Extra-high voltage bushing inner shielding electrode structure optimization method and related device

Abstract

The invention belongs to the technical field of design optimization of extra-high-voltage direct-current wall bushings, and discloses an optimization method for a shielding electrode structure in an extra-high-voltage bushing and a related device.Different parameter combinations of the shielding electrode structure in the extra-high-voltage bushing serve as input, calculation is carried out through a trained BP neural network, and the design of the shielding electrode structure in the extra-high-voltage bushing is optimized; obtaining the field intensity value of the extra-high voltage bushing corresponding to each parameter combination; constructing an optimization variable by using the parameter combination, and optimizing a target function formed by the field intensity values by taking minimization of the maximum field intensity value in the internal electric field and the external electric field of the extra-high voltage bushing and the overall uniformity of the electric field distribution of the extra-high voltage bushing as an optimization target to obtain the corresponding parameter combination meeting the optimization target; and the structure optimization of the shielding electrode in the extra-high voltage bushing is realized. According to the method, the structure of the shielding electrode in the extra-high-voltage bushing can be optimized, and the obtained structure optimization result can provide theoretical guidance and key data support for the design of the inner shielding structure of the extra-high-voltage direct-current SF6 gas insulation bushing.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of design optimization of UHV DC wall bushing, and particularly relates to a UHV bushing inner shielding electrode structure optimization method and device. BACKGROUND

[0002] As the only channel connecting valve hall and DC field in the converter station, UHV DC wall bushing is the core equipment bearing full voltage and full current of the system. Once flashover occurs, it will cause single-pole outage of the line, damage the body and surrounding precious equipment, and the consequences are very serious. In terms of internal insulation, the wall bushing of 110kV and above voltage grade is mostly of capacitive structure, that is, a capacitor core is used as the main insulation, and the remaining filling medium is insulating oil or SF6 compressed gas. Common capacitor core materials are oil-impregnated paper (OIP) and resin-impregnated paper (RIP). Another insulation structure form of the wall bushing is SF6 gas insulation structure, that is, SF6 gas is used as the main insulation, and an aluminum alloy electrode is used to shield the electric field at the three junctions. Compared with the capacitor core, the inner shielding structure has relatively weak regulation and control ability on the internal and external electric field of the bushing. In addition to the surface of the center conductor, strong electric field regions will appear on the surface of the inner shielding electrode and its nearby umbrella skirt, so the field strength control difficulty of the SF6 gas insulation bushing is greater than that of the capacitive bushing (the ratio of the design field strength of the electrode surface in the bushing to the gas pressure under lightning impulse voltage is about 60kV / (mm MPa)). In addition, since the center conductor of this type of bushing (i.e., SF6 gas insulation bushing) loses the support of the capacitor core, its mechanical problem is more prominent than that of the capacitive bushing.

[0003] Capacitive wall bushing has strong electric field control ability, and the operation experience of each voltage grade product is relatively rich because of its early development. However, the process of winding, drying and impregnating of the capacitor core is difficult, and the price of imported bushing is very expensive. In addition, the structure of capacitive bushing is complex, and there are many influencing factors, which directly leads to frequent accidents in actual operation. The inner shielding electrode of SF6 gas insulated wall bushing has relatively weak control ability to the inner and outer electric field, so the design margin of the field strength is usually large. For example, the diameter of the hollow composite insulator used in 800kV DC SF6 gas insulated wall bushing is about 1.3-1.4 times that of the capacitive wall bushing, and the length is about 1.1-1.2 times. At the same time, the lower manufacturing cost of the metal shielding electrode relative to the capacitor core is also one of the reasons for the large design margin of the bushing. So far, SF6 gas insulated bushing accidents rarely occur in actual operation products. In view of the high reliability of SF6 gas insulated structure, and its obvious advantages in installation, cost and localization compared with capacitive structure, it has been gradually applied and popularized in recent years. At present, the external insulation design of UHV DC SF6 gas insulated wall bushing is relatively unified, which is FREP / HTV hollow composite insulator, but the internal insulation design of different manufacturers still has certain differences, so it is necessary to carry out research on the optimization of the insulation structure design of UHV DC SF6 gas insulated wall bushing.

[0004] The existing bushing inner shielding electrode optimization design idea is similar to SF6 equipment or vacuum equipment such as circuit breaker and current transformer, which is to suppress the maximum field strength on the electrode surface. However, the axial size of the SF6 gas insulated bushing inner shielding electrode is significantly larger, which leads to edge effect interference with the electric field distribution on the electrode surface. Existing researches have confirmed that the SF6 partial discharge starting field strength in non-uniform electric field is affected by the curvature and area effect of the electrode surface, and the design parameters obtained by the traditional design method of simply relying on the electrode surface electric field control cannot provide accurate data support for the inner shielding structure design of UHV DC SF6 gas insulated bushing. SUMMARY

[0005] The technical problem to be solved by the present application is how to better realize the structure parameter design of the UHV bushing inner shielding electrode, so as to provide accurate data support for the inner shielding structure design of the UHV DC SF6 gas insulated bushing.

