Asymmetric grounding grid topology intelligent generation and optimization method for high resistance area

Abstract

The application discloses a kind of asymmetric ground net topology intelligent generation and optimization method suitable for high resistance area.The ground net is abstracted as graph theory model, and mixed coding is used for the discrete connection of node continuous coordinates and adjacency matrix;The node position and connection are optimized by using swarm intelligence algorithm, and the high gravity node mechanism is introduced to guide the automatic matching of topological transient current discharge law of gas insulated switchgear;Through Laplacian matrix connectivity check and minimum spanning tree repair, the topological 100% connectivity is ensured;The double objective function of material economy index and transient security index is constructed to carry out multi-objective optimization.The application breaks through the rule grid constraint, realizes asymmetric topology, and greatly reduces the consumption of grounding material under the premise of meeting transient security.

Description

Technical Field

[0001] This invention relates to the field of power system grounding technology, and more specifically, to a method for intelligent generation and optimization of asymmetric grounding grid topology adapted to high-resistivity regions. Background Technology

[0002] When designing grounding systems for gas-insulated switchgear substations in areas with high soil resistivity, the grounding grid must simultaneously meet two core requirements: firstly, reducing power frequency grounding resistance, and secondly, suppressing high-frequency transient ground potential rise. The high-frequency ground currents induced by the rapid transient overvoltages generated during the operation of disconnecting switches or circuit breakers in gas-insulated switchgear exhibit a strong skin effect and localized concentration. The majority of the energy is concentrated in a small area around the grounding down conductor of the gas-insulated switchgear, causing a sharp increase in transient ground potential in this area, seriously threatening the insulation safety of secondary equipment and the stable operation of the system.

[0003] Currently, existing grounding grid design technologies have significant limitations in addressing the aforementioned issues. The conventional approach is a regular rectangular grid design based on equally spaced equipotential bonding, relying on adjusting conductor spacing to achieve resistance reduction and voltage equalization. This design thinking is confined within the framework of a "regular grid." When further reducing grounding resistance and suppressing transient ground potential rise is needed, engineers often have no choice but to blindly densify the grid uniformly. However, excessively dense metallic conductors generate strong electromagnetic inductance and current shielding effects, i.e., the proximity effect, resulting in the large amount of newly added conductor material contributing very little to reducing potential rise, causing significant economic waste. Furthermore, shallow optimization schemes with unequal-spacing grids exist, which, while breaking the "equal spacing" principle, still maintain a physical topology of "horizontal, mutually orthogonal straight lines," failing to break free from the regular framework and unable to target and suppress localized transient potential rises. In high-resistivity areas, designers have long faced a fundamental contradiction: "for safety, infinite density is necessary" versus "dense conductors generate shielding effects, causing enormous waste."

[0004] Therefore, there is an urgent need in the existing technology for a grounding grid design method that can break through the topological constraints of regular grids and achieve a fundamental reconstruction of the topological connection relationship of the grounding grid, so as to save grounding materials to the maximum extent while meeting transient safety constraints. Summary of the Invention

[0005] This invention aims to provide a method for intelligent generation and optimization of asymmetric grounding grid topology adapted to high-resistivity regions, in order to solve the technical problems of existing grounding grid designs being limited to regular grid frameworks, unable to target and match transient current spatial discharge patterns, and unable to achieve topology-level material consumption reduction while ensuring transient safety.

[0006] This invention proposes an intelligent generation and optimization method for asymmetric grounding grid topologies adapted to high-resistivity regions. This method breaks through conventional geometric constraints, abstracting the grounding grid into a graph theory model. Through a hybrid coding mechanism and an improved heuristic intelligent algorithm, the computer automatically generates the asymmetric grounding grid topology best suited to the current electrical environment. The swarm intelligence optimization algorithm includes, but is not limited to, evolutionary algorithms with population iteration characteristics such as particle swarm optimization, genetic algorithms, differential evolution algorithms, and gray wolf optimization algorithms. Specifically, it includes the following core components:

[0007] Abstracting the grounding grid into a graph theory model ,in This is a set of nodes, representing the floating virtual nodes in the grounding grid; Let be the set of edges, representing the metal conductor connections between nodes. Each node... This corresponds to a continuously adjustable three-dimensional coordinate variable within a three-dimensional spatial constraint boundary. The connections between nodes are represented by a Boolean adjacency matrix. Indicates matrix elements satisfy: Represents a node With nodes Lay metal conductors between them. Represents a node With nodes There are no connections between them; the grounding grid is an undirected graph, and the adjacency matrix satisfies symmetry constraints. And diagonal elements .

[0008] Based on this, a smart optimization algorithm is used to jointly iteratively update the continuous node coordinate variables and the discrete adjacency matrix variables. The continuous coordinate variables are updated using a velocity-position update formula; the discrete adjacency matrix variables utilize a mapping function to map velocity to connection probability, and when the generated random number is less than this connection probability, the algorithm is set to... Otherwise .

[0009] This hybrid coding method breaks through the traditional structural presupposition of "horizontal and vertical, orthogonal regular grids" in grounding grid design. It expands the optimization space of the topology from a finite number of scalar parameters to a combined space of continuous coordinates and discrete connections, enabling the grounding grid topology to evolve in any asymmetric form. This provides a coding foundation for achieving true topology-level global optimization. Furthermore, through continuous-discrete dual-domain joint coding, the algorithm can simultaneously optimize two orthogonal dimensions: "where the metal conductors are laid" and "how the metal conductors are connected," significantly expanding the morphological diversity of feasible topologies.

[0010] Within the aforementioned graph theory hybrid coding framework, a high-gravity node mechanism and distance penalty weight are introduced based on the spatial location of the gas-insulated combined electrical equipment (GEE) down conductors to forcibly guide nodes closer to the coordinate set of the GEE down conductors. This generates higher density metallic conductor connections in the region. Specifically, it computes nodes. Coordinate set of down conductors for gas-insulated combined electrical equipment minimum Euclidean distance Apply distance penalty weights during the connection probability determination process. ,in The distance attenuation coefficient is used to adjust the density.

[0011] This mechanism mathematically embeds the spatial discharge law of transient current "central accumulation and edge attenuation" in gas-insulated combined electrical equipment into the topology generation process. This allows the asymmetric topology automatically generated by the algorithm to accurately match the physical discharge path of the transient current. It strengthens the discharge path in the area around the down conductor of the gas-insulated combined electrical equipment where the current density is concentrated, and reduces the laying of metal conductors in the edge area where the current density is low. This effectively suppresses the rise of local transient ground potential and solves the structural misalignment problem of traditional uniform grounding grids, which is that "there is not enough metal in the central area where discharge is needed, and there is an excess of metal in the edge area where discharge is not needed."

[0012] After the algorithm iterates and mutates to generate new topological individuals, the current adjacency matrix is ​​calculated. The corresponding Laplace matrix ,in The degree matrix is ​​calculated by determining the eigenvalues ​​of the Laplacian matrix or the number of connected components of the graph. Determine physical connectivity. When detected... This indicates the existence of isolated electrical breaks in the topology, triggering a graph theory repair mechanism. The minimum spanning tree algorithm is used to establish new metallic conductor connections between paired connected components based on the minimum coordinate distance, forcibly modifying the corresponding adjacency matrix elements. Until the number of connected components in the entire graph is reached. This ensures that each generated asymmetric topology scheme meets the physical connectivity requirements of electrical engineering.

