Power Equipment Insulation Status Risk Assessment Method, System, and Computer Storage Medium

Abstract

The present invention discloses a method, a system and a computer storage medium for risk assessment of the insulation state of power equipment. The method includes: measuring partial discharge data of power equipment by using various types of sensors, and performing clustering of discharge types based on the partial discharge data; establishing a regression model between the response amplitude of each sensor and the apparent discharge energy during the discharge development process for each discharge type; calculating the cumulative value of the apparent discharge energy of the partial discharge of the power equipment based on the regression model, and using the cumulative value as an index for risk assessment to evaluate the insulation state of the power equipment. The present invention provides a reliable solution for the insulation state early warning of the on-line monitoring system.

Description

Technical Field

[0001] The present invention belongs to the field of insulation of power equipment, and relates to a method, a system and a computer storage medium for risk assessment of the insulation state of power equipment. Background Art

[0002] Gas insulated switchgear has the advantages of small floor space, long maintenance period, convenient transportation and installation, etc., and has been widely used at home and abroad since the 1960s. Internal insulation defects in GIS can cause partial discharge under the action of electric field. The occurrence of partial discharge at weak parts in the insulation of power equipment under the action of strong electric field is a common problem in high-voltage power equipment. Although partial discharge generally does not cause penetrative breakdown of insulation, it can cause local damage to dielectrics (especially organic dielectrics). If partial discharge exists for a long time, it will lead to insulation deterioration and even breakdown under certain conditions. Conducting partial discharge tests on power equipment can not only understand the insulation condition of the equipment, but also timely discover many problems related to manufacturing and installation, and determine the cause and severity of insulation faults.

[0003] During the process of instantaneous discharge or continuous discharge, energy is released to molecules, ions and electrons in space through discharge, exciting light, heat, sound and other forms of energy. According to various physical processes existing in the process of partial discharge, corresponding detection methods such as ultra-high frequency detection method, optical measurement method, ultrasonic detection method, pulse current method, etc. have emerged. However, during the process of insulation medium deterioration, the partial discharge amplitude measured by the above various methods does not increase continuously. Instead, the partial discharge amplitude is relatively low in the stage when the insulation medium is approaching penetration. Therefore, it is impossible to infer the degree of insulation deterioration only from the measured amplitude, which causes great difficulties for the insulation state assessment of power equipment. In order to accurately measure the insulation state of power equipment and avoid misjudgment problems caused by a single detection amplitude, the present invention proposes a method for risk assessment of the insulation state of power equipment.

[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] Aiming at the problems existing in the prior art, the present invention proposes a method for risk assessment of the insulation state of power equipment, which accurately and reliably measures the insulation state of power equipment.

[0006] The purpose of the present invention is achieved by the following technical solutions:

[0007] A method for risk assessment of the insulation state of power equipment, the method comprising the following steps:

[0008] Measure the partial discharge data of power equipment using multiple types of sensors, and perform clustering of discharge types;

[0009] Based on the clustering results of each discharge type, establish a regression model between the response amplitudes of each sensor and the apparent discharge energy during the discharge development process;

[0010] Based on the regression model, calculate the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment, and use the cumulative value as an index for risk assessment to evaluate the insulation status of the power equipment.

[0011] In the method for risk assessment of the insulation status of the power equipment described above, measuring the historical partial discharge data of the power equipment using multiple types of sensors and performing clustering of discharge types includes the following steps:

[0012] In the case of not using a high-frequency current sensor, use a total of N sensors of other different types to collect the partial discharge data of the power equipment, and normalize the response amplitude of each sensor to obtain :

[0013]

[0014] wherein, represents the response amplitude of the sensor numbered i under the action of a partial discharge pulse, represents the sum of the response amplitudes of N sensors;

[0015] Using the results of the normalization processing of the response amplitudes of the first N - 1 sensors as the input, perform SVM clustering on the discharge types to establish a clustering region in the (N - 1)-dimensional space.

