GIS metal particle catcher optimal site selection method based on hybrid optimization algorithm

Abstract

The invention discloses an optimal site selection method for a GIS metal particle catcher based on a hybrid optimization algorithm, and the method comprises the steps: firstly building a 500kV GIS scaling model through COMSOL, carrying out the simulation, and obtaining a three-dimensional drop point coordinate data set of a metal particle track in the model; importing the data set into MATLAB, and constructing an optimization model taking the minimum sum of Euclidean distances from the position of the catcher to all particle drop points as a target; and finally, searching an optimal point in a continuous three-dimensional space under a set boundary constraint condition by adopting a hybrid optimization algorithm consisting of a WOA (whale optimization algorithm), a PSO (particle swarm optimization algorithm) and a GWO (grey wolf optimization algorithm), so that the capturing efficiency is maximized when the particle trapper is placed at the point. According to the method, the technical problems of difficulty in site selection and low capturing efficiency of the metal particle trapper in the GIS are solved, the capturing coverage rate of the trapper on the metal particles is remarkably improved, and reliable guarantee is provided for safe and stable operation of GIS equipment.

Description

Technical Field

[0001] This invention relates to the field of high-voltage power transmission in power systems, and in particular to an optimal location method for GIS metal particle traps based on a hybrid optimization algorithm. Background Technology

[0002] Gas-insulated switchgear (GIS) is a core piece of equipment in power systems. Due to its small footprint, superior insulation, and high reliability, it is widely used in 500kV and above high-voltage transmission networks. However, during its production, assembly, and operation, metal particles are generated. Under the influence of an electric field, these particles can migrate, suspend, or even discharge, threatening insulation performance and potentially causing faults and economic losses.

[0003] To address the safety hazards posed by metal particles, current technologies often involve installing particle traps inside GIS systems to capture metal particles through physical adsorption or electric field traction. However, due to the complex internal structure of GIS and the random and uncertain trajectories of metal particles, the installation location of the trap determines the capture efficiency. Traditional site selection methods rely on empirical judgment or local experiments, resulting in high subjectivity and low capture coverage. Using a single optimization algorithm for site selection is prone to getting stuck in local optima and struggles to find the global optimum in continuous three-dimensional space. Therefore, achieving scientific site selection for traps and maximizing capture efficiency is a pressing technical challenge in the field of GIS equipment operation and maintenance. Summary of the Invention

[0004] To address the above problems, this invention provides an optimal location method for GIS metal particle traps based on a hybrid optimization algorithm.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: An optimal location method for a GIS metal particle trap based on a hybrid optimization algorithm includes the following steps: Step 1: Establish a scaled-down model of the GIS, simulate and analyze the electric field distribution inside the cavity of the scaled-down model of the GIS, obtain the electric field distribution results, and then simulate the motion characteristics of the metal particles inside the GIS to obtain the three-dimensional motion trajectory and landing point coordinates of the metal particles. Step 2: Export the three-dimensional coordinate data of the metal particle impact points obtained from the simulation, perform data preprocessing, remove outliers, and construct a standardized three-dimensional dataset of particle impact points. Step 3: Based on the MATLAB platform, define an optimization objective function with the goal of minimizing the sum of Euclidean distances from the geometric center of the trap to all particle landing points. The minimum value of the optimization objective function corresponds to the maximum value of the trap's capture efficiency. Step 4: Configure the core parameters of at least three optimization algorithms, including population size, maximum number of iterations, and search space boundary; Step 5: Call the initialization function to generate the initial population of the optimization algorithm, and use the boundary check mechanism to ensure that the position of the initial population is within a reasonable search space; Step 6: Initiate the hybrid optimization algorithm iteration process: The optimization algorithms independently search for the optimal solution, while the fitness values ​​are compared during the iteration process to achieve information exchange and collaborative optimization between the algorithms; Step 7: After the iteration terminates, extract the optimal solution obtained by the optimization algorithm and obtain the final optimal address coordinates of the catcher through weighted fusion; Step 8: Based on the final optimal location coordinates, install the metal particle trap at the corresponding location on the GIS equipment.

