A switch cabinet insulation state early warning method and device

Abstract

The application discloses a kind of switch cabinet insulation state early warning method and device, for solving the existing switch cabinet insulation state early warning method, resulting in the technical problem of low warning accuracy.Method includes based on pre-set classification threshold, the historical annual switch cabinet sample data set obtained is preprocessed, generates multiple cut switch cabinet characteristic parameter set and multiple classification switch cabinet characteristic parameter set;According to pre-set error criterion, pre-set classification threshold, pre-set aging weight factor, multiple classification switch cabinet characteristic parameter set and multiple cut switch cabinet characteristic parameter set, construct initial aging factor prediction model and initial partial discharge model;Based on pre-set error criterion and pre-set correction criterion, using initial aging factor prediction model and initial partial discharge model according to received current annual switch cabinet sample data set, determine target aging factor and target partial discharge model, and combine pre-set early warning criterion, generate switch cabinet insulation state early warning result.

Description

Technical Field

[0001] This invention relates to the field of power equipment technology, and in particular to a method and device for early warning of insulation status of switchgear. Background Technology

[0002] Switchgear is a crucial type of equipment in power distribution systems, and its safe and reliable operation has become a key indicator for evaluating the level of power distribution automation. A survey of the current operational status and safety hazards of existing switchgear reveals that insulation aging is the most common and impactful type of fault in switchgear, affecting verification. It not only impacts equipment performance but also, due to insufficient insulation clearance, can lead to partial discharge, jeopardizing the safe operation of the equipment.

[0003] As switchgear operates for extended periods, its insulation gradually deteriorates due to environmental factors. Failure to monitor the switchgear's condition promptly can lead to unpredictable accidents. When insulation defects occur, partial discharge signals are generated internally. Detecting these signals allows for early detection of insulation defects within the switchgear, enabling maintenance before more serious insulation faults develop, thus preventing significant economic losses.

[0004] Existing methods for early warning of switchgear insulation status mainly rely on established partial discharge models. However, many of these models only use time variables or operating conditions for rough assessment, which weakens the influence of insulation aging factors and makes it difficult to capture microstructural changes caused by aging. This results in significant deviations in the prediction of equipment status over long-term operation, leading to low accuracy in early warning. Summary of the Invention

[0005] This invention provides a method and apparatus for early warning of insulation status of switchgear, which solves the technical problem of low accuracy in existing methods for early warning of insulation status of switchgear.

[0006] The first aspect of this invention provides a method for early warning of the insulation status of a switchgear, comprising:

[0007] Obtain a historical annual switchgear sample dataset, and preprocess the historical annual switchgear sample dataset based on a preset classification threshold to generate multiple segmented switchgear feature parameter sets and multiple classified switchgear feature parameter sets;

[0008] An initial aging factor prediction model is constructed based on the preset classification threshold, the preset aging weight factor, and multiple sets of feature parameters of the split switchgear.

[0009] Based on the preset classification threshold and preset error criteria, the initial aging factor prediction model is used to construct an initial partial discharge model according to multiple sets of characteristic parameters of the classified switchgear.

[0010] When the current year's switchgear sample dataset is received, based on the preset error criterion and preset correction criterion, the target aging factor and target partial discharge model are determined according to the current year's switchgear sample dataset using the initial aging factor prediction model and the initial partial discharge model.

[0011] The pre-set early warning criteria are used to generate early warning results for the insulation status of the switchgear based on the target aging factor and the weight factor matrix of the target partial discharge model.

[0012] Optionally, the multiple sets of feature parameters for classified switchgear include multiple sets of feature parameters for normal switchgear, multiple sets of feature parameters for abnormal switchgear, and multiple sets of feature parameters for sensitive switchgear; the step of preprocessing the historical annual switchgear sample dataset based on a preset classification threshold to generate multiple sets of segmented switchgear feature parameters and multiple sets of classified switchgear feature parameters includes:

[0013] The historical annual switchgear sample dataset is segmented to generate multiple segmented switchgear feature parameter sets;

[0014] Based on the preset classification threshold, the feature parameter sets of the multiple segmented switchgear are classified to generate multiple feature parameter sets of normal switchgear, multiple feature parameter sets of abnormal switchgear, and multiple feature parameter sets of sensitive switchgear.

[0015] Optionally, the step of constructing an initial aging factor prediction model based on the preset classification threshold, the preset aging weight factor, and multiple sets of feature parameters of the segmented switchgear includes:

[0016] Based on the preset classification threshold, the feature parameter sets of multiple switch cabinets are normalized to generate multiple normalized feature parameter sets of switch cabinets.

[0017] The normalized switchgear feature parameter sets are weighted and output as a weighted switchgear feature parameter set.

[0018] An initial aging factor prediction model is constructed based on the weighted switchgear feature parameter set and the preset aging weight factor.

[0019] Optionally, the initial partial discharge model includes a first partial discharge model, a second partial discharge model, and a third partial discharge model; the step of constructing the initial partial discharge model based on the preset classification threshold and preset error criterion, using the initial aging factor prediction model according to multiple sets of characteristic parameters of the classified switchgear, includes:

[0020] Based on the preset classification threshold, the feature parameter sets of multiple normal switchgear, multiple abnormal switchgear, and multiple sensitive switchgear are normalized to generate multiple normalized switchgear feature parameter sets, multiple abnormalized switchgear feature parameter sets, and multiple sensitive normalized switchgear feature parameter sets.

[0021] The initial aging factor is generated using the initial aging factor prediction model based on multiple sets of characteristic parameters of the normal switchgear.

[0022] The first partial discharge model is constructed using the preset error criterion based on the initial aging factor and multiple sets of normalized switchgear characteristic parameters;

[0023] The weight factor matrix of the first partial discharge model is modified using the initial aging factor and multiple sets of abnormal normalized switchgear feature parameters to determine the second partial discharge model.

[0024] Based on the preset error criterion, the weight factor matrix of the second partial discharge model is modified using the initial aging factor and multiple sets of sensitive normalized switchgear feature parameters to determine the third partial discharge model.

[0025] Optionally, the step of constructing a first partial discharge model based on the preset error criterion according to the initial aging factor and multiple sets of normalized switchgear characteristic parameters includes:

[0026] The normalized switchgear feature parameter sets are divided into multiple sets, and multiple first-divided switchgear feature parameter sets and multiple second-divided switchgear feature parameter sets are output.

[0027] Based on the initial aging factor and multiple first-division switchgear feature parameter sets, a first intermediate partial discharge model is constructed.

[0028] The first intermediate partial discharge model is used to output multiple first transient ground voltage prediction values ​​based on multiple second partitioned switch cabinet feature parameter sets and the initial aging factor.

[0029] Based on multiple predicted values ​​of the first transient ground voltage, multiple first error values ​​are calculated;

[0030] Determine whether multiple first error values ​​satisfy the preset error criterion;

[0031] If so, then the first intermediate partial discharge model is used as the first partial discharge model.

[0032] Optionally, the step of modifying the weight factor matrix of the first partial discharge model using the initial aging factor and multiple sets of abnormally normalized switchgear feature parameters to determine the second partial discharge model includes:

[0033] The abnormal normalized switchgear feature parameter sets are divided into multiple sets, and multiple third-division switchgear feature parameter sets and multiple fourth-division switchgear feature parameter sets are output.

[0034] The first partial discharge model is updated by using multiple sets of third partition switchgear feature parameters to determine the second intermediate partial discharge model.

[0035] The second intermediate partial discharge model is used to output multiple second transient ground voltage prediction values ​​based on the initial aging factor and multiple sets of fourth partitioned switchgear characteristic parameters.

[0036] Based on multiple predicted values ​​of the second transient ground voltage, multiple second error values ​​are calculated;

[0037] Based on each of the second error values, the weight factor matrix of the second intermediate partial discharge model is updated sequentially to determine the third intermediate partial discharge model, and the number of model updates is counted in real time.

[0038] Determine whether the number of model updates has reached a preset threshold;

[0039] If so, then the third intermediate partial discharge model is used as the second partial discharge model.

[0040] Optionally, the step of modifying the weight factor matrix of the second partial discharge model based on the preset error criterion using the initial aging factor and multiple sets of sensitive normalized switchgear feature parameters to determine the third partial discharge model includes:

[0041] The multiple sets of sensitive normalized switchgear feature parameters are divided, and multiple fifth-division switchgear feature parameter sets and multiple sixth-division switchgear feature parameter sets are output.

[0042] The second partial discharge model is updated by using multiple sets of the fifth partition switchgear feature parameters to determine the fourth intermediate partial discharge model.

[0043] The fourth intermediate partial discharge model is used to output multiple third transient ground voltage prediction values ​​based on the initial aging factor and multiple sixth division switch cabinet characteristic parameter sets.

[0044] Based on the multiple predicted values ​​of the third transient ground voltage, multiple third error values ​​are calculated;

[0045] Determine whether the multiple third error values ​​satisfy the preset error criterion;

[0046] If so, then the fourth intermediate partial discharge model is used as the third partial discharge model.

