A method for detecting a fault in a buffer layer of a high voltage cable based on an ablative gas

Abstract

The application discloses a kind of based on ablation gas high-voltage cable buffer layer fault detection method, comprising the following detection steps: collecting the characteristic gas sample generated by discharge is analyzed, and sample data is obtained;Sample data is divided into test set and training set, and is normalized;Support vector machine model is trained using training set, and support vector machine model is optimized by improving grey wolf optimization algorithm, and the optimal parameter value of support vector machine model penalty factor C and kernel parameter g is obtained;Get the ablation detection model after optimization, with the volume fraction of characteristic gas as the input of the model, with the corresponding buffer layer state as the output of the model, to obtain detection result;The application is based on the characteristic gas data generated by buffer layer discharge ablation defect, constructs the support vector machine ablation detection model optimized based on improved grey wolf optimization algorithm, realizes rapid fault detection, effectively improves detection efficiency, reduces detection cost.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of high-voltage cable fault detection, and particularly relates to a high-voltage cable buffer layer fault detection method based on ablation gas. BACKGROUND

[0002] High-voltage XLPE cables are important primary power transmission devices in power grids, and play an important role in safe and stable operation of power loads. The high-voltage XLPE cables have the characteristics of small line path width, easy selection of line path, hidden buried setting, less influence of surrounding environment and pollution, high reliability of power transmission, etc. With the increasing proportion of high-voltage XLPE cables in power cables, the importance of the high-voltage XLPE cables to the power transmission system is also increasing. Whether the operation condition of the high-voltage XLPE cables is good has an important influence on the safe operation of the power grid.

[0003] Most of the high-voltage XLPE cables used in the prior art adopt a corrugated (corrugated) aluminum sheath structure, and a buffer layer is additionally arranged between the metal sheath and the insulated core. The water-blocking buffer layer is composed of semi-conductive non-woven fabric and semi-conductive fluffy cotton, and a layer of water-blocking powder is filled in the middle. If the electrical contact between the buffer layer and the aluminum sheath needs to be enhanced, copper wire braided cloth (commonly known as gold cloth) is additionally wrapped around the buffer layer. The buffer layer can absorb the radial thermal expansion of the insulation, provide protection functions such as longitudinal water blocking and heat insulation, and provide a good electrical contact interface for the insulation shield and the metal sheath.

[0004] During normal operation, the cable buffer layer will gradually ablate, appear defects and breakage under the action of heat, electricity, machinery and the like. When the ablation defects of the buffer layer are serious, the insulation shield layer will be ablated, which will eventually lead to the breakdown of the body, and when the breakage is serious, it will also lead to the occurrence of faults, causing line tripping, power failure and economic losses.

[0005] In the prior art, based on the ablation formation mechanism of the buffer layer, the methods for diagnosing the ablation defects of the cable buffer layer mainly include partial discharge detection and X-ray detection. Both of these two detection methods have the limitations of less ideal detection effect, easy interference, long detection time, high cost and many other field application limitations.

[0006] Therefore, it is necessary to develop a high-voltage cable buffer layer fault detection method based on ablation gas to solve the above problems. SUMMARY

[0007] The purpose of the present application is to overcome the shortcomings of the prior art, and provide a high-voltage cable buffer layer fault detection method based on ablation gas. Based on the characteristic gas data generated by the discharge ablation defects of the buffer layer, a support vector machine ablation detection model based on an improved grey wolf optimization algorithm is constructed, rapid fault detection is realized, the detection efficiency is effectively improved, and the detection cost is reduced.

[0008] The object of the present application is achieved in that a kind of high-voltage cable buffer layer fault detection method based on ablation gas, comprising the following detection steps:

[0009] S1, by simulating cable buffer layer ablation condition, characteristic gas sample generated by discharge is collected, and characteristic gas type and volume fraction are analyzed, characteristic gas component test result is obtained, and it is corresponding to different defect states of cable buffer layer, sample data is obtained;

[0010] S2, sample data is divided into test set and training set, and data is normalized;

[0011] S3, the training set is used to train the support vector machine model, and the support vector machine model is optimized by improving grey wolf optimization algorithm, and the optimal parameter values of the penalty factor C and the kernel parameter g of the support vector machine model are obtained;Specifically, the following steps are included:

[0012] S31, set related parameters, and initialize all parameters in population;

