Particle trap for GIS / GIL, structure optimization method of particle trap and computer device

Abstract

The invention belongs to the technical field of power equipment structure optimization, and particularly relates to a particle trap for GIS / GIL, a structure optimization method of the particle trap and a computer device. The method comprises the following steps: firstly, establishing a particle trap simulation model and a particle motion simulation model, obtaining the capture rate of particle traps to particles under different structure parameters of the particle traps by using the particle trap simulation model and the particle motion simulation model, and taking the capture rate as a data set; then training the constructed machine learning model by using the data set, wherein the trained model is an equivalent calculation model for calculating the capture rate according to the structure parameters; and finally, optimizing the structure parameters by taking the capture rate calculated by using the equivalent calculation model as a target to maximize, so as to obtain the optimal value of the structure parameters. Compared with a mode of directly utilizing a simulation model to optimize the structural parameters, the method has the advantages that the calculation times are reduced, the optimization efficiency is improved, and the capture rate of the particle traps obtained according to the optimization mode is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment structure optimization, and particularly relates to a particle trap for GIS / GIL, a method for optimizing its structure, and a computer device. Background Art

[0002] Gas Insulated Transmission Line (GIL) and Gas Insulated Switchgear (GIS) are widely used in large-capacity power transmission systems due to their advantages such as compact structure, reliable operation, and convenient installation. However, metal particles are inevitably generated during their production, transportation, installation, and operation. After being charged, the metal particles move freely back and forth between the central conductor, metal shell, and insulator under the action of the electric field force, causing local distortion of the internal electric field of the equipment and affecting the insulation performance of the equipment. Every year, there are more and more accidents where the transmission line cannot operate normally due to surface discharge of the insulator or discharge between electrode gaps caused by particles. The metal particles in DC GIL / GIS are more active, have a wider movement range, and have a greater impact on the insulation performance of the equipment. Research shows that installing a particle trap is an effective way to reduce discharge accidents caused by particles. The low electric field region formed inside the particle trap can capture charged and moving particles.

[0003] The performance of the lifting type particle trap varies under different working conditions, and the structure of the lifting type particle trap has a great influence on its capture effect. Its structure includes the total length of the trap, grid width, etc. For different types of equipment, the structural parameters of the particle trap determine its capture rate. Optimizing the structural parameters of the particle trap is crucial. Currently, there is no unified specification for the design of the particle trap structure in the prior art. The Chinese invention patent with the authorization announcement number CN115864272B and the application publication date of March 28, 2023, discloses a layout method for the particle trap structure in the GIS bus chamber. The main process of this method is as follows: First, establish a particle trap simulation model, calculate the target particle capture rate of the particle trap structure according to the movement characteristics of the target particles within a set time, and then change the distances between the wedge-shaped trap and the first extraction hole trap and the second extraction hole trap to obtain a new capture rate; compare the capture rates of the target particles at different distances between the wedge-shaped trap and the first extraction hole trap and the second extraction hole trap to obtain the optimal layout scheme for the particle trap structure. The optimization of this method for the layout scheme is based on the particle trap simulation model, that is, each iterative calculation requires using the particle trap simulation model for an iterative calculation, which is a way of optimizing while simulating. The computational complexity of the simulation model itself is very large, which undoubtedly makes the structure optimization efficiency low. Summary of the Invention

[0004] The object of the present invention is to provide a particle trap for GIS / GIL, its structure optimization method and a computer device, so as to solve the problem of large computational amount caused by directly optimizing the structure of the particle trap based on the particle trap simulation model in the prior art.

[0005] To solve the above technical problems, the present invention provides a structure optimization method for a particle trap for GIS / GIL, including the following steps:

[0006] 1) Establish a particle trap simulation model and a particle motion simulation model, and use the particle trap simulation model and the particle motion simulation model to obtain the capture rate of the particle trap for particles under different structural parameters of the particle trap as a data set;

[0007] 2) Use the data set to train the constructed machine learning model, and the trained model is an equivalent calculation model for calculating the capture rate according to the structural parameters;

[0008] 3) Optimize the structural parameters with the goal of maximizing the capture rate calculated by using the equivalent calculation model, so as to obtain the optimal values of the structural parameters and realize the structural optimization of the particle trap.

