Power transmission line path planning method, system, equipment, medium and product

Abstract

The invention relates to the technical field of power systems, and discloses a power transmission line path planning method, system, device, medium and product, and the method comprises the steps: carrying out the rasterization of the original image data of a target planning region, obtaining the raster image data, classifying the ground feature types of all raster units in the raster image data, and obtaining the ground feature types of all the raster units; the method comprises the following steps: obtaining a ground feature type of each grid unit, endowing each grid unit with an actual construction estimated cost value for constructing a power transmission line according to the ground feature types obtained by classification to obtain cost grid image data, and constructing an intelligent agent for a power transmission line path planning task based on a reinforcement learning algorithm to obtain cost grid image data. Power transmission line path planning prediction is carried out on cost raster image data through an intelligent agent, a power transmission line prediction path is obtained, dynamic interaction is carried out by utilizing a reinforcement learning algorithm and an actual environment, the adaptability of power transmission line path planning can be remarkably improved, and the efficiency and reliability of power transmission line path planning are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular, to a method, system, device, medium and product for power transmission line path planning. Background Art

[0002] Power transmission line path planning is to select an optimized path that not only complies with national policies and regulations but also meets multiple requirements such as technology, economy, and environmental protection between a determined starting point and ending point. As the core link of power grid engineering design, power transmission line design needs to comprehensively consider multiple constraints such as geographical factors, project cost, construction conditions, and operation and maintenance conditions. Its scientific and reasonable path planning is the basis for ensuring the safe and efficient transmission of electric energy. In the context of the "dual carbon" goal and the construction of a new power system, this work not only concerns the scientific layout of power grid engineering infrastructure but also is an important practice for implementing the concept of sustainable development. During the planning process, it is necessary to ensure the safe, reliable, economic and efficient operation of the power system itself, take into account the cost and convenience of construction and later operation and maintenance, fully consider the coordinated development with other industries, avoid interfering with the normal operation of regions such as communication, transportation, and agriculture, and pay attention to the integrity of the natural landscape and ecological environment along the line to achieve the harmonious unity of power engineering construction and ecological protection and promote the achievement of sustainable development goals.

[0003] In traditional power transmission line path planning, line design and planning usually rely on professional staff to determine the path scheme through multiple links such as indoor route selection and on-site investigation. In the indoor route selection stage, designers initially plan alternative route schemes based on static maps, historical terrain data, and limited meteorological data through empirical judgment and simple technical analysis. In the subsequent on-site investigation link, staff need to go deep into the field to verify each of the indoor-planned lines one by one and adjust and optimize the scheme according to the actual terrain, geological conditions, surrounding environment, etc. This process not only takes a long time but also has significant limitations due to the inherent defects of traditional methods. First, the update cycle of traditional map data is long, making it difficult to reflect changes in environmental information such as terrain, land use, and vegetation cover in a timely manner. Second, traditional path selection methods are insufficient in dealing with the coordination and dynamic changes of multiple factors. Third, on-site or field investigations require a large number of professional staff to conduct point-by-point inspections of terrain, geology, and obstacles, etc., with high labor intensity and long cycle.

[0004] In addition, the current power transmission line path planning method based on intelligent algorithms has poor dynamic adaptability, resulting in difficulty in adapting to changes in the new environment. Summary of the Invention

[0005] In view of this, the present invention provides a transmission line path planning method, system, device, medium and product, which solves the technical problem that the dynamic adaptability of the current transmission line path planning method based on intelligent algorithms is poor, resulting in the inability to adapt to the changes in the new environment.

[0006] The first aspect of the present invention provides a transmission line path planning method, including:

[0007] Obtain the original image data of the target planning area corresponding to the transmission line path planning, and rasterize the original image data to obtain raster image data;

[0008] Classify the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell;

[0009] Assign the actual construction estimated cost value for building a transmission line to each grid cell according to the classified land cover types to obtain cost raster image data;

[0010] Construct an agent for the transmission line path planning task, and based on the reinforcement learning algorithm, make a prediction of the transmission line path planning by the agent in the cost raster image data to obtain a predicted transmission line path.

[0011] Preferably, the classifying the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell includes:

[0012] Use the ArcMap image classification tool to perform interactive supervised classification on the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell.

[0013] Preferably, the method further includes: determining the actual construction estimated cost value corresponding to each land cover type; the determining the actual construction estimated cost value corresponding to each land cover type includes:

[0014] Construct a hierarchical analysis model of the land cover weighting system based on the analytic hierarchy process; the hierarchical analysis model includes a target layer, a factor layer and a scheme layer; wherein, the target layer is the transmission line path planning benefit, and the factor layer includes the avoidance cost of ecological-geographical constraints, the crossing cost of linear engineering interference, the social cost sensitive to human activities, and the meteorological impact cost of meteorological load effects, and the scheme layer includes the sub-influence factors of each factor in the factor layer;

[0015] For each factor in the factor layer, construct an initial objective judgment matrix, and perform PCA dimensionality reduction on the initial objective judgment matrix to obtain an objective judgment matrix;

[0016] For each factor in the factor layer, construct a subjective judgment matrix according to the 1-9 marking method, and conduct a consistency test on the subjective judgment matrix;

[0017] In the case where the subjective judgment matrix passes the consistency test, fuse the subjective judgment matrix with the objective judgment matrix to obtain a comprehensive judgment matrix;

[0018] Calculate the single-layer weight vector for the comprehensive judgment matrix; the single-layer weight vector is used to determine the importance weights of each factor in the same layer relative to the factor in the upper layer;

[0019] According to the single-layer weight vector, determine the comprehensive weight vector, and conduct a consistency test on the comprehensive weight vector;

[0020] In the case where the comprehensive weight vector passes the consistency test, allocate weights to each of the ground object types through the comprehensive weight vector to obtain the weights of each of the ground object types;

[0021] Perform numerical transformation on the weights of each of the ground object types to obtain the actual construction estimated cost values corresponding to each ground object type.

[0022] Preferably, assigning the actual construction estimated cost value for building a transmission line to each grid cell according to the classified ground object types to obtain cost grid image data, including:

[0023] Assign the actual construction estimated cost values corresponding to each ground object type to each grid cell according to the classified ground object types to obtain the initial cost grid image data;

[0024] Normalize each grid cell of the initial cost grid image data to obtain the cost grid image data.

[0025] Preferably, constructing an agent for the transmission line path planning task, based on the reinforcement learning algorithm, and predicting the transmission line path through the agent in the cost grid image data, including:

[0026] Construct an agent for the transmission line path planning task, and the state space and action space of the agent; wherein, the state space is the geographical location of each grid unit of the cost grid image data in the grid coordinate system, and the action space includes the four-direction discrete actions that the agent can choose;

[0027] Use the proximal policy optimization algorithm to train the policy network of the agent. When the agent starts from the starting point of the line planning on the cost grid image data, use the current state and the policy network to select the current action;

[0028] Execute the current action and determine the immediate reward value of the current action through the reward function;

[0029] Update the policy network through the immediate reward value, and use the updated policy network as the current policy network;

[0030] Judge whether the agent reaches the target end point according to the state corresponding to the current action;

[0031] If it is judged that the agent does not reach the target end point, then based on the current policy network, go to the step of selecting the current action by using the current state and the policy network until it is judged that the agent reaches the target end point, and obtain the geographical locations of all grid units of the agent in the cost grid image data;

[0032] Determine the predicted path of the transmission line according to the geographical locations of all grid units of the agent in the cost grid image data.

