Inspection path generation method and device suitable for transformer substation, terminal equipment and storage medium

Abstract

The invention discloses a routing inspection path generation method and device suitable for a transformer substation, terminal equipment and a storage medium, and belongs to the technical field of transformer substation routing inspection.The routing inspection path generation method suitable for the transformer substation comprises the steps that site data and robot routing inspection data are input into a path planning model; according to the method and the system, a model is established, so that the model can extract site features and obstacle distribution features from site data and extract robot space occupation features and robot position features from robot inspection data, when the inspection path is generated, data of moving obstacles are considered, the moving obstacles can be effectively avoided, and the inspection efficiency is improved; according to the method, when the robot inspection path is obtained, the site features and the features of all the obstacles are considered, the features of the robot are also considered, and the problem that in the prior art, it is difficult for the robot to smoothly complete the inspection task in the complex and changeable environment can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of substation inspection, and in particular to a method, device, terminal equipment and storage medium for generating an inspection path suitable for a substation. Background Art

[0002] Substation inspections generally involve checking the equipment and environment within the substation to ensure proper operation and promptly identify potential faults and safety hazards. Planning a reasonable inspection route for the robot during the inspection process allows the robot to complete its inspection tasks quickly and safely, promptly identifying equipment anomalies within the substation and ensuring stable operation.

[0003] However, traditional substation inspection path planning technology cannot fully consider mobile obstacles and only plans paths based on fixed obstacle information. It is also unable to combine the robot's own parameters with the data of mobile obstacles for analysis. As a result, the planned path is prone to collision or interruption of the inspection process when encountering mobile obstacles during actual inspections, which not only greatly prolongs the time required for inspections and seriously reduces inspection efficiency, but also makes it difficult for robots to successfully complete inspection tasks in complex and changing environments. Summary of the Invention

[0004] The embodiments of the present invention provide a method, apparatus, terminal device and storage medium for generating an inspection path suitable for a substation, which can realize the combined analysis of the robot's own parameters and obstacle data, so that the shortest obstacle avoidance path finally generated not only takes into account the site characteristics and the characteristics of each obstacle, but also takes into account the characteristics of the robot itself, so that the robot can better adapt to the complex inspection site environment during inspection, and can effectively solve the problem in the prior art that robots have difficulty in successfully completing inspection tasks in complex and changeable environments.

[0005] An embodiment of the present invention provides a method for generating an inspection path applicable to a substation, comprising:

[0006] Obtaining site data corresponding to the substation; wherein the site data includes: site size data, fixed obstacle data, and mobile obstacle data; the mobile obstacle data includes: data corresponding to obstacles that move dynamically in the site;

[0007] Obtaining robot inspection data corresponding to the inspection robot; wherein the robot inspection data includes: the robot starting position, the robot end position and the robot size;

[0008] Inputting the site data and the robot inspection data into a path planning model, so that the path planning model extracts site features and obstacle distribution features from the site data, extracts robot space occupancy features from the robot inspection data, and robot position features used to characterize the relative positions of a start point and an end point in the site; and generating a robot inspection path based on the site features, obstacle distribution features, robot space occupancy features, and robot position features.

[0009] The robot inspection path is used to represent the shortest obstacle avoidance path when the robot inspects in the substation.

[0010] Preferably, the field size data includes: field length and field width; the fixed obstacle data includes: fixed obstacle position and fixed obstacle size corresponding to each fixed obstacle; the mobile obstacle data includes: mobile obstacle position, mobile obstacle size and obstacle moving path corresponding to each mobile obstacle;

[0011] The site characteristics include: site area characteristics and site shape characteristics; the obstacle distribution characteristics include: obstacle density characteristics and relative position relationship characteristics between obstacles;

[0012] The path planning model extracts site features and obstacle distribution features from the site data, including:

[0013] Generate site area characteristics based on site length and site width through the path planning model;

[0014] Generating a site shape feature according to the site area feature, the site length, and the site width;

[0015] Generate an obstacle density feature based on the position of each fixed obstacle, the size of each fixed obstacle, the position of each mobile obstacle, and the size of each mobile obstacle;

[0016] The relative position relationship features between the obstacles are generated according to the positions of each fixed obstacle, the positions of each mobile obstacle, and the obstacle movement path of each mobile obstacle.

[0017] Preferably, extracting the robot space occupancy feature and the robot position feature for characterizing the relative positions of the starting point and the end point in the site from the robot inspection data includes:

[0018] Extract the shape features of the robot based on the robot's starting position and size;

[0019] generating a robot space occupancy feature for characterizing an area occupied by the robot in a venue according to the shape feature of the robot and the size of the robot;

[0020] According to the robot's starting position and the robot's ending position, the relative position relationship between the starting point and the ending point in the field is extracted;

[0021] According to the relative position relationship, a robot position feature is generated for characterizing the relative positions of the starting point and the end point in the field.