[0013] This repair mechanism solves the technical problem that intelligent optimization algorithms are prone to generating islands or dead ends when generating irregular grounding network topologies, leading to algorithm failure. It ensures that every asymmetric grounding topology scheme generated during algorithm iteration meets the 100% electrical connectivity requirement. By eliminating invalid solution spaces, it effectively avoids the optimization process from stalling, significantly improving the convergence speed and global search success rate of the three-dimensional irregular grounding network multi-objective optimization algorithm.

[0014] A multi-objective fitness function system considering both material economy and transient safety is constructed. Objective function one is the material economy index. The physical total length of all connected conductors in the network is accurately calculated using the adjacency matrix and node coordinates; the second objective function is the transient security index. The extreme values ​​of the time-domain transient potentials of all nodes under transient current impact at the entire station were calculated through full-wave electromagnetic transient simulation. The mathematical form of the multi-objective optimization is to simultaneously minimize material consumption and transient potential rise:

[0015]

[0016] Through a multi-objective optimization game model, the algorithm can automatically identify and eliminate redundant conductors that are ineffective in reducing transient ground potential rise, thereby significantly weakening the mutual inductance shielding effect between conductors while achieving the same electromagnetic transient protection performance. The biomimetic gradient topology generated by this method significantly reduces the consumption of grounding metal materials compared to traditional regular dense grids, achieving an optimal balance between transient ground potential rise suppression performance and grounding material consumption, resulting in significant engineering and economic benefits.

[0017] After the maximum number of iterations After iterative evolution, the multi-objective algorithm converges, outputting a set of Pareto front solutions based on non-dominated sorting logic. For the user-selected final equilibrium scheme, the algorithm executes a decoding procedure to extract the continuous variable matrix of the corresponding individuals. and discrete variable matrix ,in accordance with The element with a median value of 1 will The corresponding coordinate point pairs are converted into spatial line segment data structures, which serve as grounding grid topology schemes that can be directly used for engineering implementation.

[0018] Pareto front solution sets provide designers with a complete trade-off curve between transient safety and material economy. Designers can flexibly choose the optimal balance scheme according to the actual needs of the project, which significantly improves the engineering adaptability and decision-making scientificity of grounding grid design.

[0019] This invention also provides an intelligent generation and optimization system for asymmetric grounding grid topology, comprising: a three-dimensional constraint space modeling module for constructing a three-dimensional physical constraint space of a substation and determining the coordinate set of down conductors of gas-insulated combined electrical equipment; a graph theory topology encoding module for establishing a hybrid encoding of the node set and adjacency matrix of the grounding grid; a transient simulation evaluation module for calculating the transient ground potential rise index of each topological individual; a swarm intelligent optimization iteration module for performing swarm intelligent iterative updates combining a high-gravity node mechanism and a connectivity repair operator; and a Pareto decoding output module for outputting the Pareto front solution set and decoding it into an engineering-usable grounding grid topology scheme.

[0020] This hardware-based and systematic implementation architecture enables the method to be directly deployed in grounding grid design software platforms or engineering workstations, achieving full-process automation from modeling, coding, simulation, optimization to output, and significantly improving the efficiency and automation level of grounding grid design.

[0021] This invention also provides a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps of the above-described method for intelligent generation and optimization of asymmetric grounding grid topology adapted to high-resistivity regions. This storage medium includes, but is not limited to, non-transitory media such as ROM, RAM, hard disk, solid-state drive, USB flash drive, CD-ROM, and DVD. Through this storage medium, the method of this invention can be distributed, deployed, or embedded in grounding grid design tools in software form, facilitating practical engineering applications.

[0022] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 The overall flowchart of the intelligent generation and optimization method for asymmetric grounding grid topology provided in the embodiments of the present invention is shown.

[0025] Figure 2 This is a schematic diagram of a graph theory hybrid coding structure.

[0026] Figure 3 This is a flowchart of the graph theory adaptive topology repair operator.

[0027] Figure 4 This is a schematic diagram of the asymmetric topology generation principle based on the transient current discharge law.

[0028] Figure 5 Framework diagram for multi-objective Pareto optimization.

[0029] Figure 6 This is a comparison diagram of three grounding grid topology schemes.

[0030] Figure 7 The topology detail diagram for scheme C is shown (colored by clustering according to the downleader).

[0031] Figure 8 The convergence curve for the multi-objective optimization of NSGA-II.

[0032] Figure 9 Optimizing the Pareto frontier for multiple objectives ( - (Trade-off curve) diagram.

[0033] Figure 10 for Spectral characteristics ( (Hz wideband comparison chart) Detailed Implementation

[0034] 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 with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0035] It should be noted that, unless otherwise specified, the embodiments and features described in this invention can be combined with each other.

[0036] Implementation Method 1

[0037] like Figure 1-5 As shown, this embodiment provides an intelligent generation method for asymmetric grounding grid topology adapted to high-resistivity regions. This method breaks through conventional geometric constraints, abstracts the grounding grid into a graph theory model, and generates asymmetric grounding grid topology by computer through a hybrid coding mechanism.

[0038] Step 1: Multi-dimensional constrained space modeling based on the layout of gas-insulated combined electrical equipment and the characteristics of high-resistivity soil

[0039] First, construct the three-dimensional physical constraint space of the substation. Let the three-dimensional spatial boundary of the substation grounding grid be... ,in and These represent the minimum and maximum coordinates in space along the X-axis, respectively. and These represent the minimum and maximum coordinates in space along the Y-axis, respectively. and These represent the minimum and maximum coordinates of the space along the Z-axis, respectively. The boundary of this constrained space is jointly determined by the substation's land area boundary and the depth range of the grounding grid.

[0040] The precise physical location of the gas-insulated combined electrical equipment and the coordinate set of all high-frequency transient current injection points, i.e., the grounding down conductors, are defined as... ,in , This specifies the number of down conductors for the gas-insulated switchgear. The coordinates of each down conductor are accurately obtained from the substation's gas-insulated switchgear layout construction drawings.

[0041] Set hard constraints for the algorithm: That is, the transient ground potential rise value at any node of the entire station must not exceed the safety threshold. The safety threshold is determined based on the insulation withstand level of the secondary equipment, and is typically set between 1.5kV and 3.0kV.

[0042] Step 2: Graph Theory-Based Asymmetric Topology Node-Edge Hybrid Encoding for Grounding Grids

[0043] Abandoning the traditional grid concept of "straight-line penetration," the grounding grid is abstracted into a graph theory model. ,in For a set of nodes, Let it be the set of edges.

[0044] Define the set of floating virtual nodes within the substation ,in This represents the number of nodes. Each node corresponds to a three-dimensional coordinate variable. This coordinate is continuously adjustable within the three-dimensional physical constraint space boundary, i.e., it satisfies... , , Number of nodes The value is set according to the substation size and optimization accuracy requirements, and is typically between 20 and 200. As an engineering guideline, for 110kV substations... A value of 30-50 is recommended; for substations of 330kV and above, We recommend a range of 80-120. The larger the value, the higher the degree of freedom in topological form, but the computational cost increases accordingly.