[0006] The present application solves the above technical problems by the following technical means: a UHV bushing inner shielding electrode structure optimization method, comprising the following steps:

[0007] With different parameter combinations of the inner shielding electrode structure of the ultra-high voltage bushing as input, the BP neural network trained is used for calculation to obtain field strength values of the ultra-high voltage bushing corresponding to the parameter combinations; wherein each parameter combination contains a radius of the floating potential electrode, a radius of the grounding electrode, a length of the floating potential electrode and a length of the grounding electrode; the field strength values include a maximum value of the surface field strength of the central conductor, a minimum value of the surface field strength of the inner side of the floating potential electrode, a maximum value of the surface field strength of the outer side of the floating potential electrode, a minimum value of the surface field strength of the inner side of the grounding electrode, a maximum value of the synthetic surface field strength of the outer side of the hollow composite insulator, a maximum value of the axial field strength of the outer side of the hollow composite insulator, an average value of the synthetic surface field strength of the outer side of the hollow composite insulator and an axial field strength of the outer side of the hollow composite insulator;

[0008] With the parameter combinations as input, the BP neural network trained is used for calculation to obtain the field strength values of the ultra-high voltage bushing corresponding to the parameter combinations;

[0009] Preferably, the method for optimizing the inner shielding electrode structure of the ultra-high voltage bushing further comprises the following processes:

[0010] Random points are taken in the value range of the radius of the floating potential electrode to form a first data set;

[0011] Random points are taken in the value range of the radius of the grounding electrode to form a second data set;

[0012] Random points are taken in the value range of the length of the floating potential electrode to form a third data set;

[0013] Random points are taken in the value range of the length of the grounding electrode to form a fourth data set;

[0014] The parameter combinations are constructed through the first data set, the second data set, the third data set and the fourth data set; wherein the parameter combinations contain one data in the first data set, one data in the second data set, one data in the third data set and one data in the fourth data set;

[0015] With the parameter combinations as input, the BP neural network trained is used for calculation to obtain the field strength values of the ultra-high voltage bushing corresponding to the parameter combinations;

[0016] The processes of constructing the parameter combinations and calculating the field strength values are cycled until all the data in the first data set, the second data set, the third data set and the fourth data set are traversed.

[0017] Preferably, the process of training the BP neural network comprises:

[0018] Using different parameter combinations of the shielding electrode structure inside the UHV bushing as input, and taking the maximum value of the surface electric field of the center conductor, the minimum value of the surface electric field of the inner side of the floating potential electrode, the maximum value of the surface electric field of the outer side of the floating potential electrode, the minimum value of the surface electric field of the inner side of the grounding electrode, the maximum value of the combined surface electric field of the outer side of the hollow composite insulator, the maximum value of the axial surface electric field of the outer side of the hollow composite insulator, the average value of the combined surface electric field of the outer side of the hollow composite insulator, or the axial surface electric field of the outer side of the hollow composite insulator as output, the BP neural network is trained to obtain the trained BP neural network.

[0019] The prediction error function of the BP neural network adopts the mean absolute percentage error, and the number of hidden layer neurons of the BP neural network is ≤40.

[0020] Preferably, the objective function includes: the maximum electric field strength f1(x) inside the UHV bushing, and the arithmetic mean of the electric field strength variation amplitudes within the two SF6 air gaps. 2-1 (x), the maximum electric field strength E on the surface of the central conductor max1 Minimum electric field strength E on the inner surface of the floating potential electrode min1 The maximum electric field strength E on the outer surface of the floating potential electrode max2 and the minimum electric field strength E on the inner surface of the grounding electrode min2 standard deviation f 2-3 (x), the maximum value of the combined electric field on the outer surface of the hollow composite insulator, f 3-1 (x) or the maximum tangential electric field strength f on the outer surface of the hollow composite insulator 3-2 (x):

[0021] Among them, the maximum electric field strength f1(x) inside the UHV bushing is:

[0022] f1(x)=max(E max1 E max2 )

[0023] Where x is the optimization variable;

[0024] The arithmetic mean f of the field strength variation within the two SF6 air gaps 2-1 (x):

[0025]

[0026] Maximum electric field strength E at the surface of the central conductor max1 Minimum electric field strength E on the inner surface of the floating potential electrode min1 The maximum electric field strength E on the outer surface of the floating potential electrode max2 and the minimum electric field strength E on the inner surface of the grounding electrode min2 standard deviation f 2-3 (x):

[0027]

[0028] In the formula: E mean For E max1 E min1 E max2 and E min2 The arithmetic mean.

[0029] Preferably, a non-dominated sorting genetic algorithm II is used to optimize the combination of objective functions formed by the field strength values. The non-dominated sorting genetic algorithm II adds a constraint deviation operator g to the non-dominated sorting genetic algorithm: the constraint conditions are written as g. i ≤0, i = 1, 2, ..., n, g >0 i It characterizes the degree to which the current solution deviates from the constraints;

[0030] The non-dominated sorting genetic algorithm II adds a constraint bias operator to the sorting operation based on the non-dominated sorting genetic algorithm: the constraint bias operator g of the optimization variable x and the objective function f(x) are respectively added. i (x) and g i After normalizing (f(x)), the array is sorted twice, specifically including:

[0031] a) For two solutions (x1, f(x1)) and (x2, f(x2)) in the array: when ∑g i,nor (x1), ∑g i,nor When (x2)≤0, perform a non-dominated sort on f(x1) and f(x2); when ∑g i,nor (x1)≤0,∑g i,nor When (x2)>0, the order value of (x1,f(x1)) comes first, and vice versa; when ∑g i,nor (x1),∑g i,nor When (x2)>0, if ∑g i,nor (x1)<∑g i,nor If (x2) is an integer, then the order value of (x1, f(x1)) comes first, and vice versa.