[0022] Preferably, generating the robot inspection path according to the site characteristics, obstacle distribution characteristics, robot space occupancy characteristics, and robot position characteristics includes:

[0023] Generating local features for characterizing a path feasibility area based on the site shape features, obstacle density features, and robot size features;

[0024] Generate global features for characterizing the robot's obstacle avoidance path segments in the field based on the robot's position features and the relative position relationship features between obstacles;

[0025] Fusing the local features and the global features to generate fused features;

[0026] A robot inspection path is generated based on the fused features.

[0027] Preferably, the generation of the path planning model includes:

[0028] Use the site sample data and robot inspection sample data as training samples;

[0029] Taking several training samples and the actual robot inspection path of each training sample as input, and the predicted robot inspection path corresponding to each training sample as output, the path planning model to be trained is iteratively trained until the model converges, generating a trained path planning model.

[0030] Preferably, before each iterative training, the method further includes:

[0031] Obtain the network parameters to be updated corresponding to the last training of the path planning model; wherein, during the initial training, the network parameters to be updated are randomly generated;

[0032] Repeat the position update operation until the current update number is the same as the preset update number, and output the position corresponding to the target animal; wherein the position is used to represent the network parameters in the path planning model;

[0033] Using the position corresponding to the target animal as a network parameter for training the path planning model;

[0034] The location update operation includes:

[0035] When the current update number is less than the preset update number, a plurality of current parameters are generated based on the current network parameter, the preset parameter selection range and the network parameter to be updated; each of the current parameters is then used as the current position corresponding to each animal in the population; wherein, initially, the network parameter to be updated is used as the current network parameter;

[0036] Under the network parameter selection constraint conditions corresponding to the path planning model, the initial predation position corresponding to the predator is randomly generated;

[0037] Taking the initial predation position as the center, and according to a preset predation position range, generating a plurality of target predation positions involved in the predator's predation behavior;

[0038] For each target predation location, a range of positions to be escaped is generated based on a preset distance range and the target predation location, with the target predation location as the center; animals within the range of positions to be escaped are marked as animals to be escaped; wherein the preset predation location range is larger than the preset distance range;

[0039] For each animal to be escaped, an escape simulation is performed on the animal to be escaped by simulating an escape strategy of the animal to be escaped from a predator, and an updated position corresponding to the animal to be escaped after the escape is generated;

[0040] According to the preset model loss function, the loss function value corresponding to the current position of each current animal in the population is calculated; according to the preset model loss function, the loss function value corresponding to the updated position of each animal to be escaped in the population is calculated;

[0041] Extract several parent animals according to the loss function value corresponding to each current animal and the loss function value corresponding to each animal to be escaped;

[0042] Perform hybridization operations on each parent animal to generate several offspring animals; then calculate the loss function value corresponding to the position of each offspring animal according to the preset model loss function;

[0043] According to the loss function value corresponding to each current animal, the loss function value corresponding to each animal to be escaped, and the loss function value corresponding to each offspring animal, the animal with the smallest loss function value is marked as the target animal;

[0044] Using the position corresponding to the target animal as the current network parameter for the next position update operation;

[0045] Add the preset iteration increment to the current update iteration value.

[0046] Preferably, the updated position of the escaping animal after escaping is determined according to the following formula:

[0047] X2=X1+rand×(Low+R×(High-Low));

[0048] Among them, X2 is the updated position corresponding to the animal to be escaped after escaping, X1 is the current position corresponding to the animal to be escaped before escaping, rand is a preset random number range; Low and High are the lower and upper bounds of the local search space, respectively, and the local search space is used to limit the escape range of the animal to be escaped; R is a random number used for local search.

[0049] Based on the above method embodiments, the present invention provides corresponding device embodiments.

[0050] An embodiment of the present invention provides an inspection path generation device applicable to a substation, comprising: a site data acquisition module, a robot inspection data acquisition module, and an inspection path generation module;

[0051] The site data acquisition module is used to acquire site data corresponding to the substation; wherein the site data includes: site size data, fixed obstacle data and mobile obstacle data; the mobile obstacle data includes: data corresponding to obstacles that move dynamically in the site;

[0052] The robot inspection data acquisition module is used to acquire the robot inspection data corresponding to the inspection robot; wherein the robot inspection data includes: the robot starting position, the robot end position and the robot size;

[0053] The inspection path generation module is configured to input the site data and the robot inspection data into a path planning model, so that the path planning model extracts site characteristics and obstacle distribution characteristics from the site data, extracts robot space occupancy characteristics from the robot inspection data, and robot position characteristics used to characterize the relative positions of the starting point and the end point in the site; and generates a robot inspection path based on the site characteristics, obstacle distribution characteristics, robot space occupancy characteristics, and robot position characteristics.

[0054] The robot inspection path is used to represent the shortest obstacle avoidance path when the robot inspects in the substation.

[0055] Based on the above method embodiments, the present invention provides corresponding terminal device embodiments.

[0056] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a patrol path generation method suitable for a substation as described in the above-mentioned embodiment of the invention.

[0057] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments.