[0009] Further, in step 3), a swarm intelligence algorithm is used for solving.

[0010] Further, the swarm intelligence algorithm is an ant colony algorithm, and during the use of the ant colony algorithm, the structural parameters to be optimized need to be gridded. The process of gridding is as follows: determine the value range of the structural parameters, and then determine the number of significant digits of the values of the structural parameters; map all the structural parameters to a plane grid. A node of the plane grid represents a specific value of one significant digit of a structural parameter, and a vertical line of the plane grid represents a certain significant digit of a certain structural parameter, and a horizontal line of the plane grid represents the specific value of the significant digit.

[0011] Further, the machine learning model is a random forest model.

[0012] Further, the hyperparameters including the number of decision trees, the maximum depth of the decision tree, the minimum number of samples for splitting, and the minimum number of samples in the leaf nodes in the random forest model are optimized by using a swarm intelligence algorithm.

[0013] Further, the swarm intelligence algorithm is a particle swarm algorithm.

[0014] Further, the particle motion simulation model is a particle motion simulation model considering gas buoyancy, gas viscous resistance, and collision loss;

[0015] The gas buoyancy among them is:

[0016]

[0017] In the formula, F f is the buoyancy of the gas; ρ is the density of the gas medium; a is the radius of the metal particle; g is the acceleration due to gravity;

[0018] The viscous resistance among them is:

[0019] F visc = 6πηav

[0020] In the formula, F visc represents the viscous resistance of the gas on the particle during the flight motion after lifting; η is the gas viscosity coefficient; v is the particle motion speed;

[0021] The velocity update formula considering the collision loss is:

[0022]

[0023] In the formula, v n1 and v τ1 are respectively the normal motion velocity and the tangential motion velocity of the particle before collision; v n2 and v τ2 are respectively the normal motion velocity and the tangential motion velocity of the particle after collision; k n and k τ are respectively the normal restitution coefficient and the tangential restitution coefficient of the particle collision.

[0024] Furthermore, the structural parameters include at least one parameter among the total trap length, grid width, baffle width, width ratio of the baffle to the grid, trap thickness, trap lifting height, radial distribution angle of the total grid, and radial distribution angle of a single grid.

[0025] To solve the above technical problems, the present invention also provides a computer device, including a processor, and the processor executes a computer program to implement the steps of the above-introduced method for optimizing the structure of the particle trap for GIS / GIL.

[0026] To solve the above technical problems, the present invention also provides a particle trap for GIS / GIL, and the structural parameters of the particle trap are the parameters optimized by using the above-introduced method for optimizing the structure of the particle trap for GIS / GIL.

[0027] The present invention is an improved invention, and its beneficial effects are as follows: Different from the prior art which directly optimizes the particle trap structure based on the particle trap simulation model, the present invention obtains the capture rates of different particle traps for particles under multiple sets of structural parameter values of the particle trap by using the simulation model of the particle trap and the particle motion simulation model, and uses the data obtained from the simulation model as the data set for training the machine learning model. After the machine learning model is trained, an equivalent calculation model for calculating the capture rate according to the structural parameters is obtained, and this equivalent calculation model is used to optimize the structural parameters of the particle trap. Since the data set used to train the machine learning model is obtained through the simulation model, the accuracy of this calculation model is guaranteed. Most importantly, compared with the method of directly using the simulation model to optimize the structural parameters, using this equivalent calculation model to optimize the structural parameters reduces the number of calculations, has less computational amount, improves the optimization efficiency, meets the requirements of global optimization, enables the particle trap obtained by this optimization method to greatly improve the capture rate of the particle trap, realizes the optimization of the structural parameters of the particle trap for different types and different voltage levels, and greatly reduces the probability of discharge accidents caused by particles in GIL / GIS equipment. Description of the Drawings

[0028] Figure 1 is the flowchart of the structural optimization method of the particle trap for GIS / GIL of the present invention;