[0032] b) For two solutions (x1, f(x1)) and (x2, f(x2)) in the same sequence after sorting a): when ∑g i,nor (f(x1)), ∑g i,nor When (f(x2))≤0, sort (x1,f(x1)) and (x2,f(x2)) by crowding degree without changing their order values; when ∑g i,nor (f(x1))≤0,∑g i,norWhen (f(x2))>0, the order value of (x1, f(x1)) comes first, and vice versa; when ∑g i,nor (f(x1)), ∑g i,nor When (f(x2))>0, if ∑g i,nor (f(x1))<∑g i,nor If (f(x2)), then the order value of (x1, f(x1)) comes first, and vice versa.

[0033] Preferably, the combination of objective functions includes f1(x), f 2-1 (x) and f 2-3 (x) one of them, and f 3-1 (x) and f 3-2 (x) One of them.

[0034] Preferably, when optimizing the objective function for an 1100kV DC SF6 gas-insulated wall bushing:

[0035] The optimization variable x is:

[0036] x=(R1,R2,L1,L2) T

[0037] Where R1 is the radius of the floating potential electrode, R2 is the radius of the grounding electrode, L1 is the length of the floating potential electrode, and L2 is the length of the grounding electrode;

[0038] The range of values ​​for the optimization variable x is:

[0039]

[0040] The constraints for optimizing variable x are:

[0041]

[0042] Each objective function is no greater than its own average value;

[0043] After optimization, while meeting the design margin requirements, the optimization variable x corresponding to the combination of objective functions that minimizes the length of the floating potential electrode L1 and / or the length of the ground electrode L2 is selected as the final optimization result.

[0044] This invention also provides a system for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing, used to implement the above-mentioned method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing. The system includes:

[0045] The data generation module is used to calculate the electric field value of the UHV bushing corresponding to different parameter combinations of the shielding electrode structure inside the UHV bushing through a trained BP neural network. Each parameter combination includes the radius of the floating potential electrode, the radius of the grounding electrode, the length of the floating potential electrode, and the length of the grounding electrode. The electric field value includes the maximum electric field strength on the surface of the central conductor, the minimum electric field strength on the inner surface of the floating potential electrode, the maximum electric field strength on the outer surface of the floating potential electrode, the minimum electric field strength on the inner surface of the grounding electrode, the maximum combined electric field strength on the outer surface of the hollow composite insulator, the maximum axial electric field strength on the outer surface of the hollow composite insulator, the average combined electric field strength on the outer surface of the hollow composite insulator, and the axial electric field strength on the outer surface of the hollow composite insulator.

[0046] Optimization module: Used to construct optimization variables with the parameter combination, with the optimization objectives being to minimize the maximum electric field strength in the internal and external electric fields of the UHV bushing and the overall uniformity of the electric field distribution of the UHV bushing. The objective function formed by the electric field strength values ​​is optimized to obtain the parameter combination that satisfies the optimization objective, thereby realizing the optimization of the shielding electrode structure inside the UHV bushing.

[0047] The present invention also provides an electronic device, comprising:

[0048] One or more processors;

[0049] A storage device on which one or more programs are stored;

[0050] When the one or more programs are executed by the one or more processors, the one or more processors implement the ultra-high voltage bushing internal shielding electrode structure optimization method of the present invention as described above.

[0051] The present invention also provides a storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the method for optimizing the shielding electrode structure inside the ultra-high voltage bushing as described above.

[0052] The advantages of this invention are as follows: In the optimization method for the internal shielding electrode structure of ultra-high voltage bushings, a trained BP neural network can obtain a sufficient number of mapping relationships between different parameter combinations and field strength values. The BP neural network can replace numerical calculations to solve the objective function when executing the optimization algorithm, significantly improving the execution efficiency of the algorithm while ensuring computational accuracy. This effectively solves the optimization problem of complex internal shielding structures in bushings. This invention focuses on ultra-high voltage SF6 gas-insulated bushings, systematically analyzing the influence mechanism of edge effects and structural parameters on the surface electric field of the shielding electrode. It proposes a multi-objective optimization method for the internal shielding electrode size from two aspects: minimizing the maximum field strength in the internal and external electric fields of the bushing and improving the overall uniformity of the electric field distribution. This achieves the optimization of the structural parameters of the internal shielding electrode in ultra-high voltage bushings. The resulting structural optimization results can provide theoretical guidance and key data support for the design of the internal shielding structure of ultra-high voltage DC SF6 gas-insulated bushings. Attached Figure Description

[0053] Figure 1 This is an internal parameter identifier for the 1100kV DC SF6 gas-insulated wall bushing in this embodiment of the invention;

[0054] Figure 2 This is a curve showing the change of MAPE value with the number of neurons in the hidden layer in an embodiment of the present invention;

[0055] Figure 3(a) is a schematic diagram of a constraint processing and sorting method in an embodiment of the present invention; Figure 3(b) is a schematic diagram of another constraint processing and sorting method in an embodiment of the present invention;

[0056] Figure 4 This is a flowchart illustrating the process of solving constrained optimization problems using NSGA-II in an embodiment of the present invention.