[0006] A further improvement is made to the method for simulating the motion characteristics of metal particles inside the GIS in step one, as follows: (1.1) Based on the parameters of the scaled-down model, the electric field distribution inside the GIS cavity is simulated and analyzed to obtain the electric field distribution results; (1.2) Collision detection is carried out based on the local motion trajectory of metal particles to determine whether the metal particles collide with the conductor or shell of GIS; if no collision occurs, the force, velocity and real-time position of the metal particles are calculated, and the position and velocity of the current trajectory endpoint are used as the initial parameters for the next time step. (1.3) When the metal particles collide with the grounded shell or the high-voltage conductor, the charge of the particles is recalculated and the rebound velocity of the metal particles is determined according to the collision rebound coefficient. The final force, velocity and real-time position of the particles are calculated. After recalibrating the time step, the initial motion parameters of the next time step are updated.

[0007] (1.4) By using multi-time step iterative calculation, the various local motion trajectory segments are integrated into a complete three-dimensional motion trajectory of the metal particles until the set simulation cycle is achieved.

[0008] Further improvements include defining the motion state as "collision" if the metal particles come into contact with the conductor or outer shell of the GIS; and defining the motion state as "non-collision" if the metal particles are in the space between the grounded outer shell and the high-voltage conductor. The equation for analyzing the motion state of the metal particles is as follows:

[0009] in, Let (r, θ) be the mass of the metal particle, and (r, θ) be the particle coordinates. These are the electric force and the dielectric force, respectively. The angle between the metal particle and the plane is the polar angle θ, and the gas resistance experienced by the metal particle is... Fv Decomposed into components along the direction of the electric field force. The component perpendicular to the direction of the electric field force ; r This represents the distance between the metal particles and the center of the high-voltage conductor. This indicates the second derivative; t represents time; G is the gravitational constant; F is the net force acting on the metallic particle. The electric field force experienced by the metal particles Specifically:

[0010] in, R 1 Let be the radius of the high-voltage conductor. R 2 Let be the inner radius of the GIS cavity, and k be the polarization coefficient of the metal particle. Let ln be the vacuum permittivity, and ln be a logarithmic function. Dielectrophoretic force on the metal particle at time t Specifically:

[0011] in, Let be the relative permittivity of the insulating gas C4F7N in the cavity. Wherein is the radius of the metal particle; This represents the change in electric field gradient; The gas resistance experienced by the metal particles at time t for:

[0012]

[0013]

[0014] in, The drag coefficient, For gas density, For Reynolds coefficient, For temperature C4F7N gas density at that time The gas flow rate is... The velocity of the metal particles. Let t be the velocity of the metallic foreign object relative to the gas. The dynamic viscosity coefficient of C4F7N gas.

[0015] In a further improvement, the method for data preprocessing and outlier removal in step two is as follows: Deduplication: Remove duplicate endpoint coordinates to avoid data redundancy; Outlier removal: The 3σ criterion is used to identify and delete outlier coordinates that exceed the normal distribution range.

[0016] In a further improvement, the objective function in step three is optimized as follows:

[0017] in The three-dimensional coordinates of the geometric center of the trap catcher. Let be the three-dimensional coordinates of the i-th metal particle's landing point in the GIS, and n be the total number of metal particles. It is the sum of the Euclidean distances from the geometric center of the catcher to all landing points.

[0018] Further improvements to the boundary checking mechanism are as follows: (4.1) Spatial boundary constraints: Based on the physical size and particle landing point distribution range of the GIS scaled-down model, the upper boundary L of the optimization search space is defined. b and lower boundary U b : L b =min(min(data))-δ; U b =max(max(data))+δ; Where δ is the boundary margin, which is 10% of the difference between the extreme values ​​of the coordinates of the landing point of the metal particle, i.e., δ=0.1×(max(data(:))−min(data(:))), to ensure that the search space completely covers the particle distribution area and complies with the internal installation space limit of the GIS equipment; max(data(:)) represents the maximum value among all coordinates of the landing point of the metal particle, min(data(:)) represents the minimum value among all coordinates of the landing point of the metal particle, and data(:) represents all three-dimensional coordinate data of the landing point of the metal particle; (4.2) Variable Dimension Constraints: The optimization variables are three-dimensional coordinates (x, y, z), with dimension dim=3. The value of each dimension must satisfy L. b ≤pos d ≤U b d=x,y,z; pos d Represents all coordinates of x, y, z; (4.3) Boundary constraints of algorithm parameters: the population size ranges from 30 to 100, and the maximum number of iterations Max_iter ranges from 200 to 500, to ensure a balance between the algorithm's convergence efficiency and optimization accuracy.

