in the group distribution.

[0035] The Mahalanobis distance is calculated based on the posterior mean and the posterior covariance, and then mapped to the health index and the insulation barrier decay rate, using the following formula: ;in, This is the device parameter vector; The posterior mean of the device type layer obtained by hierarchical Bayes estimation; The posterior covariance of the device type layer is obtained by hierarchical Bayes estimation; The squared Mahalanobis distance is calculated using the posterior statistic obtained through hierarchical Bayesian estimation. For equipment-level health index; The insulation barrier attenuation rate; the posterior statistic includes the posterior mean of the equipment type layer and the posterior covariance of the equipment type layer;

[0036] The confidence interval of the health index is given based on the hierarchical Bayesian posterior uncertainty, using the following formula: ;in, The quantile of the non-central chi-square distribution at quantile γ; For degrees of freedom And the non-central parameter is The non-central chi-square distribution; Let be the dimension of the device parameter vector; Take the square of the Mahalanobis distance This is the significance constant corresponding to the confidence level.

[0037] Preferably, the equipment-level evaluation results are correlated according to the location mapping to identify equipment objects with abnormal insulation performance, and early warning information and condition-based maintenance instructions for the equipment are output, including:

[0038] Based on the location mapping, the device-level evaluation results are mapped to device numbers and installation locations to form a device-level evaluation list;

[0039] Taking the health index and insulation barrier decay rate as a basis for judgment, and combining the change trend in the credible interval and sliding time window to implement hierarchical judgment, a device object with abnormal insulation performance and an abnormal level are determined;

[0040] Early warning information containing device identification, installation location, trigger basis and review item indication is generated for the determined abnormal level;

[0041] A state maintenance instruction is given according to the abnormal level;

[0042] The early warning information and the state maintenance instruction are included in the operation and maintenance record.

[0043] Preferably, the state maintenance instruction includes maintenance opportunity indication, retest item and retest time window, operation mode adjustment indication and patrol video frequency.

[0044] A 6kV plant insulation online measurement system comprises:

[0045] A micro pulse injection and synchronous acquisition module is configured on the neutral point side of the 6kV plant power system, a parallel isolated micro pulse injection unit is arranged, a polarity reversed pseudo-random micro pulse sequence is injected into the primary circuit of the 6kV plant power system at voltage zero crossing as a time window, and the injection signal and the response signal are recorded with a synchronous clock to obtain a zero crossing synchronous reference sequence and an initial response sequence;

[0046] A phase difference sampling and self-calibration decoupling module is used to perform differential gated sampling based on three-phase phase, generate a self-calibration decoupling operator using planned closing and load transition, implement voltage mutual inductance loop and protection quantity coupling suppression processing on the initial response sequence, and obtain a decoupling response sequence;

[0047] A topology constraint and sparse regression identification module is used to establish an accessibility matrix and branch sensitivity kernel according to the primary system topology structure, perform joint frequency domain and time domain sparse regression analysis on the decoupling response sequence, identify device-level polarization time constant, leakage admittance and dielectric loss angle slice, form a device parameter vector and a position mapping of the device parameter vector;

[0048] A hierarchical Bayesian evaluation and health calculation module is used to construct a hierarchical Bayesian estimation model for the device parameter vector, calculate the insulation barrier decay rate and the health index, and give a credible interval to obtain a device-level evaluation result;

[0049] An early warning correlation and state maintenance instruction module is used to correspondingly associate the device-level evaluation result according to the position mapping, determine a device object with abnormal insulation performance, and output early warning information and a state maintenance instruction of the device.

[0050] The present application discloses the following technical effects:

[0051] The application realizes safe excitation and high-resolution acquisition of insulation characteristics under system operation state by introducing a parallel isolated micro-pulse injection unit on the neutral side of the 6kV power system.

[0052] The application realizes adaptive suppression of coupling effect between voltage mutual inductance loop and protection quantity by using planned closing and load transition event to generate self-calibration decoupling operator based on differential gate sampling method with three-phase phase as reference.

[0053] The application realizes multi-dimensional identification of polarization time constant, leakage admittance and dielectric loss angle slice by establishing reachable matrix and branch sensitivity kernel according to primary system topology and performing joint frequency domain and time domain sparse regression analysis.