[0013] S32, the elite reverse learning strategy is used to initialize population, and all values in current and reverse population are sequentially brought into support vector machine model, the fitness value of corresponding individual is calculated, and the optimal first three grey wolf positions are determined according to fitness size;

[0014] S33, the convergence factor is updated by nonlinear control parameter strategy, and the grey wolf population position is updated by Levy flight strategy, and the corresponding fitness value is calculated;

[0015] S34, dynamic inertia factor is adopted, and the grey wolf individual position is updated based on hybrid particle swarm algorithm;

[0016] S35, the improved grey wolf optimization algorithm is constructed according to the above strategy, the fitness value of corresponding individual is calculated, and comparison is made, the individual position with the highest fitness value is the search target;

[0017] S36, it is judged whether the maximum iteration number is reached, if not, return to step S32, otherwise, the optimization process is ended, the optimal (C, g) combination is output, and the optimal parameters are input into the support vector machine model, so that the support vector machine model has the best classification performance, and then the support vector machine ablation detection model based on improved grey wolf optimization algorithm optimization is constructed;

[0018] S4, the optimized ablation detection model is obtained, the volume fraction of characteristic gas is used as the input of the model, and the corresponding buffer layer state is used as the output of the model, and the detection result is obtained.

[0019] Further, the step S1 adopts a gas chromatography-mass spectrometry method to analyze the collected characteristic gas sample, and the characteristic gas types include dibutyl phthalate, cinnamaldehyde, 4-ethylbenzaldehyde and formaldehyde.

[0020] Further, the sample data in the step S1 specifically includes the gas in the normal operation state cable buffer layer, the gas in the 2mm water-blocking buffer layer discharge ablation, and the gas type and volume fraction in the 0.5mm water-blocking buffer layer discharge ablation.

[0021] Further, the step S2 specifically adopts a normalization method to pre-process the sample data, and the normalization formula is: min-max

[0022]

[0023] , wherein, , are the values of the gas data before and after normalization respectively, is the minimum value of the gas data, is the maximum value of the gas data, and the gas is one of the characteristic gases generated by discharge.

[0024] Further, the step S31 sets the related parameters, including setting the penalty factor C and the kernel parameter g in the support vector machine model optimization parameter, wherein the value range of the penalty factor C is [0, 200], and the value range of the kernel parameter g is [0, 100]; initializing the related parameters includes initializing the maximum iteration number, the population size, the inertia weight factor, the speed and position of each particle, the individual learning factor and the social learning factor.

[0025] Further, the step S32 initializes the population by using the elite reverse learning strategy, specifically including: setting the elite individual in the grey wolf population as , is the search dimension, the reverse solution is , and the following is obtained:

[0026]

[0027] , wherein, is the dynamic coefficient on , , and are dynamic boundaries; if crosses the dynamic boundary to become an infeasible solution, it is reset by a random method through the following formula:

[0028]

[0029] ​Further, the fitness function for calculating the fitness value of the corresponding individual in step S32 is the classification accuracy of the support vector machine model under 5-fold cross-validation method.

[0030] Further, the nonlinear control parameter strategy in step S33 is:

[0031]

[0032] wherein, is the distance control parameter, i.e., the convergence factor, is the minimum value, which is 0; is the maximum value, which is 2; is the current iteration number; is the maximum iteration number.

[0033] Further, the updating of the gray wolf population position based on the Levy flight strategy in step S33 is represented by the following formula:

[0034]

[0035] wherein, and are the positions of the gray wolf individuals before and after updating by the Levy flight strategy, respectively; is the dot product; is the optimal solution at this time; is the step length adjustment coefficient and the parameter is updated by the nonlinear control strategy, wherein, , ; is the current iteration number; is the maximum iteration number; is the Levy flight path,

[0036]

[0037] wherein, and are subject to normal distribution, , , is the standard deviation, which is obtained by the following formula:

[0038]

[0039]

[0040] wherein, is usually in the range of [0, 2], and here .

[0041] Further, the step S34 of updating the grey wolf individual position based on the hybrid particle swarm algorithm is represented by the following formula:

[0042]

[0043]

[0044] wherein, is the updated individual moving speed, is the current individual moving speed, are random numbers between 0 and 1, is a variable coefficient, , , are the updated positions of the top three individuals in the grey wolf population respectively, is the individual position before being updated by the hybrid particle swarm algorithm;

[0045]

[0046] wherein, is the updated individual position, is the grey wolf individual position after the chase ends;

[0047] is a dynamic inertia weight factor, and the updating strategy thereof is:

[0048]

[0049] wherein, is the current iteration number, is the total iteration number, takes the value 0.9, takes the value 0.4.