[0019] In a further improvement, the optimization algorithms in step four include Whale Algorithm (WOA), Particle Swarm Optimization (PSO), and Grey Wolf Algorithm (GWO). In step six, the iterative process of the hybrid optimization algorithm includes: The WOA algorithm updates the population position through three behaviors: encirclement predation, bubble web attack, and random search. The search strategy is dynamically switched by linearly decreasing the parameter α. The PSO algorithm adjusts the global and local search capabilities by linearly decreasing the inertia weight ω, and updates the particle velocity and position by combining the individual optimal solution PBEST and the global optimal solution GBEST, with the velocity boundary limited to 20% of the search space; The GWO algorithm updates the population position by simulating the social hierarchy of gray wolves, and uses the linear change of parameter α to balance exploration and exploitation capabilities; the social hierarchy includes alpha wolf, second-best wolf and third-best wolf.

[0020] In a further improvement, the weighting coefficients for the weighted fusion in step seven are determined based on the convergence speed and iterative stability of the three algorithms. The weighting coefficients for WOA, PSO, and GWO are set to 0.35, 0.35, and 0.3, respectively, and the final optimal coordinate is pos. opt =0.35×pos WOA +0.35×pos PSO +0.3×pos GWO , where pos WOA pos PSO pos GWO These are the optimal coordinates obtained by the three algorithms independently.

[0021] A further improvement is that the scaled-down model of the GIS is the COMSOL 500kV GIS scaled-down model.

[0022] Advantages of this invention: This invention solves the technical problems of difficult site selection and low capture efficiency of metal particle traps inside GIS, significantly improves the capture coverage of metal particles by the trap, and provides a reliable guarantee for the safe and stable operation of GIS equipment. Attached Figure Description

[0023] Figure 1 This is a scaled-down model diagram of a 500kV AC GIS system as an example.

[0024] Figure 2 The diagram shows the electric field distribution results for an example.

[0025] Figure 3 This is a force analysis diagram of metal particles as an example.

[0026] Figure 4a The example algorithm's iterative convergence curve is shown.

[0027] Figure 4b This is a diagram showing the three-dimensional spatial location results of an example.

[0028] Figure 5 The overall diagram shows the result of releasing the kinetic energy of 50 particles after the trap is installed.

[0029] Figure 6 for Figure 5 Enlarged view of the circled area.

[0030] Figure 7 The overall diagram shows the result of releasing the kinetic energy of 50 particles after the trap is installed.

[0031] Figure 8 for Figure 5 Enlarged view of the circled area.

[0032] Figure 9 The trajectory of a 10-metal aluminum particle being captured after the trap is installed.

[0033] Figure 10 The trajectory of the b 50 aluminum metal particles being captured after the trap is installed. Detailed Implementation

[0034] The technical solution of the present invention will be specifically described below through specific embodiments and in conjunction with the accompanying drawings.

[0035] Example A scaled-down COMSOL 500kV GIS model was established, with an inner radius of 0.07m for the metal casing, an outer radius of 0.01m for the high-voltage conductor, aluminum particles with a radius of 0.25mm, a normal restitution coefficient of 0.6 for the electrode casing, a normal restitution coefficient of 0.47 for the insulator, a tangential restitution coefficient of 0.6 for the electrode casing, and a tangential restitution coefficient of 1 for the insulator. An AC voltage of 500,000V was applied. sin(2 pi f x), where the frequency f is 50Hz.