[0054] The application uses hierarchical Bayesian estimation model to statistically infer the device parameter vector, forming a joint evaluation index of health index and insulation potential barrier decay rate.

[0055] The application constructs a closed-loop mechanism from "measurement - modeling - evaluation - early warning - maintenance" by associating device-level evaluation results with location mapping, which can identify and grade early warning of abnormal insulation performance of the device, and automatically generate state maintenance instructions. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below only show some of the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0057] Figure 1 The method flow chart provided for the embodiments of the present application is as shown in

[0058] Figure 2 The system structure schematic diagram provided for the embodiments of the present application is as shown in DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0060] The purpose of the present application is to provide a method and system for on-line measurement of 6kV insulation, which realizes real-time monitoring and intelligent diagnosis of the insulation characteristics of the 6kV power system, can accurately identify the insulation degradation state and issue an early warning, promotes the transition from passive repair to predictive management of equipment maintenance, and improves the safety and reliability of system operation.

[0061] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0062] Figure 1 The method flow chart provided for the embodiments of the present application is as shown in Figure 1 The present application provides a method for on-line measurement of 6kV insulation, which comprises:

[0063] Step 100: configuring a parallel isolated micro-pulse injection unit at the neutral point side of the 6kV power system, injecting a pseudo-random micro-pulse sequence with polarity reversal into the primary circuit of the 6kV power system with voltage zero-crossing as the time window, and recording the injected signal and the response signal with a synchronous clock to obtain a zero-crossing synchronization reference sequence and an initial response sequence;

[0064] Step 200: performing differential gating sampling based on three-phase phase as a reference, generating a self-calibration decoupling operator using planned closing and load transition, and implementing voltage mutual inductance loop and protection quantity coupling suppression processing on the initial response sequence to obtain a decoupling response sequence;

[0065] Step 300: According to the one-time system topology, the reachability matrix and branch sensitivity core are established, the joint frequency domain and time domain sparse regression analysis is carried out on the decoupling response sequence, the device level polarization time constant, leakage admittance and dielectric loss angle slice are identified, the device parameter vector and the position mapping of the device parameter vector are formed;

[0066] Step 400: A hierarchical Bayesian estimation model is constructed for the device parameter vector, the insulation barrier decay rate and health index are calculated, and the confidence interval is given, and the device level evaluation result is obtained;

[0067] Step 500: The device level evaluation result is correspondingly associated according to the position mapping, the device object with abnormal insulation performance is determined, and the advance warning information and state maintenance instruction of the device are output.

[0068] Specifically, in step 100 of the embodiment, a micro pulse injection unit is arranged at the neutral point side of the 6kV power system, which is electrically isolated from the primary circuit through an isolation medium, a zero-crossing synchronization reference is established on the basis of continuous monitoring of the power frequency voltage, and a group of short-time pulses with alternating polarity is aligned with the reference within the window. To ensure safe injection and protection device stability, the pulse amplitude is limited to within 2% of the rated phase voltage, preferably 1%; the duration of a single pulse is controlled between 200 microseconds and 2 milliseconds, preferably 500 microseconds; the time interval between adjacent pulses is set between 20 milliseconds and 200 milliseconds, preferably 20 milliseconds consistent with a power frequency cycle; the polarity of adjacent pulses strictly alternates, with a polarity ratio of 1 to 1. Under the above time constraints, the injection window is opened at the zero-crossing instant and the injection signal output by the micro pulse injection unit is recorded synchronously, forming an injection signal record, while the injection energy density is controlled to be below 1 millijoule per phase per injection, preferably below 0.3 millijoule, to ensure that the impact of injection on the primary circuit is controlled.

[0069] Under the same synchronous clock, the voltage and current changes of the primary circuit caused by the injection sequence are collected at the predetermined measurement point to form a response signal, and the out-of-window components are suppressed by gated sampling to obtain a response signal record. To improve the time registration accuracy, the zero-crossing synchronization error is controlled to be within 200 microseconds, preferably within 100 microseconds; the gated time window is set as a symmetric interval around the zero-crossing point with a width of 1-5 milliseconds before and after, preferably 2 milliseconds each; the sampling frequency is not less than 20 kHz, preferably 50 kHz. Then, the injection signal record and the response signal record are time-registered according to the zero-crossing synchronization reference, and the two types of records are aligned to the same time axis, only the in-window segments corresponding to the pulses are extracted, and finally the zero-crossing synchronization reference sequence and the initial response sequence are obtained, providing high signal-to-noise ratio and high time consistency input data for subsequent decoupling processing and parameter identification.

