Substation layout method, device and system and medium

Abstract

The invention relates to the technical field of transformer substations, and provides a transformer substation layout method, device and system and a medium. According to the implementation scheme, a structured data set is determined based on all equipment layout parameters and all first working conditions of a transformer substation; based on the structured data set, performing initial training on a preset conditional generative adversarial network to generate a plurality of initial layout design drawings; generating a finite element physical model based on the plurality of initial layout design drawings; solving the finite element physical model to obtain a coupling solving result; extracting performance indexes based on the coupling solution result to construct a comprehensive performance evaluation function; and on the basis of the comprehensive performance evaluation function, updating parameters of generators in the conditional generative adversarial network to obtain a target generator, and generating a target layout design drawing through the target generator. According to the embodiment of the invention, the substation layout performance can be improved.

Description

Technical Field

[0001] This invention relates to the field of substation technology, and in particular to a substation layout method, apparatus, system and medium. Background Technology

[0002] Against the backdrop of the continuous advancement of power system modernization and intelligentization, substations, as key hubs in the power transmission and distribution process, have their layout and design directly affecting system operational safety, energy efficiency, and maintenance costs. With the widespread application of tools such as Computer-Aided Design (CAD), Finite Element Analysis (FEA), and electrical simulation software, including Electrical Transient Analyzer Program (ETAP) and Power Systems Computer-Aided Design (PSCAD), substation layout has gradually acquired optimization capabilities based on simulation data.

[0003] However, even with the assistance of the aforementioned tools, existing methods are mostly based on static models and lack the ability to generate them automatically, thus failing to achieve closed-loop optimization between deployment and performance. Summary of the Invention

[0004] This invention provides a substation layout method, apparatus, system, and medium that can solve at least one of the above-mentioned technical problems.

[0005] In a first aspect, embodiments of the present invention provide a substation layout method, including: Based on the layout parameters of each equipment in the substation and the first operating conditions, a structured dataset is determined. Based on the structured dataset, a pre-defined conditional generative adversarial network is initially trained to generate multiple initial layout design diagrams. Based on the multiple initial layout design drawings, a finite element physical model is generated; The finite element physical model is solved to obtain coupled solution results; Performance indicators are extracted based on the coupled solution results to construct a comprehensive performance evaluation function; Based on the comprehensive performance evaluation function, the parameters of the generator in the conditional generative adversarial network are updated to obtain the target generator, and the target layout design diagram is generated through the target generator.

[0006] In a second aspect, embodiments of the present invention provide a substation layout device, comprising: The dataset determination module is used to determine a structured dataset based on the layout parameters of each piece of equipment in the substation and each first operating condition. The training module is used to perform initial training on a preset conditional generative adversarial network based on the structured dataset to generate multiple initial layout design diagrams. The model generation module is used to generate a finite element physical model based on multiple initial layout design drawings; The solver module is used to solve the finite element physical model and obtain coupled solution results; The function construction module is used to extract performance indicators based on the coupled solution results in order to construct a comprehensive performance evaluation function; The target layout design diagram generation module is used to update the parameters of the generator in the conditional generative adversarial network based on the comprehensive performance evaluation function, obtain the target generator, and generate the target layout design diagram through the target generator.

[0007] Thirdly, embodiments of the present invention also provide a substation layout system, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described in any one of the embodiments of the present invention.

[0008] Fourthly, embodiments of the present invention also provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method described in any one of the embodiments of the present invention.

[0009] This invention employs a structured dataset determined based on the layout parameters of various equipment and the first operating conditions of a substation. Then, a pre-defined conditional generative adversarial network (GAN) is trained using this structured dataset to generate multiple initial layout design diagrams, providing feasible solutions that satisfy the constraints for subsequent optimization. Based on this, corresponding finite element physical models are established for the multiple initial layout design diagrams, and coupled solutions are obtained to comprehensively simulate the multi-physics response characteristics of different layout schemes under real or near-real operating conditions. Furthermore, performance indicators are extracted based on the coupled solution results, and a comprehensive performance evaluation function is constructed. The real physical performance obtained from the finite element analysis is fed back to the generator of the conditional GAN. By optimizing the generator's loss function, inverse constraint learning based on physical performance is achieved, enabling the generator to actively tend towards a better-performing layout form during subsequent layout diagram generation, thus obtaining the target generator. Finally, the target layout design diagram is generated based on the target generator. This not only improves the rationality and engineering feasibility of the generated layout but also significantly enhances the comprehensive performance of the final target layout design diagram in multiple aspects, thereby achieving an overall improvement in substation layout performance.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0011] The accompanying drawings are provided for a better understanding of this solution and do not constitute a limitation of the invention. Wherein: Figure 1 This is a flowchart of a substation layout method according to an embodiment of the present invention; Figure 2 This is a structural block diagram of a substation layout device according to an embodiment of the present invention; Figure 3 This is a schematic block diagram of an electronic device used to implement the methods of embodiments of the present invention. Detailed Implementation

[0012] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0013] Figure 1 This is a flowchart of a substation layout method according to an embodiment of the present invention.

[0014] like Figure 1 As shown, the substation layout method may include: S110, Based on the layout parameters of each equipment in the substation and each first operating condition, determine the structured dataset; S120, based on a structured dataset, performs initial training on a pre-defined conditional generative adversarial network to generate multiple initial layout design diagrams; S130 generates a finite element physical model based on multiple initial layout design drawings; S140, Solve the finite element physical model to obtain coupled solution results; S150, performance indicators are extracted based on the coupled solution results to construct a comprehensive performance evaluation function; S160, based on the comprehensive performance evaluation function, update the parameters of the generator in the conditional generative adversarial network to obtain the target generator, and generate the target layout design diagram through the target generator.

[0015] For example, equipment layout parameters refer to a set of parameters used to describe the spatial arrangement, installation form, and mutual constraints of various electrical equipment within a substation. Equipment layout parameters may include, but are not limited to: equipment coordinate positions, such as the position point (x, y) of a transformer in the site coordinate system, where x represents the horizontal coordinate and y represents the vertical coordinate; equipment footprint, such as the length × width dimensions of a disconnecting switch of 3m × 2m; minimum safety distance between equipment, such as a requirement of ≥1.5m between a busbar and a circuit breaker; height parameters, such as the installation height of a reactor of 2.3m; and electrical connection relationships, such as the connection constraints between lines, busbars, and transformers.