[0036] The constructed 500kV scaled-down geometric model is as follows: Figure 1 As shown; After setting parameters and dividing the grid, the electric field distribution inside the GIS cavity is simulated and analyzed based on the parameters in the simulation model, and the electric field distribution results are obtained, such as... Figure 2 As shown; The simulation of the motion characteristics of aluminum microparticles includes the following steps: (1) Simulate the motion trajectory of a single metal particle and determine whether the metal particle jumps; if the metal particle jumps, determine the state of the metal particle and calculate the position of the metal particle, and re-determine the state of the metal particle based on the position of the metal particle; if the metal particle does not jump, the simulation calculation ends. (2) Determine whether the predetermined simulation time has been reached; if the predetermined simulation time has been reached, end the simulation calculation; if the predetermined simulation time has not been reached, repeat step (1). The aforementioned take-off condition is the critical condition for the lift-off of metal particles. When the Coulomb force on the metal particles inside the GIS is greater than or equal to the gravity, the particles begin to lift.

[0037] The method for determining the state of metal particles and calculating their positions includes the following steps: When the metal particles are in a "non-collision" state, calculate the force, velocity, and real-time position of the particles. If the metal particles collide with the grounded casing or the high-voltage conductor, the charge of the particles must be recalculated first, and then the rebound velocity must be determined according to the collision rebound coefficient. Finally, the force, velocity, and real-time position of the particles after the collision are calculated.

[0038] If the metal particles are located in the space between the grounded outer shell and the high-voltage conductor, their motion state is defined as "non-collision"; The equation for analyzing the motion state of the metal particles is as follows:

[0039] in, Let (r, θ) be the mass of the metal particle, and (r, θ) be the particle coordinates. These are the electric force and the dielectric force, respectively. The angle between the metallic foreign object and the plane is the polar angle θ. The gas drag Fv experienced by the metallic foreign object is decomposed into components along the direction of the electric force. The component perpendicular to the direction of the electric field force .

[0040] The electric field force experienced by the metal particles is specifically as follows:

[0041] The dielectric force experienced by the metal particles is specifically as follows:

[0042] in The vacuum permittivity, Let be the relative permittivity of the insulating gas C4F7N in the cavity. Wherein is the radius of the metal particle; The gas resistance experienced by the metal particles is:

[0043]

[0044]

[0045] in, The drag coefficient, For gas density, For Reynolds coefficient, For temperature C4F7N gas density at that time The gas flow rate is... The velocity of the metal particles. The dynamic viscosity coefficient of C4F7N gas; Force analysis diagram as shown Figure 3 As shown; Simulate the trajectory of aluminum particles and export the simulation data; Preprocess the exported data: 1. Deduplication: Remove duplicate endpoint coordinates to avoid data redundancy; 2. Outlier removal: The 3σ criterion is used to identify and delete outlier coordinates that exceed the normal distribution range; Table 1 shows a portion of the dataset of particle final landing points. Table 1

[0046] Define the optimization objective function:

[0047] in The three-dimensional coordinates of the geometric center of the trap catcher. Let be the three-dimensional coordinates of the i-th impact point, and n be the total number of particles. It is the sum of the Euclidean distances from the geometric center of the catcher to all landing points.

[0048] The specific steps for setting boundary conditions are as follows: (1) Spatial boundary constraints: Based on the physical size of the GIS scaled-down model and the distribution range of particle landing points, the upper and lower boundaries of the optimization search space are defined as Lb=min(min(data))-δ, Ub=max(max(data))+δ, where δ is the boundary margin, which is 10% of the extreme value difference of particle landing point coordinates, i.e. δ=0.1×(max(data(:))−min(data(:))), to ensure that the search space completely covers the particle distribution area and meets the internal installation space limitations of GIS equipment; The calculated values ​​are Lb = -0.1 and Ub = 1.1 (boundary margin is 10%). (2) Variable dimension constraints: The optimization variables are three-dimensional coordinates (x, y, z), with dimension dim=3. The value of each dimension must satisfy -0.1≤posd≤1.1 (d=x, y, z). (3) Boundary constraints of algorithm parameters: the population size is set to 50, and the maximum number of iterations Max_iter is set to 300 to ensure the balance between the convergence efficiency and the optimization accuracy of the algorithm; The hybrid optimization algorithm iteratively optimizes, independently searches, and achieves information exchange.