[0016] In the method for risk assessment of the insulation status of the power equipment described above, based on the clustering results of each discharge type, establishing a regression model between the discharge development path and the apparent discharge energy includes the following steps:

[0017] Use Gaussian process regression to describe the relationship between the (N - 1)-dimensional space composed of and the apparent discharge energy ADE, where, and, the Gaussian process is described by the mean function and the covariance function as:

[0018] ,

[0019] wherein, the covariance function adopts the squared exponential type:

[0020]

[0021] A regression model is established based on Gaussian process regression, and the specific parameters of the regression model are determined by the training data set, where the training data set is:

[0022]

[0023] Among them,

[0024] Covariance function In Take When It represents the vector composed of the response amplitudes of the sensor under the action of the i-th discharge pulse: , i ranges from 1 to n, and n is the n-th discharge pulse;

[0025] Take When Then take , It represents a vector of the same type as , j also ranges from 1 to n, but j is not equal to i;

[0026] And Are the independent variables corresponding to the vectors composed of the response amplitudes of the sensor under the action of different discharge pulses, Formally not equal to ;

[0027] When i ranges from 1 to n, It represents the apparent discharge energy measured by the high-frequency current sensor under the action of the i-th discharge pulse;

[0028] The random vector of the predictive variable distribution of the training data set is generated as:

[0029] ,

[0030] Among them, the joint distribution of the function value Y and the observed target value Is expressed as:

[0031] ,

[0032] Among them, the observed target value Is the predicted value of the discharge energy to be predicted, and this predicted value of the discharge energy is the output of the regression model;

[0033] Represents an n-order identity matrix with diagonal elements of 1 and the remaining elements of 0, Is composed of all elements in the training data set And the new observed data Constitute;

[0034] is the training dataset The mean vector of, The i-th component of is The average of all elements in the i-th row;

[0035] is The mean vector of, The i-th component of is The average of all elements in the i-th row;

[0036] is Gaussian noise, and the Gaussian noise ε follows a normal distribution ;

[0037] is The random error contained in the variance of, which reduces the impact of measurement errors existing in the training dataset;

[0038] is the covariance matrix, and the element in the i-th row and j-th column of this matrix is obtained from The i-th column element of and the j-th column element by taking the squared exponential kernel function:

[0039]

[0040] The element in the i-th row and j-th column of the covariance matrix is obtained from The i-th column element of and The j-th column element of by taking the squared exponential kernel function;

[0041] The element in the i-th row and j-th column of is obtained from The i-th column element of and The j-th column element of

[0042] The element in the i-th row and j-th column of is obtained from The i-th column element of and The j-th column element of

[0043] When is known, The conditional distribution of is the mean and the variance Normal distribution:

[0044]

[0045] where the mean and variance are:

[0046]

[0047] ,

[0048] From the conditional distribution of to obtain new observed data The corresponding mean and variance are obtained, so as to obtain the predicted value of the apparent discharge energy, that is, the observed target value .

[0049] In the power equipment insulation status risk assessment method described above, based on the regression model, calculate the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment, and use the cumulative value as an index for risk assessment to evaluate the insulation status of the power equipment, including the following steps:

[0050] Find the predicted value of the apparent discharge energy corresponding to the current discharge pulse point of the power equipment in the regression model:

[0051]

[0052] where represents the normalized result of the response amplitude of the i-th sensor under the discharge pulse signal at time t, represents the predicted value of the apparent discharge energy obtained from the discharge pulse signal at time t;

[0053] During the monitoring and early warning process, accumulate the predicted value of the apparent discharge energy and time to determine the insulation risk level:

[0054]

[0055] represents the integral of the apparent discharge energy from 0 to t;

[0056] When Risk_level < a, it is determined that the power equipment has no discharge;

[0057] When a ≤ Risk_level ≤ b, it is determined that the power equipment has a slight partial discharge;

[0058] When Risk_level > b, it is determined that a severe discharge occurs in the power equipment; among them, the threshold a is set according to 120% of the risk index under the background noise of the power equipment operation site; the threshold b is set according to the risk index at the highest temperature allowed under the rated operation state of the power equipment.