[0029] Figure 2 is the flowchart of the simulation calculation of the particle capture rate of the present invention;

[0030] Figure 3 is the force analysis diagram of the metal particles in the particle motion simulation model of the present invention

[0031] Figure 4 is the flowchart of the establishment of the optimal random forest model of the present invention;

[0032] Figure 5 is the flowchart of using the ant colony algorithm to solve the optimal structural parameters of the particle trap of the present invention;

[0033] Figure 6 is the three-dimensional structure schematic diagram of the lifted particle trap optimized by the present invention;

[0034] Figure 7 is the cross-sectional schematic diagram of the lifted particle trap optimized by the present invention;

[0035] Figure 8 is the grid result diagram of the structural parameters to be optimized of the present invention. Detailed Embodiments

[0036] The main idea of the present invention is to use an equivalent calculation model for calculating the capture rate according to structural parameters to optimize the structural parameters with the goal of maximizing the capture rate. The equivalent calculation model is obtained by training a machine learning model using a data set, and the data set includes a structural parameter data set and a capture rate data set obtained by using a simulation model of the particle trap and a particle motion simulation model, specifically the capture rate under the structural parameters of different particle traps. Instead of directly optimizing the structural parameters based on the simulation model, the number of calculations is reduced, the optimization efficiency is improved, and the requirement of global optimization is met. Based on this idea, a structural optimization method for a particle trap used in GIS / GIL, a computer device, and a particle trap used in GIS / GIL of the present invention can be realized.

[0037] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0038] An embodiment of a structural optimization method for a particle trap used in GIS / GIL:

[0039] The overall process of an embodiment of the structural optimization method for a particle trap used in GIS / GIL of the present invention is as Figure 1 shown, aiming to improve the capture rate of the particle trap for metal particles in a DC GIL / GIS device through structural optimization, and improve the insulation performance and operation reliability of the device. The specific process of the whole method is as follows:

[0040] Step 1: Establish a simulation model of a lifting-type particle trap using three-dimensional finite element simulation software. The lifting-type particle trap is installed on a DC GIL / GIS device, and determine the structural parameters to be optimized for the lifting-type particle trap.

[0041] The established simulation analysis model mainly involves a DC GIS / GIL device with a certain voltage level and the corresponding lifting-type particle trap. In this embodiment, a 320 kV DC GIL device and a lifting-type particle trap with a certain size are taken as examples. Moreover, the motion of metal particles is simulated and calculated by coupling the electrostatic field module and the fluid flow particle tracking module in the finite element simulation software COMSOL Multiphysics.

[0042] As Figure 6 、 7 shown, the structural parameters to be optimized for the lifting-type particle trap in this embodiment specifically include the total length l of the trap 1 、the grid width l 2 、the baffle width l 3 、the width ratio k of the baffle to the grid, the trap thickness d, the trap lifting height h, the radial distribution angle θ of the total grid 1 and the radial distribution angle θ of a single grid 2。And select the capture rate of metal particles by the particle trap as the index for optimizing the performance of the particle trap.

[0043] Step 2: Determine the key simulation parameters in the particle motion simulation model.

[0044] The key simulation parameters in this embodiment include the density ρ of metal particles P , the density ρ of the gas medium, the acceleration of gravity g, the relative permittivity ε of vacuum 0 , the relative permittivity ε of SF 6 , the gas viscosity coefficient η, the radius a of the metal particle, the mirror correction coefficient k r , the friction coefficient μ between the metal particle and the electrode, the normal restitution coefficient and tangential restitution coefficient k of the collision between the metal particle and other structures e , k n , k τ and other parameters. Among them, the metal particles are spherical particles made of aluminum, and the gas medium is SF 6 , and the mirror correction coefficient k e takes 0.832. The normal restitution coefficient and tangential restitution coefficient k of the collision between the metal particle and the electrode n , k τ are both 0.6, and the normal restitution coefficient and tangential restitution coefficient k of the collision with the insulator n , k τ are both 0.5.

[0045] Step 3: Establish a metal particle motion simulation model.

