An improved method and device for predicting breakdown voltage based on the shortest discharge path

Abstract

The present invention provides an improved method and device for predicting breakdown voltage based on the shortest discharge path, belonging to the field of high voltage and insulation technology. It includes using training samples to perform simulation calculations to extract electric field feature quantities related to the electric field integral ratio on the shortest discharge path, using the electric field feature quantities related to the electric field integral ratio to train an optimal SVC prediction model, establishing a finite element simulation model for the predicted air gap, and using the optimal SVC prediction model to predict the breakdown voltage of the predicted air gap, and continuously adjusting the loaded estimated voltage until the gap breaks down. Since the present invention only uses feature quantities related to the electric field integral ratio as electric field feature quantities to construct the model, the selection of electric field feature quantities is simplified, making the prediction process of the air gap breakdown voltage faster.

Description

Technical Field

[0001] The invention belongs to the technical field of high voltage and insulation, and in particular relates to an improved method and device for predicting breakdown voltage based on the shortest discharge path. Background Art

[0002] Air is the most widely used insulating medium in power transmission and transformation projects, and the breakdown voltage of air gaps is the main basis for the insulation design of electrical equipment. Currently, the breakdown characteristics of air gaps are basically obtained through experiments, but the experimental research on the breakdown characteristics of air gaps has the problems of long cycle and high cost. The applicability and accuracy of empirical, semi-empirical formulas and physical models of breakdown voltage are limited. Predicting the breakdown voltage of air gaps through machine learning has long been a goal that we hope to achieve.

[0003] Some known technologies extract electric field characteristics on the electrode surface, discharge channel, shortest path and the entire area to predict the breakdown voltage of the gap, but for complex engineering gaps, these spatial regions are difficult to define. Although there are also technologies that only extract electric field characteristics on the shortest discharge path of the gap, the defined characteristics are relatively complex, making subsequent operations cumbersome and prone to errors. Summary of the invention

[0004] In view of this, the present invention aims to solve the problem that the characteristic quantities defined in the breakdown voltage prediction method based on the shortest discharge path in the prior art are relatively complex and the operation is cumbersome.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides an improved method for predicting breakdown voltage based on the shortest discharge path, comprising the following steps:

[0007] Establish an electrostatic field simulation model for training samples, load several voltages including breakdown voltage for simulation calculation and extract the electric field characteristic quantity on the shortest discharge path. The electric field characteristic quantity on the shortest discharge path is defined as the proportion of the electric field integral greater than the specified electric field strength value on the shortest discharge path to the entire path integral.

[0008] The support vector classifier is trained based on the electric field characteristics on the shortest discharge path to obtain the optimal SVC prediction model.

[0009] A finite element simulation model of the air gap to be predicted is established, the estimated voltage is loaded to calculate the electrostatic field and extract the electric field characteristics, and the estimated voltage is adjusted using the prediction results of the optimal SVC prediction model until the air gap to be predicted breaks down.

[0010] Furthermore, the electric field characteristic quantity on the shortest discharge path specifically includes: a first high-voltage side characteristic quantity, a second high-voltage side characteristic quantity, a first low-voltage side characteristic quantity, and a second low-voltage side characteristic quantity;

[0011] The first high-voltage side characteristic quantity is defined as the proportion of the electric field integral greater than the specified electric field strength value on the high-voltage side of the shortest discharge path to the entire path integral;

[0012] The second high-voltage side characteristic quantity is defined as the ratio of the electric field integral of the high-voltage side for a specified electric field strength value on the shortest discharge path to the high-voltage side path integral;

[0013] The first low-voltage side characteristic quantity is defined as the proportion of the electric field integral on the low-voltage side of the shortest discharge path that is greater than the specified electric field strength value to the entire path integral;

[0014] The second low-voltage side characteristic quantity is defined as the ratio of the electric field integral of the low-voltage side for a specified electric field strength value on the shortest discharge path to the low-voltage side path integral.

[0015] Furthermore, several voltages including the breakdown voltage are applied as follows:

[0016] The loading voltage interval is [(1-a)U, (1+a)U], with a total of 21 voltages with a step size of 0.1aU, where U is the breakdown voltage, a is a constant, [(1-a)U, (1-0.1a)U] is the tolerance interval, denoted as -1; [U, (1+a)U] is the breakdown interval, denoted as 1.

