A distribution network connection diagram layout method, system, storage medium and computing device

Abstract

The application discloses a distribution network contact graph layout method and system, a storage medium and a computing device. The application adopts a generative adversarial network to optimize node layout, connects the optimized node layout, and thus obtains the distribution network contact graph layout. Compared with a traditional method, the application is not limited by experience and can be applied to a current increasingly complex topology structure.

Description

Technical Field

[0001] This invention relates to a distribution network interconnection diagram layout method, system, storage medium, and computing device, belonging to the field of power system automation. Background Technology

[0002] In recent years, with the continuous expansion of distribution network construction, in order to avoid network complexity caused by disordered connections between feeders, power supply units have gradually become a typical design paradigm for medium-voltage distribution network structures. A power supply unit consists of multiple feeders interconnected between one or more substations. Feeders within a unit are directly or indirectly connected, and any two feeders between different units are inaccessible through interconnection. A power supply unit is a relatively independent and stable power supply area for dispatching, operation, maintenance, and repair. Efficient and aesthetically pleasing topology visualization can greatly improve the operation and maintenance management efficiency of power supply units, and the key technology for topology visualization is graph layout methods.

[0003] Current graph layout methods are based on experience and heuristics. These algorithms are limited by experience and are only effective for topologies with fixed patterns and little data variation, making them unsuitable for today's increasingly complex topologies. Summary of the Invention

[0004] This invention provides a method, system, storage medium, and computing device for layout of distribution network interconnection diagrams, which solves the problems disclosed in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A method for arranging a power distribution network interconnection diagram includes:

[0007] Obtain connection relationship data and node types from the power distribution network interconnection diagram;

[0008] The connection relationship data and node type in the power distribution network interconnection diagram are concatenated and then input into a pre-trained generative adversarial network to obtain an optimized node layout.

[0009] According to the preset connection rules, the optimized node layout is connected to obtain the distribution network connection diagram layout.

[0010] The discriminator of the generative adversarial network is equipped with a preset layout evaluation criterion. During the training process of the generative adversarial network, the discriminator uses the layout evaluation criterion to score the node layout obtained by the generator. In the adversarial learning process between the generator and the discriminator, the generator generates node layouts that conform to the layout evaluation criterion.

[0011] The layout evaluation criteria include the evaluation criteria for the dispersion of substation nodes, the evaluation criteria for the connection length of tie switch nodes, the evaluation criteria for the clustering of substation nodes and their leaf nodes, and the evaluation criteria for the number of path focal points of tie switch nodes.

[0012] The criteria for evaluating the degree of dispersion of substation nodes are as follows:

[0013] ;

[0014] Where L1 is the evaluation score for whether the dispersion of substation nodes is appropriate, and D1 is the sum of the distances between substation nodes. Let W and H be the boundary values ​​of the distances between substation nodes, where W and H represent the width and height of the node layout, respectively. For combination operations, it means selecting 2 nodes from n substation nodes as a group;

[0015] The criteria for judging the connection length of the interconnecting switch node are as follows:

[0016] ;

[0017] ;

[0018] Where D2 is the total connection length of the tie switch node, and L2 is the evaluation score for whether the connection length of the tie switch node is appropriate;

[0019] The criteria for evaluating the clustering degree of substation nodes and their leaf nodes are as follows:

[0020] ;

[0021] ;

[0022] Where S is the total area of ​​the distribution network interconnection diagram, s i Let L1 be the area of ​​the rectangle occupied by the i-th substation node and its leaf nodes. The rectangle is the smallest cuboid that can completely enclose the i-th substation node and its leaf nodes. n is the number of substation nodes, and L3 is the evaluation score of the degree of aggregation of the substation nodes and their leaf nodes.

[0023] The criteria for evaluating the number of path focal points of a connecting switch node are as follows:

[0024] .

[0025] The mean squared distance is used as a measure of difference in the loss function of generative adversarial networks.

[0026] The loss function for generative adversarial networks is:

[0027] ;

[0028] Where D is the discriminator, G is the generator, D(X) is the output of the discriminator after the data X passes through it, D(G(z)) is the output of the discriminator after the random noise z is generated by the generator and then passed through it, and log represents the calculation of the corresponding entropy value. To match the expected value of the data distribution, Let X represent data that conforms to the data distribution data, and z represent random noise that conforms to a normal distribution. The expected value is obtained after passing the layout data generated by random noise z through a discriminator.