[0047] Optionally, the target partial discharge model includes a new second partial discharge model and a new third partial discharge model; the step of determining the target aging factor and the target partial discharge model based on the preset error criterion and preset correction criterion, using the initial aging factor prediction model and the initial partial discharge model according to the current year's switchgear sample dataset, includes:

[0048] The initial aging factor prediction model is used to output the current year's aging factor based on the current year's switchgear sample dataset, and based on the current year's aging factor, multiple insulation resistance calculation values ​​are output.

[0049] Based on multiple calculated insulation resistance values ​​and multiple measured insulation resistance values ​​in the current year's switchgear sample dataset, multiple target error values ​​are calculated.

[0050] Determine whether the multiple target error values ​​satisfy the preset error criterion;

[0051] If the conditions are met, the current year's aging factor will be used as the intermediate aging factor.

[0052] The current year's switchgear sample dataset is classified to generate multiple sets of feature parameters for normal switchgear, multiple sets of feature parameters for abnormal switchgear, and multiple sets of feature parameters for sensitive switchgear.

[0053] The first partial discharge model is used to generate multiple predicted aging factors based on multiple sets of characteristic parameters of normal switchgear in the current year;

[0054] Determine whether the intermediate aging factor and the plurality of predicted aging factors satisfy the preset error criterion;

[0055] If the conditions are met, the average of the multiple predicted aging factors is taken as the target aging factor.

[0056] The second partial discharge model and the third partial discharge model are modified by weight factor matrix based on multiple sets of current year abnormal switchgear feature parameters and multiple sets of current year sensitive switchgear feature parameters, using the preset correction criteria to determine new second partial discharge models and new third partial discharge models.

[0057] Optionally, the preset early warning criteria include a first early warning criterion and a second early warning criterion; the switchgear insulation status early warning results include annual insulation status early warning results and individual parameter insulation status early warning results; the step of generating switchgear insulation status early warning results based on the target aging factor and the weight factor matrix of the target partial discharge model using the preset early warning criteria includes:

[0058] Based on the target aging factor, determine the aging factor for future years;

[0059] Determine whether the target aging factor and the future annual aging factor meet the first early warning criterion;

[0060] If the conditions are met, an annual insulation status early warning result will be generated;

[0061] Determine whether the weight factor matrix of the new second partial discharge model and the weight factor matrix of the new third partial discharge model satisfy the second early warning criterion;

[0062] If the conditions are met, a single parameter insulation status early warning result will be generated.

[0063] A second aspect of the present invention provides a switchgear insulation status early warning device, comprising:

[0064] The acquisition module is used to acquire a historical annual switchgear sample dataset and preprocess the historical annual switchgear sample dataset based on a preset classification threshold to generate multiple segmented switchgear feature parameter sets and multiple classified switchgear feature parameter sets.

[0065] The first construction module is used to construct an initial aging factor prediction model based on the preset classification threshold, the preset aging weight factor, and multiple sets of feature parameters of the split switchgear.

[0066] The second construction module is used to construct an initial partial discharge model based on the preset classification threshold and preset error criteria, using the initial aging factor prediction model according to multiple sets of characteristic parameters of the classified switchgear.

[0067] The determination module is used to determine the target aging factor and the target partial discharge model based on the preset error criteria and preset correction criteria, using the initial aging factor prediction model and the initial partial discharge model according to the current year's switchgear sample dataset when the current year's switchgear sample dataset is received.

[0068] The early warning module is used to generate early warning results for the insulation status of the switchgear based on the target aging factor and the weight factor matrix of the target partial discharge model using preset early warning criteria.

[0069] As can be seen from the above technical solutions, the present invention has the following advantages:

[0070] The above-described technical solution of the present invention provides a method for early warning of the insulation status of switchgear. First, a historical annual switchgear sample dataset is acquired, and based on a preset classification threshold, the historical annual switchgear sample dataset is preprocessed to generate multiple sets of feature parameters for segmented switchgear and multiple sets of feature parameters for categorized switchgear. Next, an initial aging factor prediction model is constructed based on the preset classification threshold, preset aging weight factor, and multiple sets of feature parameters for segmented switchgear. Based on the preset classification threshold and preset error criterion, an initial partial discharge model is constructed using the initial aging factor prediction model and the multiple sets of feature parameters for categorized switchgear. When the current year's switchgear sample dataset is received, based on the preset error criterion and preset correction criterion, the initial aging factor prediction model and the initial partial discharge model are used to predict the insulation status of the switchgear for the current year. The process involves using a sample dataset of switchgear cabinets to determine the target aging factor and target partial discharge model. Finally, a pre-set early warning criterion is used to generate an early warning result for the switchgear insulation status based on the weight factor matrix of the target aging factor and the target partial discharge model. Based on this scheme, an initial aging factor prediction model is constructed using a pre-processed set of segmented switchgear feature parameters. An initial partial discharge model is then constructed by combining a pre-set classification threshold, a pre-set error criterion, and a set of classified switchgear feature parameters. Finally, the weight factor matrix of these two models is corrected to output the early warning result for the switchgear insulation status. This invention establishes an aging factor prediction model that can output aging factor prediction results, which is used to establish the intrinsic relationship between partial discharge and aging in the partial discharge model, thereby achieving a reliable assessment of the switchgear aging status and improving the accuracy of early warnings. Attached Figure Description

[0071] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0072] Figure 1 This is a flowchart illustrating the steps of a switchgear insulation status early warning method according to Embodiment 1 of the present invention.

[0073] Figure 2 This is a schematic diagram of the process for establishing and identifying the parameters of the partial discharge model of the switchgear provided in Embodiment 1 of the present invention;

[0074] Figure 3 This is a schematic diagram illustrating the application iterative calculation and parameter correction process of the model provided in Embodiment 1 of the present invention.

[0075] Figure 4 This is a flowchart illustrating the switchgear insulation status early warning method provided in Embodiment 1 of the present invention;

[0076] Figure 5 This is a structural block diagram of a switchgear insulation status early warning device provided in Embodiment 2 of the present invention. Detailed Implementation

[0077] This invention provides a method and apparatus for early warning of insulation status of switchgear, which solves the technical problem of low accuracy in existing methods for early warning of insulation status of switchgear.

[0078] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0079] Please see Figure 1 , Figure 1 The flowchart illustrates the steps of a switchgear insulation status early warning method provided in Embodiment 1 of the present invention.

[0080] This invention provides a method for early warning of insulation status in switchgear, comprising:

[0081] Step 101: Obtain the historical annual switchgear sample dataset, and preprocess the historical annual switchgear sample dataset based on the preset classification threshold to generate multiple segmented switchgear feature parameter sets and multiple classified switchgear feature parameter sets.

[0082] The multiple classification switchgear feature parameter sets include multiple normal switchgear feature parameter sets, multiple abnormal switchgear feature parameter sets, and multiple sensitive switchgear feature parameter sets.

[0083] It should be noted that the historical annual switchgear sample dataset is a dataset A of switchgear operation samples of the same type but with different operating years. The switchgear sample data parameters in the historical annual switchgear sample dataset are: operating years t at the time of sample collection, bus temperature a, humidity b, load rate c, bus contact resistance increment d, circuit breaker breaking times e, TEV value y (Transient Earth Voltage), and bus insulation resistance z.

[0084] Furthermore, the process of preprocessing the historical annual switchgear sample dataset based on a preset classification threshold to generate multiple sets of feature parameters for segmented switchgear and multiple sets of feature parameters for classified switchgear can be achieved by executing the following steps S11 to S12:

[0085] Step S11: Segment the historical annual switchgear sample dataset to generate multiple segmented switchgear feature parameter sets;

[0086] Step S12: Based on the preset classification threshold, classify the feature parameter sets of multiple switch cabinets to generate multiple normal switch cabinet feature parameter sets, multiple abnormal switch cabinet feature parameter sets, and multiple sensitive switch cabinet feature parameter sets.

[0087] Preset classification thresholds include anomaly thresholds and sensitivity thresholds.

[0088] It should be noted that, based on the outage maintenance cycle of busbar insulation resistance, the historical annual switchgear sample dataset A is segmented according to the time of the outage maintenance plan, resulting in multiple segmented switchgear feature parameter sets, which can be represented as A={A1, A2, ..., A...} j , ...};A j Let A be the set of all datasets (i.e., the set of characteristic parameters of the switchgear) within the j-th power outage maintenance plan. j ={z j-1 , z j , t j-1 , t j , t ji a ji b ji c ji d ji e ji y ji}, z j-1 z j The bus insulation resistance t was obtained from the (j-1)th and jth power outage maintenance tests, respectively. j-1 , t j These are the time points of the (j-1)th and jth power outage maintenance, respectively, t ji For the i-th data acquisition time within the j-th power outage maintenance plan, a ji b ji c ji d ji e ji y ji t respectively ji The parameter data collected at any given time.

[0089] Based on the anomaly threshold m and sensitivity threshold n in Table 1, the sample dataset A is... j The classification is divided into normal sample set A. j1 Abnormal sample set A j2 and sensitive sample A j3 That is, multiple normal switchgear characteristic parameter sets A j1 Multiple abnormal switchgear characteristic parameter sets Aj2 and multiple sensitive switchgear feature parameter sets A j3 Among them, if{a ji b ji c ji d ji e ji y ji} all ≤ m i (The i-th anomaly threshold) is classified as A. j1 Sample; if{a ji b ji c ji d ji e ji y ji} all ≤ n i (The i-th sensitivity threshold), and remove A. j1 The sample is classified as A. j2 Sample; the remaining samples are classified as A. j3 sample.