[0045] definition Boolean adjacency matrix To represent edge sets Matrix elements Represents a node With nodes The laying status of the metal conductors between them: Represents a node With nodes A metal conductor is laid between them; Represents a node With nodes There are no connections between them. Since the grounding grid is an undirected graph, the adjacency matrix satisfies symmetry constraints. And diagonal elements This indicates that a node is not connected to itself.

[0046] Step 3: Topology generation mechanism based on transient current spatial discharge law

[0047] An improved swarm intelligence algorithm is used to jointly iteratively update the continuous node coordinate variables and the discrete adjacency matrix variables. In each iteration, the continuous coordinates and discrete connections are updated separately. The swarm intelligence algorithm includes, but is not limited to, one or more hybrid algorithms selected from particle swarm optimization, genetic algorithm, differential evolution algorithm, ant colony optimization, or gray wolf optimization algorithm.

[0048] For continuous coordinates The new node coordinates are calculated using the velocity-position update formula. Let the node... In the The speed at the next iteration is The location is The update process is as follows:

[0049] Among them, speed The update comprehensively considers the individual historical best position, the group historical best position, and the inertia term of the node. The coordinates of the new position need to be processed by boundary reflection to ensure that they do not exceed the range of the three-dimensional physical constraint space.

[0050] For discrete adjacency matrix We use a mapping function to map velocity to connection probability. The connection probability mapping function is defined as a sigmoid function:

[0051]

[0052] in, This represents the evolution rate variable corresponding to the elements of the adjacency matrix. If the generated random numbers... Then let Otherwise, let This probabilistic mapping mechanism transforms the continuous evolution rate into discrete topological connectivity decisions.

[0053] By introducing a high-gravity node mechanism and distance penalty weights, we can force the nodes to approach the coordinate set of the down conductors of the gas-insulated combined electrical equipment. This generates a higher density of metallic conductor connections in the region. Specifically, it allows the nodes to... Coordinate set of down conductors for gas-insulated combined electrical equipment The minimum Euclidean distance is The minimum Euclidean distance is defined as the node arrive The minimum Euclidean distance among all leader line coordinates:

[0054]

[0055] Apply distance penalty weights during the connection probability determination process. ,in This is the distance attenuation coefficient. The value is typically between 0.1 and 1.0. When the node is close to the down conductor of the gas-insulated switchgear... When the weight is small, The connection probability is unaffected when the node approaches 1; when the node is far from the downline... When the weight is large, As the value approaches zero, the connection probability is significantly reduced. This mechanism forces the connection closer together. area ( (Small) Generates higher density of metallic conductor connections, thereby guiding the generated asymmetric topology to automatically match the spatial discharge law of transient current.

[0056] This mechanism transforms the physical law of transient current "central accumulation and edge decay" into a guiding signal for topology generation at the algorithm level, so that the generated asymmetric topology automatically exhibits a spatial distribution characteristic of high-density connection in the area around the down conductor of the gas-insulated combined electrical equipment and low-density connection in the area away from it.

[0057] In this embodiment, the swarm intelligence algorithm can be one or more hybrid algorithms selected from particle swarm optimization algorithm, genetic algorithm, differential evolution algorithm, ant colony optimization algorithm, or gray wolf optimization algorithm.

[0058] Step 4: Topology Connectivity Verification and Adaptive Repair

[0059] After each iteration's mutation operation, the connectivity of the generated candidate topologies is checked. The current adjacency matrix is ​​then calculated. The corresponding Laplace matrix ,in The degree matrix, the matrix diagonal elements Equal to node The degree, that is Off-diagonal elements By calculating the Laplace matrix The eigenvalues ​​or the number of connected components of the graph can be calculated using a breadth-first search algorithm. To determine physical connectivity. Topological physical connectivity is... The criterion for judgment is: the Laplace matrix. The algebraic multiplicity of the zero eigenvalue is equal to 1.

[0060] When detected This indicates the presence of isolated electrical disconnections in the topology, triggering a graph theory repair mechanism. The repair process involves first identifying all connected component subsets. Then iterate through all connected component pairs. Calculate the minimum Euclidean distance between all pairs of nodes in any pair of connected components, i.e. Select the node pair corresponding to the minimum distance. Force modification of the corresponding adjacency matrix elements Merge two connected components into one connected component; repeat the above process until the total number of connected components in the graph is reached. Minimizing the length of the newly added metal conductor can be achieved using the Prim algorithm.

[0061] After the maximum number of iterations After iterative evolution, The value range is typically from 300 to 2000 generations, determined based on population size, problem complexity, and convergence criteria. For a conventional 110kV substation, Satisfactory convergence is usually achieved. From the final population, the optimal asymmetric topology individuals that satisfy the transient ground potential rise constraint are selected. Based on the elements with a value of 1 in their continuous coordinate variable matrix and discrete adjacency matrix variables, the corresponding coordinate point pairs are transformed into spatial line segment data structures, and the output is an asymmetric grounding grid topology scheme that can be directly used in engineering.

[0062] Implementation Method 2

[0063] This embodiment, based on embodiment one, further details the specific implementation of the graph theory connectivity verification and adaptive topology repair.

[0064] Specific implementation of connectivity test

[0065] The connectivity of the grounding grid topology is a fundamental prerequisite for ensuring its electrical function. A disconnected grounding grid implies the existence of isolated metal conductor segments, preventing current flow between all grounding electrodes. This significantly increases grounding resistance and potential rise, and in severe cases, is equivalent to grounding system failure. During the generation of asymmetric topologies using swarm intelligence algorithms, the random mutation operation of the connectivity matrix easily generates infeasible solutions containing multiple isolated subnets. Therefore, rigorous connectivity checks must be performed on each generation and each individual.

[0066] For a given adjacency matrix First, calculate the graph. Laplace matrix Degree matrix The calculation method is as follows: For , All other elements are set to zero. Laplace matrix. for A real symmetric positive semi-definite matrix whose algebraic multiplicity of zero eigenvalue is equal to the figure. The number of connected components. Therefore, by calculating the Laplacian matrix. The connectivity of a topology can be accurately determined by finding the zero eigenvalue multiplicity. In practical algorithm implementation, numerically stable singular value decomposition (SVD) or Lanczos iterative algorithm can be used to calculate the eigenvalues. Let... The threshold is usually a very small positive number. Statistical characteristic values satisfy The number of connected components is the number of connected components. .

[0067] As an alternative to eigenvalue analysis, a breadth-first search algorithm can also be used to count connected components. Starting from any node in the graph, a breadth-first search is performed, marking all reachable nodes. If any unmarked nodes exist, a new starting point is chosen from among the unmarked nodes to continue the search. This process is repeated until all nodes are marked. The number of rounds of the search is the number of connected components. .

[0068] Specific implementation of adaptive topology repair

[0069] When the number of connected components is detected When an isolated subnet is deemed infeasible, it means that the current individual topology is not feasible from an electrical engineering perspective. This invention does not simply discard it, but instead repairs it into a connected topology using a graph-theoretic adaptive repair operator. The repair principle is: to achieve full graph connectivity with minimal additional material cost, i.e., to find the shortest metallic conductor path required to connect all isolated subnets.