[0016] For example, in a 220 kV substation, the layout parameters of transformer T1 include: location coordinates (12.0m, 8.5m), footprint size of 6m × 4m, minimum safe distance from the main transformer bay of 3m, and installation height of 2.0m.

[0017] For example, the first operating condition refers to the external conditions used to characterize the electrical, thermal, or mechanical environment during substation operation or simulation design, including but not limited to electromagnetic load conditions, ambient temperature conditions, load variation conditions, and wind pressure conditions. Specifically, it includes: current conditions: such as the main transformer rated current of 800 amperes (A) and short-time current of 20 kiloamperes (kA); temperature conditions: such as summer high temperature of 40 degrees Celsius (°C) and winter low temperature of -10°C; wind pressure conditions: such as wind load of 0.3 kilonewtons per square meter (kN / m²). 2); Load variation conditions: such as the peak value of the daily load curve of 1.1 per unit (pu).

[0018] For example, in the design of a 500kV substation, the first operating condition might include: main transformer full-load operating temperature of 95°C and crosswind pressure of 0.25kN / m. 2 Conductor current 2000A.

[0019] For example, a structured dataset refers to a dataset that can be used for training deep learning models by organizing various spatial, physical, and operational information in a unified data format based on equipment layout parameters and first operating conditions.

[0020] In this example, structured data can take the form of a matrix, tensor, or vector structure. Examples include: a device location matrix; a distance matrix between devices; a feature vector of operating parameters; and layout status encoding (such as one-hot encoding or custom encoding). For instance, a structured data record might be: a device location matrix P (10×10), a device safety distance matrix D (10×10), and an operating condition feature vector C = (current 1500A, temperature 35°C, wind pressure 0.3kN / m²). 2 The layout validity flag is y=1 (meaning the layout is valid).

[0021] For example, the substation in step S110 can be one substation or multiple substations.

[0022] According to the above implementation method, firstly, a structured dataset describing the spatial distribution and operating status of substation equipment is constructed based on the geometric layout parameters and different operating conditions of the equipment. Then, this structured dataset is used to initially train a conditional generative adversarial network (GAN) to generate multiple initial layout design diagrams. Based on the generated layout design diagrams, a finite element physical model is constructed, and the model is coupled and solved to obtain the performance results of each layout scheme under real physical conditions. Key performance indicators are extracted based on the coupled solution results, a comprehensive performance evaluation function is constructed, and this function is used to continuously update the parameters of the generator in the generative network, enabling the generator to output a target layout design diagram with better overall performance. Through this process, the generative deep learning model can continuously optimize the layout scheme of substation equipment under the guidance of real physical feedback, realizing a collaborative design approach that couples data-driven and physical mechanisms. This not only significantly improves the rationality and safety of the layout design but also reduces the time cost of multiple rounds of simulation calculations and manual parameter tuning, ultimately obtaining a target layout design diagram with better performance and higher reliability.

[0023] In one implementation, a structured dataset is determined based on the layout parameters of each device and each first operating condition, including: acquiring historical layout data of the substation, wherein the historical layout data of the substation includes each plan layout and each design specification; extracting parameters from each plan layout to obtain the layout parameters of each device; determining a first feature parameter matrix based on the layout parameters of each device; extracting parameters from each design specification to obtain each first operating condition; and combining the first feature parameter matrix and each first operating condition to obtain a structured dataset.

[0024] For example, the plan layout files (e.g., CAD drawings) of the substation over the years can be read from the historical engineering archive database, and the different versions of CAD drawings can be formatted to make data recognition possible. At the same time, the corresponding design specifications can be loaded from the document management database for text digitization preprocessing to form a raw dataset that can be used for subsequent parameter extraction.

[0025] The design specification can be in Portable Document Format (PDF) or Word text.

[0026] For example, the system automatically downloads CAD layout drawings from 2015 to 2023 from the archives of a 220kV substation and converts the ".dwg" files into a unified vector graphic format; at the same time, it obtains the corresponding year's "Main Wiring Design Specification" and "Electrical Primary Equipment Layout Specification" and performs Optical Character Recognition (OCR) on them into structured text.

[0027] For example, by using vector feature recognition, the equipment symbols, equipment spacing, busbar direction, frame position, equipment outline rectangle and its coordinate information in the CAD plan layout drawing of a certain equipment can be extracted to obtain the set of layout parameters corresponding to the equipment.

[0028] For example, when recognizing the plan layout, the main transformer symbol is automatically detected and its upper left corner coordinates (35.2m, 18.7m) are extracted. The spacing between adjacent reactors is identified as 6.5m and the orientation of the bus centerline is 90°.

[0029] For example, parameters such as spatial coordinates, footprint, relative distance, electrical connection relationship, channel width, device orientation and safety clearance of each device are encoded according to predefined matrix row and column rules, and missing parameters are linearly interpolated or zero-filled, thereby mapping discrete layout parameters into vector representations of a unified dimension, and finally combining them in the order of devices to form the first feature parameter matrix for structured learning.

[0030] For example, the coordinates (35.2, 18.7), dimensions (8.0, 5.5), distance from the disconnector (7.2m), and connection direction (90°) of the main transformer to the busbar are encoded into a row vector; then the reactor, outgoing line bay, and other equipment are encoded sequentially to form a feature matrix of the form 120×20 (120 devices, each with 20-dimensional parameters).

[0031] For example, the design specification text is input into a work condition information extraction process based on a hybrid rule-based and statistical mode. This process first constructs an engineering work condition dictionary based on regular expression matching rules, and uses this dictionary to locate work condition keywords such as "temperature," "wind load," "altitude," "short-circuit current," "load level," "voltage level," and "operating mode" in the text. After keyword location, the physical quantities corresponding to the keywords are parsed based on the numerical patterns within a fixed window (e.g., 10 characters before and after), and intervals, upper and lower limits, or specific values ​​are extracted. When units are present, the values ​​are normalized using a unit conversion table. When the representation of the same work condition is inconsistent in different texts, synonymous expressions such as "35 degrees Celsius," "35°C," and "temperature 35 degrees" are further standardized using a pre-set standardized mapping table, thereby generating a first set of work condition conditions with a unified format.