[0057] Figure 5(a) shows the optimization results (f1-f) in the embodiment of the present invention. 2-1 —f 3-1 Electrode size distribution (“Δ” symbol represents R1, L1, The symbols represent R2 and L2 diagrams; Figure 5(b) shows the optimization results (f1-f) in the embodiment of the present invention. 2-3 —f 3-1 Electrode size distribution (“Δ” symbol represents R1, L1, The symbols represent R2 and L2 diagrams; Figure 5(c) shows the optimization results (f1-f) in the embodiment of the present invention. 2-1 —f 3-2 Electrode size distribution (“Δ” symbol represents R1, L1, The symbols represent R2 and L2 diagrams; Figure 5(d) shows the optimization results (f1-f) in the embodiment of the present invention. 2-3 —f 3-2Electrode size distribution (“Δ” symbol represents R1, L1, The symbols represent R2 and L2 diagrams.

[0058] In the diagram, 1-center conductor, 2-floating potential electrode, 3-grounding electrode, 4-gas-insulated pipe, 5-flange. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] This embodiment takes an 1100kV DC SF6 gas-insulated through-wall bushing as an example. To facilitate model reconstruction during calculation and optimization, the shielding structure inside the UHV bushing is parametrically modeled. The parameter identifiers inside the bushing are as follows: Figure 1 As shown in the diagram. Here, R0 is the radius of the center conductor, R1 is the radius of the floating potential electrode, R2 is the radius of the grounding electrode, L1 is the length of the floating potential electrode, L2 is the length of the grounding electrode, L3 is the length of the floating potential electrode in the gas-insulated pipe, and r1 to r7 are the radii of each electrode end. For a bushing of a certain voltage level, R0 and L3 are fixed values, and r1 to r7 can also be directly set. Therefore, optimization mainly considers optimizing the radius R1 of the floating potential electrode, the radius R2 of the grounding electrode, the length L1 of the floating potential electrode, and the length L2 of the grounding electrode.

[0061] A BP (Back Propagation) neural network was used to perform nonlinear fitting on the size of the inner shielding electrode (i.e., the input data below) and the field strength values ​​at some key locations in the electric field inside and outside the wall bushing (i.e., the output data below).

[0062] The prediction error function of the BP neural network uses the Mean Absolute Percent Error (MAPE):

[0063] This embodiment uses a set of samples covering the entire internal space of the sleeve to train the BP neural network:

[0064] The input data for the BP neural network includes the radius R1 of the floating potential electrode, the radius R2 of the grounding electrode, the length L1 of the floating potential electrode, and the length L2 of the grounding electrode. Within the range of values ​​for each input data point, points are selected at intervals, as follows: Floating potential electrode radius R1 = 11, 13.5, ..., 36 cm; Grounding electrode radius R2 = 12, 14.5, ..., 37 cm; Floating potential electrode length L1 = 1.75, 2.25, ..., 10.75 m; Grounding electrode length L2 = 1, 1.5, ..., 10 m; Constraints: R2 - R1 ≥ 10, L1 - L2 ≥ 0.75. Using this input data, the output data (i.e., various field strength data) corresponding to each set of input data can be pre-calculated using formulas. These input and output data can then be used to train the BP neural network model.

[0065] The output of the BP neural network includes: the maximum electric field strength E on the surface of the central conductor. max1 Minimum electric field strength E on the inner surface of the levitating potential electrode min1 and the maximum electric field strength E on the outer surface of the levitating potential electrode max2 Minimum electric field strength E on the inner surface of the grounding electrode min2 The maximum combined electric field strength E on the outer surface of the hollow composite insulator t,max The maximum axial electric field strength E on the outer surface of the hollow composite insulator z,max The average value of the combined electric field on the outer surface of the hollow composite insulator, E t,mean Axial electric field strength E on the outer surface of hollow composite insulator z,mean The number of samples, N = 12540. Basic information about the sample output data is shown in Table 1.

[0066] Table 1

[0067]

[0068]

[0069] The backpropagation neural network was configured with 10 hidden layer neurons (HN). The samples were grouped according to the output object for training. The average MAPE values ​​after three training iterations are shown in Table 2. Grouping the output objects improved the training performance compared to before grouping. For objects with high dispersion, such as E... max2 E min2 The MAPE value (see Table 1) decreased significantly because group training reduced the nonlinearity between input and output data.

[0070] Table 2

[0071]

[0072] The MAPE value changes with the number of hidden layer neurons (HN) as follows: Figure 2 Show. From Figure 2 As can be seen, the prediction error of the BP neural network generally decreases with the increase of the number of hidden layer neurons HN, and no obvious "overfitting" phenomenon is observed in the range of HN≤40. The BP neural network with the smallest MAPE value in this training (represented by the "×" symbol in the figure) is selected to replace numerical calculation to solve the objective function.