[0070] Optionally, in step 200 of the embodiment, after obtaining the initial response sequence, a phase alignment reference is established with the phases of the three-phase voltage as the unified reference. By detecting the phase difference of the three-phase voltage zero-crossing points and correcting the phase synchronization, the phases are made to be at the equivalent time point at the zero-crossing moment, so as to eliminate the sampling deviation among the phases. On this basis, a symmetric gating time window is constructed around the zero-crossing moment, and a pairing relationship corresponding to each phase is determined, so as to ensure the one-to-one correspondence of the three-phase data in the time domain and the phase at the time of differential gating sampling. The gating time window is preferably set to an interval of 2 milliseconds before and after the zero-crossing point, and the overall time symmetry error is not more than 100 microseconds, so as to ensure that the sampling window covers the main components of the polarization response and avoids external noise interference.

[0071] In the process of planned switching-on or load transition, the embodiment uses the transient event as a natural excitation trigger source, extracts the time slices before and after the transition from the initial response sequence and forms a differential observation pair. The differential observation pair is used to represent the response change of the same branch caused by the working condition disturbance in a short time, and can reflect the coupling characteristics between the voltage mutual inductance loop and the protection quantity signal. By comparing the amplitude change and phase drift of the slices before and after the transition, the embodiment extracts the response increment reflecting the coupling channel characteristics, and provides input data for subsequent self-calibration decoupling operator parameter calculation. The typical transition time is not more than 10 milliseconds, and the sampling frequency is maintained at 50 kHz, so as to ensure that the time resolution of the differential slices is sufficient.

[0072] After obtaining the differential observation pair, the embodiment calculates the transfer characteristics of the voltage mutual inductance loop coupling and the protection quantity coupling according to the relative change ratio, phase delay and correlation degree information of the voltage and current in the observation pair, and determines the parameter set of the self-calibration decoupling operator. The parameter set is used to describe the amplitude correlation and dynamic response of various coupling channels in the system, and on this basis, the initial response sequence is subjected to decoupling operation to filter out the interference components from the mutual inductor coupling and protection channel feedback, so as to obtain a high-purity decoupled response sequence. The signal-to-noise ratio of the response sequence after decoupling is improved by 3 to 5 times, and the phase jitter is less than 0.5 degrees, which provides a stable and reliable input for subsequent frequency domain analysis and parameter identification.

[0073] Further, in step 300 of the embodiment, firstly, according to the actual wiring of the primary circuit and the switch position, a "reachable matrix" is defined as a set of structural constraints for describing the electrical connection relationship and the open state, wherein the on-off relationship of each feeder, bus section and branch is corresponded to the judgment rule of whether the connection path exists in the binary state. By reading the current positions of the disconnectors, circuit breakers and contactors, the reachable matrix corresponding to the current time is generated; when the operation mode changes (for example, the bus tie switch is switched from split-column to parallel-column or a feeder is removed for maintenance), the reachable matrix is immediately reconstructed, so that the subsequent analysis is always consistent with the actual electrical channel of the primary system, and the change time is marked for time sequence management to avoid data confusion under different operation modes.

[0074] At the branch level, the embodiment establishes a "branch sensitivity kernel" for representing the contribution intensity and phase characteristics of each branch to the decoupling response. The composition of the branch sensitivity kernel is based on known or estimable parameters such as branch ground admittance, dielectric loss characteristics and geometric distribution. Type values are given for similar equipment through field archives and existing measurements, and the dimensions of different sections are scaled uniformly to make the response contributions of each branch in the same evaluation window comparable. To improve robustness, the branch sensitivity kernel sets a minimum identifiable threshold for low-contribution branches and a saturation upper limit for high-contribution branches to ensure that subsequent regression is not dominated by extreme weights.