[0049] The WOA algorithm iteratively updates the population position and searches for the optimal location by dynamically switching between three behaviors: encirclement predation (|A|<1), random search (|A|≥1), and bubble web attack (p≥0.5) through linear decrease of parameter a. The algorithm assumes that the best individual in the current population is prey, and other whales in the population surround the best whale's position to update their own positions, which are updated by formulas (1) and (2): (1) (2) Where t is the current time; Indicates the enclosing step size. and It is a coefficient vector; This is the current location of the individual whale; It is the current leadership position; This is the updated location of the individual whale; ;in ; It is the number of iterations; It is the maximum number of iterations; ; and It is a random number in the range [0,1]. (2) Bubble Web Attack Phase When hunting, humpback whales swim towards their prey in a spiral motion. In the whale optimization algorithm, individual whales update their positions using formulas (3) and (4): (3) (4) Where b is a constant that changes the shape of the spiral, and is usually set to 1; ] It is a random number between [-1, 1]; The development of the whale optimization algorithm includes two phases: shrinking the encirclement and bubble net attack. When At that time, because the whale swims in a spiral trajectory around its prey within a narrowing encirclement, there is a 50% chance that the whale will choose to surround the prey and a 50% chance that it will choose to attack with a bubble net. This process is obtained through formula (5): (5) P is a random number in [0,1]; (3) Random search and predation phase: when At that time, whales randomly search and hunt based on their location. In WOA, whales update their location using formulas (6) and (7): (6) (7) Where t is the current time; Indicates the enclosing step size. This represents the location of a randomly selected individual whale from the current population. This is the current location of the individual whale; The updated location of the individual whale; The optimal coordinates obtained are (0.33, -0.34, -0.01), the minimum distance sum is 8.7254, and the convergence curve is as follows. Figure 4a As shown; The PSO algorithm uses inertia weights. The linearly decreasing balance between global and local search capabilities, combined with individual optimal solutions and global optimal solutions, through... Update particle velocity, by Update particle positions.

[0050] c1 and c2 are acceleration constants; r1 and r2 are random numbers uniformly distributed from 0 to 1.

[0051] Inertia weight A linear decreasing strategy is adopted, with an initial value of When max=0.9, after 300 iterations min=0.2, the decreasing formula is: This is used to balance global search and local development capabilities; Acceleration coefficients: c1=2 (individual cognitive coefficient), c2=2 (social learning coefficient), enhancing the particle's ability to learn about its own optimal and global optimal states; Velocity boundaries: Vmax=(Ub−Lb)×0.2=0.24, Vmin=−0.24, to prevent the particle velocity from being too high and causing the search to be unstable; After 300 iterations, the optimal coordinates were obtained as (0.33, -0.34, -0.01), and the minimum distance sum was 8.7254. The convergence curve is shown below. Figure 4a As shown; The global and local search capabilities of the GWO algorithm are mainly adjusted based on the control factor 'a'. ; It is the number of iterations; It is the maximum number of iterations;

[0052] Where Dα, Dβ, and Dδ represent the distance vectors between the prey and the alpha wolf, the second-best wolf, and the third-best wolf, respectively; Cα, Cβ, and Cδ represent the control vectors; Xα, Xβ, and Xδ represent the position vectors of the corresponding wolves; and X(t) represents the position vector of the gray wolf in the t-th iteration.

[0053]

[0054] 1, 2, and 3 correspond to the alpha wolf, the second-best wolf, and the third-best wolf, respectively. Candidate positions:

[0055]

[0056]

[0057] Last updated location: ; After 300 iterations, the optimal coordinates are (0.33, -0.34, 0.00), the minimum distance sum is 8.7728, and the convergence curve is shown below. Figure 4a As shown; Based on the weights (WOA: 0.35, PSO: 0.35, GWO: 0.3), the final optimal coordinates of posopt = 0.35 × posWOA + 0.35 × posPSO + 0.3 × posGWO are (0.33, -0.34, -0.01). By aligning the geometric center of the trap with the obtained optimal coordinates, and simultaneously releasing 10 and 50 aluminum particles respectively, the changes in particle kinetic energy were simulated. Figures 5-10 As shown.

[0058] After the simulation ends, particles with 0 kinetic energy are considered to have been captured.

[0059] Based on the trajectory, the simulation results for 10 and 50 particles show capture efficiencies of 80% and 94%, respectively.

[0060] The above is only one specific implementation method of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing the protection scope of the present invention.