[0050] Due to the adoption of the above technical solutions, the present application has the beneficial effects that: by simulating the ablation conditions of the cable buffer layer and collecting the characteristic gas samples generated by the discharge, the collected characteristic gas samples are analyzed by using the gas chromatography-mass spectrometry method to obtain the characteristic gas type and volume fraction data, the grey wolf optimization algorithm is improved by using the gold eagle reverse learning strategy, the nonlinear control parameter strategy and the Levy flight strategy, the grey wolf individual position is updated by using the hybrid particle swarm algorithm, the improved grey wolf optimization algorithm is used to optimize the parameters of the support vector machine model, the support vector machine ablation detection model optimized based on the improved grey wolf optimization algorithm is constructed, the volume fraction of the characteristic gas is taken as the input of the model, the corresponding buffer layer state is taken as the output of the model, the detection result is obtained, the detection precision and efficiency are effectively improved, and the detection cost is reduced. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 is a flow chart of the present application.

[0052] Figure 2 is a flow chart of step S3 in the present application. DETAILED DESCRIPTION

[0053] The technical solutions of the present application will be further specifically described below by examples in combination with the drawings.

[0054] As shown in Figure 1 , Figure 2 , a kind of high voltage cable buffer layer fault detection method based on ablation gas, comprising the following detection steps:

[0055] S1, by simulating cable buffer layer ablation conditions, the characteristic gas sample generated by discharge is collected, and the characteristic gas species and volume fraction are analyzed, the characteristic gas component test result is obtained, and it is corresponding to the different defect states of cable buffer layer, sample data is obtained;Preferably, the step S1 uses the method of gas chromatography-mass spectrometry to analyze the collected characteristic gas sample, the characteristic gas species includes dibutyl phthalate, cinnamaldehyde, 4-ethyl benzaldehyde and formaldehyde;The sample data in the step S1 specifically includes normal operation state cable buffer layer gas, 2mm water blocking buffer layer discharge ablation gas, 0.5mm water blocking buffer layer discharge ablation gas species and volume fraction.

[0056] S2, the sample data is divided into test set and training set, and the data is normalized;Preferably, the step S2 specifically adopts min-max normalization method for pretreatment of sample data, and the normalization formula is:

[0057]

[0058] , wherein, , respectively, the value of the gas data before and after normalization, is the minimum value of the gas data, is the maximum value of the gas data, wherein the gas is one of the characteristic gases generated by discharge.

[0059] S3, the training set is used to train the support vector machine model, and the support vector machine model is optimized by improving the grey wolf optimization algorithm, and the optimal parameter values of the penalty factor C and the kernel parameter g of the support vector machine model are obtained. Specifically, it includes the following steps:

[0060] S31, set related parameters and initialize all parameters in the population; specifically, setting related parameters includes setting the penalty factor C and the kernel parameter g in the support vector machine model optimization parameters, wherein the penalty factor C takes a value range of [0, 200], and the kernel parameter g takes a value range of [0, 100]; initializing related parameters includes initializing the maximum number of iterations, the population size, the inertia weight factor, the speed and position of each particle, the individual learning factor and the social learning factor.

[0061] S32, initialize the population by using the elite reverse learning strategy, sequentially bring all values in the current and reverse populations into the support vector machine model, calculate the fitness value of the corresponding individual, and determine the optimal top three gray wolf positions according to the fitness size.

[0062] In the step S32, initializing the population by using the elite reverse learning strategy specifically includes: setting the elite individual in the gray wolf population as , is the search dimension, and the reverse solution is , obtaining:

[0063]

[0064] , wherein, is the dynamic coefficient on , , and are dynamic boundaries; if goes beyond the dynamic boundary to become a non-feasible solution, resetting by a randomly generated method through the following formula:

[0065]

[0066] In the step S32, when calculating the fitness value of the corresponding individual, the fitness function is the classification accuracy of the support vector machine model under the 5-fold cross-validation method.

[0067] S33, update the convergence factor by the nonlinear control parameter strategy, update the gray wolf group position by the Levy flight strategy, and calculate the corresponding fitness value.