[0073] It should be noted that in Table 2 above, group "E" max1 E min1 E max2 E min2 This refers to a set of input data being processed by a BP neural network, which simultaneously outputs "E". max1 E min1 E max2 E min2 "These four field strength values; group "E" max1 E min1 / E max2 E min2 This refers to a training method where, after a set of input data is processed by a BP neural network, it simultaneously outputs "E". max1 E min1 "These two field strength values, another training method is to simultaneously output 'E'." max2 E min2 "These two field strength values; group "E" max1 / E min1 / E max2 / E min2 This refers to the fact that after a set of input data is processed by a BP neural network, the BP neural network outputs "E" respectively. max1 E min1 E max2 or E min2 ", outputs only one field strength value each time; group "E" t,max E z,max / E t,mean E z,mean This refers to a training method where, after a set of input data is processed by a BP neural network, it simultaneously outputs "E". t,max E z,max "These two field strength values, another training method is to simultaneously output 'E'." t,mean E z,mean "These two field strength values; group "E" t,max / E z,max / E t,mean / E z,mean This refers to the fact that after a set of input data is processed by a BP neural network, the BP neural network outputs "E" respectively. t,max Ez,max E t,mean or E z,mean It outputs only one field strength value each time.

[0074] As can be seen from the results in Table 2, after a set of input data is processed by a BP neural network, the BP neural network outputs a field strength value. Under this mapping relationship, the training effect of the BP neural network is the best.

[0075] Once training is complete, random points are selected within the range of each input data value. The trained BP neural network can then be used to calculate each field strength value. This expands the mapping relationship between input and output data, providing a sufficient data foundation for the subsequent optimization process and improving the accuracy and effectiveness of optimization.

[0076] This embodiment uses the Nondominated Sorting Genetic Algorithm II (NSGA-II) to optimize the objective function. Currently, the commonly used NSGA-II code (in Matlab environment) is written by Seshadri, whose GA algorithm uses real-number encoding, simulated binary crossover, and polynomial mutation. Because the original program cannot handle constrained optimization problems, the code is modified as follows:

[0077] (1) Calculation of the constraint deviation operator g: Write the constraint condition as g i ≤0, i = 1, 2, ..., n, clearly g is greater than 0. i It characterizes the degree to which the current solution deviates from the constraints;

[0078] (2) Incorporating constraint bias operators into the sorting operation: Based on method 1 (see Figure 3(a) for a schematic diagram, where the numbers in Figure 3(a) represent ordinal values, and the numbers on either side of the short dash represent the ordinal values ​​after two sorting operations), for the optimization problem of this invention, the constraint bias operators gi(x) and gi(f(x)) of the optimization variable x and the objective function f(x) are normalized respectively, and then the array is sorted twice (method 2, see Figure 3(b) for a schematic diagram). The specific process is as follows:

[0079] a) For two solutions (x1, f(x1)) and (x2, f(x2)) in the array: when ∑g i,nor (x1), ∑g i,nor When (x2)≤0, perform a non-dominated sort on f(x1) and f(x2); when ∑g i,nor (x1)≤0,∑g i,nor When (x2)>0, the order value of (x1, f(x1)) comes first, and vice versa; when ∑g i,nor (x1), Σgi,nor When (x2)>0, if Σg i,nor (x1)<Σg i,nor If (x2) is an integer, then the order value of (x1, f(x1)) is first, and vice versa.

[0080] b) For two solutions (x1, f(x1)) and (x2, f(x2)) in the same sequence after sorting a): when Σg i,nor (f(x1)), Σg i,nor When (f(x2))≤0, sort (x1, f(x1)) and (x2, f(x2)) according to their crowding level (without changing their order); when ∑g i,nor (f(x1))≤0,∑g i,nor When (f(x2))>0, the order value of (x1, f(x1)) comes first, and vice versa; when ∑g i,nor (f(x1)), ∑g i,nor When (f(x2))>0, if ∑g i,nor (f(x1))<∑g i,nor If (f(x2)), then the order value of (x1, f(x1)) comes first, and vice versa.

[0081] Combining the trained neural network group, the complete multi-objective optimization process is as follows: Figure 4 As shown.

[0082] The following section proposes a multi-objective optimization method for the size of the inner shielding electrode, focusing on minimizing the maximum electric field strength inside and outside the through-wall bushing, and improving the overall uniformity of the electric field distribution. Optimization variables and their value ranges are as follows:

[0083] The optimization variable x is:

[0084] x=(R1,R2,L1,L2) T

[0085] The range of values ​​for the optimization variable x is:

[0086]

[0087] Objective function 1 is the maximum electric field strength f1(x) inside the wall bushing (abbreviated as f1):

[0088] f1(x)=max(E max1 E max2 )

[0089] Objective function 2 is the evaluation function for the uniformity of the electric field distribution inside the bushing. Considering that the floating potential electrode separates the central conductor and the flange into two independent SF6 gas gaps, the objective function will use the following three functions to characterize the field strength distribution:

[0090] a) The arithmetic mean of the changes in field strength within the two SF6 air gaps, f 2-1 (x)(abbreviated as f) 2-1 ):

[0091]

[0092] b) The difference f between the amplitudes of the field strength changes within the two SF6 air gaps 2-2 (x)(abbreviated as f) 2-2 ):

[0093] f 2-2 (x)=|E max1 -E min1 -E max2 +E min2 |

[0094] c)E max1 E min1 E max2 E min2 standard deviation f 2-3 (x)(abbreviated as f) 2-3 ):

[0095]

[0096] In the formula: E mean For E max1 E min1 E max2 E min2 The arithmetic mean.