[0075] Under the joint constraints of the reachable matrix and the branch sensitivity kernel, the embodiment implements joint frequency-domain and time-domain sparse regression on the decoupling response sequence. The frequency-domain focuses on the amplitude and phase mode of the fundamental and several characteristic frequency bands, reflecting the frequency band selectivity of the dielectric loss angle slice; the time-domain focuses on the slow decay pattern of the polarization and depolarization process, reflecting the distribution of the polarization time constant. The embodiment takes the equipment as the minimum merging unit, constrains the parameters within the same equipment to be consistent, and suppresses the parameter leakage across equipment, thereby directly obtaining the estimated values of the equipment-level polarization time constant, the estimated values of the leakage admittance and the estimated values of the dielectric loss angle slice, and retaining the estimation residual and the constraint satisfaction degree as quality indicators.

[0076] To ensure the usability and verifiability of the parameter results, the embodiment performs stability test and residual consistency test on the above-mentioned estimated values in multiple adjacent time windows: when the parameter fluctuation of the same equipment in adjacent windows is below a preset threshold and the residual distribution does not show systematic deviation, the parameter estimation of the equipment is confirmed to be valid; when the operation mode changes, only the parameters that satisfy the consistency condition before and after the change are included in the equipment parameter vector. Finally, the parameter estimates of each equipment are summarized according to the equipment number and installation position to form the equipment parameter vector, and a position mapping is established for precise pointing of subsequent evaluation, early warning and maintenance instructions.

[0077] Further, in the processing of the decoupling response sequence, the embodiment first constructs a joint dictionary containing both frequency domain and time domain characteristics according to the branch sensitivity kernel. The frequency domain part selects the fundamental wave and multiple characteristic frequency bands as the amplitude and phase analysis base, typically including the fundamental wave, the 3rd, 5th and 7th frequency bands; the time domain part selects multiple exponential decay prototypes as the polarization response base, typically 64 in number, with the corresponding time constants discretely distributed between 1 millisecond and 200 milliseconds. To ensure the stability and physical reasonableness of the regression calculation, the confidence weight of the observation sample is distributed between 0.6 and 1.0, and is dynamically adjusted according to the steady state degree of device operation; the initial value of the regularization strength is between 0.001 and 0.01, and is automatically fine-tuned according to the data sparsity in subsequent calculations.

[0078] Under the joint constraint of the joint dictionary and the reachable matrix, sparse regression is performed on the observation data composed of the decoupling response sequence. To ensure clear physical meaning, the embodiment groups multiple base coefficients of the same device on the same physical quantity as a minimum constraint unit, and maintains consistency within the group during the solving process, thereby avoiding parameter leakage across devices or across physical quantities. Through this method, amplitude and phase characteristics related to dielectric loss can be extracted in the frequency domain, and slow decay characteristics related to the polarization process can be extracted in the time domain, ultimately obtaining three types of estimation results of device-level polarization time constant, leakage admittance and dielectric loss angle slice.

[0079] The embodiment further aggregates the parameter groups obtained by regression according to the device number, and establishes a parameter mapping relationship based on the installation location, thereby forming a complete device parameter vector. Field tests show that the typical results of the polarization time constant are concentrated in the 5 millisecond to 100 millisecond interval, the order of magnitude of the leakage admittance is about 1 x 10 -5 to 5 x 10 -4 The calculation result of the dielectric loss angle slice is about 0.002 to 0.010. It runs stably under the condition that the sample number is 200 to 500 and the joint base number is 120 to 180, and the calculation residual variance does not exceed 3%, which can effectively reflect the insulation difference and dielectric characteristics of each device, and provide accurate physical parameter basis for subsequent health assessment and trend warning.

[0080] Specifically, in step 400 of the embodiment, the device parameter vector composed of the three estimated values of the device-level polarization time constant, leakage admittance and dielectric loss angle slice is taken as the observation quantity, and statistical estimation is performed under a hierarchical Bayesian model. First, the historical measurement results of the same type of device are taken as the group prior, and the posterior statistics of the device type layer are obtained through iterative convergence, which are used to represent the parameter distribution law of the device under actual operating conditions. Then, the observed parameters of each device are compared with the posterior statistics of the type layer to calculate the deviation of each device in the group distribution, thereby describing the difference of the individual device relative to the average level of the type. The posterior statistics include the posterior mean and the posterior covariance of the device type layer, which together constitute a quantitative expression of the uncertainty of the insulation state.