[0059] In the power equipment insulation state risk assessment method described above, the discharge types include corona discharge, surface discharge, and floating potential discharge.

[0060] In the power equipment insulation state risk assessment method described above, the types of sensors include optical sensors, ultra-high frequency sensors, and ultrasonic sensors.

[0061] In addition, the present invention also discloses a power equipment insulation state risk assessment device, which includes:

[0062] A clustering unit, which measures the partial discharge data of the power equipment by using various types of sensors and performs clustering of the discharge types;

[0063] A modeling unit, which establishes a regression model between the response amplitude of each sensor and the apparent discharge energy during the discharge development process based on the clustering results of each discharge type;

[0064] An evaluation unit, which calculates the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment based on the regression model, and uses the cumulative value as an index for risk assessment to evaluate the insulation state of the power equipment.

[0065] In addition, the present invention also discloses a power equipment insulation state risk assessment system, which includes a processor, and the processor executes the power equipment insulation state risk assessment method described in any one of the foregoing.

[0066] In the power equipment insulation state risk assessment system described above, the system further includes: a plurality of sensors connected to the processor.

[0067] In addition, the present invention also discloses a computer storage medium, which stores computer-executable instructions for executing the method described in any one of the foregoing.

[0068] Beneficial effects

[0069] The method provided by the present invention proposes a method for identifying discharge types at different scales by using multi-parameter data; and proposes a dynamic evaluation method for the insulation state based on the energy accumulation process of discharge development, which can accurately characterize the severity of partial discharge and provides a reliable solution for the insulation state early warning of the on-line monitoring system. Brief description of the drawings

[0070] By reading the detailed description in the preferred specific embodiments below, various other advantages and benefits of the present invention will become clear to those of ordinary skill in the art. The accompanying drawings of the specification are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Obviously, the drawings described below are only some embodiments of the present invention, and for those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings. Moreover, throughout the drawings, the same reference numerals are used to represent the same components.

[0071] In the drawings:

[0072] Figure 1 is a flowchart of a method for risk assessment of the insulation state of electrical equipment according to an embodiment of the present invention;

[0073] Figure 2 is a ternary diagram of relative energy under three types of discharges of a method for risk assessment of the insulation state of electrical equipment according to an embodiment of the present invention;

[0074] Figure 3 is a schematic diagram of the posterior probability of discharge type identification of a method for risk assessment of the insulation state of electrical equipment according to an embodiment of the present invention;

[0075] Figs. 4(a) to 4(c) are schematic diagrams of a regression model of the partial discharge ternary energy pattern and the apparent discharge energy of a method for risk assessment of the insulation state of electrical equipment according to an embodiment of the present invention, wherein Fig. 4(a) is corona discharge, Fig. 4(b) is surface discharge, and Fig. 4(c) is floating potential discharge;

[0076] Figs. 5(a) to 5(c) are schematic diagrams of the verification diagram of the predicted value - true value of the apparent discharge energy and the relationship between the cumulative discharge time and the risk assessment level of a method for risk assessment of the insulation state of electrical equipment according to an embodiment of the present invention, wherein Fig. 5(a) is corona discharge, Fig. 5(b) is surface discharge, and Fig. 5(c) is floating potential discharge.

[0077] The present invention will be further explained below with reference to the drawings and embodiments. Specific Embodiments

[0078] The following will refer to the attached Figure 1 to Fig. 5(c) to describe the specific embodiments of the present invention in more detail. Although the specific embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments described herein. Instead, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0079] It should be noted that in the description and claims, certain terms are used to refer to specific components. Those skilled in the art should understand that technicians may use different terms to refer to the same component. The description and claims of this specification do not use the difference in terms as a way to distinguish components, but use the difference in the functions of components as the criterion for distinction. For example, the terms "comprising" or "including" mentioned throughout the specification and claims are open-ended terms, so they should be interpreted as "including but not limited to". The subsequent description in the specification is the preferred implementation manner for implementing the present invention, but the description is for the purpose of the general principles of the specification and does not limit the scope of the present invention. The protection scope of the present invention shall be subject to what is defined by the appended claims.