[0017] Furthermore, the estimated voltage is loaded to calculate the electrostatic field and extract the electric field characteristic quantity, and the estimated voltage is adjusted using the prediction result of the optimal SVC prediction model until the predicted air gap breaks down. Specifically, the method includes:

[0018] The breakdown voltage of the air gap to be predicted is estimated to be V, and the voltage range to be loaded is set to [V min , V max ];

[0019] Loading a voltage in the voltage interval to be loaded to perform electrostatic field simulation calculation and extract electric field characteristic quantities;

[0020] The electric field characteristic quantity is input into the optimal SVC prediction model for prediction. If the prediction results are all -1, the voltage interval to be loaded is reset and V max Set as the lower limit of the new voltage range to be loaded; if the prediction results are all 1, reset the voltage range to be loaded and set V min Set as the upper limit of the new voltage range to be loaded; if the prediction result includes -1 and 1, the voltage at which 1 appears for the first time will be taken as the breakdown voltage of the air gap to be predicted.

[0021] Furthermore, establishing an electrostatic field simulation model for training samples specifically includes:

[0022] According to the characteristics of the electric field distribution on the shortest discharge path of the air gap to be predicted, similar gaps are selected to form training samples, and the electrostatic field simulation model of each training sample is established in the finite element simulation software.

[0023] In a second aspect, the present invention provides an improved breakdown voltage prediction device based on the shortest discharge path, comprising:

[0024] The first model training module is used to establish an electrostatic field simulation model of the training sample, load several voltages including the breakdown voltage for simulation calculation and extract the electric field characteristic quantity on the shortest discharge path. The electric field characteristic quantity on the shortest discharge path is defined as the proportion of the electric field integral greater than the specified electric field strength value on the shortest discharge path to the entire path integral;

[0025] The second model training module is used to train the support vector classifier based on the electric field feature quantity on the shortest discharge path to obtain the optimal SVC prediction model;

[0026] The voltage prediction module is used to establish a finite element simulation model of the air gap to be predicted, load the estimated voltage to calculate the electrostatic field and extract the electric field characteristics, and use the prediction results of the optimal SVC prediction model to adjust the estimated voltage until the air gap to be predicted breaks down.

[0027] Furthermore, the first model training module includes a feature extraction module;

[0028] The feature quantity extraction module is used to extract a first high-pressure side feature quantity, a second high-pressure side feature quantity, a first low-pressure side feature quantity, and a second low-pressure side feature quantity;

[0029] The first high-voltage side characteristic quantity is defined as the proportion of the electric field integral greater than the specified electric field strength value on the high-voltage side of the shortest discharge path to the entire path integral;

[0030] The second high-voltage side characteristic quantity is defined as the ratio of the electric field integral of the high-voltage side for a specified electric field strength value on the shortest discharge path to the high-voltage side path integral;

[0031] The first low-voltage side characteristic quantity is defined as the proportion of the electric field integral on the low-voltage side of the shortest discharge path that is greater than the specified electric field strength value to the entire path integral;

[0032] The second low-voltage side characteristic quantity is defined as the ratio of the electric field integral of the low-voltage side for a specified electric field strength value on the shortest discharge path to the low-voltage side path integral.

[0033] Further, the first model training module includes: a first voltage loading module;

[0034] The first voltage loading module is used to load a total of 21 voltages with a voltage interval of [(1-a)U, (1+a)U] and a step size of 0.1aU, where U is the breakdown voltage, a is a constant, [(1-a)U, (1-0.1a)U] is the tolerance interval, denoted as -1; [U, (1+a)U] is the breakdown interval, denoted as 1.

[0035] Further, the voltage prediction module includes: a second voltage loading module and a breakdown voltage determination module;

[0036] The second voltage loading module is used to set the voltage range to be loaded as [V min , V max ], the voltage of the voltage interval to be loaded is loaded to perform electrostatic field simulation calculation and extract the electric field characteristic quantity;

[0037] The breakdown voltage determination module is used to input the electric field characteristic quantity into the optimal SVC prediction model for prediction. If the prediction results are all -1, the voltage interval to be loaded is reset and V max Set as the lower limit of the new voltage range to be loaded; if the prediction results are all 1, reset the voltage range to be loaded and set V min Set as the upper limit of the new voltage range to be loaded; if the prediction result includes -1 and 1, the voltage at which 1 appears for the first time will be taken as the breakdown voltage of the air gap to be predicted.