[0070] The specific process of the repair algorithm is as follows:

[0071] Step 1: Identify all connected components. Based on the connected component information identified by the connectivity test, ... Each node is divided into disjoint subsets The nodes within each subset form a connected subgraph.

[0072] Step 2: Calculate the minimum distance matrix between components. For any two distinct connected components... and Calculate the minimum Euclidean distance between them. And record the corresponding node pair Therefore, a The symmetric distance matrix represents the shortest connection distance between each connected component.

[0073] Step 3: Perform minimum spanning tree join. Using each connected component as a vertex and the minimum distance matrix as the edge weight matrix, Prim's algorithm is executed to construct a minimum spanning tree. The minimum spanning tree will generate... A new edge connection is created, with each edge corresponding to a pair of nodes that should be connected by a metal conductor.

[0074] Step 4: Force modification of the adjacency matrix. For each new edge determined by the minimum spanning tree. In the adjacency matrix Forced settings At the same time, set according to symmetry .

[0075] Step 5: Verify the repair results. Recalculate the Laplacian matrix. The number of zero eigenvalue multiplicity or a breadth-first search is performed to confirm the number of connected components in the repaired topology. If the connectivity requirement is not met, the process is redirected back to step one and repeated until the requirement is met.

[0076] Minimizing the length of the newly added metallic conductor is achieved using Prim's algorithm. Specifically, with... Each connected component is a vertex ( Typically 2 to 5), initial connected components As the root node of MST, the rest Each connected component is marked as unvisited; in each iteration, among all candidate edges from the set of visited connected components to the set of unvisited connected components, the edge with the smallest edge weight (i.e., the pairwise minimum Euclidean distance) is selected. ,mark For those already visited, and The nearest node pair between Corresponding adjacency matrix elements Set it to 1. Repeat the above process. Next, we obtain the MST of the connected complete graph, and... Isolated connected components are merged into a single connected component. The integration of operators in the iterative framework is improved.

[0077] The graph-theoretic adaptive repair operator is seamlessly embedded in the iterative loop of the swarm intelligence algorithm, specifically after the dual-domain position / velocity update operator and before the fitness function evaluation. That is, after each generation completes continuous coordinate updates and discrete connectivity updates, connectivity verification and adaptive repair operations are immediately performed on each generated topology individual. Only after repair is the topology individual sent to the transient simulation evaluation module for fitness calculation. This architectural design ensures that the fitness function is always evaluated within the feasible solution space, avoiding the waste of computational resources on evaluating infeasible topology individuals.

[0078] Implementation Method 3

[0079] This embodiment, based on Embodiment 1 or Embodiment 2, further details the construction of the multi-objective fitness function.

[0080] Objective function 1: Material economy index Specific construction

[0081] The economic efficiency of a grounding grid is primarily determined by the total length of the metallic conductors laid. The total physical length of all connected conductors in the network can be accurately calculated using the adjacency matrix and node coordinates. Material economic indicators. Defined as:

[0082]

[0083] in, These are elements of the adjacency matrix, with values ​​of 0 or 1. For nodes and nodes The square of the Euclidean distance between them. The unit is a unit of length, such as meter.

[0084] From an engineering perspective, since the adjacency matrix satisfies and The above double summation only covers the upper triangular region of the matrix, and the conductor length between each pair of nodes is counted only once, ensuring the accuracy of the material calculation.

[0085] Objective function 2: Transient safety index Specific construction

[0086] Transient safety indices are calculated using full-wave electromagnetic transient simulation. First, based on nodal coordinates... and adjacency matrix The spatial geometric arrangement of the grounding grid conductors is determined; then, the Sommerfeld integral equation is solved based on the method of moments (MoM) to calculate the grounding grid's spatial arrangement. Impedance matrix over a wide frequency range of Hz ,in The first factor is the angular frequency; secondly, based on the characteristics of the rapid transient overvoltage generated during the operation of gas-insulated combined electrical equipment, a time-domain transient current excitation injected into the grounding grid is constructed. Finally, the solution is obtained through the frequency domain. Then, perform an inverse Fourier transform, or directly use the finite-difference time-domain method to solve the whole-wave problem, to obtain the time-domain transient potential response of all nodes in the station. , .

[0087] It should be noted that during the algorithm iteration process, each fitness evaluation requires calling electromagnetic transient simulation calculation. To accelerate iteration convergence, one or a combination of the following acceleration strategies can be adopted: (1) Establish a surrogate model based on deep neural networks or Gaussian processes, and use a small number of accurate simulation samples for training to replace full-wave simulation for rapid evaluation; (2) Use frequency domain fast algorithms, such as adaptive cross approximation or fast multipole method, to accelerate impedance matrix filling and solution; (3) Utilize a population parallel evaluation framework to distribute the simulation tasks of multiple topological individuals to multi-core CPUs or GPUs for parallel computing. This invention does not depend on a specific electromagnetic transient simulation method and can be implemented using existing commercial software (such as CDEGS, PSCAD) or self-developed fast algorithms. Transient safety indicators Defined as the extreme value of the time-domain transient potential of all nodes in the entire station during the simulation period:

[0088]

[0089] in Represents a node exist The absolute value of the transient ground potential rise at any given moment. The unit is voltage, such as volt or kilovolt. Objective function It is the grounding grid topology and The function, through and The results are obtained after determining the physical geometry and through the electromagnetic transient simulation process described above.

[0090] Multi-objective optimization definition

[0091] The system objective is to simultaneously minimize material consumption and transient potential rise. The mathematical form of the multi-objective optimization is:

[0092]

[0093] The constraints are satisfied: , , , And the number of connected components .

[0094] because and There is a conflict between the two objectives: reducing transient potential rise usually requires increasing the amount of grounding material, while reducing material usage may lead to an increase in transient potential rise. Therefore, the optimization result cannot be fully described by a single optimal solution, necessitating the use of the Pareto multi-objective optimization framework. One solution... Dominate another solution If and only if the solution Non-inferior to the solution on all objectives And it is strictly superior to the solution at least in one objective. The Pareto front consists of all non-dominated solutions that are not dominated by other solutions. Each point on the front represents a point in the Pareto front. and An unimprovable balance between them.

[0095] Implementation Method 4

[0096] This embodiment, based on embodiment three, further details the output of Pareto front decoding and the optimal asymmetric topology scheme.

[0097] Non-dominated sorting and crowding distance calculation

[0098] go through After iterative evolution, the multi-objective algorithm converges. The value range is usually from 300 to 2000 generations, depending on the population size, problem complexity, and convergence criteria.

[0099] Perform a non-dominated sort on every individual in the final population. The sorting process is as follows: for any two individuals in the population... and Determine the dominance relationships among their multi-objective fitness values. If an individual Less than or equal to the individual in both objectives If the corresponding target value is strictly less than 0.05 at least in one target, then the definition is... Dominate Iterate through all pairs of individuals and count the number of times each individual is dominated. and the set of other individuals controlled by each individual. .