[0033] Preferably, the reward function is:

[0034]

[0035] In the formula, is the reward function, is the state, is the action, , are both weight coefficients, , are the Euclidean distances from the previous step and the current step to the target position respectively, , are the actual construction estimated cost value and the benchmark cost of the current grid cell respectively, is the repeated visit penalty term, is the reaching target reward term.

[0036] In a second aspect, the present invention also provides a transmission line path planning system, including:

[0037] An image grid module, configured to obtain the original image data of the target planning area corresponding to the transmission line path planning, and rasterize the original image data to obtain grid image data;

[0038] A grid classification module, configured to classify the land object types of each grid cell in the grid image data to obtain the land object types of each grid cell;

[0039] A cost grid recognition module, configured to assign an actual construction estimated cost value for building a transmission line to each grid cell according to the classified land object types, so as to obtain cost grid image data;

[0040] A path planning module, configured to construct an agent for the transmission line path planning task, and based on a reinforcement learning algorithm, perform transmission line path planning prediction by the agent in the cost grid image data, so as to obtain a predicted transmission line path.

[0041] In a third aspect, the present invention further provides an electronic device, where the electronic device includes a memory and a processor, and a computer program is stored in the memory. When the computer program is executed by the processor, the processor is caused to execute the steps of the transmission line path planning method as described in the first aspect.

[0042] In a fourth aspect, the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the steps of the transmission line path planning method as described in the first aspect are implemented.

[0043] In a fifth aspect, the present invention further provides a computer program product, where the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer is caused to execute the steps of the transmission line path planning method as described in the first aspect.

[0044] As can be seen from the above technical solutions, the present invention rasterizes the original image data of the target planning area to obtain raster image data, classifies the land object types of each grid cell in the raster image data to obtain the land object types of each grid cell, assigns an actual construction estimated cost value for building a transmission line to each grid cell according to the classified land object types to obtain cost grid image data, constructs an agent for the transmission line path planning task, and based on a reinforcement learning algorithm, performs transmission line path planning prediction by the agent in the cost grid image data to obtain a predicted transmission line path, so as to dynamically interact with the actual environment by using the reinforcement learning algorithm, which can significantly improve the adaptability of the transmission line path planning and improve the efficiency and reliability of the transmission line path planning. Description of the Drawings

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is an application environment diagram of a transmission line path planning method provided by an embodiment of the present invention;

[0047] Figure 2 It is a flowchart of a transmission line path planning method provided by an embodiment of the present invention;

[0048] Figure 3 It is a standardized operation flowchart of high - resolution remote sensing image data provided by an embodiment of the present invention;

[0049] Figure 4 It is a schematic diagram of high - resolution remote sensing image data provided by an embodiment of the present invention;

[0050] Figure 5 It is an initial rasterized ground object type identification and classification diagram provided by an embodiment of the present invention;

[0051] Figure 6 It is a flowchart for solving cost values of different ground object types provided by an embodiment of the present invention;

[0052] Figure 7 It is a schematic diagram of an analytic hierarchy process model of an object weighting system provided by an embodiment of the present invention;

[0053] Figure 8 It is a schematic diagram of remote sensing image data of a certain area (1km×1km) provided by an embodiment of the present invention;

[0054] Figure 9 It is a schematic diagram of a cost raster map (100×100) and a predicted path of a certain area provided by an embodiment of the present invention;

[0055] Figure 10 It is a schematic diagram of remote sensing image data of a certain area (5km×5km) provided by an embodiment of the present invention;

[0056] Figure 11 It is a schematic diagram of a cost raster map (2000×2000) and a predicted path of a certain area provided by an embodiment of the present invention;

[0057] Figure 12 It is a schematic diagram of the structure of a transmission line path planning system provided by an embodiment of the present invention;

[0058] Figure 13 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0059] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0060] The transmission line path planning method provided by the embodiments of the present application can be applied to an application environment as Figure 1 shown. Among them, the terminal 101 communicates with the server 102 through a network. The data storage system can store the data that the server 102 needs to process. The data storage system can be integrated on the server 102, or placed in the cloud or other network servers. The terminal 101 or the server 102 acquires the original image data of the target planning area corresponding to the transmission line path planning, rasterizes the original image data to obtain raster image data; classifies the ground object types of each raster unit in the raster image data to obtain the ground object types of each raster unit; assigns an actual construction estimated cost value for building a transmission line to each raster unit according to the classified ground object types to obtain a cost raster image data; constructs an agent for the transmission line path planning task, and based on the reinforcement learning algorithm, predicts the transmission line path through the agent in the cost raster image data.

[0061] The terminal 101 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, etc.

[0062] The server 102 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0063] As Figure 2 shown, the embodiments of the present application provide a transmission line path planning method. Taking the method applied to the Figure 1 terminal 101 or the server 102 in it as an example, the following steps S1 to S4 are included. Among them:

[0064] Step S1: Acquire the original image data of the target planning area corresponding to the transmission line path planning, and rasterize the original image data to obtain raster image data.

[0065] Among them, the original image data can be remote sensing image data of the target planning area.

[0066] In some examples, high-resolution remote sensing image data of the target planning area is obtained based on Google Earth Pro and ENVI (Environment for Visualizing Images) software. The full English name and Chinese translation of ENVI software are

[0067] Specifically, in the embodiments of this application, high-resolution remote sensing image data of typical land cover areas in XX Province is obtained based on the collaborative work of Google Earth Pro and ENVI software. Google Earth Pro is a free geographic information system software developed by Google, widely used in fields such as remote sensing image acquisition, three-dimensional terrain visualization, and geospatial analysis. Its high-resolution satellite image database covers the global scope and supports multi-temporal data comparison, making it one of the effective tools for land cover research. Compared with the ordinary version, Google Earth Pro provides a higher-resolution image export function and professional measurement tools. ENVI software is a professional remote sensing image processing software developed by Harris Geospatial Solutions, supporting the analysis and interpretation of multi-spectral, hyperspectral, and radar data. Its powerful format compatibility and spatial reference retention ability make it a key tool for remote sensing data preprocessing.

[0068] Figure 3 The standardized operation flow chart for obtaining high-resolution remote sensing image data in GeoTIFF format using Google Earth Pro and ENVI software is shown.

[0069] The specific operation steps include: First, download and install the latest version of the software through the official platform of Google Earth Pro, and enter "XX Province" in the software search bar to locate the research area. To ensure the timeliness and availability of the images, check and select the latest available cloud-free image data in "View" → "Historical Imagery". Then, select target areas with typical land cover characteristics within the province, including forests, construction land, water areas, roads, rivers, industrial parks, and residential areas. During the construction of the sampling frame, a strict geometric accuracy control method is adopted. By using the "Ruler" function in the "Tools" menu and selecting the "Path" measurement mode, a square sampling frame with a side length of 1 km is constructed using the vertex drawing method. During the drawing process, the shortcut key "U" is used to ensure that the view maintains a strict vertical projection mode (i.e., "Top View") to eliminate geometric errors caused by perspective distortion. After completing the drawing of the sampling frame, use the mouse wheel and dragging function to adjust the view range to ensure that the square formed by the four reference lines is completely displayed in the window. The adjusted view is as Figure 4 shown.