Claims

1. A method for optimal site selection of a GIS metal particle trap based on a hybrid optimization algorithm, characterized in that, Includes the following steps: Step 1: Establish a scaled-down model of the GIS, simulate and analyze the electric field distribution inside the cavity of the scaled-down model of the GIS, obtain the electric field distribution results, and then simulate the motion characteristics of the metal particles inside the GIS to obtain the three-dimensional motion trajectory and landing point coordinates of the metal particles. Step 2: Export the three-dimensional coordinate data of the metal particle impact points obtained from the simulation, perform data preprocessing, remove outliers, and construct a standardized three-dimensional dataset of particle impact points. Step 3: Based on the MATLAB platform, define an optimization objective function with the goal of minimizing the sum of Euclidean distances from the geometric center of the trap to all particle landing points. The minimum value of the optimization objective function corresponds to the maximum value of the trap's capture efficiency. Step 4: Configure the core parameters of at least three optimization algorithms, including population size, maximum number of iterations, and search space boundary; Step 5: Call the initialization function to generate the initial population of the optimization algorithm, and use the boundary check mechanism to ensure that the position of the initial population is within a reasonable search space; Step 6: Initiate the hybrid optimization algorithm iteration process: The optimization algorithms independently search for the optimal solution, while the fitness values ​​are compared during the iteration process to achieve information exchange and collaborative optimization between the algorithms; Step 7: After the iteration terminates, extract the optimal solution obtained by the optimization algorithm and obtain the final optimal address coordinates of the catcher through weighted fusion; Step 8: Based on the final optimal location coordinates, install the metal particle trap at the corresponding location on the GIS equipment.

2. The optimal location method for GIS metal particle traps based on a hybrid optimization algorithm as described in claim 1, characterized in that, In step one, the method for simulating the motion characteristics of metal particles inside the GIS is as follows: (1.1) Based on the parameters of the scaled-down model, the electric field distribution inside the GIS cavity is simulated and analyzed to obtain the electric field distribution results; (1.2) Collision detection is carried out based on the local motion trajectory of metal particles to determine whether the metal particles collide with the conductor or shell of GIS; if no collision occurs, the force, velocity and real-time position of the metal particles are calculated, and the position and velocity of the current trajectory endpoint are used as the initial parameters for the next time step. (1.3) When the metal particles collide with the grounded casing or the high-voltage conductor, the charge of the particles is recalculated and the rebound velocity of the metal particles is determined according to the collision rebound coefficient. The final force, velocity and real-time position of the particles are calculated. After recalibrating the time step, the initial motion parameters of the next time step are updated. (1.4) By using multi-time step iterative calculation, the various local motion trajectory segments are integrated into a complete three-dimensional motion trajectory of the metal particles until the set simulation cycle is achieved.

3. The optimal location method for GIS metal particle traps based on a hybrid optimization algorithm as described in claim 2, characterized in that, If the metal particles come into contact with the conductor or outer shell of the GIS, their motion state is defined as "collision"; if the metal particles are in the space between the grounded outer shell and the high-voltage conductor, their motion state is defined as "non-collision". The equation for analyzing the motion state of the metal particles is as follows: ； in, Let (r, θ) be the mass of the metal particle, and (r, θ) be the particle coordinates. These are the electric force and the dielectric force, respectively. The angle between the metal particle and the plane is the polar angle θ, and the gas resistance experienced by the metal particle is... Fv Decomposed into components along the direction of the electric field force. The component perpendicular to the direction of the electric field force ; r This represents the distance between the metal particles and the center of the high-voltage conductor. This indicates the second derivative; t represents time; G is the gravitational constant; F is the net force acting on the metallic particle. The electric field force experienced by the metal particles Specifically: ； in, R 1 Let be the radius of the high-voltage conductor. R 2 Let be the inner radius of the GIS cavity, and k be the polarization coefficient of the metal particle. Let ln be the vacuum permittivity, and ln be a logarithmic function. Dielectrophoretic force on the metal particle at time t Specifically: ； in, Let be the relative permittivity of the insulating gas C4F7N in the cavity. Wherein is the radius of the metal particle; This represents the change in electric field gradient; The gas resistance experienced by the metal particles at time t for: ； ； ； in, The drag coefficient, For gas density, For Reynolds coefficient, For temperature C4F7N gas density at that time The gas flow rate is... The velocity of the metal particles. Let be the velocity of the metallic foreign object relative to the gas at time t. The dynamic viscosity coefficient of C4F7N gas.