[0090] Table 1. Required sample parameters and data normalization

[0091]

[0092] Step 102: Construct an initial aging factor prediction model based on the preset classification threshold, preset aging weight factor, and multiple feature parameter sets of the switchgear.

[0093] It should be noted that the aging factor prediction model is mainly established by dividing the switchgear feature parameter set A. j Training is then conducted. By calculating the equivalent aging coefficients of each parameter during the j-th insulation resistance test and combining this with training using sample data, the model is established.

[0094] Specifically, step 102 may include the following sub-steps S21-S23:

[0095] Step S21: Normalize the feature parameter sets of multiple split switchgear based on the preset classification threshold to generate multiple normalized feature parameter sets of split switchgear.

[0096] Step S22: Perform weighted processing on multiple normalized switchgear feature parameter sets and output a weighted switchgear feature parameter set;

[0097] Step S23: Construct an initial aging factor prediction model based on the weighted switchgear feature parameter set and the preset aging weight factor.

[0098] It should be noted that, based on a preset classification threshold, the real-time monitoring parameters are normalized to obtain multiple normalized sets of switchgear feature parameters. In the aging factor prediction model, parameter a...ji b ji c ji d ji e ji y ji The value can be defined as 0, 1, or 2:

[0099] If the parameter value is less than m, the parameter value is defined as 0;

[0100] If m ≤ parameter value ≤ n, the parameter value is defined as 1;

[0101] If the parameter value is greater than n, the parameter value is defined as 2;

[0102] Furthermore, for the weighted processing of multiple normalized switchgear characteristic parameter sets, the average value of the parameter values ​​defined during the j-th insulation resistance test is calculated. That is, the weighted switchgear characteristic parameter set includes ,in, These are the average values ​​corresponding to busbar temperature (a), humidity (b), load rate (c), busbar contact resistance increment (d), circuit breaker interruption count (e), and TEV value (y), respectively. The operation rules are as follows Where N is the number of times the real-time monitoring parameters are collected during the j-th insulation resistance test. The processing procedure and The processing procedure is the same as that of the present invention, and will not be described in detail here.

[0103] Furthermore, the normalization rule for the measured bus insulation resistance values ​​in the characteristic parameter set of the switchgear is z / z0, where z0 is the original value of the insulation resistance when the switchgear is put into operation. Then, the time for the j-th insulation resistance test is t. j The test results are normalized to z. j / z0.

[0104] Furthermore, regarding the definition of the preset aging weight factor w, w is defined as follows: a w b w c w d w e w y Let be the aging weight factors for parameters a, b, c, d, e, and y, respectively. Then, the equivalent aging coefficients for each parameter during the j-th insulation resistance test are: *w a , *w b , *w c , *w d , *we , *w y .

[0105] Furthermore, regarding the definition of the aging factor, the aging factor in the j-th insulation resistance test is defined as z. *j = It can be calculated through training on sample data, and is represented as z. *j =f( *w a , *w b , *w c , *w d , *w e , *w y ).

[0106] In the model building and sample data training, the aging factor set Z={z} is selected. *j To output the observation results, select the parameter set X={ *w a , *w b , *w c , *w d , *w e , *w y} represents the input quantity, and the intermediate hidden layer is... , and Here, H represents the two key quantities that need to be identified in the hidden layer, and H is the output of the intermediate hidden layer. The function relationship between the hidden layer parameters is represented; the parametric model can be obtained through iterative solution using the sample set. ,in, Let X be the weights of the parameter set. This is a paranoid trait.

[0107] By training with the above sample data, the initial aging factor prediction model can be established, i.e., z *j =f( *w a , *w b , *w c , *w d , *w e , *w y ).

[0108] Step 103: Based on the preset classification threshold and preset error criterion, the initial aging factor prediction model is used to construct the initial partial discharge model according to the feature parameter set of multiple classified switch cabinets.

[0109] The initial partial discharge model includes the first partial discharge model, the second partial discharge model, and the third partial discharge model.

[0110] It should be noted that the following factors are considered: a) busbar temperature, b) humidity, c) load rate, d) busbar contact resistance increment, e) circuit breaker breaking count, and z) aging factor. * Using the TEV value y as the input parameter and the TEV value y as the output observation parameter, a partial discharge model of the switchgear is established, and iterative calculations are performed on the sample data. The sample data used is the normal sample set A. j1 Abnormal sample set A j2 and sensitive sample A j3 To perform bruise calculations, a hierarchical management method for sample data is used to simplify iterative solutions.

[0111] Specifically, step 103 may include the following sub-steps S31-S35:

[0112] Step S31: Normalize multiple normal switchgear feature parameter sets, multiple abnormal switchgear feature parameter sets, and multiple sensitive switchgear feature parameter sets based on preset classification thresholds to generate multiple normal normal switchgear feature parameter sets, multiple abnormal normalized switchgear feature parameter sets, and multiple sensitive normalized switchgear feature parameter sets.

[0113] Step S32: Using the initial aging factor prediction model, generate the initial aging factor based on multiple normal switchgear characteristic parameter sets;

[0114] Step S33: Using preset error criteria, construct the first partial discharge model based on the initial aging factor and multiple normalized switchgear characteristic parameter sets.

[0115] Furthermore, step S33 may include the following sub-steps S331-S336:

[0116] Step S331: Divide multiple normalized switchgear feature parameter sets and output multiple first-divided switchgear feature parameter sets and multiple second-divided switchgear feature parameter sets;

[0117] Step S332: Construct a first intermediate partial discharge model based on the initial aging factor and multiple first-division switchgear characteristic parameter sets;

[0118] Step S333: Using the first intermediate partial discharge model, based on multiple second partitioned switchgear characteristic parameter sets and initial aging factors, output multiple first transient ground voltage prediction values;

[0119] Step S334: Calculate multiple first error values ​​based on multiple first transient ground voltage prediction values;

[0120] Step S335: Determine whether multiple first error values ​​meet the preset error criteria;

[0121] Step S336: If so, then the first intermediate partial discharge model is used as the first partial discharge model.

[0122] Step S34: The weight factor matrix of the first partial discharge model is modified using the initial aging factor and multiple abnormal normalized switchgear feature parameter sets to determine the second partial discharge model.

[0123] Furthermore, step S34 may include the following sub-steps S341-S347:

[0124] Step S341: Divide multiple abnormal normalized switchgear feature parameter sets and output multiple third-division switchgear feature parameter sets and multiple fourth-division switchgear feature parameter sets;

[0125] Step S342: Update the weight factor matrix of the first partial discharge model using multiple third-division switchgear feature parameter sets to determine the second intermediate partial discharge model;

[0126] Step S343: Using the second intermediate partial discharge model, based on the initial aging factor and multiple fourth partition switch cabinet characteristic parameter sets, output multiple second transient ground voltage prediction values.

[0127] Step S344: Calculate multiple second error values ​​based on multiple second transient ground voltage prediction values;

[0128] Step S345: Based on each second error value, update the weight factor matrix of the second intermediate partial discharge model in sequence to determine the third intermediate partial discharge model, and count the number of model updates in real time.

[0129] Step S346: Determine whether the number of model updates has reached the preset threshold.

[0130] Step S347: If so, then the third intermediate partial discharge model is used as the second partial discharge model.

[0131] Step S35: Based on the preset error criterion, the weight factor matrix of the second partial discharge model is modified using the initial aging factor and multiple sensitive normalized switchgear feature parameter sets to determine the third partial discharge model.

[0132] Furthermore, step S35 may include the following sub-steps S351-S35:

[0133] Step S351: Divide multiple sensitive normalized switchgear feature parameter sets and output multiple fifth-division switchgear feature parameter sets and multiple sixth-division switchgear feature parameter sets;

[0134] Step S352: Update the weight factor matrix of the second partial discharge model using multiple fifth-division switchgear feature parameter sets to determine the fourth intermediate partial discharge model;

[0135] Step S353: Using the fourth intermediate partial discharge model, based on the initial aging factor and multiple sixth division switch cabinet characteristic parameter sets, output multiple third transient ground voltage prediction values.

[0136] Step S354: Calculate multiple third error values ​​based on multiple predicted third transient ground voltage values;

[0137] Step S355: Determine whether multiple third error values ​​meet the preset error criteria;

[0138] Step S356: If so, then the fourth intermediate partial discharge model is used as the third partial discharge model.

