Cable collection line planning optimization method based on long and narrow photovoltaic field area

Abstract

The invention relates to the technical field of photovoltaic power generation, discloses a cable collection line planning optimization method based on a long and narrow photovoltaic field area, and aims to solve the technical problem that a traditional scheme is poor in economical efficiency and electrical performance in long and narrow photovoltaic field area cable collection line planning. The method comprises the following steps: constructing a weighted network model fusing terrain and electrical constraints; dividing electrical sub-regions based on power weight clustering and determining a virtual confluence point; and by applying a genetic algorithm, through electrical and economic coupling calculation, optimizing the trunk line by taking the total cost minimization of the whole life cycle as a target. According to the technical scheme, the power station investment return rate can be improved, the cable collection line considering the electrical performance and the economic benefit can be quickly generated, and the cable collection line is particularly suitable for a long and narrow photovoltaic field area.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic power generation technology, and more specifically, it relates to a method for planning and optimizing cable collection lines in long and narrow photovoltaic fields. Background Technology

[0002] The field of new energy technology encompasses multiple branches such as solar photovoltaic power generation, wind power generation, and hydropower generation. The core of this technology is to use renewable energy to generate electricity to replace traditional fossil fuels. Photovoltaic power generation refers to the technology of directly converting light energy into electrical energy using the photovoltaic effect at the semiconductor interface. With the global pursuit of carbon neutrality and the decline in the cost of photovoltaic modules, photovoltaic power generation technology has been widely promoted and applied, covering multiple application scenarios such as centralized large-scale power plants and distributed rooftop systems. The development of this field is not only of great significance for optimizing the energy structure, but also promotes the technological progress of related industrial chains, including power electronics, energy storage systems, and other aspects.

[0003] The planning of cable collection lines in photovoltaic power plants refers to the design and layout of cable networks within the power plant to efficiently and economically collect the direct current (DC) generated by the numerous photovoltaic strings or the alternating current (AC) converted by inverters to the substation. This planning aims to comprehensively consider factors such as cable cost, power loss, and ease of construction, and determine the cable routes, types, and connection topologies through optimization algorithms and path selection. In this planning, the cable network layout directly affects the initial investment and long-term operating efficiency of the power plant. Through reasonable path planning and cable cross-sectional area selection, energy loss and construction costs can be effectively reduced, thereby improving the overall economic benefits and energy output of the photovoltaic power plant.

[0004] Current technologies rely heavily on manual experience or preliminary designs based on general graph theory algorithms during route planning, lacking adaptability to specific site morphologies and failing to effectively optimize for irregular terrains such as narrow and elongated areas. This results in excessively long cables and increased investment costs. Route planning often focuses solely on finding the shortest path, failing to effectively couple electrical performance constraints such as cable current carrying capacity, voltage drop, and power loss. This leads to inefficiencies in some lines due to excessive losses, or the over-scaling of cable specifications to meet loss requirements. Automated planning tools fail to incorporate actual construction limitations such as terrain and road obstacles as hard constraints into the algorithm model, resulting in a disconnect between the planned scheme and on-site construction conditions, lacking feasibility. The separate processes of cable selection and route layout prevent global collaborative optimization, making it difficult to achieve an optimal balance between cost and loss. These problems are particularly pronounced in photovoltaic sites with complex terrain and narrow layouts, directly limiting the improvement of power plant system efficiency and return on investment. Summary of the Invention

[0005] The purpose of this invention is to address the technical problem that existing technologies, when planning cable collection routes in long and narrow photovoltaic fields, only focus on the shortest geometric path while ignoring the combined effects of power loss, terrain constraints, and total life cycle costs, resulting in poor economic efficiency and electrical performance of the planning schemes. The invention proposes an optimization method for planning cable collection routes in long and narrow photovoltaic fields.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for planning and optimizing cable collection lines in narrow photovoltaic fields, comprising the following steps: S1: Obtain geospatial information data of the long and narrow photovoltaic field area. The geospatial information data includes the precise geographical coordinates of the photovoltaic strings, the location coordinates of the main substation, the plot boundary vector data, and the preset no-layout zone and forced passage corridor vector data. Based on the geospatial information data, generate a weighted rasterized cost map that integrates terrain construction costs and electrical constraints, and convert the weighted rasterized cost map into a non-Euclidean weighted network graph model.

[0007] S2: Based on the weighted network graph model, a density clustering algorithm that takes into account the power generation weight is applied to divide the photovoltaic strings into several electrical collection sub-regions; within each electrical collection sub-region, a unique and cost-optimal virtual collector point is determined according to the path cost calculated by the weighted network graph model, and a virtual collector point dataset containing the coordinates of each virtual collector point and the total power of the photovoltaic strings under its jurisdiction is generated.

[0008] S3: Initialize a population for trunk cable path optimization, where each individual in the population represents a trunk cable topology tree structure connecting all virtual junction points to the main substation; for each individual's topology tree structure, perform iterative electrical-physical characteristics and economic coupling calculations, evaluate the individual through a preset fitness function that aims to minimize the total life cycle cost, and generate its fitness evaluation value.

[0009] S4: Based on the fitness evaluation value, perform selection, crossover, and mutation operations on the population to generate a new generation of population; repeat steps S3 and S4 until the average fitness evaluation value of the population converges to a preset stability threshold, or the number of iterations reaches a preset maximum generation limit; select and output the individual with the best fitness evaluation value in the final converged generation, the topology tree structure represented by this individual is the optimal trunk cable convergence line, and generate a detailed cable laying path diagram and material specification list based on this line.

[0010] As a further aspect of the present invention, the weighted rasterized cost map is generated as follows: the elongated photovoltaic field is divided into a two-dimensional matrix composed of raster units of a predetermined size; for each raster unit, a basic construction cost coefficient is assigned according to whether it is located in a restricted area, a mandatory passage corridor, or a regular area; further, the terrain slope of each raster unit is calculated by combining digital elevation model data, and the basic construction cost coefficient is corrected according to a preset mapping relationship between slope and construction difficulty to obtain the final passage cost value of the raster unit, thereby constituting the weighted rasterized cost map.

[0011] As a further aspect of the present invention, the weighted network graph model is generated as follows: the center point of each grid cell in the weighted rasterized cost map is abstracted as a network node; for any network node, it is connected to its eight adjacent nodes in the adjacent directions to generate network edges; the weight value of each network edge is defined as the product of the arithmetic mean of the passage cost values ​​of the two grid cells spanned by the two nodes and the geometric distance between the two nodes; all network nodes and weighted network edges together constitute the weighted network graph model.