[0046] The process of performing the simulation calculation of the particle capture rate is as Figure 2 shown. The schematic diagram of the force analysis of metal particles in the device is as Figure 3 shown. In this embodiment, the fluid flow particle tracking module is corrected according to formulas (1)-(6) to establish a particle motion simulation model considering the change of particle charge, gravity, Coulomb force, electric field gradient force, gas buoyancy, gas viscous resistance, and collision loss. At the same time, the condition for the particle to be captured by the trap needs to be set in the fluid flow particle tracking module. Given that the number of metal particles set is 2000, the capture rate of the particle trap is solved according to the motion state of the particles after a certain time. Specifically:

[0047] The electric charge of the metal particle is jointly determined by the particle size and the field strength at the collision point. For a spherical particle with a radius of a, the electric charge q can be expressed as:

[0048]

[0049] In the formula, ε 0 is the vacuum permittivity, ε r is the relative permittivity of the medium, a is the radius of the metal particle, and E 0is the electric field strength at the collision point. After the metal particles are charged, they are respectively subjected to the Coulomb force F q , the electric field gradient force F gard , and the gravity G, which are respectively expressed by equations (2)-(4):

[0050] F q = k e qE (2)

[0051] In the formula, k e represents the mirror correction coefficient, and E is the electric field strength at the position of the particle.

[0052]

[0053]

[0054] In the formula, ρ P is the density of the metal particle, g is the acceleration due to gravity, is the gradient of the electric field strength E at the position of the particle. The high-pressure gas inside the device also generates a buoyant force F f on the metal particle:[[]]

[0055]

[0056] In the formula, ρ is the density of the gas medium inside the device. The metal particle is also subjected to the viscous resistance F visc of the gas during the flight motion after being lifted, as follows:[[]]

[0057] F visc = 6πηav (6)

[0058] In the formula, η is the gas viscosity coefficient, and v is the particle motion speed. According to the above equations, the motion of the metal particle is simulated and calculated by coupling the electrostatic field module and the fluid flow particle tracking module of the finite element software COMSOL Multiphysics. The charge of the metal particle after collision with the metal such as the shell and the conductor is determined by equation (1). The change in the charge of the metal particle after collision with the insulator is ignored, and the random reflection phenomenon during the particle motion is compensated by releasing a large number of metal particles. The energy of the particle will be lost after collision, and its speed can be updated according to the following formula:[[]]

[0059]

[0060] In the formula, v n1 , v τ1 are respectively the normal motion speed and the tangential motion speed of the metal particle before collision, and v n2 , v τ2 are respectively the normal motion speed and the tangential motion speed of the metal particle after collision, k n , kτ They are the normal restitution coefficient and the tangential restitution coefficient of the metal particle collision respectively. The capture rate of the particle trap is calculated by analyzing the motion of the set metal particles.

[0061] Step 4: According to the particle trap simulation model constructed in Step 1 and the metal particle motion simulation model constructed in Step 3, a data set required for training and testing the machine learning model is calculated through a finite element simulation software. This data set includes a structural parameter data set and a corresponding capture rate data set, that is, the capture rates under the structural parameters of different particle traps.

[0062] The structural parameters include the total trap length l 1 , the grid width l 2 , the baffle width l 3 , the width ratio k of the baffle to the grid, the trap thickness d, the trap lifting height h, the radial distribution angle θ of the total grid 1 and the radial distribution angle θ of a single grid 2 , as Figure 6 , Figure 7 shown. In this embodiment, the initial parameters of the lifting type particle trap for 320 kV DC GIL are 300 mm, 5 mm, 15 mm, 3, 2 mm, 12 mm, 50°, and 4° in sequence. Among them, the width ratio k of the baffle to the grid is closely related to the widths of the baffle and the grid, and the range of k can be used to add constraints to the structural parameters to be optimized. Considering the overall size of the device and the small electric field distortion after setting the trap, the limit range of the structural parameters to be optimized is given, as shown in Equation (8).

[0063]

[0064] Moreover, to reduce the influence of the sampling method on the structural parameter data set, the Latin hypercube sampling method is used to obtain the structural parameter data set within the range of Equation (8), and then the corresponding capture rate data set is calculated using the established finite element simulation model.