[0070] When saving the image data, select the "File" → "Save" → "Save Image" function, and configure the following key parameters in the output settings panel:

[0071] (1) Map element control: Cancel all map options (including auxiliary elements such as scale and legend) to avoid interfering with subsequent classification analysis;

[0072] (2) Watermark processing: Adjust the view scale to 0% to minimize the occupancy of the platform watermark on the effective area of the image;

[0073] (3) Resolution setting: Set the output resolution to 8174×8174 pixels (corresponding to a 1km×1km area, with a ground resolution of approximately 0.12 m / pixel);

[0074] (4) Format selection: Select JPEG as the storage format.

[0075] It can be understood that other image formats such as PNG and TIFF can also be selected as the storage format in the embodiments of the present application. However, considering that the JPEG format has a high compression rate and good image quality performance, while ensuring the integrity of image information, it can effectively reduce the size of the image file, facilitating subsequent image processing and transmission. Therefore, in the embodiments of the present application, JPEG is preferentially selected as the storage format. After selecting the JPEG format, parameters such as image quality and color mode also need to be set to ensure that the output image data meets the requirements of subsequent processing.

[0076] In addition, the raster image data contains multiple raster cells, and each raster cell corresponds to a geographical location in the target planning area. The resolution of the raster image data determines the actual geographical area represented by each raster cell. In the embodiments of the present application, the ground resolution of the raster image data is approximately 0.12 m / pixel, that is, the actual geographical area represented by each raster cell is approximately 0.0144 square meters.

[0077] Step S2: Classify the land cover types of each raster cell in the raster image data to obtain the land cover types of each raster cell.

[0078] Among them, the land cover types are such as forest, water area, road, building, etc. It includes vegetation, construction land, lakes, nature reserves, railways, highways, transmission lines, rivers, public facilities, industrial and mining enterprises, airport facilities, and civilian houses, etc.

[0079] During the land cover type classification process, it can be achieved through machine learning algorithms, such as support vector machine, decision tree, random forest and other algorithms. By analyzing and learning the features of each raster cell in the raster image data, the land cover types of each raster cell can be obtained.

[0080] Step S3: Assign the actual estimated construction cost value for building a transmission line to each grid cell according to the classified land cover types, and obtain the cost grid image data.

[0081] Among them, the actual estimated construction cost value is estimated based on the possible costs generated by different types of land cover during the construction of a transmission line. For example, forest areas may require tree felling and land clearing, so the cost is relatively high; water areas may require the erection of bridges or the use of underwater cables, and the cost will also increase; while areas with existing infrastructure such as roads and construction land have relatively low construction costs.

[0082] Step S4: Construct an agent for the transmission line path planning task. Based on the reinforcement learning algorithm, the agent performs transmission line path planning prediction on the cost grid image data to obtain the predicted transmission line path.

[0083] Among them, the reinforcement learning algorithm is a machine learning method that finds the optimal strategy through continuous trial and error and optimization. In this embodiment, the agent is designed to be able to move in the cost grid image data and select the optimal path according to the cost of the current grid cell and possible actions (such as moving up, down, left, or right). The agent learns how to plan the optimal path by performing actions and receiving feedback (i.e., rewards or punishments) from the environment.

[0084] It should be noted that in the embodiment of the present application, the original image data of the target planning area is rasterized to obtain raster image data, and the land cover types of each grid cell in the raster image data are classified to obtain the land cover types of each grid cell. The actual estimated construction cost value for building a transmission line is assigned to each grid cell according to the classified land cover types to obtain the cost grid image data. By constructing an agent for the transmission line path planning task and based on the reinforcement learning algorithm, the agent performs transmission line path planning prediction on the cost grid image data to obtain the predicted transmission line path. Therefore, by dynamically interacting with the actual environment using the reinforcement learning algorithm, the adaptability of the transmission line path planning can be significantly improved, and the efficiency and reliability of the transmission line path planning can be enhanced.

[0085] In some embodiments, classifying the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell includes:

[0086] Using the ArcMap image classification tool to perform interactive supervised classification on the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell.

[0087] Among them, the ArcMap image classification tool is used for spatial data analysis, map production, and feature recognition. Among them, the feature recognition function (such as supervised classification) usually relies on the input of images in GeoTIFF format. Therefore, it is necessary to convert the above JPEG images obtained by GoogleEarth Pro software into GeoTIFF format. Here, ENVI software is used for format conversion to ensure that the spatial resolution and coordinate system characteristics of the images remain unchanged. The specific conversion process is as follows: First, start the ENVI software and import the JPEG image to be converted through "File" → "Open". Subsequently, select "File" → "Save As" in the main menu bar to open the file saving interface, and select the loaded JPEG file in the pop-up "File Selection" dialog box. Finally, in the "Save File As Parameters" parameter setting interface, specify the output format as GeoTIFF format and set the appropriate output path.

[0088] Based on the ArcMap software in the ArcGIS platform, the embodiment of this application adopts a method combining supervised classification and raster reclassification to achieve high-precision feature information extraction and type recognition. Specifically: Import raster image data into ArcMap through georegistration to establish base map data with accurate spatial reference, and use the interactive supervised classification in the ArcMap image classification tool to intelligently identify and classify the feature types in the study area, generating an initial rasterized feature type recognition and classification map as shown in Figure 5 to obtain the feature types of each raster cell.

[0089] In some embodiments, to achieve the comprehensive benefits of the optimal path planning of intelligent transmission lines, the present invention constructs a standardized feature weighting system, and through the collaborative application of PCA (Principal Component Analysis) and AHP (Analytic Hierarchy Process), systematically quantifies the impact of various feature types on the path cost. PCA is used to extract the main influencing factors of each element and eliminate data redundancy, while AHP is used to establish an analytic hierarchy model for multi-objective decision-making to balance ecological protection and cost control. Finally, a standardized feature weighting system is constructed to provide a structured input for the intelligent transmission line path planning method. Realize dynamically weighing various constraints in path planning, and preferentially select corridors with low conflicts and low costs, thereby systematically reducing the comprehensive cost of line construction and achieving the unity of economy and sustainability.

[0090] In some embodiments, this method further includes: determining the estimated actual construction cost value corresponding to each feature type; as shown in Figure 6 determining the estimated actual construction cost value corresponding to each feature type includes:

[0091] Step S21: Construct an analytic hierarchy process (AHP) model for the weight assignment system of ground objects; the AHP model includes an objective layer, a factor layer, and a scheme layer; among them, the objective layer is the benefit of transmission line path planning, the factor layer includes the avoidance cost of ecological-geographical constraints, the crossing cost of linear project interference, the social cost sensitive to human activities, and the meteorological impact cost of meteorological load effects, and the scheme layer includes the sub-influence factors of each factor within the factor layer.