[0012] As a further aspect of the present invention, the specific implementation of the density clustering algorithm that takes into account the power generation weight is as follows: an improved DBSCAN algorithm is adopted, the core parameters of which include the neighborhood radius (Eps) and the minimum power threshold of the core object (MinP); the neighborhood radius (Eps) is defined as the minimum path cost between two photovoltaic string nodes calculated by Dijkstra's algorithm based on the weighted network graph model; the minimum power threshold of the core object (MinP) is defined as the lower limit of the sum of the rated power of all photovoltaic string nodes contained in the neighborhood radius of a photovoltaic string node; the algorithm divides all photovoltaic string nodes into core points, boundary points and noise points, and forms several electrical aggregation sub-regions.

[0013] As a further aspect of the present invention, the method for determining the virtual junction point specifically comprises: within each electrical collection sub-region, taking all photovoltaic string nodes in the region as the target node set; traversing all network nodes in the region, taking each network node as a candidate virtual junction point; for each candidate virtual junction point, calculating the sum of the minimum path costs to all nodes in the target node set through the weighted network graph model; and selecting the candidate virtual junction point with the minimum sum of the minimum path costs as the final virtual junction point of the electrical collection sub-region.

[0014] As a further aspect of the present invention, the iterative calculation of the coupling of electrical and physical characteristics and economic efficiency specifically includes the following steps: S311: For the current individual's topology tree structure, starting from the virtual junction point furthest from the main substation, perform power flow calculations segment by segment along the topology path toward the main substation, and accurately calculate the root mean square current value that each segment of the main cable needs to carry. S312: Based on the calculated root mean square current value, query and select the cable specification that meets the current carrying capacity requirement and has the smallest cross-sectional area from a preset cable standard database containing various cable specifications, including cable model, cross-sectional area, resistance per unit length, cost per unit length, and rated current carrying capacity, as the configuration for this section of the line. S313: Based on the unit length resistance of the selected cable, the length of the line segment, and the root mean square current, calculate the average annual power loss of the line segment using Joule's law (P=I²R), multiply it by the electricity price and the design life of the power station, and then use the net present value formula to calculate the total life cycle energy loss cost of the line segment. S314: Multiply the unit length cost of the selected cable by the length of the line segment to obtain the material cost, and add up the passage cost values ​​of the grid cells traversed by the line path to obtain the construction cost. The sum of the two constitutes the initial investment cost of the line segment. S315: Add the initial investment cost of all segments in the topology tree to the total life cycle power loss cost to obtain the total life cycle cost of the complete line scheme represented by the individual. The reciprocal of the total cost is the fitness evaluation value of the individual.

[0015] As a further aspect of the present invention, the specific implementation of the selection, crossover, and mutation operations is as follows: the selection operation adopts a roulette wheel selection method, where the probability of an individual being selected is proportional to its fitness evaluation value; the crossover operation adopts a Prüfer sequence-based encoding method to encode the topological tree structure of two parent individuals, generates the encoding of two child individuals by exchanging parts of the Prüfer sequence, and then decodes it into a new topological tree structure; the mutation operation randomly changes a code value in the Prüfer sequence of an individual with a preset low probability, thereby introducing a small, local change in the connection relationship in the topological tree structure.

[0016] This invention also provides a planning and optimization system for cable collection lines in narrow photovoltaic fields. The system is used to execute the above-mentioned algorithm and includes: A geospatial data parsing and modeling module is configured to: receive and parse geospatial information containing vector data of photovoltaic string coordinates, main substation location, restricted areas, and mandatory access corridors; based on the information, generate a weighted rasterized cost map including terrain slope correction, and further transform the map into a non-Euclidean weighted network graph model with raster centers as nodes and weighted access costs as edge weights, providing a basic data structure for subsequent path calculation.

[0017] A power-aware sub-region clustering module is configured to: execute an improved DBSCAN clustering algorithm on the weighted network graph model, using path cost as the distance metric and total regional power as the density metric, to automatically divide all photovoltaic string nodes into several highly cohesive and loosely coupled electrical aggregation sub-regions; and within each sub-region, through traversal calculation, determine a network node that minimizes the total path cost to all string nodes in the region, which serves as the virtual merge point for that region.

[0018] A multi-objective co-evolutionary optimization module is configured to: use a genetic algorithm framework to initialize a population consisting of multiple trunk cable topology tree structures; and incorporate a coupled computation core to perform segment-by-segment power flow calculation, cable selection, annual average power loss assessment, and initial investment cost calculation for each topology scheme represented by each individual in the population, and finally provide a precise quantitative evaluation based on a fitness function that integrates initial investment and full life cycle operating costs.

[0019] A global optimal solution generation and output module is configured to: dominate the iterative evolution process of the genetic algorithm, continuously optimize the population by executing operators such as roulette wheel selection, crossover based on Pluff sequence, and single-point mutation; monitor the change in the average fitness of the population in real time, and after meeting the convergence condition or reaching the maximum number of iterations, select the individual with the highest fitness from the final generation population; decode the optimal individual into the final cable convergence line topology, and call the path planning algorithm to generate specific and constructible cable laying paths on the weighted network graph model, while outputting a detailed bill of materials containing the cable type, length, and total cost of each segment.

[0020] A device for planning and optimizing cable collection lines in a narrow photovoltaic field, the device comprising a memory for storing information about the system and a central processing unit for performing calculations on the system.

[0021] Compared with the prior art, the advantages and positive effects of the present invention are as follows: This invention elevates route planning from an ideal two-dimensional plane to a three-dimensional physical space closely resembling engineering realities by constructing a weighted network graph model that integrates terrain construction costs and electrical constraints, ensuring the practical feasibility of the planned path. It innovatively employs a density clustering algorithm that considers power generation weights, enabling the scientific grouping of photovoltaic strings based on electrical characteristics. This results in a more rational electrical aggregation sub-region structure, laying a solid foundation for subsequent optimization of the main cable. The core optimization process utilizes a genetic algorithm and designs a complex fitness function that couples electrical physics calculations with full life-cycle economic analysis. This ensures that the optimization objective is no longer simply the shortest cable length, but rather a globally optimal balance between initial investment costs and long-term power loss costs, significantly improving the long-term return on investment for photovoltaic power plant projects. The entire algorithm system, from data modeling, region division, backbone optimization to solution output, forms a complete, closed-loop, and automated optimization process. Compared with traditional methods that rely on human experience or single-objective optimization software, this invention can quickly and accurately generate cable collection line solutions that combine electrical performance and economic benefits under complex constraints. In particular, its optimization effect is especially significant in photovoltaic field areas with special terrain such as narrow and long shapes. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall process of a route planning optimization method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the functional modules of a route planning and optimization system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating an application scenario for cable collection line planning provided by an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the principle of generating a weighted network graph model from a weighted rasterized cost map in an embodiment of the present invention; Figure 5 This is a detailed flowchart illustrating the trunk cable path evolution optimization process in an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In the description of this invention, it should be understood that the terms "upper," "lower," "left," "right," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention.