[0058] Another embodiment of the present invention provides a storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a patrol path generation method applicable to a substation as described in the above-mentioned embodiment of the invention.

[0059] The following beneficial effects are achieved by implementing the present invention:

[0060] Embodiments of the present invention provide a method, apparatus, terminal device, and storage medium for generating inspection paths for substations. First, the present invention acquires site data encompassing site dimensions, fixed obstacle data, and mobile obstacle data, and simultaneously acquires robot inspection data encompassing the robot's starting and ending positions and robot dimensions. Next, the acquired site data and robot inspection data are input into a path planning model, enabling the model to extract site characteristics and obstacle distribution characteristics from the site data, and to extract robot spatial occupancy characteristics and robot position characteristics from the robot inspection data. Finally, through a comprehensive analysis of these characteristics, the model can generate a robot inspection path, i.e., the shortest obstacle-avoiding path for the robot during its inspection of the substation. Compared to conventional technologies, the present invention considers not only fixed obstacle data but also mobile obstacle data during the path generation process. This allows the model-generated path to effectively avoid mobile obstacles, minimizing the risk of collisions between the robot and the obstacles. This significantly shortens inspection time and improves inspection efficiency. Moreover, after extracting the robot's space occupancy characteristics and robot position characteristics based on the robot's own parameters, the above-mentioned robot's own characteristics can be combined with the site characteristics and obstacle distribution characteristics for analysis, so that the robot's own parameters can be combined with the obstacle data for analysis, so that the shortest obstacle avoidance path finally generated not only takes into account the site characteristics and the characteristics of each obstacle, but also takes into account the characteristics of the robot itself, so that the robot can better adapt to the complex inspection site environment during inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 The present invention provides a flowchart of a method for generating an inspection path applicable to a substation, provided in accordance with an embodiment of the present invention.

[0062] Figure 2 It is a schematic diagram of the algorithm flow of network parameter optimization provided by one embodiment of the present invention.

[0063] Figure 3 The present invention is a schematic structural diagram of an inspection path generation device applicable to a substation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0065] like Figure 1 As shown, in order to solve the problem in the prior art that robots have difficulty in successfully completing inspection tasks in complex and changing environments, an embodiment of the present invention provides a method for generating an inspection path applicable to a substation, comprising:

[0066] Step S1: Acquire site data corresponding to the substation; wherein the site data includes: site size data, fixed obstacle data, and mobile obstacle data; the mobile obstacle data includes: data corresponding to obstacles that move dynamically in the site;

[0067] Step S2: Obtaining robot inspection data corresponding to the inspection robot; wherein the robot inspection data includes: the robot starting position, the robot end position and the robot size;

[0068] Step S3: Inputting the site data and the robot inspection data into a path planning model, so that the path planning model extracts site features and obstacle distribution features from the site data, extracts robot space occupancy features from the robot inspection data, and robot position features used to characterize the relative positions of a start point and an end point in the site; and generates a robot inspection path based on the site features, obstacle distribution features, robot space occupancy features, and robot position features.

[0069] The robot inspection path is used to represent the shortest obstacle avoidance path when the robot inspects in the substation.

[0070] Regarding step S1, in a preferred embodiment, the field size data includes: field length and field width;

[0071] The fixed obstacle data includes: the fixed obstacle position and fixed obstacle size corresponding to each fixed obstacle; the mobile obstacle data includes: the mobile obstacle position, mobile obstacle size and obstacle movement path corresponding to each mobile obstacle;

[0072] It can be understood that the present invention can obtain the size of the site, as well as the position, size and moving path of various obstacles in the site. Through the obtained site length and site width, the position and size of fixed obstacles, and the position, size and moving path of mobile obstacles, an accurate substation site model can be constructed, so that the path planning model can draw a substation environment map based on the above site, and then a patrol path that adapts to the current substation environment can be planned based on the overall spatial layout, obstacle distribution and movement of the substation.

[0073] In a preferred embodiment, the present invention can record the length, width, and height of the substation inspection scene, record the position and size of fixed obstacles, record the size of various types of robots, and record the activity area of ​​mobile obstacles (such as staff), thereby establishing a grid map corresponding to the substation and updating the environmental map corresponding to the substation in real time.

[0074] For step S2, in a preferred embodiment, the present invention can collect the dimensions of various types of robots including length and width in advance, so that based on the current substation, various data of the robots to be inspected in the current substation can be obtained.

[0075] In order to ensure the feasibility of path planning, the embodiment of the present invention can obtain the size (length and width) of the robot to determine the space occupied by the robot during movement, thereby avoiding the planned path being too narrow, which makes it impossible for the robot to pass during actual inspection.

[0076] By specifying the robot's starting and ending positions, the present invention can determine the optimal travel direction and distance from the starting point to the end point. This, combined with the site environment and obstacle distribution, allows subsequent models to find the shortest obstacle-avoiding path, reducing unnecessary robot movement and thus improving inspection efficiency. For example, if the starting and ending positions are not considered, the robot may move in a circuitous manner within the substation, wasting time and energy. However, the present invention uses precise starting and ending position information to guide the model in planning a more efficient path.