[0032] For example, from the "Primary Equipment Layout Instructions", we can identify "short-circuit current 31.5kA", "maximum operating temperature 40°C", and "wind load 0.3kN / m". 2 Text fragments such as "maximum load rate 1.1 pu" are automatically unified into a standard structured format.

[0033] For example, the first feature parameter matrix is ​​encoded with a fixed length according to the device dimension, and then concatenated with the first working condition extracted from the corresponding design specification in the feature dimension. The samples are aligned according to the correspondence between each set of drawings and the specification, and finally all samples are combined into a structured dataset for model training according to their numbers.

[0034] For example, consider the "120×20 layout feature matrix" of a substation and its corresponding parameters: "Temperature 40°C, Short-circuit current 31.5kA, Wind load 0.3kN / m²". 2 The samples with a load rate of 1.1 pu are merged into a complete sample, and the sample is stored in a structured manner to form a dataset that can be directly used by deep learning models.

[0035] According to the above implementation method, multiple floor plan layouts and corresponding design specifications are first read from the historical layout data of the substation. The system identifies the floor plan layouts and extracts parameters such as the spatial location, size, spacing, and wiring relationships of the equipment to form a first feature parameter matrix describing the layout characteristics. Simultaneously, by parsing each design specification, first operating conditions such as electrical load, current level, heat dissipation conditions, short-circuit capacity, and protection coordination requirements under different operating scenarios are extracted. Subsequently, the first feature parameter matrix and each first operating condition are combined in a row-by-row or sample-by-sample association manner to generate a structured dataset that can be used for analysis and modeling. In this way, the originally scattered drawing data and text data can be directly used as input for subsequent simulation modeling, layout optimization, risk assessment, or machine learning models, achieving a complete data structure, thereby reducing manual processing costs and improving data quality and reliability.

[0036] In one implementation, a pre-defined conditional generative adversarial network (GAN) is initially trained based on a structured dataset to generate multiple initial layout design diagrams. This includes: acquiring the GAN, which comprises a generator and a discriminator; sampling random noise vectors from a pre-defined standard normal distribution to obtain a first random noise vector; using a second operating condition from the structured dataset and the first random noise vector as inputs to the generator to generate a first layout design diagram; inputting a pre-defined first feature parameter matrix and a third operating condition as first inputs, and the first layout design diagram and the second operating condition as second inputs, respectively, to the discriminator in the GAN to perform authenticity discrimination on the first and second inputs, obtaining the discriminator's discrimination result; based on the discrimination result, performing adversarial training on the generator and discriminator using an optimizer to obtain a trained GAN; and using a second random noise vector and a fourth operating condition as inputs to the trained GAN to generate multiple initial layout design diagrams.

[0037] For example, a conditional generative adversarial network (GAN) refers to an adversarial generative model that simultaneously receives a random noise vector and operating conditions as inputs to generate a layout image that satisfies the target conditions. It includes a generator for generating the layout image and a discriminator for judging the authenticity of the layout image and its matching with the conditions. For example, the input might be "temperature 40℃, short-circuit current 31.5kA, wind load 0.3kN / m". 2 Under certain conditions, the network automatically generates a substation layout diagram that meets the requirements of high temperature and high load conditions.

[0038] In this example, a pre-built conditional generative adversarial network (GAN) model file is loaded, the network parameters of the generator and discriminator are initialized, and an adversarial training channel between them is established. Simultaneously, the relevant optimizer, generator loss function, and discriminator loss function are loaded for joint training in subsequent steps. For example, a GAN with a U-Net (U-shaped network) generator and a PatchGAN (patch-based discriminator) discriminator is loaded.

[0039] In this example, the generator's loss function can be represented by the following function expression: In the formula, The loss function for the generator; The first noise vector Second operating condition The joint expected value; The output of the generator (first layout design diagram) is the fake sample generated based on the first noise vector and the second operating condition.

[0040] In this example, the loss function of the discriminator can be expressed as a function as follows: In the formula, The loss function of the discriminator; The joint expected value of the first characteristic parameter matrix x and the third working condition c; The discriminant's judgment result on the first feature parameter matrix x under the third operating condition c; The first noise vector in the first layout design diagram and the third working condition The joint expected value; For the discriminator to design the first layout drawing The discrimination result under the third working condition c.

[0041] For example, the first random noise vector refers to a vector composed of multidimensional random numbers sampled from a preset standard normal distribution, used to provide randomness and diversity to the generator. For example, a random vector with a dimension of 128 is sampled, such as [0.12, -0.53, 0.87, ...].

[0042] In this example, a random sampling function is invoked to generate a fixed-length vector from a multidimensional normal distribution with a mean of 0 and a variance of 1, and then reshape it into the generator's input format. For example, a first random noise vector of dimension 128 is sampled from the multidimensional normal distribution N(0,1).

[0043] For example, the second operating condition refers to the set of operating parameters obtained from the structured dataset used by the constraint generator to generate the layout diagram, including operating characteristics such as temperature, wind load, and short-circuit current level. For instance, the second operating condition might be {temperature 35℃, wind load 0.3kN / m...} 2 The short-circuit current is 20kA.

[0044] In this example, the operating conditions are encoded into conditional feature vectors of the same length as the noise vector. These two vectors are then concatenated along their feature dimensions and input into the generator. Through multi-layer convolution and deconvolution operations, a first layout design that conforms to the operating condition constraints is output. For example, a 256×256 first layout design is generated under the conditions of {temperature 35℃, short-circuit current 20kA}.

[0045] For example, the first feature parameter matrix refers to a matrix formed by encoding feature parameters such as the spatial coordinates, dimensions, adjacent spacing, electrical connection relationships, and safety clearance of each device in a fixed format, which is used as the input discriminator for real samples. For example, a matrix of size 120×20, where 120 represents the number of devices and 20 represents the 20 feature dimensions of each device.

[0046] For example, the third operating condition refers to the set of real operating conditions corresponding to the first feature parameter matrix, which is used as "condition constraints for real samples" in the discriminator. For example, if extracted from real drawings, it may be {temperature 40℃, short-circuit current 31.5kA, load factor 1.1pu}.

[0047] For example, the third operating condition, the second operating condition, and the first operating condition can be the same operating condition.

[0048] For example, the discriminator's judgment result refers to the probability output of the discriminator on whether the input sample is a real sample and whether it meets the corresponding working conditions. For example, the judgment result may be 0.93 (high authenticity) or 0.12 (low authenticity).