[0038] Further, the first model training module includes: a training sample selection module and a simulation model construction module;

[0039] The training sample selection module is used to select similar gaps to form training samples according to the characteristics of the electric field distribution on the shortest discharge path of the air gap to be predicted;

[0040] The simulation model building module is used to establish the electrostatic field simulation model of each training sample in the finite element simulation software.

[0041] In summary, the present invention provides an improved method and device for predicting breakdown voltage based on the shortest discharge path, including using training samples to perform simulation calculations to extract electric field characteristics about the electric field integral ratio on the shortest discharge path, using the electric field characteristics about the electric field integral ratio to train an optimal SVC prediction model, establishing a finite element simulation model for the predicted air gap, and using the optimal SVC prediction model to predict the breakdown voltage of the predicted air gap, and continuously adjusting the loaded estimated voltage until the gap breaks down. Since the present invention only uses characteristics related to the electric field integral ratio as electric field characteristics to construct the model, it simplifies the selection of electric field characteristics, making the prediction process of the air gap breakdown voltage faster. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0043] Figure 1 A schematic flow chart of an improved method for predicting breakdown voltage based on the shortest discharge path provided by an embodiment of the present invention;

[0044] Figure 2 A schematic flow chart of a method for predicting breakdown voltage based on a standard spherical gap provided in an embodiment of the present invention;

[0045] Figure 3 A schematic diagram of defining the electric field characteristic quantity on the shortest discharge path provided by an embodiment of the present invention;

[0046] Figure 4 A graph showing the electric field distribution on the shortest discharge path of a spherical gap provided by an embodiment of the present invention;

[0047] Figure 5 A comparison chart of the experimental value and the predicted value of the breakdown voltage provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Air is the most widely used insulating medium in power transmission and transformation projects, and the breakdown voltage of air gaps is the main basis for the insulation design of electrical equipment. Currently, the breakdown characteristics of air gaps are basically obtained through experiments, but the experimental research on the breakdown characteristics of air gaps has the problems of long cycle and high cost. The applicability and accuracy of empirical, semi-empirical formulas and physical models of breakdown voltage are limited. Predicting the breakdown voltage of air gaps through machine learning has long been a goal that we hope to achieve.

[0050] Some known technologies extract electric field characteristics on the electrode surface, discharge channel, shortest path and the entire area to predict the breakdown voltage of the gap, but for complex engineering gaps, these spatial regions are difficult to define. Although there are also technologies that only extract electric field characteristics on the shortest discharge path of the gap, the defined characteristics are relatively complex, making subsequent operations cumbersome and prone to errors.

[0051] Based on this, an embodiment of the present invention provides an improved breakdown voltage prediction method and device based on the shortest discharge path, which is used to solve the problem that the feature quantities defined by the breakdown voltage prediction method based on the shortest discharge path in the prior art are relatively complex and the operation is cumbersome.

[0052] The following is a detailed introduction to an embodiment of an improved breakdown voltage prediction method based on the shortest discharge path of the present invention.

[0053] See also Figure 1 , this embodiment provides an improved breakdown voltage prediction method based on the shortest discharge path, comprising the following steps:

[0054] S100: Establish an electrostatic field simulation model for training samples, load several voltages including breakdown voltage for simulation calculation and extract the electric field characteristic quantity on the shortest discharge path, where the electric field characteristic quantity on the shortest discharge path is defined as the proportion of the electric field integral greater than the specified electric field strength value on the shortest discharge path to the entire path integral.

[0055] It should be noted that the training samples are selected using a reasonable method based on the characteristics of the electric field distribution on the shortest discharge path of the air gap to be predicted. The specific operation is to divide the known gap samples into several intervals according to the electric field non-uniformity, randomly select one in each interval as a training sample, and then establish an electrostatic field simulation model for each training sample in the finite element simulation software.