[0077] Regarding step S3, in a preferred embodiment, the path planning model of the present invention may include a feature extraction network layer and an output layer. The feature extraction network layer can extract site features and obstacle distribution features corresponding to the site data, as well as robot space occupancy features and robot position features. The output layer is used to generate the robot inspection path based on the site features, obstacle distribution features, robot space occupancy features, and robot position features.

[0078] In a preferred embodiment, the feature extraction network layer of the present invention can use a convolutional neural network (CNN) to extract various features, and then use a fully connected layer as an output layer to generate a path.

[0079] Specifically, the present invention can process the substation environment map into a grid map, assuming that the grid map size is m*n; the site data input to the model and the robot's own data can be encoded using a one-hot encoding method, that is, the obstacles in the map can be represented by an m*n one-dimensional vector, where 0 represents a movable grid and 1 represents the grid where the obstacle is located.

[0080] The robot's starting point and target point can also be represented by a one-dimensional vector of size m*n, which contains two 1s to represent the starting point and target point, and the rest are all 0s. These two one-dimensional vectors of size m*n are spliced ​​together to obtain a one-dimensional vector of size 2*m*n as the input of the model. The model output is also one-hot encoded, with a size of m*n, which is the shortest path connecting the starting point and target point of the map and avoiding obstacles (that is, the final output robot inspection path). The corresponding value of the grid passed by the path is 1, and the other positions are 0.

[0081] Furthermore, the present invention can use the convolution kernel in the convolution layer in the feature extraction layer to identify local features in the grid map, mainly including the continuity of obstacles, the relative positions of the starting point and the target point, etc., so that the model can understand the feasibility of the path.

[0082] Furthermore, the output layer of the present invention can also capture the global information of the path through the pooling layer and the fully connected layer therein, such as how to reach the target point from the starting point while avoiding obstacles. The path planning model of the present invention can learn the relative position between the starting point and the target point in the feature map, and then generate a reasonable path.

[0083] In a preferred embodiment, the site characteristics of the present invention may include: site area characteristics and site shape characteristics;

[0084] The obstacle distribution characteristics include obstacle density characteristics and relative position relationship characteristics between obstacles.

[0085] Then, when extracting the site features and obstacle distribution features from the site data, it specifically includes:

[0086] Generate site area characteristics based on site length and site width through the path planning model;

[0087] Generating a site shape feature according to the site area feature, the site length, and the site width;

[0088] Generate an obstacle density feature based on the position of each fixed obstacle, the size of each fixed obstacle, the position of each mobile obstacle, and the size of each mobile obstacle;

[0089] The relative position relationship features between the obstacles are generated according to the positions of each fixed obstacle, the positions of each mobile obstacle, and the obstacle movement path of each mobile obstacle.

[0090] Illustratively, by extracting site area and shape features, the present invention allows for a macroscopic understanding of the overall spatial characteristics of a substation site. The obstacle density feature intuitively reflects the density of obstacles within the site, allowing for the prioritization of areas with sparse obstacles during path planning, reducing the risk of collisions between robots and obstacles. Furthermore, by extracting features regarding the relative positional relationships between obstacles, the present invention enables the path planning model to predict the dynamic changes in environmental obstacles, enabling better planning of obstacle-avoidance paths.

[0091] Furthermore, when extracting the robot space occupancy feature and the robot position feature used to characterize the relative positions of the starting point and the end point in the site from the robot inspection data, it specifically includes:

[0092] Extract the shape features of the robot based on the robot's starting position and size;

[0093] generating a robot space occupancy feature for characterizing an area occupied by the robot in a venue according to the shape feature of the robot and the size of the robot;

[0094] According to the robot's starting position and the robot's ending position, the relative position relationship between the starting point and the ending point in the field is extracted;

[0095] According to the relative position relationship, a robot position feature is generated for characterizing the relative positions of the starting point and the end point in the field.

[0096] Schematically, by extracting shape features from the robot's starting position and dimensions, and then generating a robot space occupancy feature, we can better analyze the robot's actual space requirements in the field. The generated robot position features can also clarify the target direction of path planning for the path planning model.

[0097] Therefore, the robot's spatial occupancy and positional characteristics provide the path planning model with detailed information about the robot itself, enabling the model to develop a path planning solution more suitable for the robot in question, tailored to the robot's specific characteristics. Different robots vary in size and shape, occupying different spaces, and their starting and ending locations also vary depending on the specific inspection task. By extracting these characteristics, embodiments of the present invention can enhance the adaptability of path planning for different robots and different inspection tasks.

[0098] Finally, when generating the robot inspection path, the specific process of the embodiment of the present invention is as follows:

[0099] Based on the obtained site shape features, obstacle density features and robot size features, local features are generated to characterize the path feasibility area;

[0100] Generate global features for characterizing the robot's obstacle avoidance path segments in the field based on the robot's position features and the relative position relationship features between obstacles;

[0101] Fusing the local features and the global features to generate fused features;

[0102] A robot inspection path is generated based on the fused features.

