Method for determining position of box-type transformer of sand hill photovoltaic power station

Abstract

The invention discloses a method for determining the position of a box-type transformer of a sand hill photovoltaic power station, and the method comprises the steps: extracting the topographic features of a sand hill through a graph embedding model, converting the topographic features into a high-dimensional vector, and dividing a photovoltaic module layout point matrix in combination with a plane point matrix model layout algorithm. Based on parameters of rated capacity, power supply radius and the like of the box-type transformer, a layout position is determined through the steps of screening effective point locations, dividing power supply partitions, constructing an electrical performance evaluation matrix, optimally screening an optimal candidate point location, performing final verification and the like. According to the method, topographic features and equipment parameters are fused, through multi-step system optimization, the limitation that the prior art depends on experience and single index judgment is overcome, the position determination accuracy and reasonability are improved, the method adapts to the complex sand mountain environment, losses and risks are reduced, and the operation efficiency of a power station is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of transformer position determination, and in particular to a method for determining the position of a box-type transformer of a sand mountain photovoltaic power station. BACKGROUND

[0002] When a photovoltaic power station is constructed in a sand mountain area, the position of the box-type transformer directly affects the overall operation efficiency and stability of the power station. The sand mountain terrain has characteristics such as large elevation change, steep slope, and complex geological structure, and the layout of photovoltaic components needs to adapt to the terrain features to maximize the utilization of light energy, which makes the installation position of the box-type transformer not only need to meet the electrical performance requirements, but also need to consider the terrain stability, transportation convenience and reasonable power supply radius. With the expansion of the construction scale of the photovoltaic power station in the sand mountain area, how to accurately determine the position of the box-type transformer under complex terrain conditions has become a key technical problem to improve the construction quality and operation efficiency of the power station, and has important practical significance to reduce construction cost, reduce energy loss and ensure equipment safety.

[0003] The prior art has obvious deficiencies in determining the position of the box-type transformer of the sand mountain photovoltaic power station. On the one hand, the existing methods mostly rely on manual experience or simple geometric calculation for position planning, and fail to fully combine the spatial features of the sand mountain terrain and the electrical parameters of the box-type transformer, resulting in problems such as mismatch between the power supply radius and the load, and unreasonable distance from the photovoltaic component cluster, increasing line loss and equipment operation burden. On the other hand, the existing technology lacks systematic analysis of terrain stability and equipment adaptability, and only uses a single index to judge the feasibility of the position, ignoring the influence of the slope and geological conditions of the sand mountain area on the installation and long-term operation of the equipment, which is easy to cause foundation damage due to terrain changes after the transformer is installed, or difficulty in later maintenance due to poor transportation channels. SUMMARY

[0004] In order to overcome the shortcomings and deficiencies of the prior art, the present application provides a method for determining the position of a box-type transformer of a sand mountain photovoltaic power station.

[0005] The technical solution adopted by the present application is a method for determining the position of a box-type transformer of a sand mountain photovoltaic power station, comprising:

[0006] Step S1: feature extraction of sand mountain photovoltaic power station field terrain data is performed through a graph embedding model, terrain elevation, slope and slope spatial features are converted into high-dimensional vector representation, a terrain feature atlas containing photovoltaic component layout area is constructed, a plane point matrix model layout algorithm is used to gridize the potential layout point positions of the photovoltaic components in the atlas, and an initial point matrix is formed;

[0007] Step S2: Based on the rated capacity, voltage level, and no-load loss parameters of the box-type transformer, and combined with the estimated installed capacity of the photovoltaic module in the initial point matrix, the correlation weight of different points and the surrounding terrain features is calculated through the graph embedding model, the initial point matrix is screened, and the effective points that meet the capacity matching conditions are retained;

[0008] Step S3: The effective points are analyzed by the plane point matrix model arrangement algorithm, and according to the maximum power supply radius parameter of the box-type transformer, the clustering results are divided into several power supply partitions, and several candidate box transformer layout points are generated in each partition;

[0009] Step S4: The electrical connection feature vector of different candidate points and effective points of photovoltaic modules in the partition is calculated through the graph embedding model, and the electrical performance evaluation matrix of the candidate points is constructed combined with the short-circuit impedance and load loss parameters of the box-type transformer;

[0010] Step S5: The electrical performance evaluation matrix is optimized and solved by using the plane point matrix model arrangement algorithm, and considering the stability parameters of the sand mountain terrain, the box transformer candidate points with the optimal electrical performance in different partitions are screened out;

[0011] Step S6: Based on the traffic accessibility feature map generated by the graph embedding model, combined with the transportation weight and installation size parameters of the box-type transformer, the optimal candidate points in different partitions are finally verified, and the layout position of the box-type transformer is determined.

[0012] Further, in step S2, the graph embedding model calculates the correlation weight by the following formula:

[0013]

[0014] Wherein, W i，j is the correlation weight of the ith point and the jth terrain feature point, f(v i ) is the feature vector of the ith point output by the graph embedding model, f(v j ) is the feature vector of the jth terrain feature point, n is the total number of terrain feature points, S i is the estimated installed capacity of the photovoltaic module corresponding to the ith point, S rated is the rated capacity parameter of the box-type transformer.

[0015] Further, in step S3, when the plane point matrix model arrangement algorithm performs clustering analysis, the following formula is used to divide the power supply partition:

[0016]

[0017] Wherein, C mFor the mth power supply partition, p is the effective point of the photovoltaic module, P is the effective point set, dist(p, c m ) is the distance from point p to the center of the partition c m , R max is the maximum power supply radius parameter of the box transformer, S p is the photovoltaic module capacity of point p, S rated is the rated capacity parameter of the box transformer.

[0018] Further, in step S4, when constructing the electrical performance evaluation matrix, the evaluation value is calculated by the following formula:

[0019]

[0020] Wherein, E k is the electrical performance evaluation value of the kth candidate point, Z k is the box transformer short-circuit impedance calculation value corresponding to the point, Z rated is the rated short-circuit impedance parameter of the box transformer, P k is the box transformer load loss calculation value corresponding to the point, P rated is the rated load loss parameter, and α, β, γ are weight coefficients, f elec (p, k) is the length of the electrical connection feature vector of point p and candidate point k output by the graph embedding model.

[0021] Further, in step S5, when the planar point matrix model arrangement algorithm is optimized and solved, the following formula is used to screen the optimal candidate point:

[0022]

[0023] Wherein, O k is the comprehensive optimization value of the kth candidate point, E k is the electrical performance evaluation value, δ is the terrain stability influence coefficient, S slope，k is the slope parameter at the candidate point k, which is extracted from the terrain feature map by the graph embedding model.