[0139] It should be noted that you should refer to [link / reference]. Figure 2 The data normalization rules for the characteristic parameter sets of normal switchgear, abnormal switchgear, and sensitive switchgear are as follows:

[0140] If the parameter value is less than m, the parameter value is defined as 0, and the partial discharge weight factor (weight factor matrix) is defined as v0;

[0141] If m ≤ parameter value ≤ n, the parameter value is defined as: (parameter value - m) / (nm), and the partial discharge weighting factor is defined as v1;

[0142] If the parameter value is greater than n, the parameter value is defined as 1, and the partial discharge weight factor is defined as v2;

[0143] Furthermore, the partial discharge weighting factor v is defined as follows: Let v0, v1, and v2 be the partial discharge weighting factors for the characteristic parameter sets of normal switchgear, abnormal switchgear, and sensitive switchgear, respectively. Their corresponding equivalent partial discharge coefficients can be expressed as: Normal sample z * *v0; Abnormal samples (am a ) / (n a -m a )*v 1a (bm) b ) / (n b -m b )*v 1b, (cm c ) / (n c -m c )*v 1c , (dm d ) / (n d -m d )*v 1d , (em e ) / (n e -m e )*v 1e Sensitive sample v 2a v 2b v 2c v 2d v 2e , where m a m b m c m d m e These are the abnormal thresholds corresponding to bus temperature (a), humidity (b), load rate (c), and bus contact resistance increment (d), respectively. a n b n c n d n e These are the sensitivity thresholds corresponding to bus temperature (a), humidity (b), load rate (c), and bus contact resistance increment (d), respectively. 1a v 1b v 1c v 1d v 1e These are the weighting factor matrix elements corresponding to bus temperature a, humidity b, load rate c, and bus contact resistance increment d in v1, respectively. 2a v 2b v 2c v 2d v 2e These are the weight factor matrix elements corresponding to bus temperature a, humidity b, load rate c, and bus contact resistance increment d in v2, respectively.

[0144] Furthermore, the data normalization of the TEV values ​​y in the normal switchgear characteristic parameter set, the abnormal switchgear characteristic parameter set, and the sensitive switchgear characteristic parameter set follows the same processing principle as the normalization process described above, and will not be elaborated further in this invention.

[0145] Furthermore, regarding the establishment of the first partial discharge model, the aging factor z is based on the initial aging factor prediction model. * The calculation relationship is input, and the aging factor is calculated under real-time monitoring parameters. The calculation result, i.e., the initial aging factor, is input into the normal sample set A. j1 Abnormal sample set Aj2 and sensitive sample A j3 In the middle. The normal sample set A... j1 Divided into 80%A j1 Training set (i.e., multiple first-partition switchgear feature parameter sets) and 20%A j1 The validation set (i.e., multiple sets of feature parameters for the second partitioned switchgear) is used to identify the weight v0, during which parameters a=b=c=d=e=0. a, b, c, d, e, and z are defined in the feature parameter set of the first partitioned switchgear. * The input parameter set is V = {a, b, c, d, e, z}. *}, the TEV value y in the first set of characteristic parameters of the switchgear is the output observation, i.e., Y={y Aj1}, Construct the first intermediate partial discharge model , The weight factor matrix is ​​the input parameter set. As a bias term, error analysis is then performed using multiple sets of second-division switchgear feature parameters and initial aging factors. All first error values ​​are substituted into a preset error criterion. When 80% or more of the first error values ​​satisfy the preset error criterion, it is determined that the iterative optimization is complete. The first intermediate partial discharge model is then used as the first partial discharge model, and the weight factor matrix v0 of the first partial discharge model is output. The specific preset error criterion is as follows:

[0146] ;

[0147] in, This represents the error between the model's predicted value and the actual value, i.e., the j-th first error value. This is the mean of all first error values, i.e., the average error. It is the root mean square value; This represents the total number of the first error values.

[0148] Furthermore, if less than 80% of the first error values ​​satisfy the preset error criterion, the weight factor matrix of the first intermediate partial discharge model is updated using the scatter fitting method to obtain a new first intermediate partial discharge model, and the process jumps to step S333 until there are more than or equal to 80% of the first error values ​​that satisfy the preset error criterion.

[0149] Furthermore, for the establishment of the second partial discharge model, the abnormal sample set A is used. j2 Divided into 80%A j2 Training set (i.e., multiple sets of third-partition switchgear feature parameters) and 20%A j2 The validation set (i.e., multiple sets of fourth-division switchgear feature parameters) utilizes 80%A j2The training set is used to retrain the first partial discharge model, i.e., to update the weight factor matrix of the first partial discharge model, to obtain the second intermediate partial discharge model, using 20%A. j2 The validation set is used to calculate the residuals, and the residual sequence ΔY = {ΔY1, ΔY2, ..., ΔY} is defined. i ...}, where the residual ΔY i The second error value is given as the second error value of the feature parameter set of the i-th fourth partition switchgear. Based on the second error value, the weight factor matrix of the second intermediate partial discharge model is updated. The specific process is as follows: or step is the step size. To correct The constructed random function, through these two parameters, can be used to construct the direction control for iterative solution; the update value corresponding to each second error value is calculated using each second error value. Using each updated value The weight factor matrix of the second intermediate partial discharge model is updated sequentially, and the number of model updates is counted in real time until the number of model updates reaches a preset threshold. The third intermediate partial discharge model determined when the number of model updates reaches the preset threshold is taken as the second partial discharge model, and the weight factor matrix v1 of the second partial discharge model is output.

[0150] Furthermore, for the establishment of the third localization model, the sensitive sample set A is... j3 Divided into 80%A j3 Training set (i.e., multiple sets of feature parameters of the fifth partition switchgear) and 20%A j2 The validation set (i.e., the feature parameter set of multiple sixth-division switchgear) utilizes 80%A j3 The training set is used to retrain the second partial discharge model, that is, to update the weight factor matrix of the second partial discharge model to obtain the fourth intermediate partial discharge model. Then, error analysis is performed using multiple sixth partition switch cabinet feature parameter sets and initial aging factors. The obtained third error value is substituted into the preset error criterion. When there are 80% or more of the third error values ​​that meet the preset error criterion, it is determined that the iterative optimization is completed, the fourth intermediate partial discharge model is used as the third partial discharge model, and the weight factor matrix v2 of the third partial discharge model is output.

[0151] Furthermore, if less than 80% of the third error values ​​satisfy the preset error criterion, the weight factor matrix of the fourth intermediate partial discharge model is updated using the scatter fitting method to obtain a new fourth intermediate partial discharge model, and the process jumps to step S353 until more than or equal to 80% of the first error values ​​satisfy the preset error criterion.

[0152] Step 104: When the current year's switchgear sample dataset is received, based on the preset error criteria and preset correction criteria, the target aging factor and target partial discharge model are determined using the initial aging factor prediction model and the initial partial discharge model according to the current year's switchgear sample dataset.

[0153] The target partial discharge model includes a new first partial discharge model, a new second partial discharge model, and a new third partial discharge model.

[0154] It should be noted that the current year's switchgear sample dataset is the sample data B collected in year k. k The collected data is classified into real-time operation data and maintenance data. The maintenance data is used for the parameter iteration verification of the initial aging factor prediction model in year k, while the real-time operation data is used for the parameter iteration verification of the partial discharge model in year k.

[0155] Specifically, step 104 may include the following sub-steps S41-S49:

[0156] Step S41: Using the initial aging factor prediction model, output the current year's aging factor based on the current year's switchgear sample dataset, and output multiple insulation resistance calculation values ​​based on the current year's aging factor.

[0157] Step S42: Calculate multiple target error values ​​based on multiple calculated insulation resistance values ​​and multiple measured insulation resistance values ​​in the current year's switchgear sample dataset;

[0158] Step S43: Determine whether multiple target error values ​​meet the preset error criteria;

[0159] Step S44: If satisfied, the current year's aging factor is used as the intermediate aging factor.

[0160] Step S45: Classify the current year's switchgear sample dataset to generate multiple sets of feature parameters for normal switchgear, multiple sets of feature parameters for abnormal switchgear, and multiple sets of feature parameters for sensitive switchgear.

[0161] Step S46: Using the first partial discharge model, generate multiple predicted aging factors based on multiple current year normal switchgear characteristic parameter sets;

[0162] Step S47: Determine whether the intermediate aging factor and multiple predicted aging factors meet the preset error criteria;

[0163] Step S48: If satisfied, the mean of multiple predicted aging factors is taken as the target aging factor.

[0164] Step S49: Using a preset correction criterion, the weight factor matrix of the second partial discharge model and the third partial discharge model is corrected based on multiple current year abnormal switchgear feature parameter sets and multiple current year sensitive switchgear feature parameter sets to determine the new second partial discharge model and the new third partial discharge model.

[0165] It should be noted that you should refer to [link / reference]. Figure 3 Based on the aforementioned initial aging factor prediction model, the {a, b, c, d, e, y} data from the current year's switchgear sample dataset is preprocessed and then input into the initial aging factor prediction model, outputting the current year's aging factor z. *j Combined with the calculation formula z *j = The current annual aging factor z *j Substituting the operating years t at the time of sample collection in the current year's switchgear sample data into the formula, the aging factor z for the current year is calculated. *j as well as Keeping the insulation resistance constant, calculate multiple insulation resistance values ​​sequentially. (Calculated value of insulation resistance j) Using multiple insulation resistance measurements, error analysis is performed between the measured and calculated values ​​to obtain multiple target error values. These target error values ​​are then substituted into a preset error criterion. If 90% or more of the target error values ​​satisfy the preset error criterion, no model correction is required, and the current year's aging factor is used as the intermediate aging factor z. *1 If less than 90% of the target error values ​​meet the preset error criterion, then the current year's switchgear sample dataset B needs to be used. k The initial aging factor prediction model is retrained, and the aging weight factor w is corrected until there are target error values ​​greater than or equal to 90% that satisfy the preset error criterion. The process of substituting the obtained target error values ​​into the preset error criterion is as follows:

[0166] ;

[0167] in, This represents the error between the model's predicted value and the actual value, i.e., the j-th target error value. This is the mean of all target error values, i.e., the average error. It is the root mean square value; This represents the total number of target error values.