[0103] As can be understood, by first generating local features, the present invention can fully consider the spatial characteristics of the local area, the distribution of obstacles, and the robot's own ability to pass, thereby accurately determining the feasible path area. Then, by generating global features, it can grasp the robot's task requirements and the distribution of obstacles in the entire venue from a macro perspective, ensuring that the planned path can smoothly reach the destination from the starting point and avoid all obstacles.

[0104] In a substation environment, inspection conditions are complex and ever-changing, with distinct characteristics across different regions. This embodiment of the present invention enables path planning to better adapt to environmental diversity by first generating local and global features separately and then fusing them. For example, when a localized change in the environment occurs, such as the sudden appearance of a new obstacle in a certain area, the derived local features enable a rapid response, reassessing the path feasibility in that area and promptly adjusting the global path plan through the fusion mechanism. This enhances the path planning's adaptability to environmental and task changes.

[0105] In a preferred embodiment, the training process and generation process of the path planning model adopted by the present invention specifically include:

[0106] First, the site sample data and robot inspection sample data are used as training samples;

[0107] Taking several training samples and the actual robot inspection path of each training sample as input, and the predicted robot inspection path corresponding to each training sample as output, the path planning model to be trained is iteratively trained until the model converges, generating a trained path planning model.

[0108] Instructively, when dividing the dataset into a training set, a validation set, and a test set, a ratio of 70%-20%-10% is followed, and the dataset can be labeled, for example, with data such as motion paths, obstacles, and starting positions.

[0109] In a preferred embodiment, a convolutional neural network can be selected as the network architecture of the present invention, ReLU can be selected as the activation function, SGD can be selected as the optimizer, mean square error (MSE) can be selected as the loss function, and accuracy can be selected as the evaluation indicator to evaluate the performance of the model. The number of iterations is set to 100,000, L1 regularization is used to prevent overfitting, and during the training process, the network structure is pruned to reduce the complexity of the model.

[0110] In a preferred embodiment, the present invention can update the path planning model network parameters based on optimization algorithms similar to biological evolution and animal behavior simulation, thereby finding more optimal model parameters and improving model performance.

[0111] Before each iterative training of the model, the present invention can also obtain the network parameters to be updated corresponding to the last training of the path planning model, so as to update the network parameters of the last training and optimize the network parameters of the model; wherein, during the initial training, the network parameters to be updated can be randomly generated;

[0112] Repeat the position update operation until the current update number is the same as the preset update number, and output the position corresponding to the target animal; wherein the position is used to represent the network parameters in the path planning model;

[0113] Finally, the position corresponding to the target animal is used as the network parameter for training the path planning model;

[0114] The location update operation is repeated, specifically including:

[0115] The following position update operation can be repeated until the current update number is the same as the preset update number, and the position corresponding to the target animal is output:

[0116] When the current update number is less than the preset update number, a plurality of current parameters are generated based on the current network parameter, the preset parameter selection range and the network parameter to be updated; each of the current parameters is then used as the current position corresponding to each animal in the population; wherein, initially, the network parameter to be updated is used as the current network parameter;

[0117] Under the network parameter selection constraint conditions corresponding to the path planning model, the initial predation position corresponding to the predator is randomly generated;

[0118] Taking the initial predation position as the center, and according to a preset predation position range, generating a plurality of target predation positions involved in the predator's predation behavior;

[0119] For each target predation location, a range of positions to be escaped is generated based on a preset distance range and the target predation location, with the target predation location as the center; animals within the range of positions to be escaped are marked as animals to be escaped; wherein the preset predation location range is larger than the preset distance range;

[0120] For each animal to be escaped, an escape simulation is performed on the animal to be escaped by simulating an escape strategy of the animal to be escaped from a predator, and an updated position corresponding to the animal to be escaped after the escape is generated;

[0121] According to the preset model loss function, the loss function value corresponding to the current position of each current animal in the population is calculated; according to the preset model loss function, the loss function value corresponding to the updated position of each animal to be escaped in the population is calculated;

[0122] According to the loss function value corresponding to each current animal and the loss function value corresponding to each animal to be escaped, a number of parent animals are extracted; schematically, when extracting a number of parent animals according to the loss function value corresponding to each current animal (i.e., an animal not threatened by predation) and the loss function value corresponding to each animal to be escaped (i.e., an animal threatened by predation and undergoing escape simulation), a number of animals with the smallest loss function value can be directly selected as parent animals;

[0123] Perform hybridization operations on each parent animal to generate several offspring animals; then calculate the loss function value corresponding to the position of each offspring animal according to the preset model loss function;

[0124] According to the loss function value corresponding to each current animal, the loss function value corresponding to each animal to be escaped, and the loss function value corresponding to each offspring animal, the animal with the smallest loss function value is marked as the target animal;

[0125] Using the position corresponding to the target animal as the current network parameter for the next position update operation;

[0126] Add the preset iteration increment to the current update iteration value.