[0080] For the convenience of understanding the embodiments of the present invention, the following will further explain and illustrate with specific embodiments in conjunction with the accompanying drawings, and each of the accompanying drawings does not constitute a limitation on the embodiments of the present invention.

[0081] For better understanding, Figure 1 For the power equipment insulation state risk assessment method, as Figure 1 shown, the power equipment insulation state risk assessment method includes the following steps:

[0082] Step 1, measure the partial discharge data of the power equipment using various types of sensors, and perform clustering of the discharge types;

[0083] Step 2, establish a regression model between the response amplitude of each sensor and the apparent discharge energy during the discharge development process based on the clustering results of each discharge type;

[0084] Step 3, calculate the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment based on the regression model, and use the cumulative value as an index for risk assessment to evaluate the insulation state of the power equipment.

[0085] In the preferred implementation manner of the method, in Step 1, in the case of not using a high-frequency current sensor, a total of N sensors of other different types are used to collect the partial discharge data of the power equipment, and the response amplitude of each sensor is normalized to obtain :

[0086]

[0087] where represents the response amplitude of the sensor numbered i under the action of a partial discharge pulse, represents the sum of the response amplitudes of N sensors;

[0088] Based on the normalized results of the response amplitudes of the first N - 1 sensors Take the discharge type as input, perform SVM clustering, and establish a clustering region in the N-1 dimensional space.

[0089] In the preferred embodiment of the method described, in step 2,

[0090] Use Gaussian process regression to describe the relationship between the N-1 dimensional space formed and the apparent discharge energy ADE, where the apparent discharge energy is measured by a high-frequency current sensor, and

[0091] Gaussian process is described by the mean function and the covariance function as:

[0092] ,

[0093] where the covariance function adopts the squared exponential type:

[0094]

[0095] Furthermore, the specific parameters of the regression model based on Gaussian process regression are determined by the training data set, where

[0096] For the training data set:

[0097]

[0098] where

[0099] in the covariance function , take when represents the vector composed of the response amplitudes of the sensor under the action of the i-th discharge pulse: , i takes values from 1 to n, and n is the n-th discharge pulse;

[0100] take when then take , represents the vector of the same type as , j also takes values from 1 to n, but j is not equal to i;

[0101] and are the independent variables corresponding to the vectors composed of the response amplitudes of the sensor under the action of different discharge pulses, formally not equal to ;

[0102] when i takes values from 1 to n, denotes the apparent discharge energy measured by the high-frequency current sensor under the action of the \(i\)-th discharge pulse;

[0103] Furthermore, the random vector of the predictor variable distribution of the training dataset is generated as:

[0104] ,

[0105] where the joint distribution of the function value \(Y\) and the observed target value is expressed as:

[0106] ,

[0107] where the observed target value is the predicted value of the discharge energy to be predicted, and this predicted value of the discharge energy is the output of the regression model

[0108] denotes the \(n\times n\) identity matrix with diagonal elements equal to 1 and the remaining elements equal to 0. Since is the training dataset, therefore, is composed of all the elements in the training dataset and the new observed data ;

[0109] is the mean vector of the training dataset , and the \(i\)-th component of is the average of all the elements in the \(i\)-th row;

[0110] is 's mean vector, and the \(i\)-th component of is the average of all the elements in the \(i\)-th row;

[0111] is Gaussian noise, and the Gaussian noise \(\varepsilon\) follows the normal distribution ;

[0112] is the random error included in the variance of, and this random error reduces the influence brought by the measurement error existing in the training dataset;

[0113] is the covariance matrix, and the element in the \(i\)-th row and \(j\)-th column of this matrix is obtained by taking the square exponential kernel function of the \(i\)-th column element of and the \(j\)-th column element of:

[0114]

[0115] Covariance matrix The element in the i-th row and j-th column of The elements in the i-th column of and The elements in the j-th column of are obtained by taking the squared exponential kernel function;

[0116] The element in the i-th row and j-th column of The elements in the i-th column of and The elements in the j-th column of are obtained by taking the squared exponential kernel function;

[0117] The element in the i-th row and j-th column of The elements in the i-th column of and The elements in the j-th column of are obtained by taking the squared exponential kernel function;

[0118] In the known case, the conditional distribution of is a normal distribution with mean and variance

[0119]

[0120] where the mean and variance are:

[0121]

[0122] ,

[0123] Thus, from the conditional distribution of the mean and variance corresponding to the new observed data can be obtained, so as to obtain the predicted value of the apparent discharge energy.

[0124] Furthermore, for each sensor, in the case where the response amplitude corresponding to the sensor during the discharge development process is known, and in the case where the predicted value of the corresponding apparent discharge energy is obtained, according to the regression model, a regression model between the response amplitude of each sensor and the apparent discharge energy during the discharge development process is established.

[0125] In the preferred embodiment of the method described above, in step 3,

[0126] Find the predicted value of the apparent discharge energy corresponding to the current discharge pulse point of the power equipment in the regression model:

[0127]

[0128] wherein represents the normalized result of the response amplitude of the i-th sensor under the discharge pulse signal at time t, represents the predicted value of the apparent discharge energy obtained from the discharge pulse signal at time t;

[0129] During the monitoring and early warning process, the predicted value of the apparent discharge energy and time are accumulated to determine the insulation risk level:

[0130]

[0131] represents the integral of the apparent discharge energy from 0 to t;

[0132] When Risk_level < a, it is determined that the monitored power equipment has no discharge;

[0133] When a ≤ Risk_level ≤ b, it is determined that the monitored power equipment has a slight partial discharge;

[0134] When Risk_level > b, it is determined that the monitored power equipment has a severe discharge; wherein, the threshold a is set according to 120% of the risk index under the background noise of the power equipment operation site; the threshold b is set according to the risk index at the highest temperature allowed under the rated operating state of the power equipment.

[0135] In one embodiment, the discharge types include corona discharge, surface discharge, and floating potential discharge.

[0136] In one embodiment, the types of sensors include optical sensors, UHF sensors, and ultrasonic sensors.

[0137] More specifically, for the method of using the normalized sensing parameters to perform Gaussian process regression on the apparent discharge energy ADE, the "fitrgp" toolbox of matlab can be called, or it can also be implemented with the following code:

[0138] num = length(gc_x);

[0139] K = zeros(num,num);

[0140] Ks = zeros(num,length(x));

[0141] Kss = zeros(length(x),length(x));

[0142] for i=1:num

[0143] for j = 1:num

[0144] K(i,j) = exp(-((gc_x(i)-gc_x(j)).^2));

[0145] end

[0146] for j = 1:length(x)

[0147] Ks(i,j) = exp(-((gc_x(i)-x(j)).^2));

[0148] end

[0149] end

[0150] for i = 1:length(x)

[0151] for j = 1:length(x)

[0152] Kss(i,j) = exp(-((x(i)-x(j)).^2));

[0153] end s

[0154] end

[0155] sigma = 0.001;

[0156] ;

[0157] ;

[0158] In the above code, the variable gc_x represents a matrix composed of normalized sensing parameters The variable gc_y represents an array composed of the apparent discharge energy measured by a high-frequency Rogowski coil, the variable x represents the data to be predicted, and mean_y represents the predicted value of the apparent discharge energy.

[0159] In one embodiment, a pulse current sensor (high-frequency Rogowski coil) using the pulse current method, an optical sensor using the optical measurement method, a UHF sensor using the UHF method, and an ultrasonic sensor using the ultrasonic method are used to collect partial discharge signals. The discharge defect types include corona discharge, surface discharge, and floating potential discharge.