[0081] After obtaining the posterior statistics, the embodiment determines the relative position of the device in the group distribution by calculating the multi-dimensional deviation between the observed parameters and the posterior mean, combined with the correlation between parameters reflected by the posterior covariance. The smaller the deviation, the higher the consistency of the insulation state of the device with the same type of device; the larger the deviation, the more obvious the insulation performance of the device decays. In order to facilitate engineering interpretation, the embodiment maps the deviation to a health index and an insulation potential barrier decay rate, wherein the health index varies between 0 and 1, and the value closer to 1 indicates a healthier insulation state; the insulation potential barrier decay rate is equal to 1 minus the health index, which is used to intuitively reflect the degree of insulation degradation. For example, when the health index is greater than 0.8, the device is determined to be in a normal state; when the health index is between 0.3 and 0.8, it is determined to be slightly degraded; when the health index is less than 0.3, it is determined to be significantly degraded, and state maintenance needs to be arranged.

[0082] To reflect the uncertainty of the model, the embodiment calculates the confidence interval of the health index according to the hierarchical Bayesian posterior distribution, and determines the upper and lower limits by using the quantile of the non-central chi-square distribution. In a typical case, when the parameter dimension is 3 and the significance level is 0.05, the 95% confidence interval of the health index can be calculated by the quantile. When the Mahalanobis distance square is less than about 2.4, the health index is greater than 0.3, which can be considered as a normal fluctuation range; when the Mahalanobis distance square is between 5 and 9, the health index is about 0.08 to 0.01, corresponding to a warning state; when the Mahalanobis distance square exceeds 11, the health index is less than 0.01, corresponding to a serious degradation state, at this time the insulation potential barrier decay rate is close to 1, and the system automatically outputs a state maintenance instruction. The confidence interval mechanism ensures the robustness and statistical repeatability of the evaluation results, and provides a quantifiable basis for device health classification and risk prediction.

[0083] Optionally, in step 500 of this embodiment, after obtaining the device-level evaluation results, the evaluation results of each device are first mapped to its corresponding number and installation location based on the location mapping, generating a device-level evaluation list. Location mapping is used to establish the correlation between the evaluation results and the physical location of the devices, ensuring that health indices, insulation barrier decay rates, and their confidence intervals can be accurately mapped to specific devices. Through this mapping, devices exhibiting abnormalities can be quickly located during system operation, ensuring that the source and unique attribution of each parameter are clear.

[0084] Subsequently, this embodiment uses the health index and insulation barrier decay rate as the main judgment criteria, combined with the changing trends within the confidence interval and sliding time window, to classify the insulation status of the equipment. The classification judgment adopts a multi-time period trend verification method, that is, comparing the rate and direction of change of the health index within adjacent time windows. When the index shows a continuous decline or fluctuations exceeding the confidence interval, the abnormality level of the equipment is determined. The abnormality levels are divided into three categories: normal, slightly abnormal, and severely abnormal, with corresponding health index thresholds greater than 0.8, between 0.3 and 0.8, and below 0.3, respectively, reflecting different stages of insulation degradation.

[0085] For equipment identified as abnormal, this embodiment generates an early warning message, including the equipment number, installation location, triggering criteria, and verification item instructions. The triggering criteria specify the main data characteristics leading to the abnormality determination, such as a continuous decline in the health index exceeding two time windows or an insulation barrier decay rate higher than 0.7. The verification item instructions include the types of parameters that should be retested, such as leakage admittance or dielectric loss angle slices, and specify the voltage level and ambient temperature range to be maintained during retesting to ensure the comparability of measurement results. The early warning message serves to remind maintenance personnel to conduct targeted testing and data verification while the system is running.

[0086] Based on the early warning results, this embodiment generates a condition-based maintenance instruction. This instruction includes maintenance timing instructions, retesting items and time windows, operating mode adjustment instructions, and patrol schedule arrangements. Maintenance timing is determined based on the anomaly level; equipment with minor anomalies requires retesting within 7 days, while equipment with severe anomalies requires retesting within 24 hours and a power outage inspection. Retesting items and time windows ensure the stability of diagnostic results. The operating mode adjustment instruction limits equipment load to 80% of rated power during retesting. Patrol schedules stipulate that data updates must be performed at least once every 48 hours until the anomaly is resolved. Through these steps, a continuous closed-loop process from equipment parameter assessment to early warning and maintenance is achieved, enabling insulation anomalies to be identified in advance and addressed within a controllable timeframe.