[0056] Loading several voltages including the breakdown voltage specifically refers to loading a total of 21 voltages with a voltage interval of [(1-a)U, (1+a)U] and a step size of 0.1aU, where U is the breakdown voltage, a is a constant, [(1-a)U, (1-0.1a)U] is the tolerance interval, denoted by -1; and [U, (1+a)U] is the breakdown interval, denoted by 1.

[0057] After manually setting the constant a, the voltage is applied to the training sample according to the above rules and the electric field characteristic quantity on the shortest discharge path is extracted from it. The shortest discharge path refers to the straight line where the shortest distance between the two electrodes is located. According to the definition of the shortest discharge path, this embodiment specially selects the characteristic quantity of the electric field integral ratio on the shortest discharge path as the electric field characteristic quantity for subsequent calculation. Specifically, the electric field characteristic quantity is defined as the ratio of the electric field integral greater than the specified electric field strength value on the shortest discharge path to the entire path integral.

[0058] Furthermore, in this embodiment, the electric field characteristic quantity is specifically selected as a first high-voltage side characteristic quantity, a second high-voltage side characteristic quantity, a first low-voltage side characteristic quantity, and a second low-voltage side characteristic quantity. The definitions of each characteristic quantity are as follows:

[0059] The first high-voltage side characteristic quantity is defined as the ratio of the electric field integral on the high-voltage side of the shortest discharge path that is greater than the specified electric field strength value to the entire path integral. The second high-voltage side characteristic quantity is defined as the ratio of the electric field integral on the high-voltage side of the shortest discharge path for the specified electric field strength value to the high-voltage side path integral.

[0060] The first low-voltage side characteristic quantity is defined as the ratio of the electric field integral of the low-voltage side on the shortest discharge path that is greater than the specified electric field strength value to the entire path integral. The second low-voltage side characteristic quantity is defined as the ratio of the electric field integral of the low-voltage side on the shortest discharge path for the specified electric field strength value to the low-voltage side path integral.

[0061] Using the above definition, since the electric field feature quantity uses the ratio of the relevant electric field integral as the feature quantity, the numerical range of the final feature quantity is maintained at [0, 1], and there is no need to normalize the feature quantity, thus avoiding the data saturation problem during the normalization process.

[0062] S200: training a support vector classifier based on the electric field feature quantity on the shortest discharge path to obtain an optimal SVC prediction model.

[0063] It should be noted that the feature quantity obtained in step S100 is subjected to dimension reduction processing using the Pearson correlation coefficient method and then used as the input quantity. Whether the gap breaks down at each voltage is used as the output quantity, a support vector classifier (SVC) is trained, and then the key parameters of the SVC are optimized using an optimization algorithm to obtain the optimal SVC prediction model.

[0064] The optimization method for the key parameters of the SVC model can be cross-validation and grid search.

[0065] S300: Establish a finite element simulation model of the air gap to be predicted, load the estimated voltage to calculate the electrostatic field and extract the electric field characteristic quantity, and adjust the estimated voltage using the prediction result of the optimal SVC prediction model until the air gap to be predicted breaks down.

[0066] It should be noted that the specific operations of predicting the breakdown voltage using the optimal SVC prediction model include:

[0067] S301: Estimate the breakdown voltage of the air gap to be predicted as V, and set the voltage range to be loaded as [V min , V max ].

[0068] It should be noted that the loading rule for the estimated voltage is consistent with the loading rule for the training sample voltage in step S100, that is, the voltage is loaded according to the specified step size and the set voltage interval to be loaded.

[0069] S302: Loading a voltage in the voltage range to be loaded to perform electrostatic field simulation calculation and extract electric field feature quantities.

[0070] It should be noted that the definition of the electric field feature quantity extracted in this step is the same as the definition of the electric field feature quantity extracted during the simulation calculation of the training sample.

[0071] S303: Input the electric field characteristic quantity into the optimal SVC prediction model for prediction. If the prediction results are all -1, reset the voltage interval to be loaded and set V max Set as the lower limit of the new voltage range to be loaded; if the prediction results are all 1, reset the voltage range to be loaded and set V min Set as the upper limit of the new voltage range to be loaded; if the prediction result includes -1 and 1, the voltage at which 1 appears for the first time will be taken as the breakdown voltage of the air gap to be predicted.