[0092] Among them, as Figure 7 shown, this AHP model decomposes the complex geographical environment constraints into four sub-systems: the ecological-geographical constraint system (static constraint) focuses on the avoidance costs of four types of ground objects, namely vegetation, construction land, lakes, and nature reserves; the linear project interference system (spatial conflict) pays attention to the crossing costs of four types of ground objects, namely railways, highways, transmission lines, and rivers; the human activity sensitive system (socio-economy) evaluates the social costs of four types of ground objects, namely public facilities, industrial and mining enterprises, airport facilities, and residential houses; the meteorological load effect system (dynamic impact) focuses on the meteorological impact costs of four types of ground objects, namely temperature, wind speed, humidity, and icing.

[0093] Step S22: For each factor in the factor layer, construct an initial objective judgment matrix and perform PCA dimensionality reduction on the initial objective judgment matrix to obtain an objective judgment matrix.

[0094] Among them, for a certain sub-system in the AHP model, N experts score M types of ground objects to form an initial judgment matrix X as shown in Equation (1):

[0095] (1)

[0096] The comprehensive index formula corresponding to this initial judgment matrix is as shown in Equation (2):

[0097] (2)

[0098] In the formula, is the score value of the i-th group, represents the coefficient vector of each index in the i-th group evaluation, and satisfies: , .

[0099] Through PCA dimensionality reduction, the N groups of score values are converted into k groups of score values . Then solve the coefficient vector , and the specific derivation formula is as follows:

[0100] It is known that the comprehensive index is as shown in Equation (3):

[0101] (3)

[0102] Wherein, is the transpose of the coefficient vector .

[0103] To maximize its variance, that is, is maximized, where represents the solution of variance, and W is the covariance of X. Assume that the eigenvalues of W are , and the corresponding orthonormal basis is , then as shown in Equation (4):

[0104] = (4)

[0105] At this time, is called the j-th principal component, and the contribution rate of the j-th principal component is defined as , as shown in Equation (5):

[0106] (5)

[0107] Wherein, is the eigenvalue corresponding to the j-th principal component, and k is the number of groups of scoring values.

[0108] Denote as the cumulative contribution rate of the first k largest principal components, and its calculation formula is as shown in Equation (6):

[0109] (6)

[0110] When the cumulative contribution rate of k principal components exceeds 85%, that is, , it is considered that the k principal components basically contain the original scoring information. Finally, retain these k principal components, solve the first k largest eigenvectors and the eigenvector , and then after normalizing the eigenvectors, construct the comprehensive evaluation vector , and is a linear combination of the principal component , that is, as shown in Equation (7):

[0111] (7)

[0112] In this way, the objective scoring values of each ground object type can be obtained, and then by comparing the elements in pairwise, the final objective judgment matrix can be constructed, as shown in Equation (8):

[0113] (8)

[0114] Step S23: For each factor in the factor layer, construct a subjective judgment matrix according to the 1-9 marking method, and conduct a consistency test on the subjective judgment matrix.

[0115] Among them, in combination with the hierarchical analysis model of the ground object weighting system, pairwise comparison is carried out for each ground object type in each layer to construct the corresponding pairwise judgment matrix to quantify its relative importance. In the 1-9 marking method: 1 means the two are equally important, 3 means the former is slightly more important, 5 means significantly more important, 7 means strongly more important, 9 means extremely more important, and even numbers between adjacent odd numbers represent intermediate states. The marking values and their meanings are shown in Table 1.

[0116] Table 1 Marking values and their meanings

[0117]

[0118] Taking 4 ground object types in a certain sub-system as an example, a subjective judgment matrix A1 as shown in formula (9) is constructed using the 1-9 marking method:

[0119] (9)

[0120] In the formula, each element x, y, z, p, e, q represents the subjective comparison marking value between pairwise ground object types.

[0121] This subjective judgment matrix is a positive and negative matrix, and the elements at each position in the matrix satisfy the following formula (10):

[0122]

[0123] Subjective judgment may have logical contradictions, and it is necessary to conduct a consistency test on the judgment matrix to ensure the credibility of the results. Therefore, then conduct a consistency check on the subjective judgment matrix. According to the characteristic equation of the identity matrix E Solve the judgment matrix for the maximum eigenvalue , and then solve the consistency index, as shown in formula (11):

[0124] (11)

[0125] In the formula, n is the matrix order.

[0126] Subsequently, according to the matrix order n, look up the corresponding table of the consistency index (RI) values shown in Table 2 (an existing reference value table, no need to calculate the values in the table by yourself) to obtain the random consistency index (RI), and calculate the consistency ratio (CR). Its calculation formula is as shown in formula (12):

[0127] (12)

[0128] Table 2 Corresponding Table of Consistency Index RI Values

[0129]

[0130] If CR < 0.1, it is considered that the judgment matrix passes the consistency test; otherwise, the marked values need to be adjusted and the matrix needs to be reconstructed. This step effectively avoids the circular contradiction of "A is superior to B, B is superior to C, but C is superior to A", ensuring the scientificity and reliability of the analysis results.

[0131] Step S24: When the subjective judgment matrix passes the consistency test, fuse the subjective judgment matrix and the objective judgment matrix to obtain a comprehensive judgment matrix.

[0132] Among them, by calculating the arithmetic mean of the objective judgment matrix and the subjective judgment matrix , construct the comprehensive judgment matrix B, and the specific calculation formula is shown in Equation (13):

[0133] (13)

[0134] Among them, and are the corresponding elements of the subjective and objective judgment matrices respectively. Therefore, the comprehensive judgment matrix B of a certain sub-system is shown in Equation (14):

[0135] (14)

[0136] In the formula, are the elements at each position in the comprehensive judgment matrix B. Since both the subjective and objective judgment matrices pass the consistency test, the comprehensive judgment matrix can also pass the consistency test, and at this time, there is no need to perform a consistency test on the comprehensive judgment matrix anymore. Continuing the above steps, the comprehensive judgment matrices of each feature type in each plan layer for the factor layer and the factor layer for the target layer can be obtained.

[0137] Step S25: Calculate the single-layer weight vector for the comprehensive judgment matrix; the single-layer weight vector is used to determine the importance weights of each factor in the same layer relative to the factors in the previous layer;

[0138] Among them, the single-layer weight vector is solved using the eigenvalue method, and the solution formula is shown in Equation (15):

[0139] (15)

[0140] In the formula, is the eigenvalue of the comprehensive judgment matrix B, is the eigenvalue corresponding eigenvector.

[0141] Taking the indicators in a certain subsystem as an example, the maximum eigenvalue of the comprehensive judgment matrix B is solved , and the weight vector w. Similarly, all single-layer weight vectors can be obtained, that is, the weight vectors of all ground object types in the scheme layer with respect to the factor layer and the weight vectors of the factor layer with respect to the target layer

[0142] Step S26: Determine the comprehensive weight vector based on the single-layer weight vector and conduct a consistency test on the comprehensive weight vector

[0143] Among them, based on the single-layer weight vector, the comprehensive weight vector of all ground object types in the scheme layer with respect to the target layer is solved, and the solution formula is shown in formulas (16) and (17):

[0144] (16)

[0145] (17)

[0146] Among them is the weight value of each factor in the i-th layer with respect to the target layer is the weight value of all elements decomposed in the i-th layer with respect to the th element in the j-th layer is the weight value of the factor in the i-th layer with respect to the j-th element in the i-1 layer as the criterion represents the weight value of the m-th element in the i-th layer with respect to the j-th element in the i-1 layer

[0147] Step S27: Under the condition of the consistency test of the comprehensive weight vector, weight allocation is carried out for each ground object type through the comprehensive weight vector to obtain the weights of each ground object type

[0148] Among them, weight allocation for each ground object type through the comprehensive weight vector can obtain the weight values of each ground object type with respect to the target layer from top to bottom, that is, combining the weight vector w of the factor layer with respect to the target layer with the above formula, and finally obtaining the weight values of all ground object types in the analytic hierarchy model constructed based on the ground object weight system with respect to the target layer, as shown in Table 3

[0149] Table 3 Weight values of different ground object types in the four subsystems

[0150]

[0151] Step S28: Perform numerical transformation on the weights of each ground object type to obtain the estimated actual construction cost values corresponding to each ground object type

[0152] Among them, the weights of various feature types, as dimensionless relative importance coefficients, have significant limitations in engineering applications. To improve the data usability, the present invention multiplies the weight value by 100 and then rounds it to an integer, which is used as the actual construction estimated cost value corresponding to each feature type respectively.