[0029] The connection rules include the connection rules between the root node and leaf nodes, and the connection rules between leaf nodes.

[0030] The rules for connecting the root node and leaf nodes:

[0031] The root node is only connected to its own leaf nodes;

[0032] Draw a plane coordinate system with the root node as the origin; if the leaf node is located in quadrant 1 or 2 relative to the root node, the inflection point of the leaf node and the root node is on the positive half of the y-axis, and the line connecting them is horizontal and vertical; if the leaf node is located in quadrant 3 or 4 relative to the root node, the inflection point of the leaf node and the root node is on the negative half of the y-axis, and the line connecting them is horizontal and vertical.

[0033] Rules for connecting leaf nodes:

[0034] 1) If the leaf nodes are orthogonal to each other and there are no other nodes in the direct path, then connect them directly; otherwise, go to 2).

[0035] 2) If the leaf nodes are located in non-orthogonal positions to each other, and there are no other nodes on the path of the connection according to the rules for connecting the root node and the leaf node, then connect them according to the rules for connecting the root node and the leaf node; otherwise, go to 3).

[0036] 3) If the connection path according to the rules for connecting the root node and the leaf node passes through other nodes, then add a turning point on the path to make the connection horizontal and vertical and without any nodes blocking it.

[0037] A power distribution network interconnection diagram layout system, comprising:

[0038] The acquisition module retrieves connection relationship data and node types from the distribution network interconnection diagram;

[0039] The node layout optimization module concatenates the connection relationship data and node types in the distribution network interconnection diagram and inputs them into a pre-trained generative adversarial network to obtain an optimized node layout.

[0040] The connection module connects the optimized node layout according to preset connection rules to obtain the distribution network connection diagram layout.

[0041] In the node layout optimization module, the discriminator of the generative adversarial network is equipped with a preset layout evaluation criterion. During the training process of the generative adversarial network, the discriminator uses the layout evaluation criterion to score the node layout obtained by the generator. In the adversarial learning process between the generator and the discriminator, the generator generates a node layout that conforms to the layout evaluation criterion. The loss function of the generative adversarial network uses the mean squared distance as a measure of difference.

[0042] A computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a power distribution network interconnection diagram layout method.

[0043] A computing device includes one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a distribution network tie-down diagram layout method.

[0044] The beneficial effects achieved by this invention are as follows: 1. This invention uses generative adversarial networks to optimize node layout and connects the optimized node layout to obtain the distribution network connection diagram layout. Compared with traditional methods, it is not limited by experience and can be applied to increasingly complex topologies. 2. The discriminator of the generative adversarial network of this invention is equipped with layout evaluation criteria. The evaluation criteria can be modified according to specific business needs. The setting of diverse evaluation criteria enables the system to have personalized design capabilities. 3. This invention uses the mean squared error distance as the loss function, enabling the generator of the generative adversarial network to generate layouts that better conform to the evaluation criteria, thereby realizing the constraint of the evaluation function on the generator and controlling the learning direction of the generator. Attached Figure Description

[0045] Figure 1 A flowchart of the distribution network interconnection diagram layout method;

[0046] Figure 2 A logical block diagram of the distribution network interconnection diagram layout method;

[0047] Figure 3 A schematic diagram illustrating the principle of generating adversarial network training;

[0048] Figure 4 This is a diagram showing the connections between the root node and the leaf nodes.

[0049] Figure 5 A diagram showing the connections between leaves;

[0050] Figure 6 Choose a logic diagram for the connections between leaf nodes;

[0051] Figure 7 The graph layout generated by the original Generative Adversarial Network (GAN);

[0052] Figure 8 The graph layout generated by a variant of the generative adversarial network WGAN;

[0053] Figure 9 To generate a graph layout using an adversarial network through this invention;

[0054] Figure 10 The graph layout is generated by a multilayer perceptron (MLP). Detailed Implementation

[0055] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0056] like Figure 1 As shown, a method for arranging a distribution network interconnection diagram includes the following steps:

[0057] Step 1: Obtain the connection relationship data and node types in the distribution network interconnection diagram;

[0058] Step 2: After concatenating the connection relationship data and node types in the distribution network interconnection diagram, input the data into a pre-trained generative adversarial network to obtain an optimized node layout;

[0059] Step 3: Connect the optimized node layout according to the preset connection rules to obtain the distribution network connection diagram layout.