[0049] In this example, the parameter matrix of the real sample and its corresponding operating conditions are encoded and input into the discriminator as the real input. The layout map generated by the generator is encoded in the same way as its conditional input and then fed into the discriminator as the fake input. The discriminator extracts features through convolution and outputs the probability of authenticity for the two types of inputs. For example, the discriminator may output 0.95 for real samples and 0.27 for generated samples.

[0050] In this example, the discriminator's loss is calculated to update its parameters, classifying real samples as true and generated samples as false. Simultaneously, the generator's loss is calculated to update its parameters, ensuring generated samples are recognized as true by the discriminator. Through multiple rounds of alternating training, the network's generation capability is gradually improved. This results in the trained conditional generative adversarial network.

[0051] For example, the second random noise vector refers to the random vector used to generate the initial layout design diagram, which is obtained by resampling from the standard normal distribution after the conditional generative adversarial network has been trained. For example, the dimension is 128, such as [1.12, -0.23, 0.41, ...].

[0052] For example, the fourth operating condition refers to the target operating condition input to the generator when the conditional generative adversarial network generates the layout graph after training. This can be a set of target operating condition parameters set by the user. For instance, the user can set {temperature 42℃, wind load 0.4kN / m²}. 2 The short-circuit current is 25kA.

[0053] For example, the initial layout design drawing refers to the substation equipment plan layout drawing generated by the trained conditional generative adversarial network based on random noise and operating conditions, which is used for subsequent physical simulation or optimization steps. For instance, the network automatically generates a layout sketch that includes the positional relationships of the main transformer, switchgear, and reactors. This layout sketch can be a 256×256 grayscale image or a red-green-blue (RGB) image.

[0054] In this example, multiple sets of random noise vectors are continuously sampled from a standard normal distribution as a second random noise vector. These, along with the target operating conditions, are encoded and input into the trained generator. Through forward inference, multiple layout diagrams (i.e., initial layout design diagrams) are generated at once for subsequent multi-scheme evaluation. For example, in a scenario with a temperature of 42℃ and a wind load of 0.4kN / m²... 2} generate 20 candidate substation layouts with different styles under the given conditions.

[0055] In one implementation, a finite element physical model is generated based on multiple initial layout design drawings, including: extracting equipment model data from an equipment database based on the multiple initial layout design drawings; converting each equipment model data to obtain each finite element geometric entity; defining equipment boundaries and connection relationships between each finite element geometric entity to obtain a first geometric structure model; extracting material parameters corresponding to each finite element geometric entity from an equipment material database based on the equipment type of each finite element geometric entity, and assigning each material parameter to the corresponding finite element geometric entity to update the first geometric structure model to obtain a second geometric structure model; and performing finite element mesh generation on the second geometric structure model to obtain a finite element physical model.

[0056] For example, equipment model data refers to a collection of data stored in an equipment database that describes the structural shape, external dimensions, and functional unit composition of electrical equipment. Examples include: 3D CAD files of transformers, winding dimensions, tank dimensions, and heat sink arrangement data; model assembly drawings of disconnecting switches, main disconnector length, and post insulator height; and core geometric parameters, coil radius, and number of turns of reactors.

[0057] In this example, by parsing the equipment identification information in each initial layout design drawing, the equipment type, model and corresponding equipment ID (ID is also the identifier) ​​involved in the layout drawing can be obtained. Then, the corresponding equipment CAD model file, assembly structure data and external dimension parameters can be read from the equipment database using the equipment ID as the index. At the same time, the equipment model file is converted into a unified format for subsequent geometric modeling.

[0058] For example, a finite element geometric entity refers to a geometric unit obtained by converting equipment model data into a software format suitable for finite element modeling, including but not limited to: "tank entity", "winding entity" and "core entity" converted from transformer CAD files; "core column entity" and "coil body entity" converted from reactor models; and "insulator entity", "conducting arm entity" and "contact knife entity" converted from switchgear models.

[0059] In this example, the original CAD model is transformed into a geometric entity that can participate in finite element modeling by performing structural analysis, patch construction, redundant feature removal, and topology repair on each device model. For example, after converting the main transformer CAD model into finite element format, geometric bodies such as the core entity, low-voltage winding entity, high-voltage winding entity, and oil tank entity can be obtained for subsequent mesh generation.

[0060] For example, the first geometric structure model refers to the set of geometric structures obtained by assembling multiple finite element geometric entities based on their spatial positions, boundary conditions, and connection relationships in a unified three-dimensional coordinate system.

[0061] For example, the "transformer tank entity", "high voltage bushing entity", and "radiator entity" are combined in three-dimensional space according to the relative positional relationship provided in the initial layout drawing, and the contact boundary is defined for the connection surface between the tank and the radiator to form the overall structure after assembly.

[0062] In this example, all finite element geometric entities can be imported into a unified three-dimensional coordinate system. The translation and rotation matrices of each device entity are determined by reading the relative position information of the initial layout design drawing. Based on the connection rule library in the device database, corresponding boundary conditions are set for the contact interfaces, self-locking interfaces, bolt connection interfaces, etc. between different devices. At the same time, collision detection and gap adjustment are performed during the import process to generate a first geometric structure model with complete assembly and well-defined boundaries.

[0063] For example, after assembling the transformer entity, disconnector entity, and busbar entity, the busbar end and the disconnector conductive arm are defined as rigid connection boundaries according to the connection rules, and the bottom of the disconnector and the foundation are defined as fixed support boundaries, thus obtaining a structural model with real assembly relationships.

[0064] For example, the second geometric structure model refers to a geometric model with completed material property assignment formed by allocating material parameters (such as Young's modulus, dielectric constant, thermal conductivity, etc.) from the material database to the corresponding geometric entities based on the first geometric structure model.

[0065] For example, in a transformer assembly model, the "silicon steel sheet material parameters" are assigned to the core entity, the "mineral insulating oil parameters" are assigned to the fluid domain inside the tank, and the "epoxy resin material parameters" are assigned to the bushing entity, forming an overall structure that already possesses material properties.