[0153] It can be understood that in the embodiments of the present application, by performing the above numerical transformation on the weights of various feature types, the physical meaning of "cost per unit distance" is given to the weight value, realizing the dimensional consistency with the engineering economic indicators, and meeting the requirements for the accuracy of cost parameters in the economic evaluation guidelines related to power engineering; the integerized data can enhance the calculation stability of the planning model and at the same time meet the efficient convergence characteristics of reinforcement learning in the discrete state space; it is compatible with the raster integer weighting operation of ArcGIS software, avoiding complex conversion of data types.

[0154] Under the premise of retaining the original weight distribution law, this method realizes the collaborative optimization of engineering applicability and calculation robustness. After converting the original feature type weight values into actual construction estimated cost values, the actual construction estimated cost values of different feature types in the four major subsystems shown in Table 4 are obtained. In addition, when identifying and classifying feature types in ArcMap, there are a small number of NoData areas at the edges of the exported weighted raster images of features. The main causes include: the pixels in the edge areas of the high-resolution remote sensing images of the delineated areas exported by Google Earth Pro may not be effectively covered, or the ArcMap interactive supervised classification tool does not fully consider the characteristics of the edge pixels during the processing, or there are slight deviations in the geographic coordinate information of the input image and the output settings. For such edge NoData areas caused by incomplete data coverage, tool processing characteristics or coordinate differences, the present invention sets the cost values of these areas to the default value of 127 when exported by ArcMap to ensure the integrity of the raster data and the consistency of subsequent weighting calculations.

[0155] It should also be noted that the embodiments of the present application solve the technical problems of insufficient quantification of multi-source constraints and poor engineering adaptability in traditional path planning. By constructing a "dual-driven by subjective and objective" decision-making framework: first, PCA is used to reduce the dimension of the expert scoring matrix, and key influencing factors are extracted to generate an objective judgment matrix; a subjective judgment matrix is constructed in combination with the 1-9 marking method; the subjective and objective judgment matrices are fused pairwise to form a comprehensive judgment matrix with both data objectivity and empirical reliability, effectively overcoming the limitations of a single weighting method; further, a conversion method between the feature type weight value and the actual construction estimated cost value is proposed, linearly transforming the dimensionless weight into an integer cost coefficient, realizing the dimensional consistency with the engineering economic indicators and deep compatibility with the ArcGIS platform, and improving the calculation efficiency and convergence speed of the intelligent transmission line path planning method in the discrete state space.

[0156] In some embodiments, actual construction estimated cost values for building transmission lines are assigned to each grid cell according to the classified ground object types, obtaining cost grid image data, including:

[0157] Step S301: Assign the actual construction estimated cost values corresponding to each ground object type to each grid cell according to the classified ground object types, obtaining initial cost grid image data.

[0158] Among them, the initial cost grid image data is imported into ArcGIS software for subsequent processing. In ArcGIS, the cost distance tool can be used to calculate the shortest path from the source point to each grid cell and its cumulative cost based on the initial cost grid image data. This process fully considers the influence of different ground object types on the construction cost of transmission lines, making the path planning more in line with the actual situation.

[0159] Step S302: Standardize each grid cell of the initial cost grid image data to obtain cost grid image data.

[0160] Among them, the standardization is to set the pixel size and grid size through the export settings of ArcMap, and export a standardized ground object weighted grid image in GeoTIFF format with geographic coordinate information and a unified standard size. The standardized cost grid image data is the image data exported after assigning the actual construction estimated cost value to each grid.

[0161] In the cost grid image data, a two-dimensional cost matrix is generated that is completely aligned with the original grid map. The dimension of this matrix strictly corresponds to the actual geographical range. For example, a 2000×2000 matrix unit represents a 5km×5km planning area (assuming each grid cell corresponds to an actual area of 2.5m×2.5m). Each element value in the matrix accurately reflects the actual construction estimated cost value of the corresponding geographical location. This matrix uses 32-bit floating-point values to accurately represent the comprehensive construction cost of each grid cell, achieving efficient storage and processing of large-scale geographical data while ensuring calculation accuracy. The system realizes the digital expression and dynamic management of various engineering constraint conditions (such as prohibiting crossing ecological protection areas and construction difficulties in special terrains) through the matrix operation mechanism, providing a structured environmental input for the subsequent Proximal Policy Optimization (PPO) algorithm. This modeling method not only completely retains the spatial characteristics of the original geographical data but also transforms the complex path planning problem into a quantifiable reinforcement learning task through numerical conversion.

[0162] In some embodiments, an agent for the transmission line path planning task is constructed. Based on the reinforcement learning algorithm, the agent predicts the transmission line path through the cost grid image data, and obtains the predicted transmission line path, including:

[0163] Step S401: Construct an agent for the transmission line path planning task, as well as the state space and action space of the agent. Among them, the state space is the geographical location of each grid unit of the cost grid image data in the grid coordinate system, and the action space includes the four-direction discrete actions that the agent can choose.

[0164] In terms of the design of the state space and action space, the state space directly adopts the absolute position of the grid coordinate system, which is defined as a four-dimensional vector containing the current grid position coordinates (x, y) and the target grid position coordinates (x', y'), and precisely corresponds to the grid cells in the actual geographic information system.

[0165] The action space adopts a four-direction discrete model (up, down, left, right) that conforms to the basic movement requirements of power engineering. After each action is executed, the agent will strictly move a distance of one grid unit. This design not only ensures the accuracy of path search but also meets the basic requirements for path continuity in transmission line planning and design. By introducing a real-time detection mechanism for grid cell attributes, the passability and construction cost of the next grid are dynamically evaluated in each movement step, enabling the algorithm to precisely control the engineering details of path planning while maintaining a simple action space. This design method effectively avoids situations that do not conform to engineering reality, such as path jumping or cross-grid connections, through grid movement with a fixed step size, and its reliability and practicality have been verified in multiple actual planning cases.

[0166] Step S402: Use the Proximal Policy Optimization algorithm to train the policy network of the agent. When the agent starts from the starting point of the line planning on the cost grid image data, the current state and the policy network are used to select the current action.