[0127] Illustratively, network parameters include: number of network layers, number of neurons, learning rate, and batch size. The network parameter selection constraints that satisfy the path planning model are: 7-9 network layers, 500-800 neurons, a learning rate of 0.01-0.03, and a batch size of 30-64.

[0128] It is understood that by simulating the natural selection and predation-escape behavior of animals, the present invention can rapidly explore the parameter space and find network parameter configurations with a low loss function value. Furthermore, by introducing escape strategies and hybridization operations, it is possible to escape from local optimal solutions and find more globally optimal solutions.

[0129] Therefore, each time the present invention updates the network parameters, that is, updates the position of the animals, it simulates the group behavior of the animals, realizes the simulation of the animals escaping from predators and hybridization operations, and can continuously ensure that the quality of the animals in the population is continuously improved, thereby achieving efficient optimization of the network parameters of the path planning model.

[0130] Illustratively, in a substation environment, where multiple dynamic factors exist, such as moving obstacles, the present embodiment updates parameters through multiple iterations, each adjusting the parameters based on the current environmental characteristics (reflected in the loss function) and simulated animal behavior. As the environment changes, the loss function value changes accordingly, prompting the model to reevaluate and optimize the parameters, allowing the path planning model to quickly adapt to the new environment and plan a more reasonable path.

[0131] In a preferred embodiment, when generating a plurality of current parameters according to the preset parameter selection range and the network parameters to be updated, the present invention can be solved according to the following position update formula:

[0132] X P1 =X i +rand×(DX i );

[0133] Among them, X P1 is the current parameter, D is the position of the dominant animal, which represents the current optimal model parameter, that is, the position of the target animal corresponding to the final output in the last update (that is, the current network parameter); X i It is a position within the preset parameter selection range, and rand is a random number in the range of [0-1].

[0134] In a preferred embodiment, the updated position of the escaping animal after escaping can be determined according to the following formula:

[0135] X2=X1+rand×(Low+R×(High-Low));

[0136] Among them, X2 is the updated position corresponding to the animal to be escaped after escaping, X1 is the current position corresponding to the animal to be escaped before escaping, rand is a preset random number range; Low and High are the lower and upper bounds of the local search space, respectively, and the local search space is used to limit the escape range of the animal to be escaped; R is a random number used for local search.

[0137] In a preferred embodiment, the present invention can select two animals with the smallest loss function values ​​from a population as parent individuals, and then perform a hybridization operation according to the single-point hybridization step in the genetic algorithm. For example, a crossover point is randomly selected on the chromosomes of the two parent individuals. This crossover point will determine which part of the chromosome will be exchanged. At the crossover point, the chromosomes of the two parent individuals are divided into two parts, and the parts of the two parent individuals after the crossover point are replaced, thereby forming two new offspring individuals, while the two parent individuals are still retained in the original population.

[0138] In a preferred embodiment, during each iterative optimization process of the network parameters, the present invention can adopt an improved Hippo optimization algorithm, that is, by simulating the natural behavior of hippopotamuses and introducing a hybridization step, the diversity of the hippopotamus population is increased, and the global search capability of the algorithm is improved, so that the model parameters and output results can be optimized, thereby improving the computational efficiency and accuracy of the model. The algorithm flow steps corresponding to the improved Hippo optimization algorithm are as follows: Figure 2 As shown, specifically:

[0139] Initialization: Generate an initial population of multiple hippos, each hippopotamus represents a set of model parameters, and these parameters can be randomly generated within the preset parameter range.

[0140] Fitness evaluation: Use the model loss function as the fitness function to evaluate the performance of each Hippopotamus (i.e., each set of model parameters). The smaller the loss function value, the higher the fitness of the Hippopotamus.

[0141] Hippo Position Update: Update the position of each hippopotamus (i.e., model parameters) according to the position update formula in the HOA algorithm. During the update process, the position of the optimal hippopotamus (i.e., the dominant hippopotamus) can be used as a guide and combined with random numbers for exploration.

[0142] Predator Defense: Randomly generate model parameters that meet preset conditions as predators. Then, based on the predator's location and the hippo's current position, update the hippopotamus's position near the predator to simulate the defense behavior. For example, a random walk strategy such as Lévy flight can be used to simulate the random proximity of the predator.

[0143] Escaping from a predator: For a hippopotamus threatened by a predator, a local search is performed to update its position. During the local search, a random number is used to fine-tune the hippopotamus's position within the set upper and lower bounds.

[0144] Hippo hybridization: Select the two hippos with the highest fitness from the population (including those that have escaped and those that have) as parent individuals. Perform the hybridization operation according to the single-point hybridization steps in the genetic algorithm to generate two new offspring individuals. This hybridization operation is intended to increase population diversity and enhance the algorithm's global search capabilities.

[0145] Evaluation and selection: Use the fitness function to evaluate the performance of the newly generated offspring individuals and the retained parent individuals, and select the hippopotamus with the best fitness as the dominant hippopotamus for the next generation.