[0066] In this example, based on the equipment type to which each finite element geometric entity belongs, the material parameters of the metal, insulating medium, composite material or magnetic material used by that type of equipment can be retrieved from the material database, and the corresponding material parameters can be assigned to the geometric entity, so that each entity has a complete material property field. At the same time, the consistency of the material unit system is checked during the assignment process to prevent material omissions or conflicts, so as to finally form a second geometric structure model with complete material information.

[0067] For example, for the iron core, the permeability and density parameters of silicon steel sheets are automatically retrieved from the material database; for the oil tank, the yield strength and thermal conductivity parameters of Q235 steel are retrieved; and for the insulator, the dielectric constant and coefficient of thermal expansion of epoxy resin are retrieved for the final physical solution.

[0068] For example, a finite element physical model refers to a discrete physical domain model that can be used for finite element solutions, obtained by meshing a second geometric structure model. It includes solution elements such as nodes, elements, material fields, and boundary conditions. For instance, generating a tetrahedral or hexahedral mesh for a materialized transformer geometric model makes the entire model consist of tens of thousands of elements, which can be used for thermo-electric-magnetic coupling solutions.

[0069] In this example, firstly, default target mesh element sizes and element orders (such as first-order linear or second-order parabolic elements) are defined for the entire model, and appropriate element types are assigned based on the dimensions and shapes of the geometric entities (e.g., tetrahedral or hexahedral elements for solids, shell elements for thin-walled structures, and beam elements for slender rods). Next, local mesh refinement and control are performed. Key areas such as geometric abrupt changes, load application surfaces, constraint boundaries, and component connections are identified, and smaller local mesh sizes are applied to these areas or size gradient functions are used to optimize the mesh count while ensuring computational accuracy. Then, mesh generation is performed, using algorithms such as the leading edge propagation method or Delaunay triangulation to discretize the geometric model into a finite element set. Mesh quality is then checked, evaluating mesh quality by calculating element aspect ratios, warpage angles, and other metrics. Substandard areas are optimized through mesh smoothing, node repositioning, or local re-partitioning. Finally, the material parameters assigned in the second geometric structure model are mapped from the geometric entity to the corresponding mesh element, and a finite element physical model data file containing node coordinates, element connection relationships, element types and material property sets is generated. This file can be directly output to the finite element solver to obtain the finite element physical model.

[0070] In one implementation, solving the finite element physical model to obtain coupled solution results includes: constructing an electromagnetic field in the finite element physical model based on a time-domain vector potential-scalar potential hybrid form to simulate the electromagnetic field distribution between various substation equipment in the finite element physical model, obtaining a first simulation result; constructing a thermal field in the finite element physical model based on the heat conduction equation to simulate the heat diffusion process and temperature distribution on the surface of various substation equipment in the finite element physical model, obtaining a second simulation result; constructing a structural field in the finite element physical model based on the kinematic equilibrium equation to simulate the layout structural stability of the finite element physical model, obtaining a third simulation result; setting boundary conditions according to the operating environment of each substation equipment, and coupling the first simulation result, the second simulation result, and the third simulation result using finite element analysis software to obtain coupled solution results.

[0071] For example, an electromagnetic field is constructed in the finite element physical model using a hybrid form of time-domain vector potential and scalar potential to simulate the electromagnetic field distribution between various substation devices in the finite element physical model. This simulation process can be expressed by the following functional expression: ; ; Where A represents the magnetic vector potential; Represents electric potential; Relative permeability is extracted from the equipment material database; Electrical conductivity is represented and extracted from the equipment material database; Relative permittivity is extracted from the equipment materials database; The external excitation current density is determined by the current input value and the conductor cross-sectional area; t is time. This represents charge density, determined by boundary conditions; Indicates to Get the curl; Indicates to Take the divergence; This represents the gradient.

[0072] A thermal field is constructed in the finite element model using the Fourier heat conduction equation to simulate the heat diffusion process and temperature distribution on the surface of substation equipment within the finite element model. This process can be represented by the following functional expression: ; ; Where T represents temperature distribution; t represents time; Material density is represented and extracted from the equipment material database; Specific heat capacity is extracted from the equipment and materials database; Thermal conductivity is represented and extracted from the equipment materials database; This indicates the heat source item, defined according to the device power. Indicates heat exchange in the environment; Indicates ambient temperature, determined by standard room temperature; This represents the convective heat transfer coefficient, which is determined by the standard air convection coefficient.

[0073] The structural field is constructed in the finite element physical model using the Navier-Cauchy elasticity equations to simulate the structural stability of the finite element physical model layout under long-term loads and external impacts. This process can be represented by the following functional expression: ; in, and This represents the Lamé constant, extracted from the equipment materials database; Represents the displacement vector; This represents the external load, defined as the force directly generated by the weight of the equipment.

[0074] Boundary conditions are set according to the operating environment of the substation equipment. Finite element analysis software is used to couple the simulation results of electromagnetic field, thermal field and structural field (i.e. the first simulation result, second simulation result and third simulation result in the previous example) to generate electromagnetic intensity distribution map, temperature distribution map and stress distribution map (coupled solution result).

[0075] The hybrid vector and scalar potential approach can efficiently solve for electromagnetic field distributions in complex equipment, making it particularly suitable for describing substation equipment with complex geometries and non-uniform material conditions. It fully captures the electromagnetic coupling effects between devices, providing direct feedback on electromagnetic field distribution for equipment layout optimization. This allows for effective adjustments to equipment location, distance, and overall arrangement, thereby reducing electromagnetic compatibility issues such as the impact of interference signals. Based on the Fourier heat conduction equation, a thermal field simulation model is established to accurately simulate heat generation, diffusion, and dissipation during equipment operation using the finite element method. The thermal field simulation accurately simulates the temperature field distribution of the equipment, revealing which devices have insufficient heat dissipation capacity or excessive temperature gradients, providing directions for improvement in equipment cooling design. This simulation is applicable to substation equipment. This study investigates the heat transfer relationship between station equipment and the environment, optimizes equipment layout to improve overall heat dissipation efficiency, and avoids equipment failure and reduced operating efficiency due to overheating. Based on the Navier-Cauchy elasticity equations, it models the stress distribution of equipment under static force and dynamic impact loads, detects potential stress concentration areas, ensures the long-term mechanical stability of equipment, identifies potential structural failure locations through stress distribution maps, and provides a reference for prioritizing the strengthening of certain equipment connections or layout adjustments. By setting reasonable boundary conditions (such as electromagnetic field boundaries, heat dissipation boundaries, and mechanical fixed boundaries), it achieves multi-field coupling solution of electromagnetic, thermal, and structural fields in finite element analysis software, comprehensively evaluates the multi-physics performance of the current layout design, and reveals potential deficiencies.