[0024] Reference Figure 2 This paper presents a planning and optimization system for cable collection routes in long and narrow photovoltaic power plants, which is used to execute the optimization method described in detail later. Logically, the system includes a geospatial data parsing and modeling module, a power-aware sub-region clustering module, a multi-objective co-evolutionary optimization module, and a global optimal solution generation and output module. The geospatial data parsing and modeling module is configured to receive and process various types of geospatial information to construct a cost model incorporating multiple constraints. The power-aware sub-region clustering module is configured to perform scientific electrical zoning based on the cost model. The multi-objective co-evolutionary optimization module is configured to use an evolutionary computation framework to optimize the trunk cable topology. The global optimal solution generation and output module is configured to guide the entire evolutionary process and ultimately output an executable engineering solution.

[0025] Reference Figure 1 The present invention provides a method for planning and optimizing cable collection lines in a narrow photovoltaic field, the overall process of which is shown in the schematic diagram below. Figure 1 As shown, the specific steps include: Step S1: Obtain geospatial information data of the long and narrow photovoltaic field area. The geospatial information data includes the precise geographical coordinates of the photovoltaic strings, the location coordinates of the main substation, the plot boundary vector data, and the preset no-layout zone and forced passage corridor vector data. Based on the geospatial information data, generate a weighted rasterized cost map that integrates terrain construction costs and electrical constraints, and convert the weighted rasterized cost map into a non-Euclidean weighted network graph model.

[0026] Step S2: Based on the weighted network graph model, a density clustering algorithm that takes into account the power generation weight is applied to divide the photovoltaic strings into several electrical collection sub-regions; within each electrical collection sub-region, a unique and cost-optimal virtual collector point is determined according to the path cost calculated by the weighted network graph model, and a virtual collector point dataset containing the coordinates of each virtual collector point and the total power of the photovoltaic strings under its jurisdiction is generated.

[0027] Step S3: Initialize a population for trunk cable path optimization. Each individual in the population represents a trunk cable topology tree structure connecting all virtual junction points to the main substation. For each individual's topology tree structure, perform iterative electrical and physical characteristics coupled with economics calculations. Evaluate the individual using a preset fitness function that aims to minimize the total lifecycle cost and generate its fitness evaluation value.

[0028] Step S4: Based on the fitness evaluation value, perform selection, crossover, and mutation operations on the population to generate a new generation of population; repeat steps S3 and S4 until the average fitness evaluation value of the population converges to a preset stability threshold, or the number of iterations reaches a preset maximum generation limit; select and output the individual with the best fitness evaluation value in the final converged generation, and the topology tree structure represented by this individual is the optimal trunk cable convergence line, and generate a detailed cable laying path diagram and material specification list based on this line.

[0029] In a specific embodiment, step S1 is described in detail. The core task of this step is to construct a mathematical model that can accurately reflect the cost of cable laying in the real world.

[0030] Step S101: Perform geospatial information data acquisition and preprocessing.

[0031] Specifically, the acquired geospatial information data of the photovoltaic strings can originate from an unmanned aerial survey system deployed with a Real-time Kinematic (RTK) module. After completing high-precision aerial photography of the photovoltaic field, this system can output the three-dimensional coordinates of the center point of each photovoltaic string through image interpretation and ground control point correction. Its planar positioning accuracy is better than 5 cm, and its elevation accuracy is better than 10 cm. The data structure can be defined as a list containing multiple records. The fields of each record include: string unique identifier (String_ID, data type: string), longitude (Longitude, data type: floating point), latitude (Latitude, data type: floating point), altitude (Altitude, data type: floating point), and rated DC power (Rated_Power_kW, data type: floating point). For example, a record might be {String_ID: "S-A01-01", Longitude: 113.456789, Latitude: 23.987654, Altitude:154.32, Rated_Power_kW: 48.5}. If the rated power data for some photovoltaic strings is missing in the initial dataset, the system will call a preset default value calculated based on the specifications of the mainstream photovoltaic module model used in the area (e.g., module model JKM550N-72HL4-V) and the number of modules in the string to fill in the missing data. The location coordinates of the main substation are usually a single, fixed point determined during the project design phase. The vector data for plot boundaries, restricted areas, and mandatory access corridors are usually derived from the digitization of standard 1:5000 scale topographic maps provided by local land resources management or urban and rural planning departments, or from the regional divisions determined in the project feasibility study report, in Shapefile or GeoJSON format. Before data fusion, coordinate system verification and transformation are performed to convert all source geographic coordinate data to the same projection coordinate system, such as the Gauss-Krüger projection or the Universal Transverse Mercator (UTM) projection, to ensure the accuracy of subsequent distance and area calculations.

[0032] Step S102: Generate a weighted rasterized cost map. This process transforms a continuous geographic space into a discrete raster matrix with cost attributes.

[0033] Step S1021: Perform field area rasterization. The entire two-dimensional planar projection area of ​​the photovoltaic field is divided into a two-dimensional matrix composed of raster cells of predetermined size. Determining the size of the raster cells (e.g., 2 meters × 2 meters) is a process of balancing computational accuracy and computational resource consumption. The specific value is selected based on a comprehensive consideration of the minimum operating width (approximately 1.5 meters) and turning radius of typical cable laying construction machinery (such as small tracked trenchers), as well as the original resolution of the acquired Digital Elevation Model (DEM) data (e.g., a 5-meter resolution DEM provided by the National Geomatics Center of China). Smaller raster sizes can depict the terrain more finely, but this leads to a quadratic increase in the number of nodes in the network graph model, increasing the computational burden on subsequent path search algorithms.

[0034] Step S1022: Assign a foundation construction cost coefficient. For each grid cell, based on its spatial location and the relationship with the preset area vector data, assign a dimensionless foundation construction cost coefficient C_base. In a specific embodiment, the assignment rule for this coefficient is as follows: If the center point of a grid cell is located within a normal area (i.e., not within a restricted area or a mandatory passageway), its C_base is set to the base value of 1.0.