[0024] Further, in step S6, the traffic accessibility feature map generated by the graph embedding model is finally verified by the following formula:

[0025]

[0026] Wherein, T k is the traffic accessibility verification value of the kth candidate point, f traff (k) is the traffic feature vector of the candidate point k, f std is the standard traffic accessibility feature vector, sim() is the vector similarity calculation function, and Wrated W is a rated transportation weight parameter of the box-type transformer k L is an actual transportation weight limit value at the candidate point k rated L is an installation size parameter of the box-type transformer k L is an installation space size at the candidate point k

[0027] Further, the step S3 comprises the following sub-steps:

[0028] S3.1: importing the coordinate data of the effective point into a plane point matrix model arrangement algorithm, setting the neighborhood search range of the matrix according to the maximum power supply radius parameter of the box-type transformer, and generating the local neighborhood matrix of different points; S3.2: calculating the spatial correlation characteristics of different points in the local neighborhood matrix through a graph embedding model, and determining the initial clustering center in combination with the power supply efficiency parameter of the box-type transformer; S3.3: iteratively clustering the effective points based on the initial center, and adjusting the clustering center position according to the load rate parameter of the box-type transformer after each iteration until the clustering result converges; S3.4: dividing the boundary of the converged clustering result to ensure that the total capacity of the photovoltaic components in each partition matches the rated capacity parameter of the box-type transformer, and forming a power supply partition.

[0029] Further, the step S4 comprises the following sub-steps:

[0030] S4.1: collecting three-dimensional coordinate data of different candidate points, and constructing a spatial coordinate system of the candidate points through a graph embedding model in combination with the installation height parameter of the box-type transformer; S4.2: mapping the coordinates of the effective point of the photovoltaic components in the partition to the coordinate system, calculating the electrical distance between different effective points and candidate points, and generating a preliminary connection matrix in combination with the impedance parameter of the box-type transformer; S4.3: performing sparse processing on the preliminary connection matrix through a plane point matrix model arrangement algorithm, and retaining the connection path that meets the maximum current-carrying capacity parameter of the box-type transformer; S4.4: constructing an electrical performance evaluation matrix containing path loss and voltage drop indicators based on the sparse connection path in combination with the loss parameter of the box-type transformer.

[0031] Further, the step S5 comprises the following sub-steps:

[0032] S5.1: eigenvalue decomposition is performed on the electrical performance evaluation matrix by using a planar point location matrix model arrangement algorithm to extract principal component features representing electrical performance; S5.2: the stability parameters of the sand mountain terrain are converted into feature vectors by a graph embedding model, and are fused with the principal component features to form a comprehensive feature matrix; S5.3: the comprehensive feature matrix is sorted, and the top 30% of candidate points in the comprehensive feature values are selected as key optimization objects; S5.4: the key optimization objects are secondarily evaluated in combination with the operating environment temperature parameters of the box-type transformer to determine the candidate point locations of the box-type transformer with optimal electrical performance in different partitions.

[0033] A method for determining the location of a box-type transformer of a sand mountain photovoltaic power station, which is implemented by different units, comprising:

[0034] A terrain feature atlas construction and point location matrix generation unit, which is connected to a photovoltaic module parameter acquisition sensor device, receives capacity parameters of the photovoltaic module, and outputs an initial point location matrix;

[0035] An effective point location screening unit, which is connected to the terrain feature atlas construction and point location matrix generation unit and a box-type transformer parameter database, respectively, receives the initial point location matrix and the rated capacity parameters of the box-type transformer, and outputs an effective point location set;

[0036] A power supply partition division unit, which is connected to the effective point location screening unit and a planar point location calculation module, receives the effective point location set, and outputs a power supply partition result;

[0037] An electrical performance evaluation matrix construction unit, which is connected to the power supply partition division unit and the box-type transformer parameter database, respectively, receives the power supply partition result and electrical parameters of the box-type transformer, and outputs an electrical performance evaluation matrix;

[0038] An optimal candidate point location screening unit, which is connected to the electrical performance evaluation matrix construction unit and a terrain stability analysis module, receives the electrical performance evaluation matrix and terrain parameters, and outputs an optimal candidate point location;

[0039] A final location determination unit, which is connected to the optimal candidate point location screening unit, a traffic accessibility analysis module, and a box-type transformer parameter database, respectively, receives the optimal candidate point location, traffic data, and physical parameters of the box-type transformer, and outputs a box-type transformer layout location.

[0040] Beneficial Effects: This invention proposes a method for determining the location of a box-type transformer in a sand dune photovoltaic power station. By deeply extracting the topographic features of the sand dune using a graph embedding model and combining a planar point matrix model layout algorithm with various parameters of the box-type transformer, the method achieves precise and systematic location determination. Addressing the problem of existing technologies relying on manual experience or simple calculations, this method, through multi-step model calculations and parameter fusion, fully correlates topographic spatial features with equipment electrical parameters, ensuring that the determined location matches the power supply radius and load, reducing line losses and equipment burden. Simultaneously, this method comprehensively considers factors such as terrain stability and accessibility, avoiding the limitations of single-indicator judgments through multi-dimensional verification and optimization. This ensures that the transformer installation location adapts to the sand dune slope and geological conditions, reducing the risk of foundation damage caused by terrain changes, and facilitating transportation and subsequent maintenance, thereby improving the construction quality and operational efficiency of the power station. Attached Figure Description

[0041] Figure 1 This is a flowchart of the method steps of the present invention;

[0042] Figure 2 This is a diagram showing the unit composition for implementing the method of the present invention. Detailed Implementation

[0043] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] like Figure 1 As shown, a method for determining the location of a box-type transformer in a sand dune photovoltaic power station includes:

[0045] Step S1: Extract features from the terrain data of the Shanshan photovoltaic power station site using a graph embedding model, transform the spatial features of terrain elevation, slope, and aspect into high-dimensional vector representations, construct a terrain feature map containing the photovoltaic module deployment area, and use a planar point matrix model layout algorithm to grid the potential photovoltaic module deployment points in the map to form an initial point matrix.