[0097] f 2-1 (x) and f 2-3 The evaluation results of (x) are similar. Under the premise that the grounding electrode radius R2 is fixed, the former has a smaller variation range. The floating potential electrode radius R1 corresponding to the minimum value of both is 21~27cm. At this time, the uniformity of radial field strength distribution is significantly better than in other cases. 2-2 (x) When R1 reaches its minimum value, it is 15 cm. In this case, the transition between the field strength distributions of the two SF6 air gaps is the smoothest, but the corresponding levitation potential is too high. Therefore, objective function 2 will use f 2-1 (x) or f 2-3 (x).

[0098] The maximum combined electric field strength f on the surface of the insulator is calculated. 3-1 (x)(abbreviated as f) 3-1 (or the maximum tangential electric field strength f on the insulator surface) 3-2 (x)(abbreviated as f) 3-2 As objective function 3:

[0099] f3-1 (x)=E t,max

[0100] f 3-2 (x)=E z,max

[0101] The constraints on variable x are:

[0102]

[0103] To avoid extreme cases, the objective function is limited to within the respective average values:

[0104]

[0105] Based on objective function 1 and objective function 3, the design margins of the field strength inside and outside the wall bushing are defined, as shown in Table 3.

[0106] Table 3

[0107]

[0108] Table 3 shows the results after rounding the preferred electrode dimensions (radius to 0.5 cm and length to 1 cm). The corresponding objective function values ​​are shown in [Table 3]. Figures 5(a)-5(d) .

[0109] Table 4

[0110]

[0111]

[0112] Table 4 lists the combinations of objective functions, including f1 and f2. 2-1 with f 2-3 One of them, and f 3-1 with f 3-2 One of them.

[0113] As shown in Table 4, the design margin of the optimal results under the four combinations of objective functions is around 40%. However, compared with the initial design, the dimensions of each electrode (especially the electrode length) are significantly increased, which will obviously reduce the mechanical properties of the sleeve and increase the manufacturing cost. The design with the smallest overall size, f1-f, was selected. 2-1 ―f 3-2 The optimal results were used for the design of 1100kV DC SF6 gas-insulated wall bushings.

[0114] In the above-mentioned scheme of this invention, a parameterized model system for the internal shielding structure of the bushing is constructed. Modular variable design enables flexible adjustment of geometric features, allowing for rapid model reconstruction during simulation calculations and structural optimization, significantly improving design iteration efficiency. This invention proposes using a hybrid BP neural network and NSGA-II algorithm for optimizing the dimensions of the internal shielding structure of 1100kV DC SF6 gas-insulated through-wall bushings. The BP neural network can replace numerical calculations to solve the objective function when executing the optimization algorithm, significantly improving the algorithm's execution efficiency while ensuring computational accuracy, effectively solving the problem of balancing computational efficiency and accuracy in the optimization of complex internal shielding structures. This invention proposes a group training method based on data characteristics. For highly dispersed training samples, classification enables refined training, significantly reducing the mean absolute percentage error (MAPE) and significantly weakening the nonlinearity of the input-output data. This strategy has dynamic expansion capabilities, allowing new output objects (such as floating potential percentage, radial field strength on the surface of hollow composite insulators, etc.) to be independently trained and seamlessly integrated into the neural network group, avoiding the resource consumption of global retraining in traditional methods. This invention limits the range of values ​​for the objective function based on the initial training sample data, effectively avoiding extreme cases and improving training efficiency. Based on the characteristics of the initial training sample data, a constraint model for the objective function is constructed. By adjusting boundary conditions, invalid calculations under extreme conditions are effectively avoided, thus improving the training convergence speed. This mechanism effectively shortens computation time while ensuring the reliability of the optimization results.

[0115] This invention, based on finite element method (FEM) electric field simulation, focuses on the research of UHVDC SF6 gas-insulated bushings. It systematically analyzes the influence mechanism of edge effects and structural parameters on the surface electric field of the shielding electrode. A multi-objective optimization method for the size of the inner shielding electrode is proposed, aiming to minimize the maximum field strength inside and outside the bushing and improve the overall uniformity of the electric field distribution. This invention utilizes a BP neural network combined with the NSGA-II hybrid algorithm for structural parameter optimization. The resulting structural optimization can provide theoretical guidance and key data support for the design of the inner shielding structure of UHVDC SF6 gas-insulated bushings.

[0116] Furthermore, embodiments of the present invention also provide a system for implementing the above-described method for optimizing the structure of the shielding electrode inside the ultra-high voltage bushing, the system comprising:

[0117] The data generation module is used to calculate the electric field value of the UHV bushing corresponding to different parameter combinations of the shielding electrode structure inside the UHV bushing through a trained BP neural network. Each parameter combination includes the radius of the floating potential electrode, the radius of the grounding electrode, the length of the floating potential electrode, and the length of the grounding electrode. The electric field value includes the maximum electric field strength on the surface of the central conductor, the minimum electric field strength on the inner surface of the floating potential electrode, the maximum electric field strength on the outer surface of the floating potential electrode, the minimum electric field strength on the inner surface of the grounding electrode, the maximum combined electric field strength on the outer surface of the hollow composite insulator, the maximum axial electric field strength on the outer surface of the hollow composite insulator, the average combined electric field strength on the outer surface of the hollow composite insulator, and the axial electric field strength on the outer surface of the hollow composite insulator.