[0168] Furthermore, based on the aforementioned definition of the values ​​of m and n, the sample data B for year k is processed. k The classification is decomposed into normal samples B. k1 Abnormal sample B k2 and sensitive sample B k3That is, multiple current year normal switchgear characteristic parameter sets B k1 Multiple current year abnormal switchgear characteristic parameter sets B k2 and multiple current year sensitive switchgear characteristic parameter sets B k3 This process is consistent with the above-described process of classifying multiple switchgear feature parameter sets, and will not be elaborated further in this invention.

[0169] Furthermore, combining the first partial discharge model, the sample data is used for training, employing multiple current year normal switchgear feature parameter sets B. k1 Calculate multiple predictive aging factors z *2 Specifically, the characteristic parameter set B of the normal switchgear for the current year k1 The bus temperature (a), humidity (b), load rate (c), bus contact resistance increment (d), and TEV value (y) are used as inputs to the first partial discharge model, and multiple predicted aging factors (z) are calculated through inversion. *2 Then proceed with z *1 and z *2 Error result analysis, calculation of each z *2 With z *1 The difference between them is used to determine whether the intermediate aging factor and multiple predicted aging factors meet the preset error criteria.

[0170] ;

[0171] in, Let j be the j-th predicted aging factor; is the root mean square value; n is the total number of predicted aging factors.

[0172] Furthermore, if there exists a z-value greater than or equal to 90% *2 With z *1 The differences between them all meet the preset error criteria, so no model correction is needed, and multiple predicted aging factors z are also included. *2 The mean of z is used as the target aging factor. * If there exists less than 90% of z *2 With z *1 If the difference between them satisfies the preset error criterion, then the weight factor matrix of the first partial discharge model needs to be corrected: compare multiple predicted aging factors z. *2 The mean and z *1 If multiple predictive aging factors z *2 The mean is greater than or equal to z *1If a predetermined constant is subtracted from each element of the weight factor matrix of the first partial discharge model, then the predetermined constant is added to each element of the weight factor matrix of the first partial discharge model. This completes the correction of the weight factor matrix v0 of the first partial discharge model, resulting in the corrected first partial discharge model. Several new predicted aging factors are then re-output, and the process jumps to step S47 until there is a z factor greater than or equal to 90%. *2 With z *1 The differences between them all satisfy the preset error criterion, meaning there will be a z-value greater than or equal to 90%. *2 With z *1 When the differences between them all satisfy the preset error criterion, the corrected first partial discharge model is used as the new first partial discharge model, and the existence of z values ​​greater than or equal to 90% is considered as a new first partial discharge model. *2 With z *1 The differences between them all satisfy the multiple predicted aging factors z determined when the preset error criterion is used. *2 The mean of z is used as the target aging factor. * .

[0173] Furthermore, the process of correcting the weight factor matrix for the second and third partial discharge models specifically involves using outlier sample B. k2 and sensitive sample B k3 During sample training, z * The input parameter is the partial discharge parameter y (TEV value y), and the output observation is the partial discharge parameter y (TEV value y). First, the characteristic parameter set B of multiple abnormal switchgear units for the current year is... k2 The bus temperature (a), humidity (b), load rate (c), and bus contact resistance increment (d) are combined with z. * Inputting data into the second partial discharge model, the system outputs multiple TEV values ​​y corresponding to the second partial discharge model, thereby calculating multiple error values ​​corresponding to the second partial discharge model. Similarly, multiple error values ​​corresponding to the third partial discharge model can be obtained. Error analysis is performed on the partial discharge parameter y output by the second and third partial discharge models, and anomaly sample B is defined. k2 The error judgment is not satisfied. Let P1 be the sample weight, and define sensitive sample B. k3 The error judgment is not satisfied. The sample weight is P2, and the model parameters are adjusted and determined.

[0174] Furthermore, the pre-set correction criteria are P1 > 10% and P2 > 10%, when abnormal sample B k2 The error judgment is not satisfied. When the sample weight satisfies P1 > 10%, the weight factor matrix (partial discharge weight factor {v1}) of the second partial discharge model is corrected. The correction process is as follows: extract the samples that do not meet the error criteria. Abnormal sample B k2 Compare abnormal sample B k2 The TEV value y in the data is based on the abnormal sample B. k2 The model outputs the TEV value y based on the following parameters: busbar temperature (a), humidity (b), load rate (c), and busbar contact resistance increment (d). If based on anomaly sample B... k2 a) Busbar temperature, b) Humidity, c) Load rate, d) Busbar contact resistance increment, and d) Model output TEV value y greater than abnormal sample B. k2 If the proportion of data with TEV value y is greater than or equal to 80%, then the weight factor matrix of the second partial discharge model is increased; otherwise, the weight factor matrix of the second partial discharge model is decreased. Based on the aforementioned principle, the weight factor matrix (partial discharge weight factor {v2}) of the third partial discharge model can be modified to obtain a new second partial discharge model and a new third partial discharge model. If the abnormal sample B k2 The error judgment is not satisfied. The sample weight does not satisfy P1 > 10%, and the sensitive sample B k3 The error judgment is not satisfied. If the sample weight does not satisfy P2 > 10%, then no model correction is needed. The second and third partial discharge models can be used as the corresponding new second and third partial discharge models.

[0175] It is worth mentioning that if abnormal sample B k2 The error judgment is not satisfied. The sample weight satisfies P1 > 10%, and the sensitive sample B k3 Satisfying the error judgment If the sample weight does not satisfy P2 > 10%, then the outlier threshold m is reduced, and the sample data B from year k is re-processed based on the new outlier threshold m. k The classification is then performed, and the multiple error values ​​corresponding to the second partial discharge model are recalculated. Similarly, the multiple error values ​​corresponding to the third partial discharge model can be obtained, and then the process is executed. Error analysis was performed on the partial discharge parameter y output by the second and third partial discharge models until the abnormal sample B was analyzed. k2 The error judgment is not satisfied. The sample weight does not satisfy P1 > 10%, and the sensitive sample B k3 The error judgment is not satisfied. The sample weight does not satisfy P2 > 10%.

[0176] Step 105: Using preset early warning criteria, generate early warning results for the insulation status of the switchgear based on the weight factor matrix of the target aging factor and the target partial discharge model.

[0177] The pre-set early warning criteria include the first early warning criterion and the second early warning criterion.

[0178] The switchgear insulation status early warning results include annual insulation status early warning results and individual parameter insulation status early warning results.

[0179] Specifically, step 105 may include the following sub-steps S51-S55:

[0180] Step S51: Based on the target aging factor, determine the aging factor for future years;

[0181] Step S52: Determine whether the target aging factor and the aging factor for future years meet the first warning criterion;

[0182] Step S53: If satisfied, generate the annual insulation status early warning result;

[0183] Step S54: Determine whether the weight factor matrix of the new second partial discharge model and the weight factor matrix of the new third partial discharge model satisfy the second early warning criterion;

[0184] Step S55: If satisfied, generate a single parameter insulation status early warning result.

[0185] It should be noted that the linear fitting method is used based on the target aging factor z. * Determine the aging factor z for future years *k+1 This invention uses the target aging factor z for year k. * As the object of observation, the first early warning criterion |z is adopted. *k+1 -z *k |>0.25|z *k If the target aging factor and the aging factor for future years meet the first warning criterion, the generated annual insulation status warning result indicates that the aging factor has changed significantly in that year, and an insulation status warning for year k will be issued to remind maintenance personnel to arrange a power outage maintenance plan as soon as possible. If the criteria are not met, the generated annual insulation status warning result indicates that no warning is required.

[0186] Furthermore, using the linear fitting method, the weight factor matrices for future years (year k+1) of the new second partial discharge model and the weight factor matrices for future years of the new third partial discharge model can be obtained. This invention employs a second early warning criterion for determining the early warning of parameters a, b, c, d, and e. For example, when the weight matrix elements corresponding to parameter a in the weight factor matrix of the new second partial discharge model and the weight matrix elements corresponding to parameter a in the weight factor matrix of the future years of the new second partial discharge model satisfy the second early warning criterion... If the condition is not met, the generated single-parameter insulation state warning result is determined to be an insulation state warning for parameter a; otherwise, the generated single-parameter insulation state warning result is determined to be an insulation state warning for parameter a not performed. Where, v 1k+1For the weight matrix elements corresponding to parameter 'a' in the weight factor matrix of the new second partial discharge model for future years, v 1k For the weight factor matrix elements corresponding to parameter a in the new second partial discharge model, v 1k-1 This represents the weight matrix element corresponding to parameter 'a' in the weight factor matrix for the historical year (k-1) of the new second partial discharge model. Similarly, the single-parameter insulation state early warning result generated based on the weight factor matrix of the new third partial discharge model can be obtained.