[0046] Specifically, step S1 converts the terrain data of the sand mountain photovoltaic power station site area into structured information that can be used for subsequent analysis, and preliminarily delimits the potential layout range of the photovoltaic module. This step extracts features from the terrain data through a graph embedding model, which can capture the potential associations between spatial features such as terrain elevation, slope, and slope direction. After converting these complex spatial information into high-dimensional vectors, it is convenient for computers to perform quantitative processing and analysis. The constructed terrain feature graph contains key terrain information of the photovoltaic module layout area, providing accurate geographic reference for subsequent point division. The use of a planar point matrix model layout algorithm for grid division can ensure that the distribution of potential layout points is both comprehensive and covers the available area, without increasing the complexity of subsequent calculations due to excessive density. The formation of the initial point matrix lays the data foundation for subsequent screening and optimization.

[0047] In the specific implementation process, first, high-precision terrain data of the sand mountain photovoltaic power station site area is collected, including elevation data every 5 meters, slope data (range between 0°-60°), and slope direction data (with the north direction as 0°, increasing clockwise to 360°). These data are input into the graph embedding model, which converts terrain features into 128-dimensional high-dimensional vectors through multi-layer neural network operations, with each vector corresponding to a 5m x 5m terrain grid cell. Based on these vectors, a terrain feature graph is constructed, with each node in the graph representing a grid cell, and the edge weights between nodes reflecting the similarity of terrain features. Subsequently, the planar point matrix model layout algorithm is started, which marks the area where photovoltaic modules can be laid out in the graph (excluding areas with slope greater than 30° and elevation change greater than 10 meters), and divides the area into 10m x 10m grids, with each grid center point as a potential layout point, forming an initial point matrix containing 5000-8000 points. Each element in the matrix records the coordinates, elevation, slope, and slope direction of the corresponding point.

[0048] Step S2: Based on the rated capacity, voltage level, and no-load loss parameters of the box-type transformer, combined with the estimated installed capacity of the photovoltaic module in the initial point matrix, the graph embedding model calculates the correlation weight of different points with the surrounding terrain features, and filters the initial point matrix to retain valid points that meet the capacity matching conditions.

[0049] Specifically, step S2 is to screen out effective points that meet the capacity matching condition of the box-type transformer from the initial point matrix. This step is directly related to the rationality of the subsequent power supply partition and the operation efficiency of the box-type transformer. Based on the rated capacity, voltage level, no-load loss and other parameters of the box-type transformer, the selected point can ensure that the installed capacity of the photovoltaic module corresponding to the point is matched with the carrying capacity of the box-type transformer, avoiding the situation of capacity overload or capacity waste. By calculating the correlation weight of each point and the surrounding terrain features through the graph embedding model, the influence of terrain factors on the power generation efficiency of photovoltaic modules and the operation of the box-type transformer can be further considered, so that the screening result is more scientific and reasonable. The remaining effective points will be used as the basis for subsequent power supply partition division, providing protection for improving the energy conversion efficiency of the entire photovoltaic power station and the service life of the equipment.

[0050] In specific implementation, first, the key parameters of the box-type transformer are determined, such as the rated capacity is set to 1000kVA, 2000kVA, and 3000kVA, the voltage level is 35kV / 0.4kV, and the no-load loss is between 500W-1200W. According to the illumination conditions and terrain slope of each point in the initial point matrix, the estimated capacity of the photovoltaic module that can be installed at each point (the single-point capacity range is between 50kW-200kW) is estimated. These estimated capacity data and the box-type transformer parameters are input into the graph embedding model, which calculates the correlation weight of each point and the terrain feature points within a 50-meter range. The weight value is between 0 and 1, and the higher the weight, the better the adaptability of the point to the surrounding terrain. The capacity matching condition is set as follows: the ratio of the estimated installed capacity of a single point to the rated capacity of the matched box-type transformer is between 0.6 and 1.2, and the correlation weight is not less than 0.5. Each point in the initial point matrix is checked one by one, and the points that do not meet the condition are removed. The number of effective points finally retained is usually 60%-70% of the initial point matrix, and an effective point list containing point coordinates, estimated capacity, correlation weight and other information is formed.

[0051] Step S3: using a planar point matrix model arrangement algorithm to perform cluster analysis on the effective points, and according to the maximum power supply radius parameter of the box-type transformer, the clustering results are divided into several power supply partitions, and several candidate box-type transformer layout points are generated in each partition;

[0052] Specifically, step S3 involves clustering effective locations and dividing them into power supply zones, while simultaneously generating candidate transformer substation locations. This step is crucial for achieving efficient power supply from the transformer substation. Using a planar location matrix model for clustering analysis, the dispersed effective locations can be aggregated into several relatively concentrated areas based on the spatial distribution characteristics and electrical connection requirements, facilitating centralized power supply from the transformer substation. Dividing the power supply zones based on the maximum power supply radius parameter of the transformer substation ensures that the photovoltaic modules within each zone are within the effective power supply range of the transformer substation, reducing energy loss during line transmission. The generation of several candidate transformer substation locations within each zone provides multiple alternatives for subsequent electrical performance evaluation, helping to select the optimal transformer substation location.

[0053] In the specific implementation process, the coordinate data of the effective points (accurate to 0.1 meters) are first imported into the planar point matrix model layout algorithm. Based on the maximum power supply radius parameter of the box-type transformer (set to 500 meters), the algorithm determines the neighborhood search range of each point as a circular area with a radius of 500 meters centered on that point. The algorithm calculates the spatial distance and electrical correlation between points, and uses a density clustering algorithm to cluster the effective points. During clustering, the minimum number of cluster points is set to 20, and the cluster radius is 100 meters. After clustering, based on the maximum power supply radius of the box-type transformer, each cluster result is divided into a power supply zone, ensuring that the distance from any point within a zone to the zone center does not exceed 500 meters. For each power supply zone, a candidate box-type transformer layout point is generated at a distance of 100 meters from the center in eight evenly distributed directions (spaced at 45° intervals) around its geometric center. Each zone typically generates nine candidate points. The candidate points must satisfy the following conditions: minimum average distance from all effective points within the zone, and ground slope not exceeding 15°.

[0054] Step S4: Calculate the electrical connection feature vectors of different candidate points and effective points of photovoltaic modules in the partition using the graph embedding model, and construct the electrical performance evaluation matrix of the candidate points by combining the short-circuit impedance and load loss parameters of the box transformer.

[0055] Specifically, the main task of step S4 is to construct an electrical performance evaluation matrix of the candidate point, which is an important basis for measuring the pros and cons of the electrical performance of the candidate point. By calculating the electrical connection feature vector of each candidate point and the effective point of the partition photovoltaic component through the graph embedding model, the electrical connection between the two can be fully reflected, including line length, voltage drop, power loss and other potential factors. Combined with the short-circuit impedance, load loss and other parameters of the box-type transformer, the electrical connection feature can be combined with the electrical performance of the device itself, making the evaluation matrix more targeted and practical. The electrical performance evaluation matrix constructed provides a quantitative index for subsequent optimization solution, which helps to select better candidate points from the perspective of electrical performance.