[0068] In the step S33, the nonlinear control parameter strategy is:

[0069]

[0070] , wherein, is the distance control parameter, that is, the convergence factor, is the minimum value, which takes a value of 0; is the maximum value, which takes a value of 2; is the current iteration number; The maximum number of iterations.

[0071] wherein the updating of the grey wolf colony position based on the Levy flight strategy in the step S33 is represented by the following formula:

[0072]

[0073]

[0074] wherein, and are the grey wolf individual positions before and after being updated by using the Levy flight strategy, respectively; is the dot product; is the optimal solution at this time; is a step length adjustment coefficient and the parameters are updated by using a nonlinear control strategy, wherein, , ; is the current iteration number; is the maximum iteration number; is the Levy flight path,

[0075]

[0076] wherein, and are subject to normal distribution, , , is a standard deviation, and is obtained by the following formula, respectively:

[0077]

[0078]

[0079] wherein, is usually in the range of [0, 2], and here .

[0080] S34, the grey wolf individual position is updated based on the hybrid particle swarm algorithm by using a dynamic inertia factor.

[0081] wherein the updating of the grey wolf individual position based on the hybrid particle swarm algorithm in the step S34 is represented by the following formula:

[0082]

[0083]

[0084] wherein, is the updated individual moving speed, is the current individual moving speed, are random numbers between [0, 1], is a variable coefficient, , , are the updated positions of the top three individuals in the gray wolf population, respectively, is the individual position before updating using the hybrid particle swarm algorithm;

[0085]

[0086] , wherein, is the updated individual position, is the position of the gray wolf individual after the chase ends;

[0087] is a dynamic inertia weight factor, and its updating strategy is:

[0088]

[0089] , wherein, is the current iteration number, is the total iteration number, takes the value 0.9, takes the value 0.4.

[0090] S35, according to the above strategy, the improved gray wolf optimization algorithm is constructed, the fitness values of the corresponding individuals are calculated, and comparison is made, and the position of the individual with the highest fitness value is the search target.

[0091] S36, it is judged whether the maximum iteration number is reached, if not, it returns to step S32, otherwise, the optimization process is ended, the optimal (C, g) combination is output, and the optimal parameters are input into the support vector machine model, so that the support vector machine model has the best classification performance, and then the ablation detection model based on the support vector machine optimized by the improved gray wolf optimization algorithm is constructed.

[0092] S4, the optimized ablation detection model is obtained, the volume fraction of the characteristic gas is taken as the input of the model, and the corresponding buffer layer state is taken as the output of the model, and the detection result is obtained.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can still be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement thereof should be covered in the scope of the claims of the present application.

Claims

1. A method for detecting a fault in a buffer layer of a high voltage cable based on an ablative gas, characterized in that: The method comprises the following detection steps: S1, by simulating the ablation conditions of the cable buffer layer, collecting the characteristic gas samples generated by discharge, and analyzing the types and volume fractions of the characteristic gases, obtaining the characteristic gas component test results, and corresponding to different defect states of the cable buffer layer, obtaining sample data; S2, the sample data is divided into test set and training set, and the data is normalized; S3, the training set is used to train the support vector machine model, and the support vector machine model is optimized by improving the grey wolf optimization algorithm, and the optimal parameter values of the penalty factor C and the kernel parameter g of the support vector machine model are obtained; Specifically, the following steps are included: S31, set the related parameters and initialize all parameters in the population; S32, initialize the population by using the elite reverse learning strategy, bring all values in the current and reverse population into the support vector machine model in turn, calculate the fitness value of the corresponding individual, and determine the optimal top three grey wolf positions according to the fitness value; S33, update the convergence factor by the nonlinear control parameter strategy, update the grey wolf group position by the Levy flight strategy, and calculate the corresponding fitness value; S34, update the grey wolf individual position based on the hybrid particle swarm algorithm with dynamic inertia factor; S35, according to the above strategy, the improved grey wolf optimization algorithm is constructed, the fitness value of the corresponding individual is calculated, and comparison is made, and the individual position with the highest fitness value is the search target; S36, judge whether the maximum iteration number is reached, if not, return to step S32, otherwise, end the optimization process, output the optimal (C, g) combination, and input the optimal parameters into the support vector machine model, so that the support vector machine model has the best classification performance, and then a support vector machine ablation detection model based on the improved grey wolf optimization algorithm is constructed; S4, the optimized ablation detection model is obtained, the volume fraction of the characteristic gas is taken as the input of the model, and the corresponding buffer layer state is taken as the output of the model, and the detection result is obtained.