[0146] Iteration and termination conditions: Repeat the above steps until the preset maximum number of iterations is reached or the fitness of the dominant hippopotamus meets the optimization conditions. It can be understood that in each iteration, the position of the hippopotamus in the population is updated and new individuals are introduced through hybridization operations.

[0147] Therefore, in the process of optimizing network parameters, the present invention can effectively search for the optimal solution in the parameter space by simulating the behavior of hippos in the natural environment (such as defending against predators and escaping from predators).

[0148] like Figure 3 As shown, based on the above-mentioned various embodiments of the inspection path generation method applicable to the substation, the present invention provides a corresponding device embodiment;

[0149] An embodiment of the present invention provides an inspection path generation device applicable to a substation, comprising: a site data acquisition module, a robot inspection data acquisition module, and an inspection path generation module;

[0150] The site data acquisition module is used to acquire site data corresponding to the substation; wherein the site data includes: site size data, fixed obstacle data and mobile obstacle data; the mobile obstacle data includes: data corresponding to obstacles that move dynamically in the site;

[0151] The robot inspection data acquisition module is used to acquire the robot inspection data corresponding to the inspection robot; wherein the robot inspection data includes: the robot starting position, the robot end position and the robot size;

[0152] The inspection path generation module is configured to input the site data and the robot inspection data into a path planning model, so that the path planning model extracts site characteristics and obstacle distribution characteristics from the site data, extracts robot space occupancy characteristics from the robot inspection data, and robot position characteristics used to characterize the relative positions of the starting point and the end point in the site; and generates a robot inspection path based on the site characteristics, obstacle distribution characteristics, robot space occupancy characteristics, and robot position characteristics.

[0153] The robot inspection path is used to represent the shortest obstacle avoidance path when the robot inspects in the substation.

[0154] It should be noted that the device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without paying any creative effort.

[0155] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0156] Based on the above-mentioned various embodiments of the inspection path generation method applicable to the substation, the present invention provides corresponding embodiments of the terminal equipment item.

[0157] An embodiment of the present invention provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements a patrol path generation method suitable for a substation as described in any method embodiment of the present invention.

[0158] The terminal device may be a computing terminal device such as a desktop computer, a notebook computer, a palmtop computer, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0159] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0160] The memory can be used to store the computer program, and the processor implements various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created based on the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0161] Based on the above-mentioned various embodiments of the inspection path generation method applicable to the substation, the present invention provides a corresponding storage medium embodiment.

[0162] An embodiment of the present invention provides a storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a patrol path generation method suitable for a substation as described in any method embodiment of the present invention.

[0163] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. The computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0164] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for generating an inspection path for a substation, characterized in that: include: Obtaining site data corresponding to the substation; wherein the site data includes: site size data, fixed obstacle data, and mobile obstacle data; the mobile obstacle data includes: data corresponding to obstacles that move dynamically in the site; Obtaining robot inspection data corresponding to the inspection robot; wherein the robot inspection data includes: the robot starting position, the robot end position and the robot size; Inputting the site data and the robot inspection data into a path planning model, so that the path planning model extracts site features and obstacle distribution features from the site data, extracts robot space occupancy features from the robot inspection data, and robot position features used to characterize the relative positions of a start point and an end point in the site; and generating a robot inspection path based on the site features, obstacle distribution features, robot space occupancy features, and robot position features. The robot inspection path is used to represent the shortest obstacle avoidance path when the robot inspects in the substation.

2. A method for generating an inspection path for a substation according to claim 1, characterized in that: The site dimension data includes the site length and site width; the fixed obstacle data includes the fixed obstacle position and fixed obstacle size corresponding to each fixed obstacle; the mobile obstacle data includes the mobile obstacle position, mobile obstacle size, and obstacle movement path corresponding to each mobile obstacle; The site characteristics include: site area characteristics and site shape characteristics; the obstacle distribution characteristics include: obstacle density characteristics and relative position relationship characteristics between obstacles; The path planning model extracts site features and obstacle distribution features from the site data, including: Generate site area characteristics based on site length and site width through the path planning model; Generating a site shape feature according to the site area feature, the site length, and the site width; Generate an obstacle density feature based on the position of each fixed obstacle, the size of each fixed obstacle, the position of each mobile obstacle, and the size of each mobile obstacle; The relative position relationship features between the obstacles are generated according to the positions of each fixed obstacle, the positions of each mobile obstacle, and the obstacle movement path of each mobile obstacle.

3. A method for generating an inspection path for a substation according to claim 2, characterized in that: The extracting of the robot space occupancy feature and the robot position feature for characterizing the relative positions of the starting point and the end point in the site from the robot inspection data includes: Extract the shape features of the robot based on the robot's starting position and size; generating a robot space occupancy feature for characterizing an area occupied by the robot in a venue according to the shape feature of the robot and the size of the robot; According to the robot's starting position and the robot's ending position, the relative position relationship between the starting point and the ending point in the field is extracted; According to the relative position relationship, a robot position feature is generated for characterizing the relative positions of the starting point and the end point in the field.