[0056] In specific implementation, first, the three-dimensional coordinates (X, Y, Z, Z is the altitude, the accuracy is 0.1 meters) of each candidate point and the three-dimensional coordinates of all effective points in the partition are collected. These coordinate data are input into the graph embedding model, which generates the electrical connection feature vector between each candidate point and the effective point by calculating the Euclidean distance between the points, the line laying difficulty coefficient and other parameters. The vector dimension is 20, including the normalized values of line resistance, reactance, capacitance and other parameters. At the same time, the short-circuit impedance parameter (range 4%-6%) and the load loss parameter (between 10kW-25kW, corresponding to the rated capacity) of the box-type transformer are obtained. According to the electrical connection feature vector and the box-type transformer parameters, the line loss, voltage drop and other indicators between each candidate point and each effective point are calculated, in which the line loss calculation needs to consider the wire cross-sectional area (120mm 2 -240mm 2 copper core cable) and transmission distance, and the voltage drop needs to be controlled within 5%. These indicators are integrated into a matrix, the rows of which represent the candidate points, the columns represent the electrical performance evaluation indicators (including total line loss, maximum voltage drop, short-circuit current, etc.), and each element is the specific value of the corresponding candidate point under the indicator, forming the electrical performance evaluation matrix.

[0057] Step S5: using the planar point matrix model arrangement algorithm to optimize and solve the electrical performance evaluation matrix, considering the stability parameters of the sand mountain terrain, and selecting the box transformer candidate point with the best electrical performance in different partitions;

[0058] Specifically, step S5 optimizes the electrical performance evaluation matrix to screen out the candidate point of the transformer substation with the optimal electrical performance in each subarea, which is the core step of improving the operation efficiency of the transformer substation. The optimization solution is obtained by using the plane point matrix model layout algorithm, which can comprehensively analyze the multi-dimensional indicators in the evaluation matrix and find out the candidate point that performs well in multiple indicators. By comprehensively considering the stability parameters of the sand mountain terrain, the influence of the terrain factor on the long-term operation of the transformer substation can be taken into account in the optimization process, so as to avoid equipment damage or performance decline caused by unstable terrain. The optimal candidate point screened out not only has good electrical performance, but also can adapt to the special environment of the sand mountain terrain, providing protection for the safe and stable operation of the transformer substation.

[0059] In the implementation process, first, the electrical performance evaluation matrix is input into the plane point matrix model layout algorithm. The algorithm adopts a multi-objective optimization strategy to perform weighted calculation on the line total loss, maximum voltage drop, short-circuit current and other indicators in the matrix. The weight coefficients are set according to the influence of each indicator on the system performance (line total loss weight 0.4, maximum voltage drop weight 0.3, short-circuit current weight 0.3). Through iterative calculation by the algorithm, the comprehensive electrical performance score of each candidate point is generated, and the score range is 0-100 points. The higher the score, the better the electrical performance. At the same time, the stability parameters of the sand mountain terrain are collected, including soil bearing capacity (not less than 150 kPa), seismic intensity (7-degree fortification), wind and sand erosion intensity (annual erosion depth not more than 50 mm), etc. These parameters are converted into terrain stability scores (0-100 points) through a graph embedding model. The comprehensive electrical performance score and the terrain stability score are weighted and summed in the ratio of 7:3 to obtain the comprehensive score of each candidate point. The candidate point with the highest comprehensive score is selected as the candidate point of the transformer substation with the optimal electrical performance in the subarea. If the highest score appears in a tie, the point with a higher terrain stability score is selected first.

[0060] Step S6: Based on the traffic accessibility feature map of the whole field area generated by the graph embedding model, the optimal candidate points in different subareas are finally verified in combination with the transportation weight and installation size parameters of the transformer substation to determine the layout position of the transformer substation.

[0061] Specifically, step S6 is the final link of determining the layout position of the box-type transformer, which comprehensively verifies the optimal candidate points of each partition to ensure that the selected position has feasibility in actual application. The traffic accessibility feature map of the whole field generated based on the graph embedding model can clearly reflect the traffic convenience degree of each candidate point, which is crucial for the transportation, installation and later maintenance of the box-type transformer. Combined with the transportation weight, installation size and other parameters of the box-type transformer, the verification can ensure that the site conditions of the candidate point can meet the physical installation requirements of the equipment, avoiding the limitation of site conditions leading to the failure of normal layout of the equipment. The layout position determined through the final verification can take into account the electrical performance, terrain adaptability and traffic convenience, which is the final result of the whole position determination method.

[0062] In specific implementation, first, the graph embedding model is used to analyze the traffic data of the whole field, such as road network, transportation channel width (not less than 4 meters), turning radius (not less than 15 meters), etc., to generate a traffic accessibility feature map. The traffic accessibility score of each node in the map is between 0 and 100, and the higher the score, the more convenient the transportation. The transportation weight parameter (usually 5t-15t) and installation size parameter (about 3m x 2m x 2.5m) of the box-type transformer are obtained. The optimal candidate points of each partition are verified one by one: check whether there is a passable transportation channel within 50 meters around the point, whether the carrying capacity of the channel meets the transportation weight requirement (road surface carrying capacity not less than 20t); measure the installation site size at the point, requiring the site length not less than 5 meters, the width not less than 4 meters, and the ground flatness error not more than 50mm; refer to the traffic accessibility feature map to ensure that the traffic accessibility score of the candidate point is not less than 70 points. If the candidate point meets all the verification conditions, it is determined as the layout position of the box-type transformer; if not, the other candidate points in the partition are screened in turn according to the comprehensive score until a point that meets the conditions is found.

[0063] Preferably, in step S2, the graph embedding model calculates the correlation weight by the following formula:

[0064]

[0065] wherein W i，j is the correlation weight of the ith point and the jth terrain feature point, f(v i ) is the feature vector of the ith point output by the graph embedding model, f(v j ) is the feature vector of the jth terrain feature point, n is the total number of terrain feature points, S i is the estimated installed capacity of the photovoltaic module corresponding to the ith point, S rated is the rated capacity parameter of the box-type transformer.