2. A method for detecting a fault in a buffer layer of a high voltage cable based on an ablation gas according to claim 1, characterized in that: The method of gas chromatography-mass spectrometry is used to analyze the collected characteristic gas samples in step S1, and the types of characteristic gases include dibutyl phthalate, cinnamyl aldehyde, 4-ethyl benzaldehyde and formaldehyde.

3. A method for detecting a fault in a buffer layer of a high voltage cable based on an ablation gas according to claim 1, characterized in that: The sample data in step S1 specifically includes the gas of the cable buffer layer in normal operation state, the gas of 2mm water-blocking buffer layer discharge ablation, and the gas type and volume fraction of 0.5mm water-blocking buffer layer discharge ablation.

4. A method for detecting a fault in a buffer layer of a high voltage cable based on an ablation gas according to claim 1, characterized in that: The step S2 specifically adopts min-max The normalization method is used for pre-processing the sample data, and the normalization formula is: , wherein, , are the values of the data of the gas before and after normalization, respectively, is the lowest value in the data of the gas, is the highest value in the data of the gas, wherein the gas is one of the characteristic gases generated by the discharge.

5. A method for detecting a fault in a buffer layer of a high voltage cable based on ablation gas according to claim 1, characterized in that: The related parameters in step S31 include the penalty factor C and the kernel parameter g in the support vector machine model optimization parameters, wherein the value range of the penalty factor C is [0, 200], and the value range of the kernel parameter g is [0, 100]; the initialization of the related parameters includes the initialization of the maximum iteration number, the population size, the inertia weight factor, the speed and position of each particle, the individual learning factor and the social learning factor.

6. A method for detecting a fault in a buffer layer of a high voltage cable based on ablation gas according to claim 1, characterized in that: The step S32 of initializing the population by using the elite reverse learning strategy specifically comprises: setting the elite individual in the grey wolf population as , is a search dimension, and the reverse solution thereof is , and the following is obtained: where is the dynamic coefficient on , and is the dynamic boundary; if crosses the dynamic boundary to become a non-feasible solution, it is reset by a randomly generated method through the following equation: 。 7. A method for detecting a fault in a buffer layer of a high voltage cable based on ablation gas according to claim 1, characterized in that: When calculating the fitness value of the corresponding individual in step S32, the fitness function is the classification accuracy of the support vector machine model under the 5-fold cross-validation method.

8. A method for detecting a fault in a buffer layer of a high voltage cable based on ablation gas according to claim 1, characterized in that: The nonlinear control parameter strategy in step S33 is: wherein is a distance control parameter, i.e. a convergence factor, is a minimum value, which takes the value 0; is a maximum value, which takes the value 2; is the current iteration number; is the maximum iteration number.

9. A method for detecting a fault in a buffer layer of a high voltage cable based on ablation gas according to claim 1, characterized in that: The step S33 of updating the gray wolf colony position based on the Levy flight strategy is represented by the following formula: where, and are the grey wolf individual positions before and after updating with the Levy flight strategy, respectively; is the dot product; is the optimal solution at this time; is the step length adjustment coefficient and the parameter is updated using a nonlinear control strategy, where, , ; is the current iteration number; is the maximum iteration number; is the Levy flight path, wherein and are subject to a normal distribution, , , is the standard deviation, respectively, obtained from the following equations: wherein Typically, [0, 2], here .

10. A method for detecting a fault in a buffer layer of a high voltage cable based on ablation gas according to claim 1, characterized in that: The step S34 of updating the gray wolf individual position based on the hybrid particle swarm algorithm is represented by the following formula: wherein, is the updated individual moving speed, is the current individual moving speed, are random numbers between [0, 1], is the variable coefficient, , , are the updated positions of the top three individuals in the gray wolf population, respectively, is the individual position before updating using the hybrid particle swarm algorithm; wherein, is the updated individual position, is the gray wolf individual position after the pursuit ends; is a dynamic inertia weight factor, whose updating strategy is: wherein is the current iteration number, is the total iteration number, takes the value 0.9, takes the value 0.4.

Images