[0072] It should be noted that, according to the above-mentioned definition of the voltage range, -1 indicates that the current gap has not yet broken down (corresponding to the gap tolerance range), and 1 indicates that the current gap has broken down (corresponding to the gap breakdown range). According to the relationship between -1 and 1 in the prediction result, the voltage corresponding to the breakdown of the current gap can be accurately known, thereby completing the prediction of the breakdown voltage.

[0073] The present embodiment provides an improved method for predicting breakdown voltage based on the shortest discharge path, including using training samples to perform simulation calculations to extract electric field characteristics about the electric field integral ratio on the shortest discharge path, using the electric field characteristics about the electric field integral ratio to train an optimal SVC prediction model, establishing a finite element simulation model for the predicted air gap, and using the optimal SVC prediction model to predict the breakdown voltage of the predicted air gap, and continuously adjusting the loaded estimated voltage until the gap breaks down. Since the present invention only uses characteristics related to the electric field integral ratio as electric field characteristics to construct the model, the selection of electric field characteristics is simplified, making the prediction process of the air gap breakdown voltage faster.

[0074] Furthermore, the numerical range of the electric field characteristic quantity related to the electric field integral ratio is defined as [0,1], and there is no need to normalize the characteristic quantity, thus avoiding the data saturation problem in the normalization process.

[0075] The following further introduces the above-mentioned breakdown voltage prediction method by taking the standard spherical gap as the research object.

[0076] The ball diameters D are selected to include 5cm, 6.25cm, 10cm, 12.5cm, 15cm and 25cm. The shortest gap distance d is 1cm and the longest is 12cm. A total of 102 samples are selected. Suitable training samples are selected from the ball gap samples, and the remaining ones are used as test samples. The power frequency breakdown voltage is predicted and compared with the test results. The implementation process is as follows: Figure 2 shown.

[0077] The 102 known samples are divided into 20 intervals according to the electric field inhomogeneity, and one sample is randomly selected in each interval as a training sample, for a total of 20 training samples.

[0078] The electrostatic field simulation calculation is performed using COMSOL Multiphysics software. Taking the spherical gap when D = 5 cm and d is 1, 1.2, 1.4, 1.5, and 1.6 cm respectively as an example, the electric field distribution on the shortest discharge path is as follows: Figure 4 As shown, it can be seen that the electric field intensity of the shortest discharge path of the spherical gap is distributed in a U shape.

[0079] The left and right sides of the U-shaped curve are defined as the high pressure side and the low pressure side, and characteristic quantities are defined on the high pressure side and the low pressure side, respectively. The characteristic quantity schematic diagram is shown in Figure 3 Table 1 shows all the feature quantities, a total of 84 dimensions.

[0080] Table 1. The shortest path feature set of the spherical gap

[0081] category quantity <![CDATA[S r Hx]]> 21 <![CDATA[S r HHx]]> 21 <![CDATA[S r Lx]]> 21 <![CDATA[S r LLx]]> 21

[0082] Where x represents the specific electric field strength value, ranging from 15kV / cm to 35kV / cm, with a step size of 1kV / cm. max Indicates the maximum electric field strength on the high voltage side, EL max Indicates the maximum electric field strength on the low voltage side, E min represents the minimum electric field strength, Ex represents the electric field strength when the electric field strength value is x. SH represents the high-voltage side path integral, SL represents the low-voltage side path integral, S r r stands for S r is the ratio. The mathematical expressions of the above-mentioned characteristic quantities are as follows:

[0083] The first high-voltage side characteristic quantity S r Hx = SHx / (SH+SL), which means the ratio of the electric field integral with electric field strength greater than x on the high voltage side to the integral of the entire path;

[0084] The second high-voltage side characteristic quantity S r HHx=SHx / SH, which means the ratio of the electric field integral with electric field strength greater than x on the high-voltage side to the path integral on the high-voltage side;

[0085] The first low-pressure side characteristic quantity S r Lx = SLx / (SH+SL), which represents the ratio of the electric field integral with electric field strength greater than x on the low voltage side to the integral of the entire path;

[0086] The second low-pressure side characteristic quantity S r LLx=SLx / SL, which represents the ratio of the electric field integral on the low-voltage side where the electric field strength is greater than x to the high-voltage side path integral.

[0087] Then, the above 84-dimensional feature quantities are subjected to dimensionality reduction processing, and the feature quantities greater than 0.8 are selected. The selected feature quantities are shown in Table 2.