[0060] The above method uses generative adversarial networks to optimize node layout and connects the optimized node layout to obtain the distribution network interconnection diagram layout. Compared with traditional methods, it is not limited by experience and can be applied to the increasingly complex topology structures.

[0061] The aforementioned generative adversarial networks (GANs) require pre-construction and training. Traditional GANs learn unstable layouts, failing to guarantee that the generator learns aesthetically pleasing and clear layout methods. Furthermore, traditional GAN ​​discriminators cannot finely distinguish the specific differences of each distribution, resulting in unstable and highly fluctuating generation performance. Therefore, improvements to traditional networks are necessary, as detailed below:

[0062] a) The discriminator of the generative adversarial network is equipped with preset layout evaluation criteria;

[0063] During the training of a generative adversarial network, the discriminator uses a layout evaluation criterion to score the node layouts obtained by the generator. The purpose is to use the "evaluation score" feature to distinguish between training data and generated data: for training data, the discriminator's output is close to a perfect score of 1, while for layouts generated by the generator that are not good enough, the discriminator scores them as 0. During the adversarial learning process between the generator and the discriminator, the generator is made to generate node layouts that meet the layout evaluation criterion, that is, the generated layout will receive a perfect score from the discriminator.

[0064] To generate more granular learning constraints for adversarial network design, namely the layout evaluation criteria for visualization effects, we mainly quantify the discreteness of the root node, the clustering of the root and leaf nodes, and the total complexity of global connections, etc., and transform the generated layout into an aesthetic score to influence the optimization direction.

[0065] The layout evaluation criteria mainly include the evaluation criteria for the dispersion of substation nodes, the evaluation criteria for the connection length of tie switch nodes, the evaluation criteria for the clustering of substation nodes and their leaf nodes, and the evaluation criteria for the number of path focal points of tie switch nodes.

[0066] Criteria for assessing the degree of dispersion of substation nodes:

[0067] The degree of dispersion of substation nodes is the most important factor for visualization effects. It can characterize the aesthetics and rationality of the visualization. At the same time, under the constraints of the connection length of tie switches, it can maintain an appropriate degree of sparsity in the topology layout and wiring, achieving an aesthetically pleasing overall layout.

[0068] Let n represent the number of substation nodes, the distance between substation nodes, and the sum of these values. ,in, Let W and H be the boundary values ​​of the distances between substation nodes, where W and H represent the width and height of the node layout, respectively. For combination operations, it means selecting 2 nodes from n substation nodes as a group.

[0069] Evaluation score of the degree of dispersion of substation nodes , .

[0070] Criteria for evaluating the connection length of the interconnecting switch node:

[0071] For ease of review, the visualization results should be as concise as possible, in which the length of the lines in the topology graph plays a relatively important role; the length of the lines connecting nodes can represent the distance between their positions, constrain the relative position of each leaf node, especially those connected nodes, constraining the distance between them; it is mutually constrained with the degree of dispersion of substation nodes, and at the same time constrains the aesthetics of the overall layout and the simplicity of the connections.

[0072] Total connection length of the interconnection switch node Evaluation score for the connection length of the contact switch node , .

[0073] Criteria for evaluating the clustering degree of substation nodes and their leaf nodes:

[0074] The visualization effect should be able to clearly distinguish different substation nodes. Therefore, nodes belonging to the same substation should be clustered together with that substation. Otherwise, the connection between different leaf nodes will lead to more complex and redundant intersections. At the same time, under the constraint of the degree of substation dispersion, different substation nodes and their respective leaf nodes will be clustered together to avoid the occurrence of large-span connections.

[0075] Using S as the total area of ​​the distribution network interconnection diagram, s i Let be the area of ​​the rectangle occupied by the i-th substation node and its leaf nodes. The rectangle is the smallest cuboid that can completely enclose the i-th substation node and its leaf nodes. ;

[0076] When the distribution is good, the total area occupied by substation nodes and leaf nodes should be close to or even less than the total area of ​​the layout. This means that the node distribution is not highly dispersed and the root and leaf nodes are highly clustered. Therefore, in the optimization process, we aim to make the occupied area less than the total canvas area, and evaluate the clustering degree of substation nodes and their leaf nodes by score. , .