[0167] Among them, the policy network is a deep neural network used to output the action probability distribution according to the current state (i.e., the position of the agent in the cost grid image data), and it is a convolutional neural network containing multiple convolutional layers, pooling layers, and fully connected layers. This network structure can extract the spatial features in the cost grid image data and predict the next action of the agent based on these features. During the training process, the Proximal Policy Optimization algorithm maximizes the cumulative reward of the agent from the starting point to the end point by continuously adjusting the parameters of the policy network, that is, minimizes the construction cost of the transmission line.

[0168] The embodiment of this application introduces the Proximal Policy Optimization (PPO) algorithm, which performs excellently in reinforcement learning, as the core policy learning method for the transmission line path planning task. PPO belongs to the Policy Gradient Methods. Its main advantage is that it can improve the sampling efficiency and learning effect while maintaining the stable update of the policy. Its core idea is to introduce a policy change clipping mechanism to limit the change range between the new and old policies, thereby avoiding the instability of the training process caused by drastic changes in the policy. Its core optimization objective function is as follows:

[0169] (18)

[0170] Among them, is the recorded parameter, and in reinforcement learning, the policy is improved by continuously optimizing . represents the probability ratio of the new and old policies, which is used to measure the change range of the current policy relative to the old policy for a certain action. is the advantage function, which measures the "good or bad" of the current action relative to the average policy quality. is the clipping parameter of the policy change range, which is used to avoid the instability caused by excessive policy updates. In the present invention, it is set to 0.2. is the clipping function, which limits the probability ratio within the interval, suppresses excessive policy changes, and improves the training stability. is the time step expectation, which represents the expectation (average value) of a batch of sampled time steps t and is used to approximate the gradient of the objective function.

[0171] During the training process, the environmental state is represented by a four-dimensional vector composed of the current position and the target position:

[0172] (19)

[0173] Among them, is the coordinate of the current grid, is the coordinate of the target midpoint.

[0174] The action space A (up, down, left, right) adopts a four-direction discrete action set, which meets the basic requirements of the power line engineering design for path continuity and controllability. At each time step t, the agent inputs the current state into the policy network to generate the probability distribution of four possible actions. Subsequently, the agent samples according to this distribution to select a specific action , and updates its current position accordingly. The state corresponding to the new position Rebuild it into a vector containing the current position and the target position. While the state is transferred, the environment returns the immediate reward for this step, which is used to guide the optimization and update of the policy network. Through this iterative loop of state-action-state, the agent continuously optimizes the path decision until it reaches the target or reaches the set step limit.

[0175] Step S403: Execute the current action and determine the immediate reward value of the current action through the reward function.

[0176] Among them, in order to make the agent more in line with the actual engineering requirements when searching for the path, this study designed the following comprehensive reward function:

[0177] Goal approaching reward: If the agent is closer to the target point in the current step, a positive reward is given to encourage global goal orientation. After each action is executed, if the agent shortens the Euclidean distance from the target point compared to the previous step, it is considered to be moving in the correct direction, and a positive reward is given to strengthen its goal-oriented behavior. This reward mechanism calculates the change in the Euclidean distance between the current position and the target point to achieve. When , it indicates that the agent is approaching the target, and the system gives positive feedback; otherwise, a penalty is imposed. This design can effectively guide the agent to maintain the globally optimal direction, avoiding path oscillations and ineffective movements.

[0178] Geographical cost penalty: According to the cost value of the current grid point, a penalty is set for high-cost areas to encourage the agent to prefer low-cost areas. After each action is executed, the grid cell reached by the agent has a corresponding construction cost value c, which is determined by the weighted result of the geographical information system. In order to guide the agent to avoid high-cost areas, a penalty term is introduced into the reward function, where is the benchmark cost, is the cost penalty coefficient. The higher the cost of the current grid point, the greater the penalty, so as to encourage the agent to tend to move in low-cost areas in path selection, achieving the purpose of controlling the construction cost.

[0179] Repeated visit penalty: Set a penalty for re-visiting the visited grid points to avoid path oscillations and ineffective loops. By maintaining a set of visit records to mark the grid cells that the agent has reached, if the agent moves to this grid point again, a penalty term is introduced into the reward of the current step. This design can effectively suppress meaningless state repetition, improve the efficiency and coherence of path planning, and help generate a more reasonable and coherent transmission line plan.

[0180] Goal reached reward: A large positive reward is given when the target point is successfully reached to encourage the completion of the task as soon as possible.

[0181] In summary, the reward function is as follows:

[0182]

[0183] In the formula, is the reward function, is the state, is the action, , are both weight coefficients, , are the Euclidean distances from the previous step and the current step to the target position respectively, , are the actual construction estimated cost value and the benchmark cost of the current grid cell respectively, is the repeated visit penalty term, is the reaching target reward term.

[0184] In terms of training hyperparameter settings, to adapt to large-scale map environments (such as 2000×2000 grid points), the model uses a relatively small learning rate learning_rate to improve training stability, increases the rollout step size n_steps to enhance the learning ability of long-distance path features, and at the same time increases the batch size to enhance the stability of gradient update. Finally, after multiple trainings, the model can effectively generate an optimal path that is continuous, cost-controlled, and avoids high-risk areas, providing technical support for actual transmission line planning. The values of the above parameters are shown in Table 4:

[0185] Table 4 Values of Parameters in the PPO Algorithm

[0186]

[0187] Step S404: Update the policy network through the immediate reward value, and use the updated policy network as the current policy network.

[0188] Among them, the expected cumulative reward is maximized by the gradient ascent method, that is, the parameters of the policy network are continuously adjusted to increase the probability that the agent selects an action that can obtain a higher cumulative reward in the same state. In each iteration, the agent generates an action probability distribution according to the current policy network, samples and selects an action based on this, observes the change of the environmental state and the immediate reward obtained after executing the action, and then stores this experience (state, action, reward, next state) in the experience replay buffer. When the buffer accumulates a sufficient number of experience samples, a batch of samples is randomly drawn from it for updating the policy network. In this process, the proximal policy optimization algorithm ensures the stable update of the policy by restricting the change amplitude between the old and new policies, and at the same time uses the advantage function to evaluate the quality of the current action relative to the average policy, so as to guide the optimization direction of the policy network. Through continuous iterative training, the policy network of the agent gradually converges to the optimal policy, that is, it can efficiently plan a low-cost transmission line path on the cost grid image data.

[0189] Step S405: Determine whether the agent reaches the target end point according to the state corresponding to the current action.

[0190] Step S406: If it is determined that the agent has not reached the target end point, then based on the current policy network, transfer to the step of using the current state and the policy network to select the current action until it is determined that the agent reaches the target end point, and obtain the geographical locations of all grid units in the cost grid image data.

[0191] Step S407: Determine the predicted path of the transmission line according to the geographical locations of all grid units in the cost grid image data.

[0192] It should be noted that the embodiment of the present application designs a comprehensive reward function, including target approach reward, geographical cost penalty, repeated visit penalty and target arrival reward. This reward function design enables the agent to be more in line with the actual engineering requirements when searching for a path, encourages global goal orientation, prompts the agent to preferentially select low-cost areas, avoids path oscillation and invalid loops, and at the same time encourages the task to be completed as soon as possible, improving the efficiency and practicability of path planning.