[0035] If the center point of a grid cell is located within a restricted area (e.g., a basic farmland protection zone, an ecological protection red line, a safe distance zone for high-voltage power line tower foundations, or a geologically hazardous area), its C_base is set to an infinite value in terms of utility, such as 1.0e+9. This setting ensures that any path planning algorithm will actively avoid these grid cells when accumulating costs, thus creating a de facto insurmountable obstacle in the model.

[0036] If the center point of a grid cell is located within a mandatory access corridor (e.g., a planned maintenance road or a reserved cable trench within the site), its C_base is set to a discount value less than 1.0, such as 0.3. This setting significantly reduces the path cost along the corridor, thereby strongly guiding cable paths to be preferentially laid along the corridor during the optimization process.

[0037] Step S1023: Perform terrain slope cost correction. Using digital elevation model (DEM) data, calculate the local terrain slope for each grid cell and quantify the correction of the foundation construction cost coefficient. The terrain slope S_grid (unit: degrees) can be calculated by applying a third- or fifth-order finite difference operator (such as the Sobel operator or the Horn algorithm) to the DEM data. This method can simultaneously calculate the slope value and aspect. The impact of slope on construction costs is reflected by a non-linear correction coefficient f_slope. The relationship between this coefficient and the slope S_grid can be defined by a preset piecewise function or look-up table (LUT). The specific form of this function is derived from statistical analysis of past photovoltaic project construction quota data or from consulting the experience of earthwork engineering experts. A specific function example is as follows: If 0° ≤ S_grid < 5° (flat ground): f_slope = 1.0 + 0.08 * S_grid If 5° ≤ S_grid < 15° (gentle slope): f_slope = 1.4 + 0.15 * (S_grid - 5) If 15° ≤ S_grid < 30° (slope): f_slope = 2.9 + 0.40 * (S_grid - 15) If S_grid ≥ 30° (steep slope): f_slope = 8.9 + 0.80 * (S_grid - 30) This function shows that as the slope increases, the construction difficulty and cost increase at an accelerated pace.

[0038] Step S1024: Calculate the final raster access cost. The final access cost value C_final(i, j) of each raster cell (i, j) is obtained by multiplying its basic construction cost coefficient C_base(i, j) by the terrain slope correction coefficient f_slope(i, j), i.e., C_final(i, j) = C_base(i, j) * f_slope(i, j). The C_final values ​​of all raster cells together constitute a complete weighted raster cost map.

[0039] Step S103 involves converting the weighted rasterized cost map into a weighted network graph model. This conversion aims to map the problem from raster space to graph theory space, enabling the application of mature network analysis algorithms. The principle behind this process can be found in [reference needed]. Figure 4 .

[0040] Step S1031: Perform node abstraction. Abstract the geometric center point of each raster cell in the weighted rasterized cost map into a network node N. Each node N_k stores its two-dimensional index (x_k, y_k) in the raster matrix, its geographic coordinates, and the final passage cost value C_final(k) of its raster in the data structure.

[0041] Step S1032: Perform edge generation. For any network node N_i in the graph, retrieve its eight neighboring nodes N_j in the eight neighborhood directions (up, down, left, right, upper left, lower left, upper right, lower right). If neighboring node N_j exists (i.e., it does not exceed the map boundary), then generate an undirected network edge E_ij between N_i and N_j.

[0042] Step S1033: Perform edge weight calculation. The weight W_ij of each network edge E_ij is defined as the combined cost of the path traversed to connect two nodes. Its calculation formula is: W_ij = [(C_final(i) + C_final(j)) / 2]* D_geom(i, j). Wherein, C_final(i) and C_final(j) are the passage cost values ​​of the grids where nodes N_i and N_j are located, respectively, and their arithmetic mean represents the average cost level of traversing the boundary of these two adjacent grids. D_geom(i, j) is the geometric distance between the centers of the two nodes. If N_j is a horizontal or vertical adjacent node of N_i, the value of D_geom(i, j) is equal to the grid size d; if N_j is a diagonal adjacent node, the value of D_geom(i, j) is equal to d * sqrt(2). The final generated graph model G=(V, E, W) is a non-Euclidean space representation containing all nodes, edges and their corresponding weights, where the "distance" between any two points is no longer a straight-line distance, but rather the minimum path cost that takes into account terrain and construction constraints.

[0043] Furthermore, step S2 will be described in detail. This step aims to rationally divide the large number of photovoltaic strings into several electrically independent aggregation units through scientific clustering methods, laying the foundation for subsequent topology optimization of the trunk cable.

[0044] Step S201: Perform density clustering taking into account the power generation weight. This embodiment uses an improved DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm.

[0045] Step S2011: Determine the core parameters of the algorithm. The performance of this algorithm is highly dependent on two core parameters: neighborhood radius (Eps) and minimum power threshold of the core object (MinP).

[0046] Determining the Neighborhood Radius (Eps): Eps here is not geographical distance, but rather the path cost based on the weighted network graph model. Its value is determined using an adaptive k-distance graph method. Specifically, an integer k is set (e.g., k=4, a value based on the typical number of strings connected to a combiner box, usually 4-8 neighboring strings). For each photovoltaic string node, using Dijkstra's algorithm or the more efficient A* algorithm on large graphs, the minimum path cost to all other photovoltaic string nodes in the weighted network graph model is calculated, and the cost value of the k-th nearest neighbor is found; this value is called the node's k-distance. The k-distance values ​​of all photovoltaic string nodes are sorted from largest to smallest, and a k-distance distribution graph is plotted. The horizontal axis of the graph represents the sorted node index, and the vertical axis represents the corresponding k-distance value. Typically, this curve will exhibit a clear "inflection point," with a steep downward trend before the inflection point, followed by a flattening trend. The k-distance value corresponding to this inflection point is selected as the global Eps parameter. This method can determine a reasonable neighborhood range based on the distribution density characteristics of the data itself.

[0047] Determining the Minimum Power Threshold (MinP) for Core Targets: The setting of MinP is directly related to the electrical design specifications and equipment specifications of the photovoltaic power plant. For example, if the site design uses a 16-input DC combiner box with a maximum current of 20A per input, and the rated power of a single photovoltaic string is 48.5kW, then the maximum power that a single combiner box can collect is 16 * 48.5kW = 776kW. To ensure the engineering practicality of the clustering results, MinP can be set as a reasonable proportion of the capacity of a single combiner box, for example, between 50% and 70%, i.e., a value between 388kW and 543kW. In a specific embodiment, MinP is set to 500kW. This setting aims to ensure that the sub-regions formed by clustering have the physical possibility of being electrically aggregated into one or a few combiner boxes, avoiding the generation of fragmented sub-regions with too little power and no engineering significance.

[0048] Step S2012: Perform the improved DBSCAN clustering process. This process traverses all photovoltaic string nodes and divides them into core points, boundary points, or noise points according to the following definition.