[0118] Optimization module: Used to construct optimization variables with the parameter combination, with the optimization objectives being to minimize the maximum electric field strength in the internal and external electric fields of the UHV bushing and the overall uniformity of the electric field distribution of the UHV bushing. The objective function formed by the electric field strength values ​​is optimized to obtain the parameter combination that satisfies the optimization objective, thereby realizing the optimization of the shielding electrode structure inside the UHV bushing.

[0119] The specific technical solutions involved in this system are the same as those in the above-mentioned optimization method for the shielding electrode structure inside the UHV bushing, and will not be repeated here.

[0120] The embodiments of the present invention also provide corresponding electronic devices and computer-readable storage media for implementing the solutions provided in the embodiments of the present invention.

[0121] The electronic device includes a storage device and one or more processors. The storage device stores instructions or code, and the processors execute the instructions or code to enable the device to perform the ultra-high voltage bushing internal shielding electrode structure optimization method according to any embodiment of this application.

[0122] The storage medium stores a computer program, which, when executed by a processor, implements the method for optimizing the shielding electrode structure inside the ultra-high voltage bushing as described in any embodiment of this application.

[0123] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing, characterized in that, Includes the following steps: Using different parameter combinations of the shielding electrode structure inside the UHV bushing as input, a trained BP neural network is used to calculate the electric field strength value of the UHV bushing corresponding to each parameter combination. Each parameter combination includes the radius of the floating potential electrode, the radius of the grounding electrode, the length of the floating potential electrode, and the length of the grounding electrode. The electric field strength value includes the maximum electric field strength on the surface of the central conductor, the minimum electric field strength on the inner surface of the floating potential electrode, the maximum electric field strength on the outer surface of the floating potential electrode, the minimum electric field strength on the inner surface of the grounding electrode, the maximum combined electric field strength on the outer surface of the hollow composite insulator, the maximum axial electric field strength on the outer surface of the hollow composite insulator, the average combined electric field strength on the outer surface of the hollow composite insulator, and the axial electric field strength on the outer surface of the hollow composite insulator. The optimization variables are constructed using the parameter combination. The optimization objectives are to minimize the maximum electric field strength in the internal and external electric fields of the UHV bushing and the overall uniformity of the electric field distribution of the UHV bushing. The objective function formed by the electric field strength values ​​is optimized to obtain the parameter combination that satisfies the optimization objective, thereby realizing the optimization of the shielding electrode structure inside the UHV bushing.

2. The method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing according to claim 1, characterized in that, It also includes the following processes: Within the range of values ​​for the radius of the floating potential electrode, random points are selected to form the first dataset; Within the range of the grounding electrode radius, random points are selected to form a second dataset; Within the range of values ​​for the length of the suspended potential electrode, random points are selected to form a third dataset; Within the range of the grounding electrode length, random points are selected to form the fourth dataset; A parameter combination is constructed using the first dataset, the second dataset, the third dataset, and the fourth dataset; wherein, the parameter combination includes one data point from the first dataset, one data point from the second dataset, one data point from the third dataset, and one data point from the fourth dataset; Using the parameter combination as input, the trained BP neural network is used to calculate and obtain the field strength values ​​of the UHV bushing corresponding to the parameter combination. The process of constructing parameter combinations and calculating each field strength value is repeated until all data in the first, second, third, and fourth datasets has been traversed.

3. The method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing according to claim 1 or 2, characterized in that, The process of training a BP neural network includes: Using different parameter combinations of the shielding electrode structure inside the UHV bushing as input, and taking the maximum value of the surface electric field of the center conductor, the minimum value of the surface electric field of the inner side of the floating potential electrode, the maximum value of the surface electric field of the outer side of the floating potential electrode, the minimum value of the surface electric field of the inner side of the grounding electrode, the maximum value of the combined surface electric field of the outer side of the hollow composite insulator, the maximum value of the axial surface electric field of the outer side of the hollow composite insulator, the average value of the combined surface electric field of the outer side of the hollow composite insulator, or the axial surface electric field of the outer side of the hollow composite insulator as output, the BP neural network is trained to obtain the trained BP neural network. The prediction error function of the BP neural network adopts the mean absolute percentage error, and the number of hidden layer neurons of the BP neural network is ≤40.