4. A method for generating an inspection path for a substation according to claim 3, characterized in that: Generating a robot inspection path according to the site characteristics, obstacle distribution characteristics, robot space occupancy characteristics, and robot position characteristics includes: Generating local features for characterizing a path feasibility area based on the site shape features, obstacle density features, and robot size features; Generate global features for characterizing the robot's obstacle avoidance path segments in the field based on the robot's position features and the relative position relationship features between obstacles; Fusing the local features and the global features to generate fused features; A robot inspection path is generated based on the fused features.

5. A method for generating an inspection path for a substation according to claim 4, characterized in that: The generation of the path planning model includes: Use the site sample data and robot inspection sample data as training samples; Taking several training samples and the actual robot inspection path of each training sample as input, and the predicted robot inspection path corresponding to each training sample as output, the path planning model to be trained is iteratively trained until the model converges, generating a trained path planning model.

6. A method for generating an inspection path for a substation according to claim 5, characterized in that: Before each iterative training, it also includes: Obtain the network parameters to be updated corresponding to the last training of the path planning model; wherein, during the initial training, the network parameters to be updated are randomly generated; Repeat the position update operation until the current update number is the same as the preset update number, and output the position corresponding to the target animal; wherein the position is used to represent the network parameters in the path planning model; Using the position corresponding to the target animal as a network parameter for training the path planning model; The location update operation includes: When the current update number is less than the preset update number, a plurality of current parameters are generated based on the current network parameter, the preset parameter selection range and the network parameter to be updated; each of the current parameters is then used as the current position corresponding to each animal in the population; wherein, initially, the network parameter to be updated is used as the current network parameter; Under the network parameter selection constraint conditions corresponding to the path planning model, the initial predation position corresponding to the predator is randomly generated; Taking the initial predation position as the center, and according to a preset predation position range, generating a plurality of target predation positions involved in the predator's predation behavior; For each target predation location, a range of positions to be escaped is generated based on a preset distance range and the target predation location, with the target predation location as the center; animals within the range of positions to be escaped are marked as animals to be escaped; wherein the preset predation location range is larger than the preset distance range; For each animal to be escaped, an escape simulation is performed on the animal to be escaped by simulating an escape strategy of the animal to be escaped from a predator, and an updated position corresponding to the animal to be escaped after the escape is generated; According to the preset model loss function, the loss function value corresponding to the current position of each current animal in the population is calculated; according to the preset model loss function, the loss function value corresponding to the updated position of each animal to be escaped in the population is calculated; Extract several parent animals according to the loss function value corresponding to each current animal and the loss function value corresponding to each animal to be escaped; Perform hybridization operations on each parent animal to generate several offspring animals; then calculate the loss function value corresponding to the position of each offspring animal according to the preset model loss function; According to the loss function value corresponding to each current animal, the loss function value corresponding to each animal to be escaped, and the loss function value corresponding to each offspring animal, the animal with the smallest loss function value is marked as the target animal; Using the position corresponding to the target animal as the current network parameter for the next position update operation; Add the preset iteration increment to the current update iteration value.

7. A method for generating an inspection path for a substation according to claim 6, characterized in that: The updated position of the animal after escaping is determined according to the following formula: X2=X1+rand×(Low+R×(High-Low)); Among them, X2 is the updated position corresponding to the animal to be escaped after escaping, X1 is the current position corresponding to the animal to be escaped before escaping, rand is a preset random number range; Low and High are the lower and upper bounds of the local search space, respectively, and the local search space is used to limit the escape range of the animal to be escaped; R is a random number used for local search.

8. A patrol route generation device suitable for a substation, characterized in that: include: Site data acquisition module, robot inspection data acquisition module and inspection path generation module; The site data acquisition module is used to acquire site data corresponding to the substation; wherein the site data includes: site size data, fixed obstacle data and mobile obstacle data; the mobile obstacle data includes: data corresponding to obstacles that move dynamically in the site; The robot inspection data acquisition module is used to acquire the robot inspection data corresponding to the inspection robot; wherein the robot inspection data includes: the robot starting position, the robot end position and the robot size; The inspection path generation module is configured to input the site data and the robot inspection data into a path planning model, so that the path planning model extracts site characteristics and obstacle distribution characteristics from the site data, extracts robot space occupancy characteristics from the robot inspection data, and robot position characteristics used to characterize the relative positions of the starting point and the end point in the site; and generates a robot inspection path based on the site characteristics, obstacle distribution characteristics, robot space occupancy characteristics, and robot position characteristics. The robot inspection path is used to represent the shortest obstacle avoidance path when the robot inspects in the substation.

9. A terminal device, characterized in that: The invention comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a method for generating an inspection path applicable to a substation as claimed in any one of claims 1 to 7 is implemented.

10. A storage medium, characterized in that: The storage medium includes a stored computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the inspection path generation method applicable to a substation according to any one of claims 1 to 7.

Images