4. The optimal location method for GIS metal particle traps based on a hybrid optimization algorithm as described in claim 1, characterized in that, In step two, the method for data preprocessing and outlier removal is as follows: Deduplication: Remove duplicate endpoint coordinates to avoid data redundancy; Outlier removal: The 3σ criterion is used to identify and delete outlier coordinates that exceed the normal distribution range.

5. The optimal location method for GIS metal particle traps based on a hybrid optimization algorithm as described in claim 1, characterized in that... In step three, the objective function is optimized as follows: ； in The three-dimensional coordinates of the geometric center of the trap catcher. Let be the three-dimensional coordinates of the i-th metal particle's landing point in the GIS, and n be the total number of metal particles. It is the sum of the Euclidean distances from the geometric center of the catcher to all landing points.

6. The method for modeling electromagnetic interference in a high-frequency driving circuit considering near-field coupling characteristics according to claim 1, characterized in that, The boundary check mechanism is as follows: (4.1) Spatial boundary constraints: Based on the physical size and particle landing point distribution range of the GIS scaled-down model, the upper boundary L of the optimization search space is defined. b and lower boundary U b : L b =min(min(data))-δ; U b =max(max(data))+δ; Where δ is the boundary margin, which is 10% of the difference between the extreme values ​​of the coordinates of the landing point of the metal particle, i.e., δ=0.1×(max(data(:))−min(data(:))), to ensure that the search space completely covers the particle distribution area and complies with the internal installation space limit of the GIS equipment; max(data(:)) represents the maximum value among all coordinates of the landing point of the metal particle, min(data(:)) represents the minimum value among all coordinates of the landing point of the metal particle, and data(:) represents all three-dimensional coordinate data of the landing point of the metal particle; (4.2) Variable Dimension Constraints: The optimization variables are three-dimensional coordinates (x, y, z), with dimension dim=3. The value of each dimension must satisfy L. b ≤pos d ≤U b d=x,y,z; pos d Represents all coordinates of x, y, z; (4.3) Boundary constraints of algorithm parameters: the population size ranges from 30 to 100, and the maximum number of iterations Max_iter ranges from 200 to 500, to ensure a balance between the algorithm's convergence efficiency and optimization accuracy.

7. The method for modeling electromagnetic interference in a high-frequency driving circuit considering near-field coupling characteristics according to claim 1, characterized in that, In step four, the optimization algorithms include Whale Algorithm (WOA), Particle Swarm Optimization (PSO), and Grey Wolf Algorithm (GWO). In step six, the iterative process of the hybrid optimization algorithm includes: The WOA algorithm updates the population position through three behaviors: encirclement predation, bubble web attack, and random search. The search strategy is dynamically switched by linearly decreasing the parameter α. The PSO algorithm adjusts the global and local search capabilities by linearly decreasing the inertia weight ω, and updates the particle velocity and position by combining the individual optimal solution PBEST and the global optimal solution GBEST, with the velocity boundary limited to 20% of the search space; The GWO algorithm updates the population position by simulating the social hierarchy of gray wolves, and uses the linear change of parameter α to balance exploration and exploitation capabilities; the social hierarchy includes alpha wolf, second-best wolf and third-best wolf.

8. The method for modeling electromagnetic interference in a high-frequency driving circuit considering near-field coupling characteristics according to claim 1, characterized in that, The weighting coefficients for the weighted fusion in step seven are determined based on the convergence speed and iterative stability of the three algorithms. The weighting coefficients for WOA, PSO, and GWO are set to 0.35, 0.35, and 0.3, respectively, and the final optimal coordinate is pos. opt =0.35×pos WOA +0.35×pos PSO +0.3×pos GWO , where pos WOA pos PSO pos GWO These are the optimal coordinates obtained by the three algorithms independently.

9. A modeling method for electromagnetic interference of a high-frequency driving circuit considering near-field coupling characteristics according to claim 1, characterized in that, The scaled-down model of the GIS is the COMSOL 500kV GIS scaled-down model.

Images