[0077] Criteria for evaluating the number of path focal points of a connecting switch node:

[0078] Visualization should prioritize clear lines for easy tracing. Excessive intersections between nodes lead to cluttered lines and hinder review; therefore, reducing intersections improves layout visibility. Based on the range of evaluation factors, for any layout's score L, it is necessary to satisfy... .

[0079] When a layout conforms more closely to the layout evaluation criteria, its score is closer to 1, indicating a more reasonable and aesthetically pleasing layout. Conversely, when a layout does not conform to the evaluation criteria, its score is closer to 0, indicating an extremely unreasonable layout.

[0080] This invention uses the evaluation score of the layout as a constraint for training and the evaluation score as the basis for the discriminator to distinguish the differences between different distributions in a fine-grained manner. This allows the training to be optimized in the expected direction and ultimately obtains the layout capability guided by the evaluation method.

[0081] b) Due to the difference in granularity, this invention also changes the loss function of the generative adversarial network, using the mean squared distance as a measure of difference, that is, using the mean squared distance as the loss function, which can adapt to more precise training optimization control.

[0082] The mean squared distance is the absolute value of the square of the difference between the target value and the predicted value, and its mathematical expression is as follows:

[0083] ;

[0084] Where m represents a total of m samples. Let y represent the predicted label for the j-th sample. j This represents the true label of the j-th sample. This statistical parameter is the mean of the sum of squared errors between corresponding points in the predicted data and the original data.

[0085] The improved generative adversarial network is a system consisting of two models: a discriminator D. The generator G and the discriminator G are used to generate target data based on random noise, with the ultimate goal of generating data that the discriminator will classify as real. The discriminator, on the other hand, determines whether the data is real or generated, aiming to distinguish between the two types of data as much as possible. The purpose of the dual-model approach is for the generator to confuse the discriminator as much as possible, while simultaneously allowing the discriminator to determine the source of the input data. The two models are in a competitive relationship, each trying to improve itself by defeating the other. The generator can obtain feedback from the discriminator regarding whether its generated data matches the distribution of the positive dataset, while the discriminator can obtain more training samples from the generator.

[0086] The training process of a Generative Adversarial Network (GAN) can be viewed as solving a problem of finding the maxima and minima of a value function. Training a GAN essentially involves finding the optimal parameters for the generator and discriminator, such that the value function has a minimum and maximum value. The adversarial loss function of a GAN can be expressed by the formula:

[0087] ;

[0088] Where D is the discriminator, G is the generator, D(X) is the output of the discriminator after the data X passes through it, D(G(z)) is the output of the discriminator after the random noise z is generated by the generator and then passed through it, and log represents the calculation of the corresponding entropy value. To match the expected value of the data distribution, Let X represent data that conforms to the data distribution data, which here means training data with a relatively reasonable layout, and z represent random noise that conforms to a normal distribution. The expected value is obtained after passing the layout data generated by random noise z through a discriminator.

[0089] The specific objectives are: 1. To find an optimal discriminator that can better distinguish the distribution differences between real data and data generated by the generator; 2. To find an optimal generator whose generated data feature distribution is closer to that of the real data, meaning the generated data and the real data are closer in high-dimensional space and have smaller differences, making it difficult for the discriminator to distinguish them. Combining the above methods for finding maxima, the overall process can be understood as finding the minimum and maximum values ​​of a value function.

[0090] Improved generative adversarial networks (GANs) focus on enabling the generator to learn layout patterns with certain structural constraints based on reasonable real-world layouts. This prevents the generator from producing layouts that are too unexpected or violate basic constraints, such as multiple nodes not being in the same position. If aesthetically pleasing layouts are required, then visually appealing layout criteria need to be incorporated as constraints into the training optimization process. This allows the generator to gradually approach the layout generation pattern that best conforms to the layout criteria under these loss constraints.