[0065] Step 5: Establish a machine learning model, train and test the machine learning model using the data set obtained in Step 4, and finally obtain an equivalent calculation model for calculating the capture rate according to the structural parameters.

[0066] In this embodiment, the machine learning model used is a random forest model, and its hyperparameters include the number of decision trees (n_estimators), the maximum depth of the decision tree (max_depth), the minimum number of samples for splitting (min_samples_split), and the minimum number of samples in the leaf node (min_samples_leaf). Moreover, to improve the accuracy of the model, a swarm intelligence algorithm (specifically, the particle swarm algorithm is selected in this embodiment) is used to optimize the above hyperparameters, and the obtained optimal random forest model is used as the equivalent calculation model for the capture rate simulation calculation of the lifting type particle trap for DC GIL / GIS. The process is as follows Figure 4 as shown below:

[0067] 1) Divide the data set obtained in step four. Divide the structural parameter data set and the capture rate data set calculated using the finite element simulation calculation model into a training set and a test set. To obtain a random forest model that can completely equivalent the particle trap simulation model and the particle motion simulation model to achieve capture rate calculation, add the particle swarm optimization algorithm to optimize the selected hyperparameters. The training set is used for the construction of the random forest model, and the test set is used for the hyperparameter optimization of the particle swarm algorithm.

[0068] 2) Construct the optimal random forest model. Construct the random forest model based on the Python library function, and determine the optimization range of the selected hyperparameters, as shown in Equation (9). Based on the random forest model, construct the fitness function for particle swarm optimization. Use the root mean square error of the test set of the random forest model as the fitness function to optimize the hyperparameters, and obtain the optimal random forest model for the equivalent finite element simulation calculation model.

[0069]

[0070] In step six, with the maximization of the particle trap capture rate as the objective function, use the equivalent calculation model obtained in step five to globally optimize the structural parameters, obtain the optimal structural parameters, and realize the design optimization of the particle trap structure.

[0071] The solution algorithm used for the global optimization of the structural parameters is a swarm intelligence algorithm. Specifically, the ant colony algorithm is selected in this embodiment. The process is as shown in the figure as follows Figure 5 as shown below:

[0072] 1) Initialize the parameters. Initialize the ant colony size m = 150, the pheromone importance factor α = 1, the heuristic function importance factor β = 2.5, the pheromone evaporation factor ρ = 0.1, the maximum number of iterations iter_max, the initial value of the number of iterations iter = 0, the random forest equivalent model f(x), and the pheromone concentration at the node (x i , y i,j ) at time t is τ(x i, y i,j , t), and at the initial moment τ(x i , y i,j , 0) = τ 0 , and the maximum and minimum values of pheromone are τ max = 1, τ min = 0.0024.

[0073] 2) Parameter gridification. Determine the range of the structural parameters to be optimized and gridify the parameter variable region. The range of the structural parameters to be optimized is the same as that selected in step 4, including seven structural parameters to be optimized. Divide the structural parameters to be optimized into N equal parts, and adopt the method that each node in the grid represents one significant digit of a structural parameter to make the N - equal - divided structural parameters correspond to the grid nodes one by one. The number of significant digits of the seven structural parameters to be optimized is 3, 3, 3, 3, 3, 2, 1 in sequence, with a total of 18 significant digits. The plane grid corresponding to the seven structures to be optimized in this embodiment is as shown in Figure 8 . The plane grid has 18 vertical lines in the x - direction, and each vertical line represents one significant digit of a certain parameter. Each vertical line has 10 nodes in the y - direction, representing the specific value of this significant digit. There are a total of 180 nodes. The specific number of nodes set in the y - direction is determined by the value range of the structural parameters to be optimized, and the value range is determined by Equation (8). This step decomposes the continuous problem and reduces the difficulty of solving.