[0066] Specifically, step S2 specifies the specific implementation of the graph embedding model to calculate the correlation weight, which improves the accuracy of effective point selection by comprehensively considering the correlation degree and capacity matching degree of the point and the terrain feature point. In actual operation, first, it is necessary to specify that the feature vector dimension output by the graph embedding model is 128 dimensions, each dimension corresponds to a specific attribute of the terrain or point, the spatial correlation strength of the point and the terrain feature point is quantified by vector dot product operation, and the dot product result is converted by exponential function and normalized to obtain the correlation weight ratio based on the terrain feature alone. At the same time, the ratio of the estimated installed capacity of the photovoltaic module to the rated capacity of the box-type transformer is introduced as a capacity matching coefficient, wherein the rated capacity covers three levels of 1000kVA, 2000kVA and 3000kVA, the estimated installed capacity is calculated according to the point illumination time (2800-3200 hours per year), the component conversion efficiency (18%-22%) and the available area (10 meters x 10 meters grid corresponds to 100 square meters, and 50-60 components can be installed), and the single-point capacity ranges between 50kW-200kW. Multiply the terrain correlation weight and the capacity matching coefficient to get the final correlation weight, which reflects the adaptability of the point and the surrounding terrain (such as the point weight of the slope less than 25° and the slope facing south is usually higher than 0.6), and ensures the capacity supply-demand balance (when the ratio is between 0.6-1.2, the positive contribution of the coefficient to the weight is significant), and by setting the weight threshold (not less than 0.5), the points with unsuitable terrain or unmatched capacity can be effectively eliminated, providing high-quality basic data for subsequent power supply partition division.

[0067] Preferably, in step S3, when the planar point matrix model arrangement algorithm performs cluster analysis, the following formula is used to divide the power supply partition:

[0068]

[0069] wherein C m is the mth power supply partition, p is the effective point of the photovoltaic module, P is the set of effective points, dist(p, c m ) is the distance from the point p to the partition center c m , R max is the maximum power supply radius parameter of the box-type transformer, S p is the capacity of the photovoltaic module of the point p, and S rated is the rated capacity parameter of the box-type transformer.

[0070] Specifically, step S3 ensures that the partition result not only meets the power supply capacity of the box-type transformer, but also realizes efficient power supply of the photovoltaic module by determining the quantitative standard of cluster analysis. In implementation, first, the clustering parameters of the planar point matrix model arrangement algorithm are determined, including the minimum clustering point number 20 (corresponding to the installed capacity of 1000-1200 kW, matching 1000 kVA box transformer), clustering radius 100 meters (ensuring that the point positions in the same region are close, reducing line loss), and the algorithm aggregates the effective point positions with close spatial distance and close electrical connection into the same cluster by calculating the Euclidean distance (accuracy to 0.1 meters) and electrical correlation degree (calculated based on wire resistance, transmission power and other parameters) between the point positions. After clustering, the partition boundary is set according to the maximum power supply radius 500 meters of the box-type transformer, to ensure that the straight line distance from any point position in the partition to the center of the partition does not exceed the value, so as to control the line transmission loss within 5%. At the same time, the capacity constraint condition is introduced, which requires that the total capacity (sum of estimated capacities of all point positions) of all photovoltaic modules in each partition does not exceed the rated capacity (1000 kVA, 2000 kVA or 3000 kVA) of the matched box-type transformer. When the total capacity exceeds 1.2 times of the rated capacity, the clustering result needs to be split, and the capacity deviation rate (not more than ±10%) and center distance (not less than 300 meters) of the split sub-partition are calculated to ensure that the split sub-partition still meets the power supply radius and capacity requirements. This division method not only avoids the decline of power supply efficiency caused by too large partition, but also prevents equipment failure caused by capacity overload, laying a reasonable regional foundation for subsequent box transformer point selection.

[0071] Preferably, in step S4, when constructing the electrical performance evaluation matrix, the evaluation value is calculated by the following formula:

[0072]

[0073] wherein E k is the electrical performance evaluation value of the kth candidate point position, Z k is the box transformer short-circuit impedance calculation value corresponding to the point position, Z rated is the rated short-circuit impedance parameter of the box-type transformer, P k is the box transformer load loss calculation value corresponding to the point position, P rated is the rated load loss parameter, and a, b and g are weight coefficients, f elec (p, k) is the module length of the electrical connection feature vector of the point position p and the candidate point position k output by the graph embedding model.

[0074] Specifically, the electrical performance evaluation matrix construction of step S4 is to integrate the electrical parameters of the box-type transformer and the point connection characteristics to form a comprehensive performance evaluation system. In the implementation process, first, the three-dimensional coordinates (X, Y accuracy to 0.1 meters, Z accuracy to 0.5 meters) of the candidate point and the effective point are collected, the straight-line distance (used to estimate the line length) and the elevation difference (affecting the laying difficulty) between the two are calculated, and the graph embedding model generates a 20-dimensional electrical connection feature vector based on these data, covering the normalized values (range 0-1) of the line resistance (calculated according to the wire cross-sectional area 120mm 2 -240mm 2 and length), reactance (related to wire spacing and laying method), capacitance (affected by environmental humidity and temperature), etc. At the same time, the short-circuit impedance (4%-6%) and load loss (10kW-25kW) parameters of the box-type transformer are obtained, where the short-circuit impedance is used to evaluate the current limiting capability when the system is short-circuited, and the load loss reflects the energy loss when running at full load. When calculating the evaluation value, set the weight coefficients α=0.3 (short-circuit impedance proportion), β=0.3 (load loss proportion), and γ=0.4 (electrical connection feature proportion), multiply the ratio of the short-circuit impedance calculation value to the rated value (controlled between 0.8-1.2), the ratio of the load loss calculation value to the rated value (controlled between 0.7-1.3), and the average value of the electrical connection feature vector length (reflecting the overall connection quality, the higher the better) by the corresponding weight, and then sum them up to get the electrical performance evaluation value of each candidate point. The lower the evaluation value, the better the electrical performance of the candidate point. By integrating the evaluation values of all candidate points into a matrix, an intuitive quantitative comparison basis can be provided for subsequent optimization and selection, ensuring that the selected point performs well in short-circuit characteristics, energy loss, and connection quality.

[0075] Preferably, in step S5, when the plane point matrix model arrangement algorithm is optimized and solved, the following formula is used to select the optimal candidate point:

[0076]

[0077] where O k is the comprehensive optimization value of the kth candidate point, E k is the electrical performance evaluation value, δ is the terrain stability influence coefficient, S slope，k is the slope parameter at the kth candidate point, which is extracted from the terrain feature map by the graph embedding model.