[0187] It is worth mentioning that if the single parameter insulation status early warning result generated based on the weight factor matrix of the new third partial discharge model and the single parameter insulation status early warning result generated based on the weight factor matrix of the new second partial discharge model are determined to be the same parameter insulation status early warning, then the k-year insulation status early warning will be issued to remind the operation and maintenance personnel to arrange a power outage maintenance plan as soon as possible.

[0188] Furthermore, based on the parameter correction relationship in the aforementioned model, the model parameters are updated to complete the prediction model output for year k+1 (i.e., the new first partial discharge model, the new second partial discharge model, and the new third partial discharge model), which can be used for aging factor prediction and partial discharge prediction in year k+1. The assessment and early warning of the switchgear insulation status in year k+1 can be completed without power outage maintenance.

[0189] For comparison of technical effectiveness, existing technologies can be used as a reference. In the investigation of potential insulation aging hazards in existing switchgear, partial discharge monitoring and insulation characteristic testing are relatively mature applications. However, while TEV or ultrasonic testing results can be used to assess partial discharge, many factors influence it; for example, partial discharge can occur even without obvious aging. Therefore, partial discharge values ​​cannot be directly applied to assess the degree of insulation aging. Insulation characteristic testing can be used to comprehensively assess the state of insulation aging, but this process needs to be carried out during power outage maintenance. For continuously operating equipment, real-time prediction and early warning are not possible.

[0190] Based on the above, the shortcomings of the existing technology are as follows: (1) In the real-time monitoring of the insulation performance of switchgear, partial discharge monitoring can be used to investigate insulation hazards under a certain degree of aging, but it cannot be quantitatively evaluated; at the same time, partial discharge is a special event under a specific working scenario, and is affected by many factors, so it cannot be directly applied to the quantitative evaluation of the degree of aging. (2) In terms of the quantitative evaluation of the degree of aging of switchgear, insulation resistance testing is the most effective and direct method, but this method needs to be tested in the scenario of power outage maintenance, and it cannot realize the aging evaluation and early warning under real-time operating conditions. (3) In the existing insulation prediction models of switchgear, a large number of studies focus on data training and model establishment, but whether the model is applicable to equipment used in different scenarios on site still needs to be verified. If the dynamic parameter correction cannot be achieved, a single model cannot meet the reliable evaluation application of insulation status for different types and scenarios.

[0191] To address the above problems, this invention provides a method for early warning of switchgear insulation status. It selects insulation resistance test results from maintenance and repair, and partial discharge data under real-time operating conditions. Through the establishment of an aging factor prediction model and a partial discharge model, combined with sample training using existing operating data, it achieves high-precision switchgear insulation status assessment and early warning. For details, please refer to... Figure 4 The process can be roughly divided into six parts: 1) Collection and classification of switchgear operation sample data; 2) Establishment of aging factor prediction model and output of aging factor calculation relationship; 3) Establishment of switchgear partial discharge model and identification of partial discharge factor; 4) Calculation and error analysis of switchgear aging factor in year k; 5) Application training and parameter correction of switchgear partial discharge model in year k; 6) Insulation status early warning in year k and output of model in year k+1. Regarding the collection and classification of switchgear operation sample data, this part mainly involves collecting switchgear operation sample dataset A of the same type but with different operating years. Based on the test time point of bus insulation resistance in the power outage maintenance plan, the collected dataset is divided into {A}. j Based on the anomaly threshold m and the sensitivity threshold n, the sample set is classified into a normal sample set A. j1 Abnormal sample set A j2 and sensitive sample A j3 A j For establishing the initial aging factor prediction model, normal sample set A j1 Abnormal sample set A j2 and sensitive sample A j3 Used for establishing a partial discharge model for switchgear. This section outputs the relationship between the initial aging factor prediction model and the aging factor calculation. This part mainly involves normalizing the collected sample data, based on the definition of the aging factor and the A of the sample set. jThe process involves several steps: First, data training to establish an initial aging factor prediction model for the switchgear. After establishing this model, the calculation relationships for the aging factors are output. Second, the establishment of the switchgear partial discharge model and identification of partial discharge factors are performed. This part mainly involves establishing the switchgear partial discharge model based on multi-dimensional parameter inputs (humidity, heat, electricity, mechanical properties, and aging), combined with hierarchical iteration of sample data. Through iterative calculation of sample data, the partial discharge factors of each input parameter in the partial discharge model are identified. Third, the calculation and error analysis of the switchgear aging factor for year k are performed. This part mainly involves applying the aforementioned model and algorithm. Real-time operation and maintenance data of the switchgear for year k are collected, and the collected data are input into the initial aging factor prediction model and the switchgear partial discharge model to calculate the aging factor of the switchgear for year k. Error analysis is then performed on the two calculation results. Fourth, the application training and parameter correction of the switchgear partial discharge model for year k are performed. This part mainly involves correcting the relevant parameters of the model based on the calculation error results of the switchgear aging factor for year k. Finally, the output of the insulation status early warning for year k and the model for year k+1 is completed. Based on the parameter correction results, an early warning of the insulation status in year k is provided. At the same time, the model parameters are updated to produce the model output for year k+1.

[0192] Furthermore, this invention establishes a multi-dimensional parameter-based aging factor prediction model, selecting aging factors as the observation object, evaluating the influence of each parameter on the aging factor, and achieving continuous output of aging factor calculation results. It also outputs aging factor prediction results for establishing the intrinsic relationship between partial discharge (PD) and aging in the PD model, making PD results more targeted for aging degree assessment. Simultaneously, this invention adopts a stratified sample management method, classifying samples according to the influence thresholds of each parameter on insulation aging and PD, to simplify iterative calculations and parameter corrections in the model. Moreover, by establishing a basic model for aging factor prediction and PD prediction, it can be used for output prediction of general sample data. This invention also incorporates the collection of engineering application data for iterative calculation and correction of model parameters to improve the model's adaptability to engineering applications.

[0193] In summary, this invention establishes an aging factor prediction model, modeling discrete insulation resistance detection results and outputting aging factors for training partial discharge models, enabling reliable assessment of switchgear aging status. Simultaneously, to improve the efficiency of model establishment and iterative calculation, a sample classification approach is adopted for hierarchical iterative calculation, reducing the computational burden of model parameter identification and allowing for more targeted parameter correction strategies. Furthermore, regarding model application adaptability, a parameter correction method based on data iterative calculation is proposed. Using natural years as units, error calculations are performed between the basic model and real-time acquired data, and the application verification of the model is achieved through parameter correction.

[0194] In this embodiment of the invention, a method for early warning of switchgear insulation status is provided. The method involves acquiring a historical annual switchgear sample dataset, preprocessing the dataset based on a preset classification threshold to generate multiple sets of feature parameters for segmented switchgear and multiple sets of feature parameters for categorized switchgear; constructing an initial aging factor prediction model based on the preset classification threshold, preset aging weight factor, and multiple sets of feature parameters for segmented switchgear; and constructing an initial partial discharge model based on the preset classification threshold and preset error criteria, using the initial aging factor prediction model and the multiple sets of feature parameters for categorized switchgear. When the current year's switchgear sample dataset is received, the method uses the initial aging factor prediction model and the initial partial discharge model based on the preset error criteria and preset correction criteria to predict the insulation status of the switchgear in the current year. The process involves using a sample dataset of switchgear cabinets to determine the target aging factor and target partial discharge model. Pre-set early warning criteria are used to generate early warning results for the switchgear insulation status based on the weight factor matrix of the target aging factor and target partial discharge model. Based on this scheme, an initial aging factor prediction model is constructed using a pre-processed set of segmented switchgear feature parameters. An initial partial discharge model is then constructed by combining a pre-set classification threshold, a pre-set error criterion, and a set of classified switchgear feature parameters. Finally, the weight factor matrix of these two models is corrected to output the early warning results for the switchgear insulation status. This invention establishes an aging factor prediction model that can output aging factor prediction results, which is used to establish the intrinsic relationship between partial discharge and aging in the partial discharge model, thereby achieving reliable assessment of the switchgear aging status and improving early warning accuracy.

[0195] Please see Figure 5 , Figure 5 This is a structural block diagram of a switchgear insulation status early warning device provided in Embodiment 2 of the present invention.

[0196] The present invention provides a switchgear insulation status early warning device, comprising:

[0197] The acquisition module 501 is used to acquire historical annual switch cabinet sample datasets and preprocess the historical annual switch cabinet sample datasets based on preset classification thresholds to generate multiple segmented switch cabinet feature parameter sets and multiple classified switch cabinet feature parameter sets.

[0198] The first construction module 502 is used to construct an initial aging factor prediction model based on a preset classification threshold, a preset aging weight factor, and multiple feature parameter sets of the split switchgear.

[0199] The second construction module 503 is used to construct an initial partial discharge model based on a preset classification threshold and a preset error criterion, using an initial aging factor prediction model and multiple sets of characteristic parameters of the switchgear.

[0200] The determination module 504 is used to determine the target aging factor and target partial discharge model based on the current year's switchgear sample dataset when it receives the current year's switchgear sample dataset, using the initial aging factor prediction model and the initial partial discharge model based on the preset error criteria and preset correction criteria.