[0074] 3) Construct the solution space, set a certain number of ants, calculate the probability of an ant moving to the next node and determine the next node the ant will move to in the way of roulette. Each ant needs to pass through 18 nodes and calculate the selected values of the seven structural parameters to be optimized. Each time an ant's choice of the moving path is determined by a probability function, P k (x i , y i,j , t) represents the probability that ant k transfers to node (x i , y i,j ) at time t, τ(x i , y i,j , t) represents the pheromone remaining on node (x i , y i,j ) at time t. At the initial moment, the pheromone on all nodes is the same. The transfer probability calculation formula is:

[0075]

[0076] In the formula, η(x i , y i,j , t) is the heuristic function, representing the expected degree of an ant transferring to node (x i , y i,j ); α represents the importance factor of pheromone, β represents the importance factor of the heuristic function; η(xi , y i,j , t) = f(x) * , f(x) * is the mean of the random forest models corresponding to all ants that select this node.

[0077] 4) Update the pheromone. Calculate the particle trap capture rate under all ant selection parameters through the random forest model, record the optimal parameter combination and the optimal path selected by the corresponding ants, and update the pheromone on each node. Update the pheromone using Equation (11), and at the same time judge whether the pheromone concentration exceeds the range. If τ > τ max , then τ = τ max ; if τ < τ min , then τ = τ min .

[0078] τ(x i , y i,j , t + 1) = (1 - ρ) * τ(x i , y i,j , t) + ρ * τ 0 (11)

[0079] In the formula, ρ represents the pheromone evaporation factor.

[0080] 5) If the number of iterations is less than the maximum number of iterations, clear the path records of the ants and return to step 3). Otherwise, terminate the calculation and output the optimal parameter set and the corresponding capture rate.

[0081] Under the premise of meeting the number of iterations, the optimization results of all parameters are obtained and the capture rate of the particle trap is calculated. Table 1 shows the parameter comparison before and after optimization.

[0082] Table 1 Parameter Comparison Before and After Optimization

[0083]

[0084]

[0085] Based on the random forest model, good results have been achieved in optimizing the structural parameters of the lifting type particle trap for high-voltage DC GIL / GIS. The ant colony algorithm disassembles the structural parameters to be optimized into a certain number of arrays of 0-9, transforms the connection function problem into an n-level decision-making problem, greatly reduces the difficulty of solving, improves the calculation efficiency, and overcomes the drawback of large computational amount of traditional optimization algorithms. The capture rate of the optimized lifting type particle trap has been greatly improved, from 25.6% to 64.83%. Obviously, it can greatly reduce the probability of particle-induced discharge faults and greatly improve the insulation performance and operation reliability of high-voltage DC GIL / GIS equipment.

[0086] An embodiment of a computer device:

[0087] An embodiment of a computer device of the present invention includes a memory, a processor, an internal bus, and a computer program stored in the memory. The processor and the memory complete mutual communication and data interaction through the internal bus. The processor executes the computer program to implement the steps of the method introduced in an embodiment of the method for optimizing the structure of a particulate trap of the present invention. Among them, the processor can be a programmable logic device FPGA, a processor based on the ARM architecture, and other processing devices; the memory can be various memories that store information in an electrical energy manner, such as RAM, ROM, etc., or can also be a memory using other methods.

[0088] An embodiment of a particulate trap for GIS / GIL:

[0089] An embodiment of a particulate trap for GIS / GIL of the present invention, and the specific values of the structural parameters of the particulate trap for GIS / GIL are obtained by using the method for optimizing the structure of a particulate trap for GIS / GIL of the present invention. The principle and specific implementation process of this method have been introduced in detail in an embodiment of the method for optimizing the structure of a particulate trap, and will not be elaborated in this embodiment.

[0090] In summary, the present invention has established a three-dimensional finite element simulation model of a lifting particulate trap for DC GIL / GIS, and based on this, a data set for constructing a random forest model has been calculated. Further, a random forest equivalent model of the three-dimensional finite element simulation model of the lifting particulate trap for DC GIL / GIS has been constructed, and the structural parameters of the lifting particulate trap have been optimized based on the ant colony algorithm, greatly improving the capture rate of the particulate trap. The present invention can optimize the structures of lifting particulate traps for different types and voltage levels of GIL / GIS, can improve the capture rate of the corresponding particulate traps, reduce the probability of discharge accidents caused by particulate matter in the equipment, thereby reducing the economic losses and casualties caused by line failures, maintenance, etc. caused by particulate matter, and ensuring the property safety and life safety of the people.