[0078] Specifically, the optimal candidate point selection process of step S5 realizes more scientific point optimization by comprehensively considering electrical performance and terrain stability. In implementation, the planar point matrix model arrangement algorithm first processes the electrical performance evaluation matrix, which contains line total loss (range 5 kW-30 kW), maximum voltage drop (1%-5%), short-circuit current (8 kA-20 kA), and other indicators. The algorithm calculates the comprehensive electrical performance score by using the weighted summation method, wherein the line total loss weight is 0.4 (as it directly affects the operation cost), the maximum voltage drop weight is 0.3 (related to power supply quality), and the short-circuit current weight is 0.3 (related to equipment safety). The score range is 0-100 points, and the higher the score, the better the electrical performance (for example, a point with a score of 80 points or more, the line loss is usually less than 10 kW, and the voltage drop is less than 3%). At the same time, the sand mountain terrain stability parameters are processed by the graph embedding model, including soil bearing capacity (measured value 150 kPa-300 kPa, points with less than 150 kPa have a stability score deduction of 30 points or more), seismic intensity (7 degrees of fortification, local 8 degrees of region score reduction of 20 points), and wind and sand erosion intensity (annual erosion depth 30 mm-80 mm, points exceeding 50 mm have a score deduction of 15 points). These parameters are converted into terrain stability scores of 0-100 points. The final comprehensive optimization value is obtained by multiplying the reciprocal of the electrical performance score (reflecting that the higher the electrical performance, the higher the value) and the terrain stability correction term (1 minus the product of the slope influence coefficient and the slope value, the slope influence coefficient is set to 0.01, and the correction term of a point with a slope of 15° is 0.85). The point with the highest optimization value is the optimal candidate. This method not only ensures the electrical efficiency of the point, but also reduces the risk of equipment damage caused by terrain problems (for example, points with a stability score of 80 points or more have a foundation settlement probability of less than 5% in five years).

[0079] Preferably, in step S6, the traffic accessibility feature map generated by the graph embedding model is finally verified by the following formula:

[0080]

[0081] wherein T k is the traffic accessibility verification value of the kth candidate point, f traff (k) is the traffic feature vector of the candidate point k, f std is the standard traffic accessibility feature vector, sim() is the vector similarity calculation function, W rated is the rated transportation weight parameter of the box-type transformer, W k is the actual transportation weight limit value at the candidate point k, L rated is the installation size parameter of the box-type transformer, L k is the installation space size at the candidate point k.

[0082] Specifically, the final verification process of step S6 provides a quantitative verification standard based on traffic accessibility and device parameters to ensure that the selected point has actual layout feasibility. In implementation, the graph embedding model first analyzes the traffic data of the whole field, including road width (5-8 meters for main roads and 3-5 meters for branch roads), turning radius (20-30 meters for main roads and 15-20 meters for branch roads), and road surface type (asphalt road, gravel road, etc., different types correspond to different traffic coefficients), generates a traffic feature vector, and calculates the similarity (using cosine similarity, range 0-1) with the standard traffic accessibility feature vector (based on ideal transportation conditions) to obtain the traffic accessibility basic score (0-100 points). For example, point locations near main roads with sufficient turning radius usually have a similarity of more than 0.8 and a score of more than 80 points. At the same time, the physical parameters of the box-type transformer are introduced for correction. The ratio of the transportation weight parameter (5t-15t) to the actual transportation weight limit value of the candidate point (calculated according to the road bearing, such as the gravel road usually limiting within 10t) reflects the transportation feasibility (ratio greater than 1 indicates overweight and needs to be deducted); the ratio of the installation size parameter (3m x 2m x 2.5m) to the installation space size (4-6 meters in length and 3-5 meters in width) evaluates the site adaptability (ratio greater than 1 indicates insufficient space). Multiply the traffic accessibility basic score by the two ratios to get the final traffic accessibility verification value, which needs to be greater than 0.6 (corresponding to a score of more than 60 points). For example, point locations with a verification value of more than 0.8 can not only meet the transportation of vehicles (15t load and 18m turning radius) but also accommodate device installation (including operating space), which can ensure the smooth implementation of the whole process from transportation to installation of the box-type transformer and avoid project delays or cost increases due to site problems in the later stage.

[0083] Preferably, the step S3 comprises the following sub-steps:

[0084] S3.1: Import the coordinate data of the effective point into the plane point matrix model layout algorithm, set the neighborhood search range of the matrix according to the maximum power supply radius parameter of the box-type transformer, and generate the local neighborhood matrix of different point locations; S3.2: Calculate the spatial correlation features of different point locations in the local neighborhood matrix through the graph embedding model, and determine the initial clustering center combined with the power supply efficiency parameter of the box-type transformer; S3.3: Iterative clustering of effective point locations based on the initial center, adjusting the clustering center position according to the load rate parameter of the box-type transformer after each iteration, until the clustering result converges; S3.4: Boundary division is performed on the converged clustering result to ensure that the total capacity of photovoltaic modules in each partition matches the rated capacity parameter of the box-type transformer, forming a power supply partition.

[0085] Specifically, step S3 specifies the specific implementation process of the power supply partitioning through four sub-steps, aiming to realize the precision of partitioning by combining the parameters of the box-type transformer and the terrain characteristics. In step S3.1, the coordinate data of the effective point (accurate to 0.01 meters) is imported into the plane point matrix model layout algorithm, and according to the maximum power supply radius parameter of the box-type transformer (set to 500 meters), the algorithm automatically generates a circular neighborhood search range with a radius of 500 meters centered on each point, and then constructs a local neighborhood matrix, which contains the coordinates and relative distance information of all other effective point within the range, providing basic data support for subsequent clustering. In step S3.2, the local neighborhood matrix is processed by the graph embedding model to extract the spatial correlation characteristics of each point (such as slope consistency and elevation continuity), and combined with the 85% optimal power supply efficiency parameter of the box-type transformer, the point with the most significant correlation characteristics is selected as the initial clustering center. The number of initial centers is determined according to the total number of effective points, with 1 center corresponding to every 20-30 points. In step S3.3, during the iterative clustering process, the clustering center position is adjusted according to the 60%-100% load rate parameter of the box-type transformer after each iteration. When the center coordinate variation of three consecutive iterations is less than 0.5 meters, it is determined that the clustering result converges, ensuring that the clustering area is compatible with the load capacity of the box-type transformer. In step S3.4, when dividing the boundary of the converged clustering result, the ratio of the total capacity of photovoltaic components in each partition to the rated capacity of the box-type transformer (1000 kVA, 2000 kVA, 3000 kVA) is calculated to ensure that the ratio is between 0.8 and 1.2. At the same time, the Voronoi diagram algorithm is used to divide the partition boundary, so that the distance difference between any point on the boundary and the centers of the adjacent two partitions does not exceed 50 meters. The final power supply partition not only meets the electrical performance requirements, but also adapts to the spatial distribution characteristics of the sand mountain terrain.