[0076] In one implementation, performance indicators are extracted based on the coupled solution results to construct a comprehensive performance evaluation function. This includes: obtaining the coupled solution results, which include an electromagnetic intensity distribution map, a temperature distribution map, and a stress distribution map; extracting the maximum electric field strength and maximum magnetic field strength from the electromagnetic intensity distribution map as a first performance indicator; extracting the maximum temperature change from the temperature distribution map as a second performance indicator; extracting the maximum equivalent stress from the stress distribution map as a third performance indicator; and integrating the first performance indicator, the third performance indicator, and the third performance indicator based on a weighted voting strategy to obtain the comprehensive performance evaluation function.

[0077] For example, firstly, the post-processing module reads the electromagnetic field simulation result data file (electromagnetic intensity distribution map), identifies and iterates through the scalar or vector amplitudes of the electric and magnetic field intensities of all nodes in the entire solution domain, compares the values ​​of all nodes using a global search algorithm, and filters out the maximum values, recording them as the maximum electric field intensity and the maximum magnetic field intensity, respectively, which together constitute the first performance index. Next, the thermal simulation result data (temperature distribution map) is accessed, a reference temperature (such as ambient temperature or initial temperature) is set, the absolute difference between the temperature of all nodes and this reference temperature is calculated, and the maximum value of this difference is found by iterating through all nodes, recorded as the maximum temperature change, as the second performance index. Then, by reading the structural mechanics analysis results (stress distribution map), the equivalent stress values ​​of all elements or nodes are obtained according to the yield criterion used (such as the von Mises criterion), and the maximum value is determined through global comparison, recorded as the maximum equivalent stress, as the third performance index.

[0078] Finally, the three performance indicators are integrated based on a weighted voting strategy to obtain a comprehensive performance evaluation function. Specifically, firstly, the first, second, and third performance indicators need to be normalized to eliminate the influence of dimensions. Then, according to the relative importance of each indicator in the system security and reliability evaluation, appropriate weight coefficients are assigned to them (e.g., weight w1 for the first performance indicator, weight w2 for the second performance indicator, and weight w3 for the third performance indicator, satisfying w1+w2+w3=1). Finally, the comprehensive performance evaluation value is calculated using a linear weighted sum model, i.e. In the formula, This is a comprehensive performance evaluation function; As the first performance indicator weight; The primary performance indicator; As the weight of the second performance metric; This is the second performance indicator; As the weight of the third performance indicator; This is the third performance indicator.

[0079] According to the above implementation method, this method extracts unified indicators from the electromagnetic, thermal, and mechanical field distribution diagrams obtained by coupled solution, transforming key physical quantities such as maximum electric field strength, maximum magnetic field strength, maximum temperature change, and maximum equivalent stress into quantifiable performance indicators. A weighted voting strategy is then used to integrate the multi-physics field indicators to form a comprehensive performance evaluation function. This automates, structures, and unifies the transformation of complex multi-field distribution results into a single evaluation indicator that can be used for optimization, comparison, and decision-making. This eliminates the need for manual interpretation of graphs in multi-physics field analysis, improving analysis efficiency and objectivity. Simultaneously, it supports rapid quantitative comparison between different design schemes, promoting an automated closed-loop process for layout design, structural optimization, and safety assessment.

[0080] In one implementation, based on a comprehensive performance evaluation function, the loss function of the generator in the conditional generative adversarial network is adjusted to obtain a target generator, and a target layout design map is generated through the target generator. This includes: constructing a target loss function for the generator based on the comprehensive performance evaluation function and the generator's loss function; iteratively updating the generator's parameters with the generator's target loss function as the optimization objective to obtain the target generator; generating multiple candidate layout design maps based on the target generator by inputting multiple different third random noise vectors; and extracting the design map from the multiple candidate layout design maps through multiphysics simulation to obtain the target layout design map.

[0081] For example, firstly, the comprehensive performance evaluation function is used as the reward signal in reinforcement learning, or its negative value is used as a penalty term through the policy gradient method, and linearly combined with the generator's inherent loss function (such as the cross-entropy loss used for conditional generation), where a tradeoff coefficient is introduced to adjust the relative importance between performance reward and original generation quality, thereby forming a complete objective loss function.

[0082] Next, the generator parameters are iteratively updated with the target loss function as the optimization objective to obtain the target generator. In the training loop, the gradient of the target loss function with respect to all trainable parameters of the generator is calculated using the backpropagation algorithm. Then, stochastic optimization algorithms such as Adam or SGD (stochastic gradient descent) are used to update the generator's weights and bias parameters according to the calculated gradient direction. This process is repeated until the loss function converges or the preset number of iterations is reached. At this point, the generator parameters are fixed, which is the target generator.

[0083] Then, based on the obtained target generator, multiple candidate layout design diagrams are generated by inputting multiple different third random noise vectors. The operation is to sample a set of noise vectors that conform to a specific distribution (such as Gaussian distribution) and are different from each other in the latent space. These vectors, along with any necessary conditional information (such as device constraints), are input into the trained target generator. The forward propagation calculation of the target generator will output a series of diverse candidate layout design diagram data that conform to the original data distribution.

[0084] Finally, all candidate layout designs generated in the previous step are used as inputs for multiphysics simulation in sequence. The complete process described above, from equipment data extraction, geometric modeling, material allocation, mesh generation to multiphysics (electromagnetic-thermal-stress) simulation, is automatically executed. The comprehensive performance evaluation function value corresponding to each candidate layout is calculated. By comparing all these evaluation values, the candidate layout design that makes the comprehensive performance evaluation function optimal (usually the maximum or minimum, depending on the function definition) is selected and determined as the target layout design.

[0085] According to the above implementation method, the generator's loss function is enhanced based on a comprehensive performance evaluation function. This allows the generator to consider not only the adversarial feedback of the discriminator when updating parameters, but also the comprehensive performance of the layout scheme in multi-physics indices such as electromagnetics, temperature, and stress, thus obtaining a performance-oriented target generator. Based on this, a large number of candidate layout designs are generated by inputting multiple different random noise vectors into the target generator. Multiphysics simulation is then used to automatically select the target layout design that is optimal in terms of safety, stability, and overall performance. This significantly improves the physical rationality and engineering feasibility of the layout design, reduces human trial-and-error costs, improves design efficiency and reliability, and makes automated layout generation more closely aligned with actual engineering requirements.