[0100] All times dominated The individuals constitute the Pareto frontier level .Will After temporarily removing individuals from the hierarchy, the number of times they are dominated is recounted to obtain the secondary frontier. This process continues until all individuals are stratified. Within the same frontier level, each target is sorted by its crowding level; a larger crowding distance indicates that the individual is more sparse in the target space and has a higher priority for being selected and retained.

[0101] Final solution output and implementation format

[0102] For the final balancing scheme selected by the user based on project requirements, the algorithm executes a decoding procedure. The specific steps are: extracting the continuous variable matrix for the corresponding individual variables. and discrete variable matrix ; Traversal All of the above satisfy and element pairs ,Will The corresponding coordinates and It is transformed into a spatial line segment data structure, which includes the starting point coordinates, ending point coordinates, conductor specifications, and burial depth information.

[0103] The final output grounding grid topology scheme is presented in a standard format and includes the following information categories: overall description of the grounding grid, including the total number of nodes, the total number of conductor segments, the estimated total length of materials, the expected transient ground potential rise and safety margin; a node coordinate table, listing the three-dimensional coordinate values ​​of all nodes; a conductor connection table, listing the first and last node numbers and corresponding lengths of all conductor segments; and a construction reference drawing, providing spatial wireframe model data that can be imported into computer-aided design software.

[0104] Implementation Method 5

[0105] This embodiment provides an intelligent generation and optimization system for asymmetric grounding grid topology adapted to high-resistivity regions. This system is used to execute the intelligent generation and optimization method for asymmetric grounding grid topology described in any of the above embodiments. This system is deployed in a grounding grid design software platform or engineering workstation and can be implemented in software or partially or entirely in hardware.

[0106] This system comprises the following five core modules:

[0107] 3D Constrained Space Modeling Module. This module is responsible for constructing a 3D physical constraint space based on the actual geographical information of the substation and the layout data of the gas-insulated switchgear. Input data includes substation land boundary data, coordinate data of the grounding down conductors of the gas-insulated switchgear, soil resistivity stratification data, and transient ground potential rise safety threshold. The module outputs a standardized 3D constrained space description file, including the spatial boundary coordinate range and the set of leader line coordinates. And safety thresholds. This module also integrates high-resistivity soil resistivity data processing functions, supporting seasonal variation correction and stratified modeling of soil resistivity.

[0108] Graph Theory Topology Encoding Module. This module is responsible for building the graph theory model of the grounding grid and performing hybrid encoding. The module receives constraint space data output from the 3D constraint space modeling module and the number of nodes set by the user. The module performs the following function: randomly initializes within the three-dimensional constrained space. The coordinates of each node are used to generate an initial node set. Generate the initial adjacency matrix. This includes generating a connected initial topology using the minimum spanning tree algorithm, applying a higher initial connection probability near the down conductor of the gas-insulated combined electrical equipment using a high-gravity node mechanism, and performing connectivity checks and necessary repairs on the initial encoding. The module output is a standardized topology-encoded data packet in the following format: .

[0109] Transient simulation evaluation module. This module is responsible for calculating the transient ground potential rise index and material economy index for each topological individual. The module integrates an electromagnetic calculation engine based on the Method of Moments (MoM), supporting broadband impedance matrix calculation and time-domain simulation. To accelerate the large number of repetitive calculations during the iteration process, this module has a built-in parallel evaluation interface and an optional surrogate model for fast prediction. The module receives the topology encoding scheme from the swarm intelligence optimization iteration module and outputs the scheme's... Indicator values ​​and Indicator value.

[0110] The swarm intelligence optimization and iterative module is the core computational module of the system, responsible for executing swarm intelligence iterative updates that combine a high-gravity node mechanism and a connectivity repair operator. The module embeds several optional swarm intelligence algorithm frameworks, including particle swarm optimization, genetic algorithms, and differential evolution algorithms, allowing users to choose according to the specific characteristics of the problem. The module's core sub-functions include: population initialization and encoding preprocessing; a continuous coordinate update submodule, performing velocity-position update operations; a discrete connection update submodule, performing probabilistic mapping operations based on a sigmoid function; a high-gravity node guidance submodule, calculating the distance from each node to the gas-insulated combined electrical equipment down conductor and applying penalty weights; and a connectivity verification and adaptive repair submodule, detecting and repairing topological connectivity.

[0111] The Pareto decoding output module is responsible for performing non-dominated sorting on the final population after iterative convergence, outputting the Pareto front solution set, and performing topology decoding according to user selection. The core sub-functions of this module include: a non-dominated sorting submodule, which determines the dominance relationship and front stratification of the final population; a Pareto front display submodule, which graphically displays the distribution of front points; and a topology decoding submodule, which decodes the selected encoding scheme into a standardized grounding grid topology and outputs a construction drawing format file.

[0112] Data flow relationships between modules

[0113] like Figure 1 As shown, the output data of the 3D constrained space modeling module is passed to the graph theory topology encoding module. The initial population encoding data output by the graph theory topology encoding module is passed to the swarm intelligence optimization iteration module. The swarm intelligence optimization iteration module completes the fitness calculation of each individual by calling the transient simulation evaluation module. The iterative convergence result of the swarm intelligence optimization iteration module is passed to the Pareto decoding output module to complete the final solution output.

[0114] Key technical parameters of this system

[0115] Recommended configuration: Number of nodes The population size is 50 to 200 topological individuals, with a maximum number of iterations of 30 to 100. The distance attenuation coefficient ranges from 500 to 2000 generations. The connectivity threshold ranges from 0.3 to 0.8. for Transient ground potential rise safety threshold The voltage ranges from 1.5kV to 3.0kV. All of the above parameters are adjustable and can be flexibly adjusted according to specific engineering conditions.

[0116] Hardware deployment method of this system

[0117] This system relies on a computing server with multi-core parallel processing capabilities. The population individual evaluation of the swarm intelligence optimization iteration module is inherently parallel, allowing population individuals to be distributed to different computing cores for parallel evaluation, significantly shortening the computation time of a single iteration.

[0118] Through the collaborative work of the above five modules, this system achieves full-process automation from constraint space modeling, topology coding, dual-objective evaluation of transient security and material economy, swarm intelligence optimization iteration to scheme decoding output. It can provide a scientific and efficient auxiliary decision-making tool for the design of substation grounding grids for gas-insulated combined electrical equipment in high-resistivity areas.

[0119] Implementation Method Six

[0120] This embodiment provides a computer-readable storage medium storing computer program instructions that, when executed by one or more processors, can implement all or part of the steps of the intelligent generation and optimization method for asymmetric grounding grid topology adapted to high-resistivity regions described in any of the foregoing embodiments.

[0121] Specifically, the storage medium can be a non-volatile storage medium, such as magnetic storage media (hard disks, magnetic tapes, etc.), optical storage media (DVDs, CD-ROMs, etc.), solid-state storage media (SSDs, USB flash drives, etc.), and various read-only memories (ROMs, EPROMs, EEPROMs, etc.) or random access memories (RAMs). The processor can be a general-purpose central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices.