[0086] Preferably, the step S4 includes the following sub-steps:

[0087] S4.1: Collect three-dimensional coordinate data of different candidate point positions, and construct a spatial coordinate system of the candidate point positions by a graph embedding model combined with the installation height parameter of the box-type transformer; S4.2: Map the coordinates of the effective point positions of the photovoltaic components in the partition to the coordinate system, calculate the electrical distance between different effective point positions and the candidate point positions, and generate a preliminary connection matrix combined with the impedance parameter of the box-type transformer; S4.3: Perform sparse processing on the preliminary connection matrix by a plane point matrix model layout algorithm, and retain the connection paths that meet the maximum current-carrying capacity parameter of the box-type transformer; S4.4: Based on the sparse connection paths, construct an electrical performance evaluation matrix containing path loss and voltage drop indicators combined with the loss parameter of the box-type transformer.

[0088] Specifically, step S4 constructs the electrical performance evaluation matrix through four sub-steps, fully integrates the parameters of the box-type transformer and the spatial relationship of the point positions, and improves the comprehensiveness and accuracy of the evaluation. After collecting the three-dimensional coordinates (X, Y, Z axis accuracy is 0.01 meters, 0.01 meters, and 0.1 meters) of the candidate point positions in step S4.1, combined with the installation height parameter of 1.2 meters of the box-type transformer, the spatial coordinate system with the candidate point position as the origin, the X axis pointing east, the Y axis pointing north, and the Z axis perpendicular to the ground is constructed through the graph embedding model, ensuring that all point position coordinates are calculated in a unified coordinate system. In step S4.2, after mapping the coordinates of the effective point positions of the photovoltaic components in the sub-area to the coordinate system, the electrical distance (including the comprehensive influence of horizontal distance and vertical distance) between each effective point position and the candidate point position is calculated using the three-dimensional distance formula, and combined with the short-circuit impedance parameter of 4%-6% of the box-type transformer, a preliminary connection matrix is generated, and the matrix elements are the impedance values between each point position, reflecting the impedance loss in the process of power transmission. In step S4.3, when the preliminary connection matrix is processed by the planar point matrix model arrangement algorithm, according to the maximum carrying capacity parameter of 630A-1250A of the box-type transformer, the connection paths with carrying capacity exceeding the limit value are removed, and the efficient paths with carrying capacity within 70%-90% of the rated value are retained, so that the matrix sparsity is controlled between 30%-50%, and the invalid calculation amount is reduced. In step S4.4, based on the connection paths after sparsification, combined with the load loss parameter of 10kW-25kW of the box-type transformer, the power loss (calculated according to the copper loss formula) and voltage drop (derived according to Ohm's law) of each path are calculated, and these indicators are integrated into the electrical performance evaluation matrix, with the matrix rows representing the candidate point positions, the columns representing the evaluation indicators, and the element values representing the specific quantitative results, providing multi-dimensional electrical performance basis for subsequent optimization and selection.

[0089] Preferably, the step S5 comprises the following sub-steps:

[0090] S5.1: performing eigenvalue decomposition on the electrical performance evaluation matrix using the planar point matrix model arrangement algorithm to extract principal component features representing electrical performance; S5.2: converting the stability parameters of the sand mountain terrain into feature vectors through the graph embedding model, and fusing the feature vectors with the principal component features to form a comprehensive feature matrix; S5.3: sorting the comprehensive feature matrix to select the top 30% of the candidate point positions as the key optimization objects; S5.4: performing secondary evaluation on the key optimization objects in combination with the operating environment temperature parameter of the box-type transformer to determine the box-type transformer candidate point positions with optimal electrical performance in different sub-areas.

[0091] Specifically, the implementation process of step S5 realizes accurate screening of the optimal candidate point by four sub-steps, taking into account the dual requirements of electrical performance and terrain adaptability. In step S5.1, the plane point matrix model arrangement algorithm performs eigenvalue decomposition on the electrical performance evaluation matrix, extracts the top 3 principal component features (cumulative contribution rate not less than 85%), which correspond to line loss, voltage stability, short circuit safety and other core electrical performances respectively, reduces the high-dimensional matrix to low-dimensional feature vectors, and simplifies the calculation complexity. In step S5.2, the graph embedding model is used to convert the sand mountain terrain stability parameters (soil bearing capacity 150kPa-300kPa, earthquake intensity 7-8 degrees, wind and sand erosion intensity 30mm-80mm / year) into a 10-dimensional feature vector, which is weighted and fused with the principal component features (electrical feature weight 0.7, terrain feature weight 0.3) to form a comprehensive feature matrix, realizing the coordinated consideration of electrical performance and terrain stability. In step S5.3, the comprehensive feature matrix is normalized by row (mapping the numerical value to the range of 0-1), sorted by comprehensive feature value from high to low, and the top 30% of candidate points are selected as the key optimization objects, which perform well in electrical performance and terrain adaptability, and are the narrowed range for final screening. In step S5.4, combined with the operating environment temperature parameters of the box-type transformer of-30°C to 50°C, the key optimization objects are evaluated again. Through the installation of an environmental temperature sensor to collect the annual average extreme temperature value of the point, the points exceeding the temperature range are eliminated, and the temperature correction coefficient (between 0.9-1.1) of the remaining points is calculated. The comprehensive feature value is corrected, and finally the point with the highest corrected comprehensive feature value is selected as the optimal candidate point of each sub-area, ensuring that the selected point can maintain good electrical performance and long-term stable operation in the complex sand mountain environment.