[0201] The early warning module 505 is used to generate early warning results of the switchgear insulation status based on the target aging factor and the weight factor matrix of the target partial discharge model using preset early warning criteria.

[0202] Furthermore, the multiple sets of characteristic parameters for classified switchgear include multiple sets of characteristic parameters for normal switchgear, multiple sets of characteristic parameters for abnormal switchgear, and multiple sets of characteristic parameters for sensitive switchgear; the acquisition module 501 is specifically used for:

[0203] The historical annual switchgear sample dataset is segmented to generate multiple segmented switchgear feature parameter sets;

[0204] Based on a preset classification threshold, multiple feature parameter sets of segmented switchgear are classified to generate multiple feature parameter sets of normal switchgear, multiple feature parameter sets of abnormal switchgear, and multiple feature parameter sets of sensitive switchgear.

[0205] Furthermore, the first building module 502 is specifically used for:

[0206] Based on a preset classification threshold, multiple feature parameter sets of the split switchgear are normalized to generate multiple normalized feature parameter sets of the split switchgear.

[0207] Multiple normalized switchgear feature parameter sets are weighted and output as a weighted switchgear feature parameter set;

[0208] An initial aging factor prediction model is constructed based on the weighted switchgear feature parameter set and the preset aging weight factor.

[0209] Furthermore, the initial partial discharge model includes a first partial discharge model, a second partial discharge model, and a third partial discharge model; the second building module 503 includes:

[0210] The first submodule is used to normalize multiple normal switchgear feature parameter sets, multiple abnormal switchgear feature parameter sets and multiple sensitive switchgear feature parameter sets based on a preset classification threshold, and generate multiple normalized switchgear feature parameter sets, multiple abnormalized switchgear feature parameter sets and multiple sensitive normalized switchgear feature parameter sets.

[0211] The second submodule is used to generate the initial aging factor based on multiple normal switchgear feature parameter sets using the initial aging factor prediction model.

[0212] The third submodule is used to construct the first partial discharge model based on the initial aging factor and multiple normalized switchgear characteristic parameter sets using a preset error criterion.

[0213] The fourth submodule is used to modify the weight factor matrix of the first partial discharge model using the initial aging factor and multiple abnormal normalized switchgear feature parameter sets, and to determine the second partial discharge model.

[0214] The fifth submodule is used to modify the weight factor matrix of the second partial discharge model based on the preset error criteria, using the initial aging factor and multiple sensitive normalized switchgear feature parameter sets, and to determine the third partial discharge model.

[0215] Furthermore, the third submodule is specifically used for:

[0216] Multiple normalized switchgear feature parameter sets are divided, and multiple first-division switchgear feature parameter sets and multiple second-division switchgear feature parameter sets are output.

[0217] Based on the initial aging factor and multiple first-division switchgear characteristic parameter sets, a first intermediate partial discharge model is constructed;

[0218] The first intermediate partial discharge model is adopted to output multiple first transient ground voltage prediction values ​​based on multiple second partition switch cabinet characteristic parameter sets and initial aging factors.

[0219] Based on multiple first transient ground voltage prediction values, multiple first error values ​​are calculated;

[0220] Determine whether multiple first error values ​​meet the preset error criteria;

[0221] If so, then the first intermediate partial discharge model will be used as the first partial discharge model.

[0222] Furthermore, the fourth submodule is specifically used for:

[0223] Multiple abnormal normalized switchgear feature parameter sets are divided, and multiple third-division switchgear feature parameter sets and multiple fourth-division switchgear feature parameter sets are output.

[0224] The weight factor matrix of the first partial discharge model is updated by using multiple third-division switchgear feature parameter sets to determine the second intermediate partial discharge model.

[0225] The second intermediate partial discharge model is adopted to output multiple second transient ground voltage prediction values ​​based on the initial aging factor and multiple fourth partition switch cabinet characteristic parameter sets.

[0226] Based on multiple predicted values ​​of the second transient ground voltage, multiple second error values ​​are calculated;

[0227] The weight factor matrix of the second intermediate partial discharge model is updated sequentially based on each second error value to determine the third intermediate partial discharge model, and the number of model updates is counted in real time.

[0228] Determine whether the number of model updates has reached the preset threshold;

[0229] If so, then the third intermediate partial discharge model will be used as the second partial discharge model.

[0230] Furthermore, the fifth submodule is specifically used for:

[0231] Multiple sensitive normalized switchgear feature parameter sets are divided, and multiple fifth-division switchgear feature parameter sets and multiple sixth-division switchgear feature parameter sets are output.

[0232] The weight factor matrix of the second partial discharge model is updated by using multiple fifth-division switchgear feature parameter sets to determine the fourth intermediate partial discharge model.

[0233] The fourth intermediate partial discharge model is adopted to output multiple predicted values ​​of the third transient ground voltage based on the initial aging factor and multiple sixth division switch cabinet characteristic parameter sets.

[0234] Based on multiple predicted values ​​of the third transient ground voltage, multiple third error values ​​are calculated;

[0235] Determine whether multiple third error values ​​meet the preset error criteria;

[0236] If so, then the fourth intermediate partial discharge model will be used as the third partial discharge model.

[0237] Furthermore, the target partial discharge model includes a new second partial discharge model and a new third partial discharge model; the determination module 504 is specifically used for:

[0238] The initial aging factor prediction model is used to output the current year's aging factor based on the current year's switchgear sample dataset, and based on the current year's aging factor, it outputs multiple insulation resistance calculation values.

[0239] Based on multiple calculated insulation resistance values ​​and multiple measured insulation resistance values ​​in the current year's switchgear sample dataset, multiple target error values ​​are calculated.

[0240] Determine whether multiple target error values ​​meet the preset error criteria;

[0241] If the conditions are met, the current year's aging factor will be used as the intermediate aging factor.

[0242] The current year's switchgear sample dataset is classified to generate multiple sets of feature parameters for normal switchgear, multiple sets of feature parameters for abnormal switchgear, and multiple sets of feature parameters for sensitive switchgear.

[0243] The first partial discharge model is used to generate multiple predicted aging factors based on multiple sets of characteristic parameters of normal switchgear in the current year;

[0244] Determine whether the intermediate aging factor and multiple predicted aging factors meet the preset error criteria;

[0245] If the conditions are met, the mean of multiple predicted aging factors will be used as the target aging factor.

[0246] The second and third partial discharge models are modified by using a pre-set correction criterion based on the characteristic parameter sets of multiple abnormal switchgear and multiple characteristic parameter sets of sensitive switchgear in the current year, so as to determine the new second and third partial discharge models.

[0247] Furthermore, the preset early warning criteria include a first early warning criterion and a second early warning criterion; the switchgear insulation status early warning results include annual insulation status early warning results and individual parameter insulation status early warning results; the early warning module 505 is specifically used for:

[0248] Based on the target aging factor, determine the aging factor for future years;

[0249] Determine whether the target aging factor and the aging factor for future years meet the first early warning criterion;

[0250] If the conditions are met, an annual insulation status early warning result will be generated;

[0251] Determine whether the weight factor matrix of the new second partial discharge model and the weight factor matrix of the new third partial discharge model satisfy the second early warning criterion;

[0252] If the conditions are met, a single parameter insulation status early warning result will be generated.

[0253] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, modules, and sub-modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0254] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0255] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0256] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A switch cabinet insulation state early warning method, characterized in that, The method comprises the following steps: acquire historical annual switch cabinet sample data sets, and preprocess the historical annual switch cabinet sample data sets based on preset classification thresholds to generate a plurality of split switch cabinet feature parameter sets and a plurality of classification switch cabinet feature parameter sets; construct an initial aging factor prediction model according to the preset classification thresholds, a preset aging weight factor and a plurality of the split switch cabinet feature parameter sets; based on the preset classification thresholds and a preset error criterion, an initial aging factor prediction model is used to construct an initial partial discharge model according to a plurality of the classification switch cabinet feature parameter sets; when a current annual switch cabinet sample data set is received, based on the preset error criterion and a preset correction criterion, the initial aging factor prediction model and the initial partial discharge model are used to determine a target aging factor and a target partial discharge model according to the current annual switch cabinet sample data set; a preset warning criterion is used to generate a switch cabinet insulation state warning result according to the target aging factor and a weight factor matrix of the target partial discharge model; the construction process of a first partial discharge model in the initial partial discharge model is specifically as follows: a plurality of normal normalized switch cabinet feature parameter sets are generated by normalizing a plurality of normal switch cabinet feature parameter sets in the plurality of classification switch cabinet feature parameter sets based on the preset classification thresholds; an initial aging factor is generated by using the initial aging factor prediction model according to a plurality of the normal switch cabinet feature parameter sets; a first partial discharge model is constructed by using the preset error criterion according to the initial aging factor and a plurality of the normal normalized switch cabinet feature parameter sets; the first partial discharge model in the initial partial discharge model is specifically as follows: ； wherein, is a weight factor matrix for the input parameter set; is a bias term; is an input parameter set from the normal normalized switchgear characteristic parameter set; is an output observation, representing a transient ground voltage prediction value from the normal normalized switchgear characteristic parameter set; is a functional relationship between the hidden layer parameters; the preset error criterion is specifically as follows: ； wherein, is the error of the transient voltage prediction value and the actual value; is the average value of the error; is the root mean square value; is the total number of error values.