[0091] Traditional generative adversarial networks (GANs) compare the distribution differences between two datasets. Adversarial learning allows the generator to learn certain data distribution characteristics, while the discriminator learns the differences between the two data distributions. In layout generation tasks, the difference in evaluation scores can be viewed as a difference in the distributions of the two datasets, but the granularity of this difference cannot guide the generator to learn a clearer generation pattern.

[0092] The model S scores the layout; the better it meets the evaluation criteria, the higher the score. The loss is calculated using the mean squared error distance (MSE). The discriminator is solved as follows:

[0093] ;

[0094] The generator is solved as follows:

[0095] ;

[0096] Where MSE represents the mean squared error (MSE) of the loss calculation method, z represents random noise, and x represents data that conforms to the true distribution. This indicates that the generator G is selected to obtain the minimum value in equation (—);

[0097] During adversarial learning, the discriminator continuously approaches S, eventually acquiring the ability to judge S. Simultaneously, to ensure the generated layout score is close to the maximum of 1, the generator learns to generate layouts that best meet the judging criteria. This constrains the generator through the judging function, thus controlling the generator's learning direction.

[0098] The layout generation algorithm based on generative adversarial networks (GANs) using mean squared error loss first extracts and condenses the features of structured data. Then, through constraints imposed by a reasonable evaluation function, it trains and learns a layout generation method that conforms to the evaluation function, ultimately achieving a reasonable layout for topological data. This algorithm yields a novel layout approach that surpasses heuristic algorithms, working effectively in various scenarios and with complex topological data. It improves the efficiency and effectiveness of topological data visualization, and the system's generation results change with the evaluation criteria, thus enabling the generation of personalized layout targets.

[0099] like Figure 3 As shown, the improved generative adversarial network (GAN) is trained using node distribution maps as training samples. Specifically, the power distribution network topology can be represented as an embedding as input to the model. During training, the discriminator strives to improve its ability to distinguish between real and fake data, while the generator tries to generate real data. The discriminator uses both real and fake data for training to update weights, while the generator can only use the complete model (a frozen discriminator with fake data as input, but labeled as real) and uses the discriminator's error to feed back to update weights. The training process of a GAN is essentially a min-max process, where the discriminator minimizes the error in distinguishing between real and fake data, while the generator maximizes the probability of the discriminator making an incorrect judgment.

[0100] like Figure 2 As shown, the connection relationships and nodes in the distribution network interconnection diagram to be laid out are obtained, where the connection relationships are represented by graph embedding. These matrices representing the topology structure are input into a trained generative adversarial network to obtain an optimized node layout. Finally, the optimized node layout is connected according to preset connection rules to obtain the distribution network interconnection diagram layout.

[0101] The connection rules include the connection rules between the root node and leaf nodes, and the connection rules between leaf nodes.

[0102] The rules for connecting the root node and leaf nodes:

[0103] Due to the connectivity between nodes in a visualization graph, the root node is only connected to its own leaf nodes, so the connection between the root node and the leaf nodes is determined by their relative relationship.

[0104] like Figure 4 As shown, draw a plane coordinate system with the root node as the origin; if the leaf node is located in quadrant 1 or 2 relative to the root node, the inflection point of the leaf node and the root node is on the positive half-axis of the y-axis, and the line connecting them is horizontal and vertical; if the leaf node is located in quadrant 3 or 4 relative to the root node, the inflection point of the leaf node and the root node is on the negative half-axis of the y-axis, and the line connecting them is horizontal and vertical.

[0105] Figure 5 The rules for connecting leaf nodes:

[0106] Considering the significant and unpredictable variations in the layout learned during training, three priority levels of connection schemes are considered to achieve better connection results even in relatively irregular layouts. These priorities decrease sequentially, as follows: Figure 6 Specifically as follows:

[0107] 1) If the leaf nodes are orthogonal to each other and there are no other nodes in the direct path, then connect them directly; otherwise, go to 2).

[0108] 2) If the leaf nodes are located in non-orthogonal positions to each other, and there are no other nodes on the path of the connection according to the rules for connecting the root node and the leaf node, then connect them according to the rules for connecting the root node and the leaf node; otherwise, go to 3).

[0109] 3) If the connection path according to the rules for connecting the root node and the leaf node passes through other nodes, then add a turning point on the path to make the connection horizontal and vertical and without any nodes blocking it.