[0087] Corresponding to the above methods, such as Figure 2 As shown, this embodiment also provides a 6kV online insulation measurement system for plant use, including:

[0088] A micro-impulse injection and synchronous acquisition module is configured in parallel with an isolated micro-impulse injection unit at a neutral point of a 6kV power system, and a polarity-reversed pseudo-random micro-impulse sequence is injected into a primary loop of the 6kV power system at a voltage zero-crossing time window, and a synchronous clock is used to record the injected signal and a response signal, to obtain a zero-crossing synchronous reference sequence and an initial response sequence.

[0089] A phase difference sampling and self-calibration decoupling module is used to perform differential gated sampling based on three-phase phases, to generate a self-calibration decoupling operator by planned closing and load transition, to implement coupling suppression processing of a voltage mutual inductance loop and a protection quantity on the initial response sequence, to obtain a decoupling response sequence.

[0090] A topology constraint and sparse regression identification module is used to establish an accessibility matrix and a branch sensitivity kernel according to a primary system topology structure, to perform joint frequency domain and time domain sparse regression analysis on the decoupling response sequence, to identify device-level polarization time constants, leakage admittance and dielectric loss angle slices, to form a device parameter vector and a position mapping of the device parameter vector.

[0091] A hierarchical Bayesian evaluation and health calculation module is used to construct a hierarchical Bayesian estimation model for the device parameter vector, to calculate an insulation barrier decay rate and a health index, and to give a confidence interval, to obtain a device-level evaluation result.

[0092] A warning correlation and state maintenance instruction module is used to correspondingly correlate the device-level evaluation result according to the position mapping, to determine an insulation performance abnormal device object, and to output an early warning information and a state maintenance instruction of the device.

[0093] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0094] The principles and implementation manners of the present application are described by using specific examples in this paper. The above embodiment description is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range will be changed. In conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for on-line measurement of plant service 6 kV insulation, characterized by, The method comprises the following steps: A micro-pulse injection unit is configured in parallel with the neutral point of the 6kV power system, and a pseudo-random micro-pulse sequence with polarity reversal is injected into the primary circuit of the 6kV power system in the time window of voltage zero crossing, and the injection signal and the response signal are recorded with a synchronous clock to obtain a zero-crossing synchronous reference sequence and an initial response sequence; Difference gating sampling is performed based on the three-phase phase, self-calibration decoupling operators are generated by planned closing and load transition, coupling suppression processing of the voltage mutual inductance circuit and the protection quantity is performed on the initial response sequence to obtain a decoupling response sequence; A reachable matrix and a branch sensitivity kernel are established according to the topology of the primary system, and joint frequency domain and time domain sparse regression analysis is performed on the decoupling response sequence to identify the device-level polarization time constant, leakage admittance and dielectric loss angle slice, form a device parameter vector and a position mapping of the device parameter vector; A hierarchical Bayesian estimation model is constructed for the device parameter vector, the insulation barrier decay rate and the health index are calculated, and a confidence interval is given to obtain a device-level evaluation result; The device-level evaluation result is correspondingly associated according to the position mapping to determine the device object with abnormal insulation performance, and early warning information and state maintenance instructions of the device are output.

2. The method of claim 1, wherein the factory 6 kV insulation on-line measurement method is characterized by, A micro-pulse injection unit is configured in parallel with the neutral point of the 6kV power system, and a pseudo-random micro-pulse sequence with polarity reversal is injected into the primary circuit of the 6kV power system in the time window of voltage zero crossing, and the injection signal and the response signal are recorded with a synchronous clock to obtain a zero-crossing synchronous reference sequence and an initial response sequence, comprising: A zero-point detection reference is established by the micro-pulse injection unit based on the power frequency voltage of the primary circuit to determine the time mark and phase relationship of the zero-crossing transient to obtain a zero-crossing synchronous reference; A pseudo-random micro-pulse sequence with polarity reversal is generated by the micro-pulse injection unit under the constraint of safe injection parameters, and is aligned with the zero-crossing synchronous reference within the window to form a to-be-injected sequence; The to-be-injected sequence is injected into the primary circuit through the parallel isolation channel of the micro-pulse injection unit, and the injection signal is recorded using the synchronous clock to obtain an injection signal record; The voltage response and current response of the primary circuit to the pseudo-random micro-pulse sequence are collected at the predetermined measurement point under the sampling reference consistent with the synchronous clock to form a response signal, and the out-of-window components are suppressed by gated sampling to obtain a response signal record; The injection signal record and the response signal record are time-registered according to the zero-crossing synchronous reference to extract the in-window sequence to obtain the zero-crossing synchronous reference sequence and the initial response sequence.