[0160] In one embodiment, the data obtained by the optical measurement method, the UHF method, and the ultrasonic method are normalized:

[0161]

[0162]

[0163]

[0164] represents the amplitude measured by the ultrasonic sensor represents the amplitude measured by the optical sensor represents the amplitude measured by the UHF sensor, and correspondingly represents the result after corresponding normalization processing

[0165] Plot the results in the David triangle, such as Figure 2 the relative energy ternary diagram under the three types of discharges shown Figure 2 In it, the coordinate calculation formula for each point is:

[0166] ,

[0167] Taking Figure 2 the coordinates of the data points in it as the dataset variables, and the discharge types, including surface discharge, corona discharge, and floating potential discharge, as the label data, use the support vector machine to cluster the discharge types. In order to obtain the confidence probability of the clustering results, the following method is used for output transformation:

[0168] .

[0169] The obtained posterior probability is as Figure 3 shown

[0170] For each discharge type, taking the coordinates of the data points in the ternary diagram as the dataset variables and the apparent discharge energy measured by the high-frequency Rogowski coil as the predictive variable, use Gaussian process regression to establish a regression model between the partial discharge ternary energy pattern and the apparent discharge energy:

[0171] ,

[0172] The covariance matrix is calculated using the Gaussian kernel function:

[0173]

[0174] The apparent discharge energy ADE follows an n-variate Gaussian joint distribution on the dataset, and for new data points, it follows an n+1-variate Gaussian joint distribution:

[0175]

[0176] Then, on the new data points, its conditional distribution is a univariate Gaussian distribution:

[0177]

[0178] Among them, the calculation formulas for the mean and variance are as follows:

[0179]

[0180]

[0181] Therefore, the mean and variance of the Gaussian distribution at any point of ADE can be obtained, where the variance can provide a confidence interval, and the mean is used as the predicted value of ADE. Taking the predicted value of the apparent discharge energy as the z-axis, the regression surface is plotted in a three-dimensional coordinate system, as Figures 4(a) to 4(c) shown, where Figs. 4(a), 4(b), and 4(c) are the results under corona discharge, surface discharge, and floating discharge respectively. From Figures 4(a) to 4(c) it can be seen that under the three different types of discharges, the discharge data points all fall on the fitting surface obtained by Gaussian process regression, which proves the effectiveness of the Gaussian process regression model in this scheme.

[0182] In order to measure the effect of the risk assessment method described in the present invention, a predicted value-true value verification diagram of the apparent discharge energy and the relationship between the cumulative discharge time and the risk assessment level are plotted, as Figures 5(a) to 5(c) shown, where Figs. 5(a), 5(b), and 5(c) are the results under corona discharge, surface discharge, and floating discharge respectively. From Figures 5(a) to 5(c) it can be seen that the pulse mode with weaker ADE may have relatively high risk indicators, which indicates that it is difficult to conduct risk assessment only through the magnitude of ADE; the secondary coordinates (upper and right coordinate axes) are represented by a bar chart to show the correlation between the risk level and the risk indicator, indicating that there is basically a positive correlation between the two evaluation indicators, which proves the effectiveness of the risk assessment method proposed in the present invention.

[0183] In addition, in one embodiment, the present invention also provides a power equipment insulation state risk assessment system, which includes a processor, and the processor executes the power equipment insulation state risk assessment method described in any one of the foregoing.

[0184] In the power equipment insulation state risk assessment system described above, the system further includes: a plurality of sensors connected to the processor.

[0185] Preferably, the sensor includes a partial discharge sensor.

[0186] Preferably, the processor includes a single-chip microcomputer.

[0187] Preferably, the processor includes a wireless communication unit.

[0188] In addition, the present invention also discloses a power equipment insulation state risk assessment device, and the device includes:

[0189] A clustering unit that uses various types of sensors to measure the partial discharge data of power equipment and performs clustering of discharge types;

[0190] A modeling unit that establishes a regression model between the response amplitude of each sensor and the apparent discharge energy during the discharge development process based on the clustering results of each discharge type;

[0191] An evaluation unit that calculates the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment based on the regression model, and uses the cumulative value as an index for risk assessment to evaluate the insulation status of the power equipment.