[0110] Optimized node layouts are obtained using the original Generative Adversarial Network (GAN), a variant of GAN (WGAN), the network of this invention, and a Multilayer Perceptron (MLP). Then, the same connection rules are used to fully present the visualization effect. (See...) Figures 7-10 As can be seen from the figure, this invention can generate visually appealing layouts, which can more effectively assist power distribution network personnel in operation and maintenance management.

[0111] The above method uses an improved generative adversarial network to lay out the distribution network interconnection diagram, which can achieve a beautiful and concise visualization effect that meets the evaluation criteria, and ultimately achieve efficient, automated and aesthetically pleasing end-to-end visualization of distribution network data.

[0112] Based on the same technical solution, this invention also discloses a software system for the above method, a power distribution network interconnection diagram layout system, comprising:

[0113] The module retrieves connection relationship data and node types from the distribution network interconnection diagram.

[0114] The node layout optimization module concatenates the connection relationship data and node types in the distribution network interconnection diagram and inputs them into a pre-trained generative adversarial network to obtain an optimized node layout.

[0115] In the node layout optimization module, the discriminator of the generative adversarial network is equipped with a preset layout evaluation criterion. During the training process of the generative adversarial network, the discriminator uses the layout evaluation criterion to score the node layout obtained by the generator. In the adversarial learning process between the generator and the discriminator, the generator generates a node layout that conforms to the layout evaluation criterion. The loss function of the generative adversarial network uses the mean squared distance as a measure of difference.

[0116] The connection module connects the optimized node layout according to preset connection rules to obtain the distribution network connection diagram layout.

[0117] The data processing flow and steps of each module in the above system are consistent with those of the corresponding methods, and will not be described again here.

[0118] Based on the same technical solution, the present invention also discloses a computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform a distribution network interconnection diagram layout method.

[0119] Based on the same technical solution, the present invention also discloses a computing device, including one or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing a distribution network interconnection diagram layout method.

[0120] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0122] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0123] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0124] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. A method for arranging a power distribution network interconnection diagram, characterized in that, include: Obtain connection relationship data and node types from the power distribution network interconnection diagram; The connection relationship data and node type in the power distribution network interconnection diagram are concatenated and then input into a pre-trained generative adversarial network to obtain an optimized node layout. According to the preset connection rules, the optimized node layout is connected to obtain the distribution network connection diagram layout; The discriminator of the generative adversarial network is equipped with a preset layout evaluation criterion. During the training process of the generative adversarial network, the discriminator uses the layout evaluation criterion to score the node layout obtained by the generator. During the adversarial learning process between the generator and the discriminator, the generator generates node layouts that conform to the layout evaluation criterion. The layout evaluation criteria include the evaluation criteria for the dispersion of substation nodes, the evaluation criteria for the connection length of tie switch nodes, the evaluation criteria for the clustering of substation nodes and their leaf nodes, and the evaluation criteria for the number of path focal points of tie switch nodes. The criteria for evaluating the degree of dispersion of substation nodes are as follows: ； Where L1 is the evaluation score for whether the dispersion of substation nodes is appropriate, and D1 is the sum of the distances between substation nodes. Let W and H be the boundary values ​​of the distances between substation nodes, where W and H represent the width and height of the node layout, respectively. For combination operations, it means selecting 2 nodes from n substation nodes as a group; The criteria for judging the connection length of the interconnecting switch node are as follows: ； ； Where D2 is the total connection length of the tie switch node, and L2 is the evaluation score for whether the connection length of the tie switch node is appropriate; The criteria for evaluating the clustering degree of substation nodes and their leaf nodes are as follows: ； ； Where S is the total area of ​​the distribution network interconnection diagram, s i Let L1 be the area of ​​the rectangle occupied by the i-th substation node and its leaf nodes. The rectangle is the smallest cuboid that can completely enclose the i-th substation node and its leaf nodes. n is the number of substation nodes, and L3 is the evaluation score of the degree of aggregation of the substation nodes and their leaf nodes. The criteria for evaluating the number of path focal points of a connecting switch node are as follows: 。 2. The distribution network interconnection diagram layout method according to claim 1, characterized in that, The mean squared distance is used as a measure of difference in the loss function of generative adversarial networks.