[0088] Table 2 Feature quantities after dimensionality reduction

[0089]

[0090] The 21 randomly selected training samples are shown in Table 3, where U t is the breakdown voltage of the gap.

[0091] Table 3 Training sample set

[0092]

[0093]

[0094] The training sample set in Table 3 is used to train the SVC model, and the optimal SVC prediction model is established by parameter optimization through 3-fold cross validation and grid search algorithm. The breakdown voltage of the remaining test samples is predicted, and the prediction results are as follows: Figure 5 As shown. It can be seen that the experimental values ​​and predicted values ​​of most samples are in good agreement. The mean absolute percentage error (MAPE) is used to evaluate the prediction error. MAPE can be obtained according to formula (1):

[0095]

[0096] N represents the total number of test samples, i represents the i-th sample, and U t and U p The experimental value and predicted value of breakdown voltage are shown respectively. The calculated results show that MAPE is 3.91%, which means that the prediction accuracy of this prediction method is good.

[0097] The above is a detailed introduction to an embodiment of an improved method for predicting breakdown voltage based on the shortest discharge path of the present invention. The following is a detailed introduction to an embodiment of an improved device for predicting breakdown voltage based on the shortest discharge path of the present invention.

[0098] This embodiment provides an improved breakdown voltage prediction device based on the shortest discharge path, including: a first model training module, a second model training module and a voltage prediction module.

[0099] In this embodiment, the first model training module is used to establish an electrostatic field simulation model of the training sample, load several voltages including the breakdown voltage for simulation calculation and extract the electric field characteristic quantity on the shortest discharge path. The electric field characteristic quantity on the shortest discharge path is defined as the proportion of the electric field integral on the shortest discharge path that is greater than the specified electric field strength value to the entire path integral.

[0100] It should be noted that the first model training module includes a feature extraction module.

[0101] The feature extraction module is used to extract a first high-pressure side feature, a second high-pressure side feature, a first low-pressure side feature and a second low-pressure side feature.

[0102] Among them, the first high-voltage side characteristic quantity is defined as the proportion of the electric field integral of the high-voltage side on the shortest discharge path that is greater than the specified electric field strength value to the entire path integral; the second high-voltage side characteristic quantity is defined as the proportion of the electric field integral of the high-voltage side on the shortest discharge path for the specified electric field strength value to the high-voltage side path integral; the first low-voltage side characteristic quantity is defined as the proportion of the electric field integral of the low-voltage side on the shortest discharge path that is greater than the specified electric field strength value to the entire path integral; the second low-voltage side characteristic quantity is defined as the proportion of the electric field integral of the low-voltage side on the shortest discharge path for the specified electric field strength value to the low-voltage side path integral.

[0103] Furthermore, the first model training module also includes a first voltage loading module.

[0104] The first voltage loading module is used to load a total of 21 voltages with a voltage interval of [(1-a)U, (1+a)U] and a step size of 0.1aU, where U is the breakdown voltage, a is a constant, [(1-a)U, (1-0.1a)U] is the tolerance interval, denoted as -1; [U, (1+a)U] is the breakdown interval, denoted as 1.

[0105] Furthermore, the first model training module also includes a training sample selection module and a simulation model construction module;

[0106] The training sample selection module is used to select similar gaps to form training samples according to the characteristics of the electric field distribution on the shortest discharge path of the air gap to be predicted;

[0107] The simulation model building module is used to establish the electrostatic field simulation model of each training sample in the finite element simulation software.

[0108] In this embodiment, the second model training module is used to train the support vector classifier based on the electric field feature quantity on the shortest discharge path to obtain the optimal SVC prediction model.

[0109] In this embodiment, the voltage prediction module is used to establish a finite element simulation model of the air gap to be predicted, load the estimated voltage to perform electrostatic field calculation and extract the electric field characteristic quantity, and adjust the estimated voltage using the prediction result of the optimal SVC prediction model until the air gap to be predicted breaks down.

[0110] It should be noted that the voltage prediction module includes: a second voltage loading module and a breakdown voltage determination module.