[0091] The specific implementation manners are given above, but the present invention is not limited to the described implementation manners. The basic idea of the present invention lies in the above basic scheme. For those of ordinary skill in the art, according to the teachings of the present invention, it does not require creative labor to design various deformed models, formulas, and parameters. Changes, modifications, substitutions, and variations made to the implementation manners without departing from the principle and spirit of the present invention still fall within the protection scope of the present invention.

Claims

1. A method for optimizing the structure of a particle trap for GIS / GIL, characterized in that: The steps include: 1) Establish a particle trap simulation model and a particle motion simulation model, and use the particle trap simulation model and the particle motion simulation model to obtain the particle capture rate of the particle trap under different particle trap structural parameters and use them as a data set; 2) using the data set to train the constructed machine learning model, the trained model is an equivalent calculation model for calculating the capture rate according to the structural parameters; 3) Optimizing the structural parameters with the goal of maximizing the capture rate calculated using the equivalent calculation model, thereby obtaining the optimal values ​​of the structural parameters to achieve structural optimization of the particle trap.

2. The structural optimization method for a particle trap for GIS / GIL according to claim 1, characterized in that: In step 3), a swarm intelligence algorithm is used to solve the problem.

3. The structural optimization method of the particle trap for GIS / GIL according to claim 2, characterized in that: The swarm intelligence algorithm is an ant colony algorithm, and in the process of using the ant colony algorithm, the structural parameters to be optimized need to be gridded. The process of gridding is: determining the value range of the structural parameters, and then determining the number of valid digits of the structural parameter values; all structural parameters are mapped to a plane grid, a node of the plane grid represents the specific value of a valid digit of a structural parameter, and a vertical line of the plane grid represents a valid digit of a certain structural parameter, and a horizontal line of the plane grid represents the specific value of the valid digit.

4. The structural optimization method for a particle trap for GIS / GIL according to claim 1, characterized in that: The machine learning model is a random forest model.

5. The structural optimization method of the particle trap for GIS / GIL according to claim 4, characterized in that: The hyperparameters in the random forest model, including the number of decision trees, the maximum depth of the decision trees, the minimum number of sample splits, and the minimum number of leaf node samples, are optimized using a swarm intelligence algorithm.

6. The structural optimization method of the particle trap for GIS / GIL according to claim 5, characterized in that: The swarm intelligence algorithm is a particle swarm algorithm.

7. The structural optimization method for a particle trap for GIS / GIL according to claim 1, characterized in that: The particle motion simulation model is a particle motion simulation model that takes into account gas buoyancy, gas viscous resistance and collision loss; The gas buoyancy is: In the formula, F f is the gas buoyancy; ρ is the density of the gas medium; a is the radius of the metal particle; g is the gravitational acceleration; The viscous resistance is: F visc =6πιav In the formula, F visc It indicates that the particle is subject to the viscous resistance of the gas during its flight after being lifted; η is the gas viscosity coefficient; v is the particle movement speed; The speed update formula after considering collision loss is: In the formula, v n1 、v τ1 are the normal velocity and tangential velocity of the particles before collision, respectively; v n2 、v τ2 are the normal velocity and tangential velocity of the particles after collision, respectively; k n , k τ are the normal restitution coefficient and tangential restitution coefficient of particle collision, respectively.

8. The method for optimizing the structure of a particle trap for GIS / GIL according to any one of claims 1 to 7, characterized in that: The structural parameters include at least one of the total trap length, grid width, baffle width, baffle to grid width ratio, trap thickness, trap lift height, radial distribution angle of the total grid, and radial distribution angle of a single grid.

9. A computer device comprising a processor, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 8.

10. A particle trap for GIS / GIL, characterized in that: The structural parameters of the particle trap are parameters optimized by the method described in any one of claims 1 to 8.

Images