2. The switchgear insulation condition early warning method according to claim 1, characterized in that, the plurality of classification switch cabinet feature parameter sets further comprise a plurality of abnormal switch cabinet feature parameter sets and a plurality of sensitive switch cabinet feature parameter sets; the preprocessing of the historical annual switch cabinet sample data sets based on the preset classification thresholds to generate a plurality of split switch cabinet feature parameter sets and a plurality of classification switch cabinet feature parameter sets comprises: the historical annual switch cabinet sample data sets are split to generate a plurality of split switch cabinet feature parameter sets; based on the preset classification thresholds, a plurality of normal switch cabinet feature parameter sets, a plurality of abnormal switch cabinet feature parameter sets and a plurality of sensitive switch cabinet feature parameter sets are generated by classifying a plurality of the split switch cabinet feature parameter sets.

3. The switchgear insulation condition early warning method according to claim 1, characterized in that, the construction of the initial aging factor prediction model according to the preset classification thresholds, a preset aging weight factor and a plurality of the split switch cabinet feature parameter sets comprises: a plurality of normalized split switch cabinet feature parameter sets are generated by normalizing a plurality of the split switch cabinet feature parameter sets based on the preset classification thresholds; a weighted switch cabinet feature parameter set is output by weighting a plurality of the normalized split switch cabinet feature parameter sets; an initial aging factor prediction model is constructed according to the weighted switch cabinet feature parameter set and the preset aging weight factor.

4. The switchgear insulation condition early warning method according to claim 2, characterized in that, The initial partial discharge model further comprises a second partial discharge model and a third partial discharge model; the initial aging factor prediction model is used to construct an initial partial discharge model based on the preset classification threshold and the preset error criterion, comprising: normalizing the plurality of abnormal switch cabinet characteristic parameter sets and the plurality of sensitive switch cabinet characteristic parameter sets based on the preset classification threshold, to generate a plurality of abnormal normalized switch cabinet characteristic parameter sets and a plurality of sensitive normalized switch cabinet characteristic parameter sets; using the initial aging factor and the plurality of abnormal normalized switch cabinet characteristic parameter sets to correct the weight factor matrix of the first partial discharge model to determine a second partial discharge model; based on the preset error criterion, using the initial aging factor and the plurality of sensitive normalized switch cabinet characteristic parameter sets to correct the weight factor matrix of the second partial discharge model to determine a third partial discharge model.

5. The switchgear insulation condition early warning method according to claim 4, characterized in that, The initial aging factor prediction model is used to construct a first partial discharge model based on the initial aging factor and the plurality of normal normalized switch cabinet characteristic parameter sets, comprising: dividing the plurality of normal normalized switch cabinet characteristic parameter sets to output a plurality of first divided switch cabinet characteristic parameter sets and a plurality of second divided switch cabinet characteristic parameter sets; constructing a first intermediate partial discharge model based on the initial aging factor and the plurality of first divided switch cabinet characteristic parameter sets; using the first intermediate partial discharge model to output a plurality of first transient ground voltage prediction values based on the plurality of second divided switch cabinet characteristic parameter sets and the initial aging factor; calculating a plurality of first error values based on the plurality of first transient ground voltage prediction values; determining whether the plurality of first error values meet the preset error criterion; if yes, the first intermediate partial discharge model is used as the first partial discharge model.

6. The switchgear insulation condition early warning method according to claim 4, characterized in that, The initial aging factor prediction model is used to construct a first partial discharge model based on the initial aging factor and the plurality of normal normalized switch cabinet characteristic parameter sets, comprising: dividing the plurality of abnormal normalized switch cabinet characteristic parameter sets to output a plurality of third divided switch cabinet characteristic parameter sets and a plurality of fourth divided switch cabinet characteristic parameter sets; using the plurality of third divided switch cabinet characteristic parameter sets to update the weight factor matrix of the first partial discharge model to determine a second intermediate partial discharge model; using the second intermediate partial discharge model to output a plurality of second transient ground voltage prediction values based on the initial aging factor, the plurality of fourth divided switch cabinet characteristic parameter sets; calculating a plurality of second error values based on the plurality of second transient ground voltage prediction values; updating the weight factor matrix of the second intermediate partial discharge model based on each of the second error values in turn to determine a third intermediate partial discharge model, and real-time statistics of the number of model updates; determining whether the number of model updates reaches a preset number threshold; if yes, the third intermediate partial discharge model is used as the second partial discharge model.

7. The switchgear insulation condition early warning method according to claim 4, characterized in that, The weight factor matrix of the second partial discharge model is corrected based on the preset error criterion, using the initial aging factor and a plurality of the sensitive normalized switch cabinet characteristic parameter sets, to determine a third partial discharge model, including: The plurality of sensitive normalized switch cabinet characteristic parameter sets are divided to output a plurality of fifth divided switch cabinet characteristic parameter sets and a plurality of sixth divided switch cabinet characteristic parameter sets; The weight factor matrix of the second partial discharge model is updated using the plurality of fifth divided switch cabinet characteristic parameter sets to determine a fourth intermediate partial discharge model; The fourth intermediate partial discharge model is used to output a plurality of third transient ground voltage prediction values according to the initial aging factor and the plurality of sixth divided switch cabinet characteristic parameter sets; A plurality of third error values are calculated according to the plurality of third transient ground voltage prediction values; It is judged whether the plurality of third error values meet the preset error criterion; If yes, the fourth intermediate partial discharge model is taken as the third partial discharge model.

8. The switchgear insulation condition early warning method according to claim 4, characterized in that, The target partial discharge model includes a new second partial discharge model and a new third partial discharge model; the target aging factor and the target partial discharge model are determined based on the preset error criterion and the preset correction criterion, using the initial aging factor prediction model and the initial partial discharge model according to the current annual switch cabinet sample data set, including: The initial aging factor prediction model is used to output a current annual aging factor according to the current annual switch cabinet sample data set, and based on the current annual aging factor, a plurality of insulation resistance calculation values are output; A plurality of target error values are calculated based on a plurality of the insulation resistance calculation values and a plurality of insulation resistance measurement values in the current annual switch cabinet sample data set; It is judged whether the plurality of target error values meet the preset error criterion; If yes, the current annual aging factor is taken as an intermediate aging factor; The current annual switch cabinet sample data set is classified to generate a plurality of current annual normal switch cabinet characteristic parameter sets, a plurality of current annual abnormal switch cabinet characteristic parameter sets, and a plurality of current annual sensitive switch cabinet characteristic parameter sets; A plurality of predicted aging factors are generated using the first partial discharge model according to the plurality of current annual normal switch cabinet characteristic parameter sets; It is judged whether the intermediate aging factor and the plurality of predicted aging factors meet the preset error criterion; If yes, the mean of the plurality of predicted aging factors is taken as the target aging factor; The second partial discharge model and the third partial discharge model are corrected in weight factor matrix using the preset correction criterion according to the plurality of current annual abnormal switch cabinet characteristic parameter sets and the plurality of current annual sensitive switch cabinet characteristic parameter sets to determine a new second partial discharge model and a new third partial discharge model.

9. The switchgear insulation condition early warning method according to claim 8, characterized in that, The preset warning criterion includes a first warning criterion and a second warning criterion; the switch cabinet insulation state warning result includes an annual insulation state warning result and a single parameter insulation state warning result; The switch cabinet insulation state warning result is generated using the preset warning criterion according to the weight factor matrix of the target aging factor and the target partial discharge model, including: The future annual aging factor is determined based on the target aging factor; determining whether the target aging factor and the future annual aging factor satisfy a first early warning criterion; if yes, generating an annual insulation state early warning result; determining whether the weight factor matrix of the new second partial discharge model and the weight factor matrix of the new third partial discharge model satisfy a second early warning criterion; if yes, generating a single-parameter insulation state early warning result.

10. A switch cabinet insulation state early warning device applied to the switch cabinet insulation state early warning method of claim 1, characterized in that, The method comprises the following steps: an acquisition module is configured to acquire historical annual switch cabinet sample data sets, and perform preprocessing on the historical annual switch cabinet sample data sets based on a preset classification threshold, to generate a plurality of split switch cabinet feature parameter sets and a plurality of classification switch cabinet feature parameter sets; a first construction module is configured to construct an initial aging factor prediction model according to the preset classification threshold, a preset aging weight factor, and a plurality of the split switch cabinet feature parameter sets; a second construction module is configured to construct an initial partial discharge model based on the preset classification threshold and a preset error criterion, and using the initial aging factor prediction model to construct the initial partial discharge model according to a plurality of the classification switch cabinet feature parameter sets; a determination module is configured to, when a current annual switch cabinet sample data set is received, determine a target aging factor and a target partial discharge model based on the preset error criterion and a preset correction criterion, and using the initial aging factor prediction model and the initial partial discharge model to determine the target aging factor and the target partial discharge model according to the current annual switch cabinet sample data set; an early warning module is configured to generate a switch cabinet insulation state early warning result according to the target aging factor and the weight factor matrix of the target partial discharge model using a preset early warning criterion.

Images