[0122] When the instructions are loaded into the memory of a computer or embedded device and executed by a processor, the computer or device is able to perform the following operations: establish a graph theory hybrid coding model for the grounding grid; jointly optimize node coordinates and adjacency matrices using a swarm intelligence algorithm; introduce a high-gravity node mechanism to guide topology matching and transient current discharge law; perform connectivity verification and use the Prim algorithm for minimum spanning tree repair; calculate material economy indicators and transient safety indicators; perform multi-objective Pareto optimization; and output the grounding grid topology scheme.

[0123] This storage medium can be provided to users separately or distributed as part of the grounding grid design software. Using this medium, engineers can perform intelligent generation and optimization design of asymmetric topology for substation grounding grids in high-resistivity areas on a general-purpose computer without connecting to specific hardware.

[0124] Implementation Method Seven

[0125] This embodiment uses a 110kV GIS substation in Shaanxi Province as a specific engineering case to verify in detail the implementation effect of the aforementioned method of the present invention.

[0126] I. Calculation Example Scenario Setup

[0127] 1.1 Project Background

[0128] This implementation example uses a 110kV GIS substation in Shaanxi Province as the research object. Located in the Guanzhong Plain, the substation is situated in an area dominated by loess soil, which has a significantly higher resistivity than the alluvial plains of eastern Shaanxi, representing a typical high-resistivity soil environment. The substation's main wiring is a single busbar segmented system, equipped with two main transformers and four 110kV outgoing lines. All high-voltage equipment utilizes GIS (Gas Insulated Switchgear).

[0129] Operation of GIS equipment (opening and closing of disconnecting switches, operation of circuit breakers) generates rapid transient overvoltages (VFTOs), with a spectrum concentrated between 100kHz and 100MHz, a rise time of approximately 1ns, and a fall time of approximately 50ns. High-frequency current flows through the GIS metal casing and into the grounding grid via the grounding down conductor, inducing severe high-frequency transient ground potential rise (TGPR) locally, threatening the insulation safety of secondary equipment. In areas with high soil resistivity, conventional, evenly spaced, regular grid grounding grids are insufficient to effectively suppress TGPR. Traditional methods rely on denser grids, resulting in material waste and electromagnetic shielding effects between adjacent conductors, causing the marginal contribution of newly added conductors to TGPR reduction to decay rapidly.

[0130] 1.2 Soil electrical parameters

[0131] This example uses a two-layer horizontal stratified soil model: the top 0-5m layer is weathered loess with a resistivity ρ1 = 500 Ω·m; below 5m is the original loess layer with a resistivity ρ2 = 3000 Ω·m. The equivalent apparent resistivity ρ_eff = 2000 Ω·m, which falls within the high-resistivity soil range (ρ > 1000 Ω·m) defined in GB / T50065-2011.

[0132] 1.3 Site geometry and location of GIS equipment down conductors

[0133] The grounding grid is designed for a rectangular site. burial depth The conductor is made of galvanized flat steel with a cross-section of 50mm² (radius). The site has four GIS equipment downlines, with their plane coordinates as follows: , , , (Unit: m; corresponding to main transformer incoming line #1, main transformer incoming line #2, outgoing line #1, and outgoing line #2). These four down conductors are the main injection points for high-frequency transient currents and the core area for the automatic asymmetric topology aggregation of conductor density in this invention.

[0134] 1.4VFTO transient excitation parameters

[0135] The VFTO transient current excitation in this example is constructed using a double exponential function: The rising time constant Falling edge time constant Amplitude (Baseline is) Phase voltage peak value). Excitation spectrum coverage. .

[0136] II. Three Scheme Settings

[0137] To verify the superiority of the method of this invention, this example constructs three grounding grid topology schemes for comparison: Scheme A is a traditional equal-spacing regular grid, serving as a baseline; Scheme B is an unequal-spacing exponentially compressed grid, reflecting the shallow optimization approach; Scheme C is the asymmetric graph theory topology of this invention.

[0139] Option A: Evenly spaced regular grid. (According to...) The nodes are evenly spaced throughout the site, with 13 columns arranged horizontally. ), arranged in 17 rows vertically ( There are 221 nodes in total. Each node is connected only to its four nearest neighbors (east, south, west, and north), forming a horizontally and vertically aligned orthogonal grid.

[0140] Option B: Unequal-spacing exponentially compressed mesh. In the central region ( , (Covering the areas where 4 GIS downlines are located) The nodes are densely spaced, and in the outer ring area (edge ​​of the site) according to... – The nodes are spaced sparsely. This scheme embodies the shallow optimization concept of "dense at the center and sparse at the edges".

[0141] Option C: The asymmetric topology of this invention. The total number of nodes is generated using the method described above. The node positions are continuously adjustable within the three-dimensional constrained space of the site, using an adjacency matrix. This represents the metallic conductor connections between nodes. Population size: 100; maximum number of iterations: [not specified]. Distance attenuation coefficient After multi-objective optimization by NSGA-II, the nodes form a locally high-density spider web-like connection near the four GIS down conductors, while the conductor density is significantly reduced in the edge area of ​​the site, showing a clear non-orthogonal, gradual biomimetic topology.

[0142] III. Algorithm Implementation Steps

[0143] Step 1: 3D Constraint Space Modeling Based on GIS Layout and High-Resistivity Soil Characteristics. Establish a three-dimensional physical constraint space. .exist Internal random generation There are candidate nodes, each node satisfy Constraints; Coordinates of 4 GIS Leader Lines As a high-gravity node.

[0144] Step 2: Graph-based asymmetric topology node-edge hybrid encoding of the grounding grid. The grounding grid is encoded as a graph. ,in It is a 100-dimensional continuous node coordinate matrix. for Boolean adjacency matrix. Continuous variables (node ​​coordinates) are updated using the velocity-position formula:

[0145]

[0146] Discrete variables (adjacency matrix) are mapped using the Sigmoid function:

[0147]

[0148] continuous velocity variable Mapped to connection probabilities, determined through random judgment. The value of .

[0149] Step 3: High-gravity node mechanism based on GIS downline spatial location. For each node... Calculate the minimum Euclidean distance from it to the four GIS downlines:

[0150]

[0151] Apply distance penalty weights during the connection probability determination process:

[0152]

[0153] in The closer a node is to the GIS down conductor, the greater the probability of its adjacent matrix elements being connected. This guides the algorithm to automatically generate high-density metal conductor connections near the GIS down conductor, while automatically sparsening them in the far edge region.

[0154] Step 4: Graph Theory Connectivity Verification and Adaptive Repair Based on Prim's Algorithm. After the mutation operation in each algorithm iteration, a connectivity verification is performed on the generated candidate topologies: the adjacency matrix is ​​calculated. The corresponding Laplace matrix ( (for degree matrix), through Zero eigenvalue multiplicity or breadth-first search algorithm to calculate the number of connected components in a graph .when When this occurs, a graph theory adaptive repair mechanism is triggered: Prim's algorithm (using connected components as vertices and the minimum Euclidean distance between components as edge weights) is used to establish the shortest connection edges between each pair of connected components, forcibly modifying the corresponding adjacency matrix elements. Repeat the above process. Next, until the number of connected components in the entire graph is reached. Individuals sent for fitness assessment after repair are guaranteed to have 100% electrical connectivity.