3. The method of claim 1, wherein the factory 6 kV insulation on-line measurement method is characterized by, Difference gating sampling is performed based on the three-phase phase, self-calibration decoupling operators are generated by planned closing and load transition, coupling suppression processing of the voltage mutual inductance circuit and the protection quantity is performed on the initial response sequence to obtain a decoupling response sequence, comprising: A phase alignment reference is established based on the three-phase phase to determine the symmetric gating time window around the zero crossing and the three-phase pairing relationship; The initial response sequence is extracted before and after the transition triggered by the planned closing and load transition, and a difference observation pair is formed; The transfer characteristic of the voltage mutual inductance loop coupling and the protection quantity coupling is estimated according to the difference observation pair, and the parameter set of the self-calibration decoupling operator is determined; The decoupling operation is performed on the initial response sequence according to the self-calibration decoupling operator, and the decoupling response sequence is obtained.

4. The method of claim 1, wherein the factory 6 kV insulation on-line measurement method is characterized by, The reachability matrix and branch sensitivity kernel are established according to the topology of the primary system, and the joint frequency domain and time domain sparse regression analysis is performed on the decoupling response sequence to identify the device-level polarization time constant, leakage admittance and dielectric loss angle slice, and form the device parameter vector and the position mapping of the device parameter vector, including: The reachability matrix is generated according to the switching position and connectivity of the primary circuit, and the reachability matrix is updated when the operating mode changes; The branch sensitivity kernel is established according to the branch ground admittance and dielectric loss characteristics, and the dimension of different sections is unified; Under the joint constraints of the reachability matrix and the branch sensitivity kernel, the joint frequency domain and time domain sparse regression is performed on the decoupling response sequence to obtain the estimated value of the device-level polarization time constant, the estimated value of the leakage admittance and the estimated value of the dielectric loss angle slice; The effectiveness of each estimated value is confirmed by the stability test and residual consistency test across the time window, the estimated values are summarized as the device parameter vector according to the device number and installation position, and the position mapping of the device parameter vector is established.

5. The method of claim 4, wherein the factory 6 kV insulation on-line measurement method is characterized by, Under the joint constraints of the reachability matrix and the branch sensitivity kernel, the joint frequency domain and time domain sparse regression is performed on the decoupling response sequence to obtain the estimated value of the device-level polarization time constant, the estimated value of the leakage admittance and the estimated value of the dielectric loss angle slice, including: The joint dictionary containing the frequency domain basis and the time domain basis is constructed according to the branch sensitivity kernel, and the basis set that can participate in regression is determined under the reachability relationship given by the reachability matrix; Under the constraints of the reachable relation and the joint dictionary, sparse regression is performed on an observation vector formed by the decoupled response sequence, where a coefficient sub-vector divided by equipment and physical quantity is taken as a minimum unit of sparse constraint, and a formula is as follows: ; wherein, is an observation vector formed by the decoupled response sequence; is a joint dictionary matrix constructed according to the branch sensitivity kernel; is a to-be-solved basis coefficient vector; is an optimal solution of the basis coefficient vector; is an index set of a coefficient sub-vector divided by equipment and physical quantity; is a coefficient sub-vector with an index being ; and is a diagonal weight matrix with observation sample confidence as weight; is a sparse regularization coefficient. The obtained coefficient sub-vectors are merged according to the device number and mapped as each estimated value of the device-level parameter, and the formula is: ; wherein, is the estimated value of the device-level polarization time constant; is the estimated value of the device-level leakage admittance; is the estimated value of the device-level dielectric loss angle slice; are the coefficient sub-vector index sets corresponding to the polarization time constant, the leakage admittance and the dielectric loss angle slice, respectively; is the discrete candidate value of the polarization time constant base; is the discrete candidate value of the dielectric loss angle slice base; are the base coefficients corresponding to the index set, respectively.