[0049] Distance metric: The distance dist(N_a, N_b) between any two photovoltaic string nodes N_a and N_b is defined as the minimum cumulative path cost calculated by Dijkstra's algorithm on the weighted network graph model generated in step S103.

[0050] Neighborhood definition: The Eps neighborhood N_Eps(N_a) of a node N_a is the set of all photovoltaic string nodes N_x whose distance dist(N_a, N_x) from N_a is not greater than Eps.

[0051] Core point definition: If the sum of the rated power of all photovoltaic string nodes (including itself) in the Eps neighborhood of a photovoltaic string node N_a is greater than or equal to MinP, then the node N_a is defined as a core point.

[0052] Clustering logic: The algorithm starts with an arbitrary unvisited core point, creates a new cluster, and adds the core point itself and all points within its Eps neighborhood (including other core points and boundary points) to the cluster. Then, using a breadth-first or depth-first approach, it recursively merges the neighborhoods of newly discovered core points into the cluster until the cluster can no longer expand. After constructing a cluster, a new unvisited core point is selected from the remaining unvisited core points, and the above process is repeated. All photovoltaic string nodes not included in any cluster are ultimately marked as noise points.

[0053] Step S2013: Processing Noise Points. For photovoltaic strings identified as noise points by the algorithm, a post-processing allocation step is performed. Specifically, for each noise point, the minimum weighted path cost to each core point in all formed electrical aggregation sub-regions is calculated. Then, the noise point is allocated to the sub-region that minimizes its path cost. To prevent abnormal allocation, a "maximum home distance threshold" can be set. If the minimum cost from a noise point to all sub-regions exceeds this threshold, the point will be marked as an "isolated point" requiring manual review and decision-making, and an alarm message will be generated.

[0054] Step S202: Determine the virtual busbar point for each electrical aggregation sub-region. The virtual busbar point is the theoretically optimal point for the aggregation of all string cables within the sub-region, and its location directly affects the length and cost of the subsequent trunk cable.

[0055] Step S2021: Define the candidate point set. For each electrical aggregation sub-region C_k divided in step S201, the set of all network nodes (not limited to photovoltaic string nodes) in the weighted network graph model covered by the geographical range of the sub-region is defined as the candidate virtual merge point set S_cand_k of the sub-region.

[0056] Step S2022: Perform iterative path cost summation calculation. Traverse each candidate node N_cand in the candidate point set S_cand_k. For each N_cand, treat it as a potential sink center and calculate the sum of minimum path costs from it to the set of all photovoltaic string nodes {N_pv_1, N_pv_2, ..., N_pv_m} within that sub-region C_k. The total path cost TC(N_cand) is calculated as: TC(N_cand) = Σ_{i=1 to m} [Dijkstra(N_cand, N_pv_i)]. Where Dijkstra(A, B) represents the minimum path cost from node A to node B in the weighted network graph model. This calculation requires performing Dijkstra's algorithm multiple times for each candidate point.

[0057] Step S2023: Select the optimal virtual combiner point. After calculating the total path cost for all candidate nodes in S_cand_k, select the candidate node N_cand_optimal that minimizes TC(N_cand) as the final virtual combiner point N_vcp_k for the electrical aggregation sub-region C_k. Physically, this point represents the construction of a combiner box within this sub-region, minimizing the total laying cost of branch cables connecting all strings in the region to this combiner box, considering terrain and construction constraints.

[0058] As an alternative, the method for determining the virtual combiner point in step S2022 can incorporate power weighting. In this case, the formula for calculating the total cost is modified to: TC_weighted(N_cand) = Σ_{i=1 to m} [P_i *Dijkstra(N_cand, N_pv_i)]. Here, P_i is the rated power of the photovoltaic string node N_pv_i. This approach ensures that the calculated total cost reflects not only the construction cost of the route but also indirectly the current-carrying capacity of the cable. The result tends to locate the virtual combiner point closer to the high-power string, which helps reduce voltage drop and power loss caused by long-distance heavy current transmission in subsequent cable selection.

[0059] As a computationally simpler alternative, the virtual merge point can be determined based on its geometric centroid. First, the arithmetic mean of the geographic coordinates of all photovoltaic string nodes within the sub-region is calculated to obtain the coordinates of a geometric centroid (X_c, Y_c). Then, in the weighted network graph model, the node whose geographic coordinates are closest to this geometric centroid in the Euclidean distance is found and directly designated as the virtual merge point. This approach significantly reduces computational complexity because it avoids extensive Dijkstra path searches. However, its drawback is that it completely ignores the heterogeneous costs introduced by factors such as terrain and restricted areas. The selected point may be geometrically central, but not optimal in terms of actual path cost; for example, the point might be located on a small steep slope or in a difficult construction area.

[0060] Step S203: Generate a virtual combiner point dataset. After determining the virtual combiner points for all sub-regions, the system generates a structured virtual combiner point dataset for subsequent trunk cable optimization steps. This dataset is a list where each element represents a virtual combiner point, and its data structure can be defined as: {VCP_ID: integer, Sub_Region_ID: integer, Coordinate: {X_proj: float, Y_proj: float}, Total_Power_kW: float, Member_PV_String_IDs: list[string]}. Here, VCP_ID is the unique identifier of the virtual combiner point, Sub_Region_ID is the identifier of its sub-region, Coordinate is its coordinates in the projected coordinate system, Total_Power_kW is the sum of the rated power of all photovoltaic strings within its sub-region, and Member_PV_String_IDs is a list of IDs for all photovoltaic strings within that sub-region. This dataset constitutes a crucial intermediate layer connecting the photovoltaic strings (branch cables) and the main substation (trunk cable).

[0061] Further, step S3 is described in detail. This step utilizes the aforementioned virtual merge point dataset to initiate an evolutionary optimization process for the trunk cable topology. The core of this process is the application of a genetic algorithm, treating the trunk cable topology as individuals within a biological population, and iteratively searching for the optimal solution by simulating the mechanisms of natural selection and genetic variation. (Refer to...) Figure 5 The diagram details the internal loop process of trunk cable path evolution optimization.

[0062] Step S301: Perform individual coding and population initialization.