[0086] Figure 2 This is a structural block diagram of a substation layout device according to an embodiment of the present invention.

[0087] like Figure 2 As shown, the substation layout device may include: The dataset determination module 510 is used to determine a structured dataset based on the layout parameters of each piece of equipment in the substation and each first operating condition. Training module 520 is used to perform initial training on a preset conditional generative adversarial network based on the structured dataset to generate multiple initial layout design diagrams. The model generation module 530 is used to generate a finite element physical model based on multiple initial layout design drawings; Solver module 540 is used to solve the finite element physical model and obtain coupled solution results; The function construction module 550 is used to extract performance indicators based on the coupled solution results in order to construct a comprehensive performance evaluation function; The target layout design diagram generation module 560 is used to update the parameters of the generator in the conditional generative adversarial network based on the comprehensive performance evaluation function, obtain the target generator, and generate the target layout design diagram through the target generator.

[0088] In one implementation, the dataset determination module includes: The data acquisition unit is used to acquire historical layout data of the substation, wherein the historical layout data of the substation includes various floor plans and various design specifications; The first parameter extraction unit is used to extract parameters from each of the plan layout diagrams to obtain the layout parameters of each of the devices; The parameter matrix determination unit is used to determine the first feature parameter matrix based on the layout parameters of each of the devices; The second parameter extraction unit is used to extract parameters from each of the design specifications to obtain each of the first working conditions. The combination unit is used to combine the first feature parameter matrix and each of the first working conditions to obtain the structured dataset.

[0089] In one implementation, the training module includes: A conditional generative adversarial network acquisition unit is used to acquire the conditional generative adversarial network, wherein the conditional generative adversarial network includes a generator and a discriminator; The sampling unit is used to sample a random noise vector from a preset standard normal distribution to obtain a first random noise vector; The first generation unit is used to take the second working condition and the first random noise vector in the structured dataset as input to the generator to generate the first layout design diagram. The discrimination unit is used to input the preset first feature parameter matrix and the third working condition as the first input, and the first layout design diagram and the second working condition as the second input, respectively into the discriminator in the conditional generative adversarial network, so as to perform authenticity discrimination on the first input and the second input and obtain the discrimination result of the discriminator; An adversarial training unit is used to perform adversarial training on the generator and the discriminator through an optimizer based on the discrimination result, so as to obtain a trained conditional generative adversarial network. The second generation unit is used to take the second random noise vector and the fourth operating condition as input to the trained conditional generative adversarial network to generate multiple initial layout design diagrams.

[0090] In one implementation, the model generation module includes: The equipment model data extraction unit is used to extract each equipment model data from the equipment database based on multiple initial layout design drawings. The conversion unit is used to convert the data of each of the device models to obtain each finite element geometric entity; A definition unit is used to define the device boundaries and connection relationships between the various finite element geometric entities to obtain the first geometric structure model; The model update unit is used to extract the material parameters corresponding to each finite element geometric entity from the equipment material database based on the equipment type of each finite element geometric entity, and assign each material parameter to the corresponding finite element geometric entity to update the first geometric structure model and obtain the second geometric structure model. The meshing unit is used to perform finite element meshing on the second geometric structure model to obtain the finite element physical model.

[0091] In one embodiment, the solving module includes: The first simulation unit is used to construct an electromagnetic field in the finite element physical model based on a time-domain vector potential-scalar potential hybrid form, so as to simulate the electromagnetic field distribution between various substation equipment in the finite element physical model and obtain the first simulation result. The second simulation unit is used to construct a thermal field in the finite element physical model based on the heat conduction equation, so as to simulate the heat diffusion process and temperature distribution on the surface of each substation equipment in the finite element physical model and obtain the second simulation results. The third simulation unit is used to construct a structural field in the finite element physical model based on the equation of motion equilibrium, so as to simulate the stability of the layout structure of the finite element physical model and obtain the third simulation result. The coupled solution unit is used to set boundary conditions according to the operating environment of each substation equipment, and to perform coupled solution on the first simulation result, the second simulation result and the third simulation result based on finite element analysis software to obtain the coupled solution result.

[0092] In one implementation, the function construction module includes: A coupled solution result acquisition unit is used to acquire the coupled solution results, wherein the coupled solution results include an electromagnetic intensity distribution map, a temperature distribution map, and a stress distribution map; The first extraction unit is used to extract the maximum electric field strength and the maximum magnetic field strength from the electromagnetic intensity distribution map as the first performance index. The second extraction unit is used to extract the maximum temperature change from the temperature distribution map as a second performance indicator. The third extraction unit is used to extract the maximum equivalent stress from the stress distribution map as a third performance index. An integration unit is used to integrate the first performance indicator, the third performance indicator, and the third performance indicator based on a weighted voting strategy to obtain the comprehensive performance evaluation function.

[0093] In one embodiment, the target layout design drawing generation module includes: The construction unit is used to construct the target loss function of the generator based on the comprehensive performance evaluation function and the generator's loss function; An iterative update unit is used to iteratively update the parameters of the generator with the target loss function of the generator as the optimization objective, so as to obtain the target generator; The first layout design drawing generation unit is used to generate multiple candidate layout design drawings based on the target generator by inputting multiple different third random noise vectors. The second layout design drawing generation unit is used to extract the design drawing from multiple candidate layout design drawings through multiphysics simulation to obtain the target layout design drawing.

[0094] The specific functions and examples of each module and submodule of the system in this embodiment of the invention can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0095] The acquisition, storage, and application of user personal information involved in the technical solution of this invention all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0096] This invention also provides a substation layout system, comprising: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method described in any one of the embodiments of the present invention.

[0097] The beneficial effects of the substation layout system in this embodiment of the invention are equivalent to the beneficial effects of the substation layout method described above, and will not be repeated here.

[0098] This invention also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method described in any one of the embodiments of this invention.

[0099] The beneficial effects of the storage medium of the present invention are equivalent to the beneficial effects of the substation layout method described above, and will not be repeated here.