[0192] In addition, the present invention also discloses a computer storage medium storing computer-executable instructions for executing any of the methods described above.

[0193] It should be noted that except that the apparent discharge energy is measured by a high-frequency current sensor during the process of establishing the regression model, the types of sensors in the present invention also include optical sensors, ultra-high frequency sensors, and ultrasonic sensors. Although the embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the above specific embodiments and application fields. The above specific embodiments are merely illustrative and guiding, rather than restrictive. Those of ordinary skill in the art can also make many forms under the inspiration of this specification and without departing from the scope protected by the claims of the present invention, and these all belong to the scope of protection of the present invention.

Claims

1. A method for risk assessment of the insulation state of electrical equipment, characterized in that, The method includes the following steps: Measure the partial discharge data of the power equipment using multiple types of sensors and perform clustering of discharge types; Based on the clustering results of each discharge type, establish a regression model between the response amplitude of each sensor and the apparent discharge energy during the discharge development process; Based on the regression model, calculate the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment, and use the cumulative value as an index for risk assessment to evaluate the insulation status of the power equipment; Among them, calculating the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment based on the regression model, and using the cumulative value as an index for risk assessment to evaluate the insulation status of the power equipment includes the following steps: Find the predicted value of the apparent discharge energy corresponding to the current discharge pulse point of the power equipment in the regression model: ， wherein represents the normalized result of the response amplitude of the i-th sensor under the discharge pulse signal at time t, represents the predicted value of the apparent discharge energy obtained from the discharge pulse signal at time t; During the monitoring and early warning process, accumulate the predicted value of the apparent discharge energy and time to determine the insulation risk level: ， Represents the integral of the apparent discharge energy from 0 to t; When Risk_level < a, it is determined that the power equipment has no discharge; When a ≤ Risk_level ≤ b, it is determined that the power equipment has a slight partial discharge; When Risk_level > b, it is determined that the power equipment has a severe discharge; among them, the threshold a is set at 120% of the risk index under the background noise at the operation site of the power equipment; the threshold b is set according to the risk index at the highest temperature allowed under the rated operation state of the power equipment.

2. The method according to claim 1, characterized in that, Measure the historical partial discharge data of the power equipment using multiple types of sensors and perform clustering of discharge types, including the following steps: Without using a high-frequency current sensor, a total of N sensors of different types are used to collect partial discharge data of power equipment, and the response amplitudes of each sensor are normalized to obtain : ， Among them, represents the response amplitude of the sensor numbered i under the action of a partial discharge pulse, represents the sum of the response amplitudes of N sensors; The result after normalizing the response amplitudes of the previous N - 1 sensors Taking this as the input, perform SVM clustering on the discharge types to establish a clustering region in the N - 1 dimensional space.