6. The method of claim 1, wherein the 6 kV insulation on-line measurement method is characterized by, The hierarchical Bayesian estimation model is constructed for the device parameter vector, the insulation barrier decay rate and the health index are calculated, and the confidence interval is given to obtain the device-level evaluation result, including: The device parameter vector composed of the estimated value of the device-level polarization time constant, the estimated value of the leakage admittance and the estimated value of the dielectric loss angle slice is taken as the observation, and the population prior of the same type of device is converged under the hierarchical Bayesian framework to obtain the posterior mean and the posterior covariance of the device type layer, which is used to describe the deviation degree of the device in the population distribution; A Mahalanobis distance is calculated based on the posterior mean and the posterior covariance, and the Mahalanobis distance is mapped to a health index and an insulation barrier decay rate, with a formula being: ; wherein, is the device parameter vector; is a device type layer posterior mean obtained through hierarchical Bayesian estimation; is a device type layer posterior covariance obtained through hierarchical Bayesian estimation; is a square of the Mahalanobis distance calculated under the posterior statistics obtained through hierarchical Bayesian estimation; is a device level health index; is an insulation barrier decay rate; the posterior statistics include the device type layer posterior mean and the device type layer posterior covariance; A confidence interval of the health index is given according to the hierarchical Bayesian posterior uncertainty, which is formulated as: ; wherein, is a quantile of a non-central chi-squared distribution at quantile point γ; is a non-central chi-squared distribution with degrees of freedom and non-central parameter ; and is the dimension of the equipment parameter vector; takes the squared Mahalanobis distance is a significance constant corresponding to a confidence level.

7. The method of claim 1, wherein the factory 6 kV insulation on-line measurement method is characterized by, The device-level evaluation result is correspondingly associated according to the position mapping to determine the device object with abnormal insulation performance, and the advance warning information and the state maintenance instruction of the device are output, including: The device-level evaluation list is formed according to the position mapping of the device-level evaluation result; The insulation barrier decay rate and the health index are taken as the judgment basis, and the hierarchical judgment is implemented combined with the confidence interval and the change trend in the sliding time window to determine the device object with abnormal insulation performance and the abnormal level. Generate early warning information containing device identification, installation location, trigger basis and review item indication according to the determined abnormal level; Give state maintenance instructions according to the abnormal level; Include the early warning information and the state maintenance instructions in the operation and maintenance records.

8. The method of claim 1, wherein the 6 kV insulation on-line measurement method is characterized by, The state maintenance instructions include maintenance opportunity indication, retest items and retest time window, operation mode adjustment indication and patrol video frequency.

9. An on-line measurement system for 6 kV plant insulation, characterized in that, Comprise: Micro pulse injection and synchronous acquisition module, configured with parallel isolated micro pulse injection unit at the neutral point side of the plant 6kV power system, injects a pseudo-random micro pulse sequence with polarity reversal into the primary circuit of the plant 6kV power system with voltage zero crossing as the time window, and records the injection signal and the response signal with a synchronous clock to obtain the zero crossing synchronization reference sequence and the initial response sequence; Phase difference sampling and self-calibration decoupling module, for differential gated sampling based on three-phase phase, using planned closing and load transition to generate self-calibration decoupling operator, to implement voltage mutual inductance loop and protection quantity coupling suppression processing on the initial response sequence, to obtain decoupling response sequence; Topology constraint and sparse regression identification module, for establishing reachability matrix and branch sensitivity kernel according to the primary system topology structure, performing joint frequency domain and time domain sparse regression analysis on the decoupling response sequence, identifying device level polarization time constant, leakage admittance and dielectric loss angle slice, forming device parameter vector and position mapping of the device parameter vector; Hierarchical Bayesian evaluation and health calculation module, for constructing hierarchical Bayesian estimation model for the device parameter vector, calculating insulation barrier decay rate and health index, and giving confidence interval to obtain device level evaluation result; Early warning correlation and state maintenance instruction module, for corresponding correlation of the device level evaluation result according to the position mapping, determining the insulation performance abnormal device object, and outputting the early warning information and the state maintenance instruction of the device.

Images