[0100] Figure 3A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present invention is shown. Electronic device 800 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 800 may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0101] like Figure 3 As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. The RAM 803 may also store various programs and data required for the operation of the electronic device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0102] Multiple components in electronic device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of displays, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows electronic device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0103] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as a substation layout method. For example, in some embodiments, the substation layout method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the substation layout method described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured to perform the substation layout method by any other suitable means (e.g., by means of firmware).

[0104] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0105] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0106] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0107] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0108] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0109] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0110] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0111] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this invention should be included within the scope of protection of this invention.

Claims

1. A substation layout method, characterized in that, include: Based on the layout parameters of each equipment in the substation and the first operating conditions, a structured dataset is determined. Based on the structured dataset, a pre-defined conditional generative adversarial network is initially trained to generate multiple initial layout design diagrams. Based on the multiple initial layout design drawings, a finite element physical model is generated; The finite element physical model is solved to obtain coupled solution results; Performance indicators are extracted based on the coupled solution results to construct a comprehensive performance evaluation function; Based on the comprehensive performance evaluation function, the parameters of the generator in the conditional generative adversarial network are updated to obtain the target generator, and the target layout design diagram is generated through the target generator.

2. The method according to claim 1, characterized in that, The structured dataset determined based on the layout parameters of each device and the first operating condition includes: Obtain historical layout data of substations, including various floor plans and design specifications; Parameters are extracted from each of the aforementioned floor plan layouts to obtain the layout parameters of each of the aforementioned devices; Based on the layout parameters of each of the aforementioned devices, a first feature parameter matrix is ​​determined; Parameters are extracted from each of the design specifications to obtain each of the first operating conditions. The first feature parameter matrix and each of the first operating conditions are combined to obtain the structured dataset.

3. The method according to claim 1, characterized in that, The process involves initial training of a pre-defined conditional generative adversarial network based on the structured dataset to generate multiple initial layout design diagrams, including: Obtain the conditional generative adversarial network, wherein the conditional generative adversarial network includes a generator and a discriminator; The first random noise vector is obtained by sampling random noise vectors from a preset standard normal distribution; The second working condition in the structured dataset and the first random noise vector are used together as input to the generator to generate the first layout design diagram. The first feature parameter matrix and the third working condition are used as the first input, and the first layout design diagram and the second working condition are used as the second input. These are respectively input to the discriminator in the conditional generative adversarial network to judge the authenticity of the first input and the second input, and obtain the discriminator's judgment result. Based on the discrimination result, the generator and the discriminator are subjected to adversarial training through an optimizer to obtain the trained conditional generative adversarial network; The second random noise vector and the fourth operating condition are used as inputs to the trained conditional generative adversarial network to generate multiple initial layout design diagrams.

4. The method according to claim 1, characterized in that, The process of generating a finite element physical model based on multiple initial layout design drawings includes: Based on multiple initial layout design drawings, extract the model data of each device from the device database; The data of each of the device models are converted to obtain each finite element geometric entity. The device boundaries and connection relationships between each of the finite element geometric entities are defined to obtain the first geometric structure model; Based on the equipment type of each finite element geometric entity, material parameters corresponding to each finite element geometric entity are extracted from the equipment material database, and each material parameter is assigned to the corresponding finite element geometric entity to update the first geometric structure model and obtain the second geometric structure model. The second geometric structure model is meshed using finite element methods to obtain the finite element physical model.

5. The method according to claim 1, characterized in that, The process of solving the finite element physical model to obtain coupled solution results includes: Electromagnetic fields are constructed in the finite element physical model based on a hybrid form of time-domain vector potential and scalar potential to simulate the electromagnetic field distribution between various substation equipment in the finite element physical model and obtain the first simulation result. A thermal field is constructed in the finite element physical model based on the heat conduction equation to simulate the heat diffusion process and temperature distribution on the surface of each substation equipment in the finite element physical model, and a second simulation result is obtained. Based on the equations of motion equilibrium, a structural field is constructed in the finite element physical model to simulate the stability of the layout structure of the finite element physical model, and a third simulation result is obtained. Boundary conditions are set according to the operating environment of each substation equipment, and the first simulation result, the second simulation result, and the third simulation result are coupled and solved using finite element analysis software to obtain the coupled solution result.

6. The method according to claim 1, characterized in that, The step of extracting performance indicators based on the coupled solution results to construct a comprehensive performance evaluation function includes: Obtain the coupled solution results, wherein the coupled solution results include electromagnetic intensity distribution map, temperature distribution map and stress distribution map; The maximum electric field strength and the maximum magnetic field strength are extracted from the electromagnetic intensity distribution map and used as the first performance index. The maximum temperature change is extracted from the temperature distribution map and used as a second performance indicator. The maximum equivalent stress is extracted from the stress distribution diagram and used as the third performance index. The first performance metric, the third performance metric, and the third performance metric are integrated based on a weighted voting strategy to obtain the comprehensive performance evaluation function.

7. The method according to claim 1, characterized in that, The process of updating the parameters of the generator in the conditional generative adversarial network based on the comprehensive performance evaluation function to obtain the target generator, and generating a target layout design diagram using the target generator, includes: Based on the comprehensive performance evaluation function and the generator's loss function, construct the generator's target loss function; The parameters of the generator are iteratively updated using the target loss function of the generator as the optimization objective to obtain the target generator; Based on the target generator, multiple candidate layout design diagrams are generated by inputting multiple different third random noise vectors; The target layout design drawing is obtained by extracting design drawings from multiple candidate layout design drawings through multiphysics simulation.

8. A substation layout device, characterized in that, include: The dataset determination module is used to determine a structured dataset based on the layout parameters of each piece of equipment in the substation and each first operating condition. The training module is used to perform initial training on a preset conditional generative adversarial network based on the structured dataset to generate multiple initial layout design diagrams. The model generation module is used to generate a finite element physical model based on multiple initial layout design drawings; The solver module is used to solve the finite element physical model and obtain coupled solution results; The function construction module is used to extract performance indicators based on the coupled solution results in order to construct a comprehensive performance evaluation function; The target layout design diagram generation module is used to update the parameters of the generator in the conditional generative adversarial network based on the comprehensive performance evaluation function, obtain the target generator, and generate the target layout design diagram through the target generator.

9. A substation layout system, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-7.

Images