[0193] In some embodiments, after predicting the predicted path of the transmission line, it is applied to the path planning prediction of the transmission line in the real map data, and compared and evaluated with the classic ant colony algorithm to verify the application effect of the PPO model in a complex geographical environment, especially the path planning ability in the face of various engineering constraint conditions. By comparing with the ant colony algorithm, the advantages and disadvantages of the PPO algorithm in path optimization, calculation efficiency, adaptability, etc. can be fully evaluated.

[0194] Select a certain area (1km×1km) in XX City. First, obtain high-resolution remote sensing image data as shown in Figure 8 using Google Earth Pro and ENVI software. Then, use ArcMap to realize the rasterized recognition and classification of land cover types, calculate the cost weights of the rasters, and finally use the PPO path prediction model for path planning and prediction. The rasterized map (100×100) and prediction results of this area are shown in Figure 9 . The red line is the predicted path from the starting point to the target point given by the PPO model, and the blue line is the predicted path of the ant colony algorithm.

[0195] From Figure 9 , it can be seen that the total cost of the path predicted by the PPO algorithm is 2872, while the total cost of the path predicted by the ant colony algorithm is 4007, reflecting that the PPO model has a more prominent cost control ability in overall path planning. This result shows that the PPO algorithm can effectively guide the agent to avoid high-cost areas, thus generating a more economical and reasonable path. In contrast, the ant colony algorithm is more likely to fall into local optimality, resulting in the path passing through areas with higher-cost land cover, thereby significantly increasing the overall construction cost. In terms of path length, the path generated by the PPO algorithm is 2.00 km, slightly better than the 2.10 km predicted by the ant colony algorithm, further demonstrating its advantage in path compactness. In terms of computational efficiency, the time taken for a single path prediction by the PPO algorithm is 35 seconds, which is basically equivalent to 31 seconds of the ant colony algorithm. It should be noted that in this invention patent, the total cost and total length of the path are calculated by accumulating the cost values and actual lengths of each raster cell passed by the path, ensuring the accuracy and comparability of the evaluation results.

[0196] In addition, in order to further evaluate the generalization ability and practicality of the PPO path prediction model, this study selected a larger test area - a certain section in XX City, XX Province (5 km×5 km), and conducted path prediction experiments under the actual land cover distribution and cost constraints of this area. As shown in Figure 10 , the terrain of this area is more complex and the distribution of land cover types is more diverse, posing a higher challenge to the path planning algorithm. The raster map (2000×2000) formed after data processing of this map is shown in Figure 11 .

[0197] From Figure 11It can be seen that in the path planning experiment of the 5km × 5km area in XX City, the PPO algorithm demonstrates good cost control and path optimization capabilities. The total cost of its predicted path is 70,451, the path length is 10.72km, and the calculation time is 36 seconds, all of which are superior to the comparison algorithm. In contrast, the total cost of the path generated by the ant colony algorithm in the same environment is as high as 91,877, the path length is 12.46km, and the prediction time reaches 8 minutes and 44 seconds, reflecting problems such as low search efficiency and degraded path quality in large-scale complex terrains. This result further verifies the robustness and adaptability of the PPO algorithm in dealing with large-scale and multi-constraint path planning problems. The PPO model can generate a feasible solution with lower cost and shorter path in a shorter time, effectively avoiding high-cost areas, demonstrating the application potential and advantages of reinforcement learning in actual power line planning tasks.

[0198] Based on the same inventive concept, an embodiment of the present application further provides a transmission line path planning system for implementing the transmission line path planning method involved above.

[0199] The implementation solution provided by this system to solve the problem is similar to the implementation solution described in the above method. Therefore, the specific limitations in one or more embodiments of the transmission line path planning system provided below can refer to the limitations on the transmission line path planning method in the above text and will not be elaborated here.

[0200] As Figure 12 shown, an embodiment of the present application provides a transmission line path planning system, including:

[0201] An image raster module 100, configured to obtain original image data for the target planning area corresponding to the transmission line path planning, and rasterize the original image data to obtain raster image data;

[0202] A raster classification module 200, configured to classify the land cover types of each raster unit in the raster image data to obtain the land cover types of each raster unit;

[0203] A cost raster recognition module 300, configured to assign an actual construction estimated cost value for building a transmission line to each raster unit according to the classified land cover types to obtain cost raster image data;

[0204] A path planning module 400, configured to construct an agent for the transmission line path planning task, and based on the reinforcement learning algorithm, predict the transmission line path through the agent in the cost raster image data.

[0205] In some embodiments, the raster classification module 200 is configured to:

[0206] Use the ArcMap image classification tool to perform interactive supervised classification on the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell.

[0207] In some embodiments, the system further includes: a cost determination module for determining the actual construction estimated cost values corresponding to each land cover type; determining the actual construction estimated cost values corresponding to each land cover type includes:

[0208] Construct an analytic hierarchy process model for the land cover weighting system based on the analytic hierarchy process; the analytic hierarchy process model includes an objective layer, a factor layer, and a scheme layer; wherein, the objective layer is the transmission line path planning benefit, the factor layer includes the avoidance cost of ecological-geographical constraints, the crossing cost of linear engineering interference, the social cost sensitive to human activities, and the meteorological impact cost of meteorological load effects, and the scheme layer includes the sub-influence factors of each factor within the factor layer;

[0209] For each factor in the factor layer, construct an initial objective judgment matrix and perform PCA dimensionality reduction on the initial objective judgment matrix to obtain an objective judgment matrix;

[0210] For each factor in the factor layer, construct a subjective judgment matrix according to the 1-9 marking method and perform a consistency test on the subjective judgment matrix;

[0211] When the subjective judgment matrix passes the consistency test, fuse the subjective judgment matrix with the objective judgment matrix to obtain a comprehensive judgment matrix;

[0212] Calculate the single-layer weight vector for the comprehensive judgment matrix; the single-layer weight vector is used to determine the importance weights of each factor in the same layer relative to the factor in the upper layer;

[0213] According to the single-layer weight vector, determine the comprehensive weight vector and perform a consistency test on the comprehensive weight vector;

[0214] When the comprehensive weight vector passes the consistency test, allocate weights to each land cover type through the comprehensive weight vector to obtain the weights of each land cover type;

[0215] Perform numerical transformation on the weights of each land cover type to obtain the actual construction estimated cost values corresponding to each land cover type.

[0216] In some embodiments, the cost raster recognition module 300 is used for:

[0217] Assign the actual construction estimated cost values corresponding to each land cover type to each grid cell for constructing a transmission line according to the classified land cover types to obtain the initial cost raster image data;

[0218] Normalize each grid cell of the initial cost grid image data to obtain the cost grid image data.

[0219] In some embodiments, the path planning module 400 is configured to:

[0220] Construct an agent for the transmission line path planning task, as well as the state space and action space of the agent; wherein, the state space is the geographical location of each grid unit of the cost grid image data in the grid coordinate system, and the action space includes the four-direction discrete actions that the agent can choose.

[0221] Use the proximal policy optimization algorithm to train the policy network of the agent. When the agent starts from the starting point of the line planning on the cost grid image data, use the current state and the policy network to select the current action.

[0222] Execute the current action and determine the immediate reward value of the current action through the reward function.

[0223] Update the policy network through the immediate reward value, and use the updated policy network as the current policy network.

[0224] Judge whether the agent reaches the target end point according to the state corresponding to the current action.