[0155] Step 5: Calculation of the broadband impedance matrix based on the method of moments (MoM). For each candidate topology, solve the Sommerfeld integral equation using the method of moments (MoM) to calculate the grounding grid's impedance matrix. Impedance matrix over a wide frequency range :

[0156]

[0157] in This is a partial inductance matrix of a conductor. For frequency-varying resistance matrices that take into account the skin effect, This is the ground admittance matrix calculated using the Sommerfeld integral. The time-domain node potential response is obtained by inverse frequency-time transform (FFT / IFFT). And extract the extreme values ​​of transient ground potential rise for the entire station. This example uses the commercial software CDEGSHIFREQ module to solve the MoM problem.

[0158] Step Six: Multi-objective optimization and Pareto front decoding based on NSGA-II. Constructing material economic indicators. (Total physical length of all metallic conductors) and transient safety indicators (Transient ground potential rise extreme value) are two conflicting optimization objectives, which are jointly optimized using the non-dominated sorting genetic algorithm NSGA-II:

[0159]

[0160] NSGA-II key parameters: population size Maximum iteration algebra Crossover probability 0.9, polynomial mutation probability 0.1, crowding distance coefficient .go through After evolution, the output is a Pareto front consisting of 25 non-dominated solutions. During engineering implementation, the TGPR safety threshold constraint is applied. Select from Pareto fronts The smallest solution is taken as the final solution.

[0161] IV. Result Comparison

[0162] 4.1 Comparison of Topological Forms

[0163] like Figure 6 As shown, Scheme A presents a typical horizontal and vertical orthogonal grid with equal spacing, and the conductor density is the same near the four GIS down conductors and the edge area; Scheme B forms a dense grid with a spacing of 2m in the central area and a sparse grid with a spacing of 4-6m in the outer area, which has begun to reflect the idea of ​​"dense in the center and sparse at the edge", but the physical topology is still an orthogonal grid; Scheme C presents obvious non-orthogonal and asymmetrical features: a locally high-density spider web-like metal conductor connection is formed near the four GIS down conductors, while the conductor density is significantly reduced in the edge area of ​​the site, forming a gradual biomimetic topological structure, which completely breaks through the framework limitation of the regular grid.

[0164] 4.2 Scheme C Topology Details

[0165] like Figure 7 As shown, the topological details of scheme C show that a high density of local connections (spider web structure) is formed within each downlead cluster, and the clusters are connected by long "bridging" conductors; this structure can efficiently conduct the high-frequency transient current injected into each downlead while minimizing the total material consumption.

[0166] 4.3 NSGA-II Convergence Curve

[0167] like Figure 8 As shown, schemes A and B have no convergence process due to their fixed topology; scheme C evolves via NSGA-II. The initial convergence rate (2200m) rapidly decreased to approximately 1500m within 200 generations (initial convergence), and then stabilized at 1320m after 600 generations (fine convergence). The voltage decreased from 3.2kV ​​to approximately 1.9kV over 250 generations, and then stabilized at 1.8kV after 600 generations. During the convergence process, the two targets decreased in tandem, with no obvious conflicting oscillations.

[0168] 4.4 Pareto Frontier

[0169] like Figure 9 As shown, in – On the plane, 25 non-dominated solutions form a continuous tradeoff curve; the positions of schemes A, B, and C on the graph are respectively... , , All are located at or near the Pareto front. TGPR safety threshold. Option A's TGPR exceeds the threshold (unsafe), Option B is near the threshold (critical, no margin), and Option C is significantly below the threshold (safe, with margin). (Margin). In engineering design, the margin is usually selected from the Pareto front under conditions that satisfy safety constraints. The smallest solution, in this example, is to choose scheme C as the final implementation scheme.

[0170] 4.5 Spectral characteristics

[0171] like Figure 10 As shown, in Wide frequency range: In the power frequency band below 1kHz, the three schemes The difference is small (approximately 0.4–0.7 kV), mainly reflecting the difference in power frequency grounding resistance; in the 1 kHz–1 MHz intermediate frequency band, Scheme A... The voltage rises to approximately 3.2kV ​​(peak value) in scheme A, approximately 2.1kV in scheme B, and approximately 1.8kV in scheme C; in the high-frequency range above 1MHz, the equivalent resistance increases due to the skin effect. It rose further, but Option C remained at the lowest level.

[0172] 4.6 Comprehensive Comparison

[0173] Table 1 summarizes the comprehensive comparison results of the three schemes.

[0174]

[0175] V. Conclusion

[0176] This implementation method, through simulation verification of the grounding grid design of a 110kV GIS substation in Shaanxi Province, yields the following conclusions:

[0177] (1) The asymmetric grounding grid topology intelligent generation and optimization method described in this invention has advantages in material economy ( ) and transient security ( Significant synergistic optimization was achieved on two conflicting objectives: Solution C achieved significant synergistic optimization on two conflicting objectives. Under the condition of total conductor length (saving 40.0% compared to scheme A and 28.6% compared to scheme B), it was achieved The transient ground potential rise value was reduced by 43.8% compared to Scheme A and by 14.3% compared to Scheme B, and the safety threshold was met. Requirement, to be retained Engineering margin.

[0178] (2) The asymmetric grounding grid generated by Scheme C exhibits obvious irregular and non-orthogonal characteristics: a local high-density spider web-like connection of metal conductors is formed around the GIS down conductor, while the conductor density is significantly reduced in the edge area of ​​the site far from the down conductor, forming a gradual biomimetic topology. This morphology precisely matches the spatial discharge law of "central accumulation and edge attenuation" of VFTO transient current, effectively solving the structural misalignment problem of "insufficient metal in the central area that needs discharge and excessive metal in the edge area that does not need discharge" in traditional uniform grounding grids.

[0179] (3) Calculation of broadband impedance matrix based on method of moments (MoM) The combination of NSGA-II multi-objective optimization algorithm provides a complete automated solution for the design of grounding grids in GIS substations in high-resistivity areas. It can be directly applied to the optimization design of grounding grids in GIS substations of various voltage levels from 110kV to 500kV in actual engineering practice, and has significant engineering economic benefits and promotion and application value.

[0180] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural modifications made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for intelligently generating asymmetric grounding grid topology adapted to high-resistivity regions, characterized in that, Includes the following: Abstracting the grounding grid into a graph theory model ,in This is a set of nodes, representing the floating virtual nodes in the grounding grid; Let be the set of edges, representing the metal conductor connections between nodes; Define a set of nodes Each node Corresponding to a constraint boundary in three-dimensional space Internally continuously adjustable three-dimensional coordinate variables ; definition Boolean adjacency matrix Represents the metal conductor connections between nodes; matrix elements Represents a node With nodes Connection status between: This indicates that a metal conductor is laid between the two nodes. This indicates that the two nodes are not connected; the adjacency matrix satisfies the symmetry constraint. And diagonal elements ; The node coordinate variables and the adjacency matrix variables are jointly iteratively updated using a swarm intelligence optimization algorithm: the spatial position of the nodes is updated using a velocity-position update operation for continuous coordinate variables. The evolution velocity variable is mapped to a connection probability using a mapping function on the discrete adjacency matrix variable, and the connection state is determined by random determination based on this connection probability. The value of is used to generate an asymmetric grounding grid topology.