[0111] The second voltage loading module is used to set the voltage range to be loaded as [V min , V max ], the voltage of the voltage interval to be loaded is loaded to perform electrostatic field simulation calculation and extract the electric field characteristic quantity;

[0112] The breakdown voltage determination module is used to input the electric field characteristic quantity into the optimal SVC prediction model for prediction. If the prediction results are all -1, the voltage interval to be loaded is reset and V max Set as the lower limit of the new voltage range to be loaded; if the prediction results are all 1, reset the voltage range to be loaded and set V min Set as the upper limit of the new voltage range to be loaded; if the prediction result includes -1 and 1, the voltage at which 1 appears for the first time will be taken as the breakdown voltage of the air gap to be predicted.

[0113] The breakdown voltage prediction device provided in this embodiment is used to implement the breakdown voltage prediction method of the aforementioned embodiment. The specific configuration of each module is based on the implementation of the method and will not be repeated here.

[0114] The present embodiment provides an improved breakdown voltage prediction device based on the shortest discharge path, including a first model training module that uses training samples to perform simulation calculations to extract electric field feature quantities related to the electric field integral ratio on the shortest discharge path, a second model training module that uses the electric field feature quantities related to the electric field integral ratio to train an optimal SVC prediction model, and a voltage prediction module that establishes a finite element simulation model for the predicted air gap, and uses the optimal SVC prediction model to predict the breakdown voltage of the predicted air gap, and continuously adjusts the loaded estimated voltage until the gap breaks down. Since the first model training module and the voltage prediction module of the present invention only use feature quantities related to the electric field integral ratio as electric field feature quantities to construct the model, the selection of electric field feature quantities is simplified, making the prediction process of the air gap breakdown voltage faster.

[0115] Furthermore, the numerical range of the electric field characteristic quantity related to the electric field integral ratio is defined as [0,1], and there is no need to normalize the characteristic quantity, thus avoiding the data saturation problem in the normalization process.

[0116] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An improved breakdown voltage prediction method based on the shortest discharge path, characterized in that: The following steps are involved: Establish an electrostatic field simulation model for training samples, load several voltages including breakdown voltage for simulation calculation and extract the electric field characteristics on the shortest discharge path; Training a support vector classifier based on the electric field feature quantity on the shortest discharge path to obtain an optimal SVC prediction model; Establishing a finite element simulation model of the air gap to be predicted, loading an estimated voltage to perform electrostatic field calculation and extract electric field characteristic quantities, and adjusting the estimated voltage using the prediction result of the optimal SVC prediction model until the air gap to be predicted breaks down; The electric field characteristic quantity on the shortest discharge path is specifically selected as a first high-voltage side characteristic quantity, a second high-voltage side characteristic quantity, a first low-voltage side characteristic quantity, and a second low-voltage side characteristic quantity; The first high-voltage side characteristic quantity is defined as the proportion of the electric field integral greater than the specified electric field strength value on the high-voltage side of the shortest discharge path to the entire path integral; The second high-voltage side characteristic quantity is defined as the ratio of the electric field integral on the high-voltage side of the shortest discharge path that is greater than the specified electric field strength value to the high-voltage side path integral; The first low-voltage side characteristic quantity is defined as the proportion of the electric field integral on the low-voltage side of the shortest discharge path that is greater than the specified electric field strength value to the entire path integral; The second low-voltage side characteristic quantity is defined as the ratio of the electric field integral on the low-voltage side greater than the specified electric field strength value on the shortest discharge path to the low-voltage side path integral; The loading of several voltages including the breakdown voltage is specifically: The loading voltage interval is [(1-a)U, (1+a)U], with a total of 21 voltages with a step size of 0.1aU, where U is the breakdown voltage, a is a constant, [(1-a)U, (1-0.1a)U] is the tolerance interval, and the corresponding output result of the SVC prediction model is recorded as -1; [U, (1+a)U] is the breakdown interval, and the corresponding output result of the SVC prediction model is recorded as 1.