3. The method according to claim 2, characterized in that, Based on the clustering results of each discharge type, establish a regression model of the apparent discharge energy in the discharge development path, including the following steps: Use Gaussian process regression to describe the relationship between the N-1 dimensional space formed and the apparent discharge energy ADE, where the Gaussian process is described by a mean function and a covariance function as follows: ； and are the independent variables corresponding to the vectors formed by the response amplitudes of the sensor under different discharge pulses, is not equal in form to ; Among them, the covariance function Adopts the squared exponential type: ， Establish a regression model based on Gaussian process regression, and the specific parameters of the regression model are determined by the training data set, where the training data set is: ， Among them, Covariance function In Take When It represents the vector composed of the response amplitudes of the sensor under the action of the i-th discharge pulse: , i ranges from 1 to n, and n is the n-th discharge pulse; When taking at that time then take , indicating a vector of the same type as ; j also takes values from 1 to n, but j is not equal to i When i takes values from 1 to n, represents the apparent discharge energy measured by the high-frequency current sensor under the action of the i-th discharge pulse; The random vector of the distribution of the predictor variables in the training data set is generated as: ， Among them, the joint distribution of the function value Y and the observed target value is expressed as: ， Among them, the observed target value is the predicted value of the discharge energy to be predicted, and the predicted value of the discharge energy is the output of the regression model; denotes an \(n\times n\) identity matrix with diagonal elements equal to 1 and the remaining elements equal to 0, which is composed of all elements in the training data set and the new observed data ; is the training data set The mean vector of The i-th component of is the average value of all elements in the i-th row; is the mean vector of, the i-th component of the average of all elements in the i-th row; is Gaussian noise, and the Gaussian noise ε follows a normal distribution ; For the random error contained in the variance, which reduces the impact brought by the measurement error existing in the training dataset; is the covariance matrix, and the element in the \(i\)-th row and \(j\)-th column of this matrix is obtained from the \(i\)-th column element of and the \(j\)-th column element by taking the squared exponential kernel function: ， Covariance matrix The element in the i-th row and j-th column is obtained from the i-th column element of and the j-th column element of by taking the squared exponential kernel function; The element in the $i$-th row and $j$-th column of is the $i$-th column element of and is the $j$-th column element of obtained by taking a squared exponential kernel function; The element in the i-th row and j-th column of is composed of the elements in the i-th column of and the elements in the j-th column of by taking the squared exponential kernel function; Given that is known, the conditional distribution is a normal distribution with mean and variance : ， where the mean value , variance is as follows: ， ， From the conditional distribution, new observed data is obtained, and the corresponding mean and variance are obtained, so as to obtain the predicted value of the apparent discharge energy, that is, the observed target value .

4. The method according to claim 1, wherein The discharge types include corona discharge, surface discharge, and floating potential discharge.

5. The method according to claim 1, wherein The types of sensors include optical sensors, ultra-high frequency sensors, and ultrasonic sensors.

6. A risk assessment device for the insulation state of electrical equipment, characterized in that, The device includes: A clustering unit that measures the partial discharge data of the power equipment using multiple types of sensors and performs clustering of discharge types; A modeling unit that, based on the clustering results of each discharge type, establishes a regression model between the response amplitude of each sensor and the apparent discharge energy during the discharge development process; An evaluation unit that, based on the regression model, calculates the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment, and uses the cumulative value as an index for risk assessment to evaluate the insulation status of the power equipment; Among them, calculating the cumulative value of the apparent discharge energy of the current partial discharge of the power equipment based on the regression model, and using the cumulative value as an index for risk assessment to evaluate the insulation status of the power equipment includes the following steps: Find the predicted value of the apparent discharge energy corresponding to the current discharge pulse point of the power equipment in the regression model: ， wherein represents the normalized result of the response amplitude of the i-th sensor under the discharge pulse signal at time t, represents the predicted value of the apparent discharge energy obtained from the discharge pulse signal at time t; During the monitoring and early warning process, accumulate the predicted value of the apparent discharge energy and time to determine the insulation risk level: ， Represents the integral of the apparent discharge energy from 0 to t; When Risk_level < a, it is determined that the power equipment has no discharge; When a ≤ Risk_level ≤ b, it is determined that the power equipment has a slight partial discharge; When Risk_level > b, it is determined that a serious discharge occurs in the power equipment; among them, the threshold value a is set according to 120% of the risk index under the background noise at the operation site of the power equipment; the threshold value b is set according to the risk index at the highest temperature allowed under the rated operation state of the power equipment.

7. A power equipment insulation state risk assessment system, characterized in that It includes a processor, and the processor executes the power equipment insulation state risk assessment method described in any one of claims 1-5.

8. The system according to claim 7, wherein The system further includes: a plurality of sensors connected to the processor.

9. A computer storage medium, characterized in that, The computer storage medium stores computer-executable instructions, and the computer-executable instructions are used to execute the method described in any one of claims 1 to 5.

Images