3. The distribution network interconnection diagram layout method according to claim 2, characterized in that, The loss function for generative adversarial networks is: ； Where D is the discriminator, G is the generator, D(X) is the output of the discriminator after the data X passes through it, D(G(z)) is the output of the discriminator after the random noise z is generated by the generator and then passed through it, and log represents the calculation of the corresponding entropy value. To match the expected value of the data distribution, Let X represent data that conforms to the data distribution data, and z represent random noise that conforms to a normal distribution. The expected value is obtained after passing the layout data generated by random noise z through a discriminator.

4. The distribution network interconnection diagram layout method according to claim 1, characterized in that, The connection rules include the connection rules between the root node and leaf nodes, and the connection rules between leaf nodes. The rules for connecting the root node and leaf nodes: The root node is only connected to its own leaf nodes; Draw a plane coordinate system with the root node as the origin; if the leaf node is located in quadrant 1 or 2 relative to the root node, the inflection point of the leaf node and the root node is on the positive half of the y-axis, and the line connecting them is horizontal and vertical. If a leaf node is located in the 3rd or 4th quadrant relative to the root node, the inflection point between the leaf node and the root node is on the negative half of the y-axis, and the line connecting them is horizontal and vertical. Rules for connecting leaf nodes: 1) If the leaf nodes are orthogonal to each other and there are no other nodes in the direct path, then connect them directly; otherwise, go to 2). 2) If the leaf nodes are located in non-orthogonal positions to each other, and there are no other nodes on the path of the connection according to the rules for connecting the root node and the leaf node, then connect them according to the rules for connecting the root node and the leaf node; otherwise, go to 3). 3) If the connection path according to the rules for connecting the root node and the leaf node passes through other nodes, then add a turning point on the path to make the connection horizontal and vertical and without any nodes blocking it.

5. A distribution network interconnection diagram layout system, characterized in that, include: The acquisition module retrieves connection relationship data and node types from the distribution network interconnection diagram; The node layout optimization module concatenates the connection relationship data and node types in the distribution network interconnection diagram and inputs them into a pre-trained generative adversarial network to obtain an optimized node layout. The discriminator of the generative adversarial network is equipped with a preset layout evaluation criterion. During the training process of the generative adversarial network, the discriminator uses the layout evaluation criterion to score the node layout obtained by the generator. During the adversarial learning process between the generator and the discriminator, the generator generates node layouts that conform to the layout evaluation criterion. The layout evaluation criteria include the evaluation criteria for the dispersion of substation nodes, the evaluation criteria for the connection length of tie switch nodes, the evaluation criteria for the clustering of substation nodes and their leaf nodes, and the evaluation criteria for the number of path focal points of tie switch nodes. The criteria for evaluating the degree of dispersion of substation nodes are as follows: ； Where L1 is the evaluation score for whether the dispersion of substation nodes is appropriate, and D1 is the sum of the distances between substation nodes. Let W and H be the boundary values ​​of the distances between substation nodes, where W and H represent the width and height of the node layout, respectively. For combination operations, it means selecting 2 nodes from n substation nodes as a group; The criteria for judging the connection length of the interconnecting switch node are as follows: ； ； Where D2 is the total connection length of the tie switch node, and L2 is the evaluation score for whether the connection length of the tie switch node is appropriate; The criteria for evaluating the clustering degree of substation nodes and their leaf nodes are as follows: ； ； Where S is the total area of ​​the distribution network interconnection diagram, s i Let L1 be the area of ​​the rectangle occupied by the i-th substation node and its leaf nodes. The rectangle is the smallest cuboid that can completely enclose the i-th substation node and its leaf nodes. n is the number of substation nodes, and L3 is the evaluation score of the degree of aggregation of the substation nodes and their leaf nodes. The criteria for evaluating the number of path focal points of a connecting switch node are as follows: ； The connection module connects the optimized node layout according to preset connection rules to obtain the distribution network connection diagram layout.

6. A distribution network interconnection diagram layout system according to claim 5, characterized in that, In the node layout optimization module, the mean squared distance is used as a measure of difference in the loss function of the generative adversarial network.

7. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods according to claims 1 to 4.

8. A computing device, characterized in that, include: One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods according to claims 1 to 4.

Images