[0225] If it is judged that the agent does not reach the target end point, then based on the current policy network, go to the step of using the current state and the policy network to select the current action until it is judged that the agent reaches the target end point, and obtain the geographical location of all grid units of the agent in the cost grid image data.

[0226] Determine the predicted path of the transmission line according to the geographical location of all grid units of the agent in the cost grid image data.

[0227] In some embodiments, the reward function is:

[0228]

[0229] In the formula, is the reward function, is the state, is the action, 、 are both weight coefficients, 、 are the Euclidean distances from the previous step and the current step to the target position respectively, 、 are the actual construction estimated cost value and the benchmark cost of the current grid cell respectively, is the repeated access penalty term, is the reach target reward term.

[0230] Such asFigure 13 As shown in Figure 13 , an embodiment of the present application provides an electronic device. The electronic device 10 includes a memory 20 and a processor 30. A computer program is stored in the memory 20. When the computer program is executed by the processor 30, the processor 30 is caused to execute the steps of the transmission line path planning method in the above embodiment.

[0231] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed, the steps of the transmission line path planning method in the above embodiment are implemented.

[0232] An embodiment of the present application provides a computer program product. The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. Wherein, when the program instructions are executed by a computer, the computer is caused to execute the steps of the transmission line path planning method described in the above embodiment.

[0233] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, electronic device, computer storage medium, and computer program product can refer to the corresponding processes in the foregoing method embodiments and will not be described herein again.

[0234] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above drawings are used to distinguish similar objects and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0235] It should be understood that although the steps in the flowcharts involved in the various embodiments described above are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the various embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0236] In several embodiments provided by the present invention, it should be understood that the disclosed systems, electronic devices, computer storage media, computer program products and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, indirect couplings or communication connections of devices or units, and can be in electrical, mechanical or other forms.

[0237] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0238] In addition, the functional units in the various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0239] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in various embodiments of the present invention through a computer device (which may be a personal computer, a server, or a network device, etc.). The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (English full name: Read-Only Memory, English abbreviation: ROM), random access memories (English full name: Random Access Memory, English abbreviation: RAM), magnetic disks, or optical discs that can store program codes.

[0240] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. A method for planning the path of a transmission line, characterized in that, Including: Obtain the original image data of the target planning area corresponding to the transmission line path planning, and rasterize the original image data to obtain raster image data; Classify the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell; Assign the actual construction estimated cost value of building a transmission line to each grid cell according to the classified land cover types to obtain cost raster image data; Construct an agent for the transmission line path planning task, and based on the reinforcement learning algorithm, predict the transmission line path through the agent in the cost raster image data to obtain the predicted transmission line path.

2. The transmission line path planning method according to claim 1, wherein The classifying the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell includes: Use the ArcMap image classification tool to perform interactive supervised classification on the land cover types of each grid cell in the raster image data to obtain the land cover types of each grid cell.

3. The transmission line path planning method according to claim 1, wherein It also includes: Determine the actual construction estimated cost value corresponding to each land cover type; The determining the actual construction estimated cost value corresponding to each land cover type includes: Construct an analytic hierarchy process model of the land cover weighting system based on the analytic hierarchy process; the analytic hierarchy process model includes an objective layer, a factor layer, and a scheme layer; among them, the objective layer is the transmission line path planning benefit, and the factor layer includes the avoidance cost of ecological-geographical constraints, the crossing cost of linear engineering interference, the social cost sensitive to human activities, and the meteorological impact cost of meteorological load effects, and the scheme layer includes the sub-influence factors of each factor in the factor layer; For each factor in the factor layer, construct an initial objective judgment matrix, and perform PCA dimensionality reduction on the initial objective judgment matrix to obtain an objective judgment matrix; For each factor in the factor layer, construct a subjective judgment matrix according to the 1-9 marking method, and perform a consistency test on the subjective judgment matrix; When the subjective judgment matrix passes the consistency test, fuse the subjective judgment matrix with the objective judgment matrix to obtain a comprehensive judgment matrix; Calculate the single-layer weight vector for the comprehensive judgment matrix; the single-layer weight vector is used to determine the importance weight of each factor in the same layer relative to the factor in the upper layer; According to the single-layer weight vector, determine the comprehensive weight vector, and perform a consistency test on the comprehensive weight vector; When the comprehensive weight vector passes the consistency test, allocate weights to each land cover type through the comprehensive weight vector to obtain the weights of each land cover type; Perform numerical transformation on the weights of each land cover type to obtain the actual construction estimated cost value corresponding to each land cover type.

4. The transmission line path planning method according to claim 3, wherein The assigning the actual construction estimated cost value of building a transmission line to each grid cell according to the classified land cover types to obtain cost raster image data includes: Assign the actual construction estimated cost value corresponding to each land cover type to each grid cell according to the classified land cover types to obtain the initial cost raster image data; Normalize each grid cell of the initial cost grid image data to obtain the cost grid image data.

5. The transmission line path planning method according to any one of claims 1 to 4, characterized in that, Construct an agent for the transmission line path planning task. Based on the reinforcement learning algorithm, use the agent to predict the transmission line path in the cost grid image data, obtaining the predicted transmission line path, including: Construct an agent for the transmission line path planning task, as well as the state space and action space of the agent; wherein, the state space is the geographical location of each grid unit of the cost grid image data in the grid coordinate system, and the action space includes the four-direction discrete actions that the agent can choose. Use the Proximal Policy Optimization algorithm to train the policy network of the agent. When the agent starts from the starting point of the line planning on the cost grid image data, use the current state and the policy network to select the current action. Execute the current action and determine the immediate reward value of the current action through the reward function. Update the policy network through the immediate reward value and use the updated policy network as the current policy network. Judge whether the agent reaches the target end point according to the state corresponding to the current action. If it is judged that the agent does not reach the target end point, then based on the current policy network, go to the step of using the current state and the policy network to select the current action until it is judged that the agent reaches the target end point, obtaining the geographical locations of all grid units of the agent in the cost grid image data. Determine the predicted transmission line path according to the geographical locations of all grid units of the agent in the cost grid image data.

6. The transmission line path planning method according to claim 5, wherein, The reward function is: In the formula, is the reward function, is the state, is the action, and are both weight coefficients, and are the Euclidean distances from the previous step and the current step to the target position respectively, and are the actual construction estimated cost value and the benchmark cost of the current grid cell respectively, is the repeated access penalty term, is the reaching target reward term.

7. A transmission line path planning system, characterized in that, Including: An image grid module, used to obtain the original image data of the target planning area corresponding to the transmission line path planning, and rasterize the original image data to obtain raster image data. A raster classification module, used to classify the land object types of each grid cell in the raster image data to obtain the land object types of each grid cell. A cost grid recognition module, used to assign the actual construction estimated cost value for building a transmission line to each grid cell according to the classified land object types, obtaining the cost grid image data. A path planning module, used to construct an agent for the transmission line path planning task. Based on the reinforcement learning algorithm, use the agent to predict the transmission line path in the cost grid image data, obtaining the predicted transmission line path.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor. When the computer program stored in the memory is executed by the processor, the processor executes the steps of the transmission line path planning method according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it realizes the steps of the transmission line path planning method according to any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, wherein when the program instructions are executed by a computer, the computer is caused to execute the steps of the transmission line path planning method according to any one of claims 1-6.

Images