[0063] Step S3011: Select and implement an individual coding scheme. In a specific embodiment, a Prüfer sequence-based coding method is used to represent the topology tree structure of the trunk cable. A tree structure containing N nodes can be uniquely represented by a sequence of length N-2. In this scenario, the node set includes all M virtual junction points determined in step S2 and one main substation, for a total of N = M+1 nodes. The Prüfer sequence generation process is as follows: repeatedly find the node with degree 1 (i.e., leaf node) and the smallest label in the tree, record the label of the node connected to this leaf node in the Prüfer sequence, and then remove the leaf node and its connected edge from the tree. This process is repeated N-2 times until only two nodes and one edge remain in the tree. In this way, any complex topology tree connecting all virtual junction points and the main substation is transformed into a fixed-length linear integer sequence composed of node labels, greatly simplifying the design and implementation of crossover and mutation operators in the subsequent genetic algorithm.

[0064] Step S3012: Perform population initialization. Set a population size P_size, for example, P_size = 200. To generate each individual in the initial population, the system generates a Pruffer sequence of length N-2. Each position in the sequence is filled by randomly and repeatedly selecting a label from the set {1, 2, ..., N} of N nodes (M virtual junctions and 1 main substation). By generating P_size such completely randomized Pruffer sequences, the initial population can be formed. This randomized initialization method aims to ensure the diversity of the initial population, allowing the search to start from multiple different starting points widely distributed in the solution space, avoiding the algorithm from getting trapped in local optima too early. Each Pruffer sequence exists as an individual in the data structure, awaiting subsequent decoding and fitness evaluation.

[0065] Step S302 involves performing iterative calculations coupling electrical and physical characteristics with economic efficiency, i.e., evaluating the fitness of individuals. This step is the core of the entire optimization method, as it links an abstract topological structure (represented by a Plüffer sequence) with specific engineering and economic indicators. For each individual in the population, the following series of calculations must be performed to determine its fitness evaluation value.

[0066] Step S310, Individual Decoding. First, the Pluff sequence of an individual is decoded to restore the specific topological tree structure. The decoding process is the inverse operation of encoding; using a specific algorithm, the corresponding tree connection relationship can be uniquely reconstructed from a Pluff sequence. The result after decoding is an adjacency list or adjacency matrix, clearly describing which two nodes (virtual junction points or main substations) have a trunk cable connection.

[0067] Step S311: Perform segment-by-segment power flow calculation. For the topology tree structure decoded by the current individual, perform a power aggregation calculation from the end to the root node (main substation). Specifically, the main substation is the root of the tree, and all virtual merging points are the nodes of the tree. For each leaf node with a degree of 1 (i.e., the end virtual merging point), the power flowing out is equal to the total power of its sub-region, which can be directly obtained from the virtual merging point dataset generated in step S203. Then, traverse the topology path upwards towards the main substation level by level. For any virtual merging point that is not a leaf node, the power flowing through the cable segment it connects to upwards is equal to the sum of the total power of its own sub-region and the power of all downstream branches connected to the root node through it. Repeat this process until the total apparent power S (in kVA) that needs to be transmitted for each connection (i.e., each edge) in the topology tree is calculated. Based on the apparent power S, the system rated voltage V (e.g., 10kV) and the preset power factor cos(φ) (e.g., 0.98), the root mean square current value I that each section of the main cable needs to carry is accurately calculated. The calculation formula is I = S / (sqrt(3) * V * cos(φ)).

[0068] Step S312: Perform segmented cable selection. Based on the root mean square current value I calculated for each segment of the line in step S311, query and select the optimal cable specification from a pre-set cable standard database. This database is a structured data table that records detailed parameters of various standard cable models, including: cable model (e.g., YJV22-10kV-3x50), nominal cross-sectional area (mm²), resistance per unit length (Ω / km, at a specific operating temperature), reactance per unit length (Ω / km), rated current carrying capacity (A, long-term allowable current carrying capacity under specific laying conditions), and material cost per unit length (yuan / meter). For each cable segment, the selection logic is as follows: First, filter out all cable models in the database whose rated current carrying capacity is greater than or equal to the calculated current I for that segment; then, among the candidate models that meet the current carrying capacity requirement, select the one with the smallest nominal cross-sectional area. This strategy ensures that the selected cable has the lowest initial material cost while meeting the requirements for safe operation (without overheating).

[0069] Step S313: Calculate the total lifecycle energy loss cost. Based on the electrical parameters and line characteristics of the selected cable, quantify the economic cost corresponding to the energy loss caused by resistive heating during the entire design life of the power station (e.g., 25 years). For each cable segment in the topology tree, its annual average active power loss P_loss (in kW) is calculated using Joule's law, with the formula P_loss = 3 * I² * R. Where I is the root mean square current value of the cable segment, and R is the total resistance of the cable segment. The total resistance R is obtained by multiplying the unit length resistance value obtained from the database by the actual laying length of the line segment. The line length is not the straight-line distance between two points, but rather the path length corresponding to the minimum path cost connecting two nodes (e.g., virtual merge point A and virtual merge point B) calculated by running Dijkstra's algorithm on the weighted network graph model generated in step S103. The average annual energy loss (in kWh) equals P_loss multiplied by the equivalent full-load operating hours per year (e.g., 1300 hours based on local solar resources). Multiplying the average annual energy loss by the local grid-connected electricity price (in yuan / kWh) yields the annual energy loss cost. Finally, using the Net Present Value (NPV) formula, the annual energy loss cost for the next 25 years is discounted to the total present value at the beginning of the project, using a predetermined discount rate (e.g., 8%). This discounted total is the total life-cycle energy loss cost (Cost_loss) for this section of the line.

[0070] Step S314: Calculate the initial investment cost. This cost mainly consists of two parts: material cost and construction cost. For each cable segment in the topology tree, its material cost, Cost_material, is obtained by multiplying the unit length cost of the cable type selected in step S312 by the actual laying length of the segment calculated in step S313. Its construction cost, Cost_construction, is obtained by accumulating the final passage cost value, C_final, of all grid cells traversed by the minimum cost path of the cable segment. Specifically, the path recorded by the Dijkstra algorithm during pathfinding is a series of continuous grid cells. The C_final values ​​of these grid cells (which incorporate multiple factors such as terrain, restricted areas, and forced passages) are summed, and then multiplied by a calibration coefficient (unit: yuan / cost unit) to convert the dimensionless cost value into an actual monetary amount, thus obtaining the construction cost of the segment. The sum of the two, Cost_initial = Cost_material + Cost_construction, constitutes the initial investment cost of the segment.

[0071] Step S315: Calculate the total cost and fitness score of an individual. The initial investment cost (Cost_initial) and the total lifecycle energy loss cost (Cost_loss) of all segments in the complete topology tree represented by the current individual are summed to obtain the total lifecycle cost (TC_LCC) of that individual (i.e., the line scheme). TC_LCC = Σ(Cost_initial) + Σ(Cost_loss). To unify the optimization problem into a maximization problem, the fitness score of an individual is defined as the reciprocal of the total cost, i.e., Fitness = 1 / TC_LCC. Therefore, a line scheme with a lower total lifecycle cost will obtain a higher fitness score, thus having a higher probability of survival and reproduction in subsequent genetic operations.