4. The method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing according to claim 1, characterized in that, The objective function includes: the maximum electric field strength f1(x) inside the UHV bushing, and the arithmetic mean of the electric field strength variation amplitudes within the two SF6 air gaps. 2-1 (x), the maximum electric field strength E on the surface of the central conductor max1 Minimum electric field strength E on the inner surface of the floating potential electrode min1 The maximum electric field strength E on the outer surface of the floating potential electrode max2 and the minimum electric field strength E on the inner surface of the grounding electrode min2 standard deviation f 2-3 (x), the maximum value of the combined electric field on the outer surface of the hollow composite insulator, f 3-1 (x) or the maximum tangential electric field strength f on the outer surface of the hollow composite insulator 3-2 (x): Among them, the maximum electric field strength f1(x) inside the UHV bushing is: f1(x) = max(E max1 ,AND max2 ) Where x is the optimization variable; The arithmetic mean f of the field strength variation within the two SF6 air gaps 2-1 (x): Maximum electric field strength E at the surface of the central conductor max1 Minimum electric field strength E on the inner surface of the floating potential electrode min1 The maximum electric field strength E on the outer surface of the floating potential electrode max2 and the minimum electric field strength E on the inner surface of the grounding electrode min2 standard deviation f 2-3 (x): In the formula: E mean For E max1 E min1 E max2 and E min2 The arithmetic mean.

5. The method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing according to claim 4, characterized in that, The non-dominated sorting genetic algorithm II is used to optimize the combination of objective functions formed by the field strength values. The non-dominated sorting genetic algorithm II adds a constraint deviation operator g to the non-dominated sorting genetic algorithm: the constraint conditions are written as g. i ≤0, i = 1, 2, ..., n, g >0 i It characterizes the degree to which the current solution deviates from the constraints; The non-dominated sorting genetic algorithm II adds a constraint bias operator to the sorting operation based on the non-dominated sorting genetic algorithm: the constraint bias operator g of the optimization variable x and the objective function f(x) are respectively added. i (x) and g i After normalizing f(x), the array is sorted twice, specifically including: a) For two solutions (x1, f(x1)) and (x2, f(x2)) in the array: when ∑g i,nor (x1), ∑g i,nor When (x2)≤0, perform a non-dominated sort on f(x1) and f(x2); when ∑g i,nor (x1)≤0,∑g i,nor When (x2)>0, the order value of (x1, f(x1)) comes first, and vice versa; when ∑g i,nor (x1), ∑g i,nor When (x2)>0, if ∑g i,nor (x1)<∑g i,nor If (x2) is an integer, then the order value of (x1, f(x1)) comes first, and vice versa. b) For two solutions (x1, f(x1)) and (x2, f(x2)) in the same sequence after sorting a): when ∑g i,nor (f(x1)), ∑g i,nor When (f(x2))≤0, sort (x1, f(x1)) and (x2, f(x2)) by crowding degree without changing their order values; when ∑g i,nor (f(x1))≤0,∑g i,nor When (f(x2))>0, the order value of (x1,f(x1)) comes first, and vice versa; when ∑g i,nor (f(x1)),∑g i,nor When (f(x2))>0, if ∑g i,nor (f(x1))<∑g i,nor If (f(x2)), then the order value of (x1,f(x1)) comes first, and vice versa.

6. The method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing according to claim 5, characterized in that, The combination of objective functions includes f1(x), f 2-1 (x) and f 2-3 (x) one of them, and f 3-1 (x) and f 3-2 (x) One of them.

7. The method for optimizing the structure of the shielding electrode inside an ultra-high voltage bushing according to claim 6, characterized in that, For 1100kV DC SF6 gas-insulated wall bushings, when optimizing the objective function: The optimization variable x is: x=(R1,R2,L1,L2) T Where R1 is the radius of the floating potential electrode, R2 is the radius of the grounding electrode, L1 is the length of the floating potential electrode, and L2 is the length of the grounding electrode; The range of values ​​for the optimization variable x is: The constraints for optimizing variable x are: Each objective function is no greater than its own average value; After optimization, while meeting the design margin requirements, the optimization variable x corresponding to the combination of objective functions that minimizes the length of the floating potential electrode L1 and / or the length of the ground electrode L2 is selected as the final optimization result.

8. A system for optimizing the structure of an internal shielding electrode in an ultra-high voltage bushing, characterized in that, include: The data generation module is used to calculate the electric field value of the UHV bushing corresponding to different parameter combinations of the shielding electrode structure inside the UHV bushing through a trained BP neural network. Each parameter combination includes the radius of the floating potential electrode, the radius of the grounding electrode, the length of the floating potential electrode, and the length of the grounding electrode. The electric field value includes the maximum electric field strength on the surface of the central conductor, the minimum electric field strength on the inner surface of the floating potential electrode, the maximum electric field strength on the outer surface of the floating potential electrode, the minimum electric field strength on the inner surface of the grounding electrode, the maximum combined electric field strength on the outer surface of the hollow composite insulator, the maximum axial electric field strength on the outer surface of the hollow composite insulator, the average combined electric field strength on the outer surface of the hollow composite insulator, and the axial electric field strength on the outer surface of the hollow composite insulator. Optimization module: Used to construct optimization variables with the parameter combination, with the optimization objectives being to minimize the maximum electric field strength in the internal and external electric fields of the UHV bushing and the overall uniformity of the electric field distribution of the UHV bushing. The objective function formed by the electric field strength values ​​is optimized to obtain the parameter combination that satisfies the optimization objective, thereby realizing the optimization of the shielding electrode structure inside the UHV bushing.

9. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for optimizing the shielding electrode structure inside the ultra-high voltage bushing as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, It stores a computer program, wherein when the computer program is executed by a processor, it implements the method for optimizing the shielding electrode structure inside the ultra-high voltage bushing as described in any one of claims 1-7.

Images