2. The intelligent generation method for asymmetric grounding grid topology according to claim 1, characterized in that, The mapping function for the connection probability is a sigmoid function: ,in For the elements corresponding to the adjacency matrix The evolution rate variable; if the generated random numbers Then let Otherwise .

3. The intelligent generation method for asymmetric grounding grid topology according to claim 2, characterized in that, The generation method also includes the following: introducing a high-gravity node mechanism based on the spatial location of the down conductor of the gas-insulated combined electrical equipment; and defining the coordinate set of the down conductor of the gas-insulated combined electrical equipment. ,in ; Calculate each node To the set of coordinates of the lead-in line minimum Euclidean distance Apply distance penalty weights during the connection probability determination process. , The distance attenuation coefficient and Forced guidance to approach the coordinate set of the down conductor of the gas-insulated combined electrical equipment area ( (Small) Generates higher density of metallic conductor connections, thereby guiding the generated asymmetric topology to automatically match the spatial discharge law of transient current.

4. A method for intelligent generation and optimization of asymmetric grounding grid topology adapted to high-resistivity regions, characterized in that, Based on the asymmetric grounding grid topology intelligent generation method according to any one of claims 1 to 3, it further includes graph theory connectivity verification and adaptive topology repair steps: After the mutation operation in each algorithm iteration, the connectivity of the generated candidate topologies is checked: the current adjacency matrix is ​​calculated. The corresponding Laplace matrix ,in For degree matrix, ; By calculating the Laplace matrix Zero eigenvalue multiplicity or breadth-first search algorithm to calculate the number of connected components in a graph ; When the number of connected components is detected When an isolated electrical disconnection is detected in the topology, the graph theory adaptive repair mechanism is triggered: Identify all connected component subsets ; Traverse all connected component pairs Calculate the minimum Euclidean distance between any two connected components: ； The minimum spanning tree algorithm is used to determine the new metal conductor connection edges that need to be added, and the nodes with the minimum distance are selected to forcibly modify the corresponding adjacency matrix elements. This merges two connected components. Repeat the above repair process until the total number of connected components in the entire graph is reached. .

5. The method for intelligent generation and optimization of asymmetric grounding grid topology according to claim 4, characterized in that, The minimum spanning tree algorithm is implemented using Prim's algorithm: Each connected component is used as a vertex, and the minimum Euclidean distance between connected components is used as the edge weight. Any connected component is initialized as the root node of the minimum spanning tree (MST). In each iteration, the edge with the smallest weight from all candidate edges between the visited set and the unvisited set is selected and added to the spanning tree. This process is repeated. This process merges all components, achieving topological connectivity repair with minimal additional material cost. The number of connected components.

6. The method for intelligent generation and optimization of asymmetric grounding grid topology according to claim 4, characterized in that, The connectivity verification and adaptive repair are embedded in the iterative loop of the swarm intelligence optimization algorithm, and are executed after the dual-domain position / velocity update operation and before the fitness function evaluation; that is, after each generation completes the node coordinate update and adjacency matrix connection update, connectivity verification and adaptive repair are immediately performed on the generated topology individuals, and the number of connected components is confirmed after repair. Individuals were sent to the fitness assessment stage.

7. A method for intelligent generation and optimization of asymmetric grounding grid topology adapted to high-resistivity regions, characterized in that, Based on the asymmetric grounding grid topology intelligent generation and optimization method according to any one of claims 4 to 6, it further includes the steps of constructing a multi-objective fitness function and performing multi-objective optimization: Constructing material economic indicators As the first objective function: ； Constructing transient safety indicators As the second objective function: ； in, To ensure that each node is under transient current excitation throughout the station The time-domain transient ground potential rise response was obtained through electromagnetic transient simulation calculation; The mathematical form of multi-objective optimization is to simultaneously minimize material consumption and transient ground potential rise: ； The two objective functions are jointly optimized by a multi-objective swarm intelligent optimization algorithm, which automatically identifies and eliminates redundant metal conductors that are ineffective in reducing transient ground potential rise.

8. The method for intelligent generation and optimization of asymmetric grounding grid topology according to claim 7, characterized in that, The transient security index The calculation includes: based on node coordinates and adjacency matrix Determine the spatial geometric arrangement of the grounding grid conductors; solve the Sommerfeld integral equation based on the method of moments (MoM) to calculate the grounding grid under 10... 0 ~10 8 Impedance matrix over a wide frequency range of Hz Based on the characteristics of rapid transient overvoltages generated during the operation of gas-insulated combined electrical equipment, a time-domain transient current excitation for injection into the grounding grid is constructed. The time-domain transient potential response of all nodes in the entire station is obtained through frequency-time domain transformation. And extract the transient ground potential rise extreme values ​​of the entire station.

9. The method for intelligent generation and optimization of asymmetric grounding grid topology according to claim 7, characterized in that, go through After iterative evolution, the final population is hierarchically sorted according to non-dominated sorting logic, outputting a set of Pareto front solutions; the following decoding operation is performed: extract the continuous variable matrix corresponding to the user-selected equilibrium scheme. and discrete variable matrix ,in accordance with The element with a median value of 1 will The corresponding coordinate point pairs are converted into spatial line segment data structures as the output of the grounding grid topology scheme.

10. A smart generation and optimization system for asymmetric grounding grid topology adapted to high-resistivity regions, characterized in that, include: The 3D constrained space modeling module is used to construct a 3D physical constrained space based on the actual geographical information of the substation and the layout data of the gas-insulated switchgear, and to determine the coordinate set of the down conductors of the gas-insulated switchgear. ; The graph theory topology coding module is used to build a graph theory model of the grounding grid and generate a set of nodes. and adjacency matrix The mixed encoding is performed, and the initial encoding is checked for connectivity and repaired as necessary. The transient simulation evaluation module is used to receive the topology coding scheme and calculate the material economy index of each topology individual. and transient safety indicators ; The transient simulation evaluation module uses the method of moments (MoM) to solve the grounding grid at 10 0 ~10 8 Impedance matrix over a wide frequency range of Hz ; The swarm intelligence optimization and iteration module is used to perform swarm intelligence iterative update operations. This swarm intelligence optimization and iteration module has a built-in continuous coordinate update submodule that performs speed-position update operations. The discrete connection update submodule performs probabilistic connection determination based on the mapping function; The high-gravity node guiding submodule calculates the set of coordinates from each node to the downleader. Find the minimum Euclidean distance and apply distance penalty weights. The connectivity verification and repair submodule detects the number of topologically connected components and triggers minimum spanning tree repair using the Prim algorithm when an island is detected. The Pareto decoding output module is used to perform non-dominated sorting and front layering on the final population after iterative convergence, output the Pareto front solution set, and perform topology decoding according to the user-selected balancing scheme. It transforms the optimized node coordinates and adjacency matrix into a spatial line segment data structure and outputs it as the final grounding network topology scheme.

11. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the instruction is executed by the processor, it implements the steps of the method for intelligent generation and optimization of asymmetric grounding grid topology adapted to high-resistivity regions as described in any one of claims 1 to 9.

Images