[0072] Next, step S4 will be described in detail. This step executes the core operation of the genetic algorithm based on the fitness evaluation values ​​calculated for all individuals in the population, in order to generate a better offspring population.

[0073] Step S401: Perform the selection operation. The selection operation simulates the natural law of "survival of the fittest." In a specific embodiment, Roulette Wheel Selection is used. The specific implementation of this method is as follows: First, calculate the total fitness of all individuals in the entire population: Fitness_total = Σ(Fitness_i). Then, calculate the proportion of each individual i's fitness value to the total fitness: p_i = Fitness_i / Fitness_total. This proportion is the probability that the individual will be selected. Next, generate a random number r from 0 to 1. Starting with the first individual, the probability of selection is accumulated sequentially, forming a cumulative probability distribution. The first individual whose cumulative probability is greater than or equal to the random number r is selected and enters the mating pool. This process is repeated P_size times to generate a parent population of the same size as the original population, used for crossover operations. Individuals with higher fitness occupy a larger sector area on the roulette wheel, and have a greater chance of being selected by the pointer (random number).

[0074] Step S402: Perform the crossover operation. Crossover simulates gene recombination in organisms and is a primary pathway for generating new individuals. This embodiment employs a single-point crossover method based on the Pruffer sequence. Specifically, two individuals are randomly selected as parents from the mating pool generated in step S401. A relatively high crossover probability p_c (e.g., 0.85) is set. A random number is generated; if it is less than p_c, crossover is performed on these two parents. A random integer within the range of the Pruffer sequence length (N-2) is selected as the crossover point. The Pruffer sequences of the two parent individuals are cut at the crossover point, and the latter half is exchanged. For example, if parent A's sequence is [a1, a2, ..., ak | ak+1, ..., aN-2] and parent B's sequence is [b1, b2, ..., bk | bk+1, ..., bN-2], the two child sequences generated after crossover are [a1, a2, ..., ak | bk+1, ..., bN-2] and [b1, b2, ..., bk | ak+1, ..., aN-2]. These two newly generated Pluffer sequences represent two new child individuals that inherit some topological characteristics from their parents. If the generated random number is greater than p_c, crossover is not performed, and the two parent individuals are directly copied to the next generation.

[0075] Step S403: Perform the mutation operation. The mutation operation introduces randomness into the algorithm, helping to escape local optima and maintain population diversity. Specifically, for each individual in the new population generated after crossover, a low mutation probability p_m (e.g., 0.05) is assigned. A random number is generated; if it is less than p_m, mutation is performed on that individual. The mutation method involves randomly selecting a position in its Plüffer sequence and randomly replacing the original node label at that position with another different label from the node label set {1, 2, ..., N}. This small change will cause a change in a local connection relationship in the topology tree structure after decoding. For example, a virtual merge point originally connected to node X may be connected to node Y after mutation.

[0076] Step S404: Generate a new generation of population and iterate. The new set of individuals generated after selection, crossover, and mutation operations constitutes the new generation of population. To prevent the loss of the currently found optimal solution during iteration, an elitism strategy can be introduced. That is, before each generation's selection operation, the top 1% or 2% of individuals with the highest fitness in the current population are directly copied to the next generation population without participating in crossover and mutation. Subsequently, steps S3 (decoding and fitness evaluation of each individual in the new population) and S4 (selection, crossover, and mutation) are repeated to form a closed-loop evolutionary process.

[0077] Step S405: Determine the termination condition and output the optimal solution. This iterative process continues until one of the preset termination conditions is met. Condition 1: Convergence. The algorithm monitors the changes in the average fitness value or the optimal fitness value of the population in real time. If, within a continuous G_conv generation (e.g., 30 generations), the growth rate of the average fitness value of the population is lower than a very small threshold ε (e.g., 0.01%), the algorithm is considered to have converged, and the optimal solution has basically stabilized. Condition 2: Reaching the maximum number of iterations. To prevent the algorithm from running indefinitely, a maximum number of iterations G_max (e.g., 500 generations) is set. When the number of iterations reaches this limit, the algorithm is forcibly terminated. When any termination condition is triggered, the algorithm selects and outputs the individual with the highest fitness value from the final generation of the population.

[0078] Step S406: Generate the final solution. The topology tree structure obtained after decoding the Pluff sequence corresponding to the individual with the best fitness is the optimal trunk cable convergence route determined by this invention. The system will generate a detailed engineering output file based on this optimal topology structure. Specifically, for each edge in the optimal topology (i.e., each trunk cable segment), the Dijkstra algorithm is called again to generate its specific laying path defined by a series of geographic coordinate points on the weighted network graph model. By integrating the paths of all segments, a cable laying path map (e.g., DXF or KML format) that can be directly imported into Geographic Information System (GIS) or Computer-Aided Design (CAD) software can be generated. At the same time, based on the cable selection, length, cost, and other information calculated for the optimal individual during the last fitness evaluation, a detailed material specification list is automatically generated. The list is presented in tabular form, listing key data for each cable segment, including the start and end points, cable type, cross-sectional area, length, unit price of materials, total price of materials, construction cost, average annual power loss, and total life-cycle loss cost, providing accurate data support for project procurement, budgeting, and construction.

[0079] A device for planning and optimizing cable collection lines in a narrow photovoltaic field, the device comprising a memory for storing information about the system and a central processing unit for performing calculations on the system.

[0080] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for planning and optimizing a cable collection line based on a long and narrow photovoltaic field area, characterized by, Includes the following steps: S1: Obtain geospatial information data of the photovoltaic field area, including the coordinates of photovoltaic strings, the location of the main substation, the vector data of restricted areas and mandatory access corridors, generate a weighted raster cost map that integrates terrain construction costs, and convert it into a non-Euclidean weighted network graph model. S2: Based on the weighted network graph model, a density clustering algorithm that takes into account the power generation weight is applied to divide the photovoltaic string into several electrical collection sub-regions, and the virtual collector point with the best cost is determined in each sub-region to generate a virtual collector point dataset; S3: Initialize the trunk cable topology tree population that connects all virtual junction points to the main substation. For the topology tree structure represented by each individual in the population, perform coupled calculations of electrical and physical characteristics and economic efficiency. Generate individual fitness evaluation values ​​through a fitness function that aims to minimize the total life cycle cost. S4: Based on the individual fitness evaluation value, perform selection, crossover and mutation operations on the population to generate a new generation population. After the population converges, select and output the individual with the best fitness evaluation value to obtain the optimal trunk cable collection line.