2. The improved breakdown voltage prediction method based on the shortest discharge path according to claim 1, characterized in that: The step of loading the estimated voltage to calculate the electrostatic field and extract the electric field characteristic quantity, and adjusting the estimated voltage by using the prediction result of the optimal SVC prediction model until the predicted air gap breaks down specifically includes: The breakdown voltage of the air gap to be predicted is estimated to be V, and the voltage range to be loaded is set to [V min , V max ]; Loading the voltage in the voltage interval to be loaded to perform electrostatic field simulation calculation and extract electric field characteristic quantities; The electric field characteristic quantity is input into the optimal SVC prediction model for prediction. If the prediction results are all -1, the voltage interval to be loaded is reset and V max Set as the new lower limit of the voltage range to be loaded; if the prediction results are all 1, reset the voltage range to be loaded and set V min is set as the new upper limit of the voltage range to be loaded; if the prediction result includes -1 and 1, the voltage at which 1 appears for the first time is taken as the breakdown voltage of the air gap to be predicted.

3. The improved breakdown voltage prediction method based on the shortest discharge path according to claim 1, characterized in that: The electrostatic field simulation model of the training sample is established specifically including: According to the characteristics of the electric field distribution on the shortest discharge path of the air gap to be predicted, similar gaps are selected to form training samples, and an electrostatic field simulation model of each training sample is established in finite element simulation software.

4. An improved breakdown voltage prediction device based on the shortest discharge path, characterized in that: include: The first model training module is used to establish an electrostatic field simulation model of the training sample, load several voltages including the breakdown voltage for simulation calculation and extract the electric field characteristic quantity on the shortest discharge path; A second model training module is used to train a support vector classifier based on the electric field feature quantity on the shortest discharge path to obtain an optimal SVC prediction model; A voltage prediction module is used to establish a finite element simulation model of the air gap to be predicted, load the estimated voltage to perform electrostatic field calculation and extract electric field characteristic quantities, and adjust the estimated voltage using the prediction result of the optimal SVC prediction model until the air gap to be predicted breaks down; The first model training module includes a feature extraction module; The feature extraction module is specifically used to extract a first high-pressure side feature, a second high-pressure side feature, a first low-pressure side feature, and a second low-pressure side feature; The first high-voltage side characteristic quantity is defined as the proportion of the electric field integral greater than the specified electric field strength value on the high-voltage side of the shortest discharge path to the entire path integral; The second high-voltage side characteristic quantity is defined as the ratio of the electric field integral on the high-voltage side of the shortest discharge path that is greater than the specified electric field strength value to the high-voltage side path integral; The first low-voltage side characteristic quantity is defined as the proportion of the electric field integral on the low-voltage side of the shortest discharge path that is greater than the specified electric field strength value to the entire path integral; The second low-voltage side characteristic quantity is defined as the ratio of the electric field integral on the low-voltage side greater than the specified electric field strength value on the shortest discharge path to the low-voltage side path integral; The first model training module includes: a first voltage loading module; The first voltage loading module is used to load a total of 21 voltages with a voltage range of [(1-a)U, (1+a)U] and a step size of 0.1aU, where U is the breakdown voltage, a is a constant, [(1-a)U, (1-0.1a)U] is the tolerance interval, and the corresponding output result of the SVC prediction model is recorded as -1; [U, (1+a)U] is the breakdown interval, and the corresponding output result of the SVC prediction model is recorded as 1.

5. The improved breakdown voltage prediction device based on the shortest discharge path according to claim 4, characterized in that: The voltage prediction module includes: a second voltage loading module and a breakdown voltage determination module; The second voltage loading module is used to set the voltage interval to be loaded as [V min , V max ], loading the voltage of the voltage interval to be loaded to perform electrostatic field simulation calculation and extract electric field characteristic quantities; The breakdown voltage determination module is used to input the electric field characteristic quantity into the optimal SVC prediction model for prediction. If the prediction results are all -1, the voltage interval to be loaded is reset and V max Set as the new lower limit of the voltage range to be loaded; if the prediction results are all 1, reset the voltage range to be loaded and set V min is set as the new upper limit of the voltage range to be loaded; if the prediction result includes -1 and 1, the voltage at which 1 appears for the first time is taken as the breakdown voltage of the air gap to be predicted.

6. The improved breakdown voltage prediction device based on the shortest discharge path according to claim 4, characterized in that: The first model training module includes: a training sample selection module and a simulation model construction module; The training sample selection module is used to select similar gaps to form training samples according to the characteristics of the electric field distribution on the shortest discharge path of the air gap to be predicted; The simulation model building module is used to establish an electrostatic field simulation model of each training sample in the finite element simulation software.

Images