2. The method for planning and optimizing cable collection lines in a narrow photovoltaic field according to claim 1, characterized in that, The weighted rasterized cost map includes raster cell access cost value, raster two-dimensional matrix index, basic construction cost coefficient, and terrain slope correction coefficient. The weighted network graph model includes network node set, network edge set, and edge weight value. The virtual merge point dataset includes virtual merge point number, sub-region identifier, coordinates, and total power under its jurisdiction. The optimal trunk cable convergence line includes segmented line connection relationship, node coordinates, and cable specifications.

3. The method for planning and optimizing cable collection lines in a narrow photovoltaic field according to claim 1, characterized in that, The specific steps for generating the weighted network graph model are as follows: S111: Obtain the vector data of the photovoltaic string coordinates, main substation location, restricted area and forced passage corridor, divide the site into grid units of predetermined size, and assign a basic construction cost coefficient to each grid unit according to the vector data to generate a basic cost grid map. S112: Based on the basic cost raster map, the terrain slope of each raster unit is calculated by combining digital elevation model data, and a terrain slope correction coefficient is generated according to the preset slope and construction difficulty mapping relationship. The basic construction cost coefficient is then corrected to obtain the final weighted raster cost map. S113: Based on the final weighted rasterized cost map, the center point of each raster cell is abstracted as a network node, and the nodes in the eight adjacent directions are connected to generate network edges. The product of the average passage cost across two raster cells and the geometric distance is used as the edge weight to generate the weighted network graph model.

4. The method for planning and optimizing cable collection lines in a narrow photovoltaic field according to claim 3, characterized in that, The specific steps for generating the virtual confluence point dataset are as follows: S211: Based on the weighted network graph model, an improved DBSCAN clustering algorithm with path cost as the distance metric and total regional power as the density metric is used to divide all photovoltaic string nodes and assign noise points to form several electrical aggregation sub-regions. S212: Call the electrical aggregation sub-region, traverse the candidate network nodes in each sub-region, calculate the sum of the minimum path costs from each candidate node to all photovoltaic string nodes in the region, select the node with the minimum cost as the virtual merge point of the region, and integrate to generate the virtual merge point dataset.

5. The method for planning and optimizing cable collection lines in a narrow photovoltaic field according to claim 4, characterized in that, The specific steps for generating the individual fitness evaluation value are as follows: S311: For the current individual representative's topology tree structure, perform power flow calculations segment by segment from the end virtual junction point toward the main substation, determine the root mean square current value, and query and select the cable specification that meets the current carrying capacity and has the smallest cross-sectional area from the cable standard database to generate a segmented cable configuration scheme. S312: Based on the segmented cable configuration scheme, calculate the total life-cycle power loss cost and initial investment cost of each segment, sum the two costs of all segments to obtain the total life-cycle cost, and use the reciprocal of the total cost as the individual fitness evaluation value.

6. The method for planning and optimizing cable collection lines in a narrow photovoltaic field according to claim 5, characterized in that, The specific steps for obtaining the optimal trunk cable collection line are as follows: S411: Based on the individual fitness evaluation value, the roulette wheel selection method is used to perform a selection operation on the population, select individuals with high fitness, and obtain the parent topology set; S412: Call the parent topology set, perform a crossover operation on the Pluft sequence encoding of the parent individuals, and generate the child topology set; S413: Based on the offspring topology set, perform mutation operations on the Plüffer sequence encoding of individuals with a preset probability to form a new generation population, and after the average fitness evaluation value of the population converges to a preset stable threshold, output the optimal trunk cable convergence line corresponding to the individual with the best fitness.

7. The method for planning and optimizing cable collection lines in a narrow photovoltaic field according to claim 1, characterized in that, The method also includes step S5: S5: Call the optimal trunk cable collection line, plan a specific laying path for each trunk cable segment on the weighted network diagram model, and integrate the cable specifications and path length to generate cable laying scheme details; The cable laying plan details include the segmented cable path coordinate sequence, cable model and specifications, laying length, and estimates of itemized and total investment costs.

8. The method for planning and optimizing cable collection lines in a narrow photovoltaic field according to claim 7, characterized in that, The specific steps for generating the cable laying plan details are as follows: S511: Call the start and end nodes of each segment in the optimal trunk cable collection line, and apply the A* algorithm or Dijkstra algorithm to search for the minimum cost path between two nodes on the weighted network graph model to obtain the coordinate set of the specific laying path. S512: Based on the coordinate set of the specific laying path, calculate the precise laying length of each segment, and combine it with the cable specifications determined in step S3 to form a segment material specification list. S513: Based on the segmented material specification list, calculate the material cost of each segment, and add up the construction cost according to the grid along the route. Finally, integrate all the data to generate the cable laying plan details.

9. A planning and optimization system for cable collection lines in a narrow photovoltaic field, characterized in that, The system is used to implement the planning and optimization method for cable collection lines in narrow photovoltaic fields as described in any one of claims 1-8, and the system includes: The geographic information modeling module acquires geospatial information data of the photovoltaic field area, generates a weighted rasterized cost map that integrates terrain and regional constraints, and converts it into a non-Euclidean weighted network graph model with weighted travel cost as the edge weight. The string clustering and partitioning module, based on the weighted network graph model, executes a power-aware density clustering algorithm to divide the photovoltaic strings into electrical collection sub-regions, and locates the virtual collector point that minimizes the internal connection cost in each region, generating a virtual collector point dataset; The topology fitness evaluation module, based on the virtual merge point dataset, performs power flow calculation, cable selection, and coupled calculation of power loss cost and initial investment cost for each topology tree individual in the genetic algorithm population, and outputs an individual fitness evaluation value based on the total life cycle cost. The population iterative evolution module, based on the individual fitness evaluation value, guides the selection of the dominant genetic algorithm, performs crossover and mutation operations based on the Pluffer sequence, continuously optimizes the population until convergence, and selects and outputs the optimal trunk cable collection line corresponding to the individual with the highest fitness. The engineering scheme generation module, based on the optimal trunk cable convergence line, calls the path planning algorithm to generate specific cable laying paths on the weighted network graph model, and outputs a detailed bill of materials containing the cable type, length and cost of each segment, forming the final cable laying scheme details.

10. An apparatus for using the cable collection line planning and optimization system for a narrow photovoltaic field as described in claim 9, the apparatus comprising a memory for storing the system and a central processing unit for performing operations on the system.

Images