[0092] The graph embedding model in the application is a technical model capable of converting complex terrain features and point space correlation of a sand mountain photovoltaic power station site into a high-dimensional vector. The essence is to quantitatively encode spatial features such as terrain elevation, slope, slope direction, soil bearing capacity, seismic intensity, wind and sand erosion intensity, and electrical connection features of photovoltaic component points and candidate points of box-type transformers through multi-layer neural network operation. In the implementation process, first, high-precision terrain data (such as elevation data at every 5-meter interval, slope data at 0°-60°, etc.) and point coordinate information (accuracy to 0.01 meters) are collected, and after inputting these data into the model, the model learns the potential correlation between the features and outputs a 128-dimensional or 20-dimensional high-dimensional vector, each vector corresponding to a specific terrain grid unit or point, and the vector contains the comprehensive feature information of the unit or point. For example, when processing the terrain stability parameter, the model will convert the soil bearing capacity (150 kPa-300 kPa) and seismic intensity (7-8 degrees) into a 10-dimensional feature vector; when processing the electrical connection feature, it will encode the line resistance, reactance and other parameters into a 20-dimensional vector. Its main functions are reflected in three aspects: first, it realizes the quantitative expression of terrain features, providing calculable basic data for subsequent point analysis; second, it captures the spatial correlation and electrical connection features between points, such as slope consistency, elevation continuity and line loss; third, it integrates multi-source data, integrating terrain, electrical, environmental and other different types of parameters into a unified feature space. The significance of the model lies in breaking through the limitations of traditional terrain analysis features that are difficult to quantify and integrate, and realizing efficient processing of complex information in the form of high-dimensional vectors, providing a scientific feature extraction and integration method for accurate determination of the position of box-type transformers of photovoltaic power stations in sand mountains, and improving the intelligence and accuracy of the position determination process.

[0093] The point position matrix model arrangement algorithm in the application is an algorithm for processing potential layout points of photovoltaic modules and spatial distribution and optimization of candidate points of box-type transformers. The algorithm organizes point position data in a matrix form, realizes screening, partitioning and evaluation of the point positions through grid division, cluster analysis, matrix operation and other operations. In implementation, first, according to parameters such as the maximum power supply radius (such as 500 meters) of the box-type transformer, the field area is divided into 10m*10m grids to form an initial point position matrix, and the matrix elements contain point position coordinates, estimated capacity and other information; then, effective point positions are clustered by using density clustering and other methods, parameters such as the minimum number of clustered points (such as 20) and the clustering radius (such as 100 meters) are set, and the point positions are divided into power supply partitions through iterative calculation (such as convergence when the variation of the central coordinates is less than 0.5 meters for three times of iteration); then, eigenvalue decomposition, sparsification processing (such as retaining paths with load carrying capacity of 70%-90% of the rated value) and other operations are performed on the electrical performance evaluation matrix to extract principal component features and calculate a comprehensive optimization value. The algorithm has the following effects: systematic management and screening of the point positions, such as elimination of unsuitable point positions with a slope greater than 30°; reasonable division of the power supply partitions to ensure that each partition meets the requirements of the rated capacity (1000kVA, 2000kVA, etc.) and the maximum power supply radius of the box-type transformer; quantitative evaluation and optimization of the electrical performance of the candidate point positions to screen out point positions with the best comprehensive performance. The significance of the algorithm lies in that it provides a standardized and streamlined processing method for spatial analysis and optimization of the point positions, efficiently processes a large amount of point position data (usually containing 5000-8000 initial point positions) through matrix operation and iterative optimization, ensures the rationality of the division of the power supply partitions, the accuracy of the electrical performance evaluation and the scientificity of the screening of the optimal point positions, effectively solves the problem of position determination caused by the dispersed distribution of point positions and the complex terrain in sand mountain areas, provides reliable algorithm support for position determination of the box-type transformer, and improves the overall design quality and operation efficiency of the photovoltaic power station.

[0094] As shown in Figure 2 A method for determining the position of a box-type transformer of a sand mountain photovoltaic power station, the method is implemented by different units, comprising:

[0095] A topographic feature map construction and point position matrix generation unit, which is connected with a photovoltaic module parameter acquisition sensor device, receives capacity parameters of the photovoltaic modules and outputs an initial point position matrix;

[0096] An effective point position screening unit, which is connected with the topographic feature map construction and point position matrix generation unit and a box-type transformer parameter database, receives the initial point position matrix and the rated capacity parameters of the box-type transformer, and outputs an effective point position set;

[0097] A power supply partition division unit, which is connected with the effective point position screening unit and a planar point position calculation module, receives the effective point position set and outputs a power supply partition result;

[0098] An electrical performance evaluation matrix construction unit connected with the power supply partition unit and the box-type transformer parameter database, receives the power supply partition result and the electrical parameters of the box-type transformer, and outputs the electrical performance evaluation matrix;

[0099] An optimal candidate point screening unit connected with the electrical performance evaluation matrix construction unit and the terrain stability analysis module, receives the electrical performance evaluation matrix and the terrain parameters, and outputs the optimal candidate point;

[0100] A final position determination unit connected with the optimal candidate point screening unit, the traffic accessibility analysis module, and the box-type transformer parameter database, receives the optimal candidate point, the traffic data, and the physical parameters of the box-type transformer, and outputs the box-type transformer layout position.

[0101] A position determination method for a box-type transformer of a sand mountain photovoltaic power station, the primary advantage of the method is to realize the deep integration of terrain features and equipment parameters, effectively overcoming the limitations of relying only on manual experience or simple calculation in the prior art. Through the high-dimensional vector conversion of the elevation and slope features of the sand mountain terrain by the graph embedding model, combined with the plane point matrix model layout algorithm, the rated capacity, power supply radius and other parameters of the box-type transformer are included in the analysis system, so that the position determination process no longer considers single factors in isolation. This integration ensures that the determined position can accurately match the load demand of the photovoltaic module cluster, avoiding the problem of line loss caused by mismatch between power supply radius and actual load, and improving the energy utilization efficiency of the power station.

[0102] Secondly, the method solves the limitations brought by single index judgment in the prior art through multi-step system optimization and verification. From the screening of the initial point matrix, the division of the power supply partition, to the construction of the electrical performance evaluation matrix, and finally to the verification combined with traffic accessibility, each step is carried out around the operating parameters of the box-type transformer and the stability requirements of the sand mountain terrain. This multi-dimensional analysis process ensures that the selected position not only meets the electrical performance requirements of the equipment, but also adapts to the complex geological conditions of the sand mountain area, reducing the risk of foundation damage caused by terrain changes, and also providing convenience for the transportation and later maintenance of the equipment.

[0103] In addition, the algorithm design of the method has strong pertinence and adaptability, further making up for the shortcomings of the prior art. The plane point matrix model arrangement algorithm can efficiently process a large amount of point data in view of the grid characteristics of the sand mountain terrain, and the graph embedding model can accurately capture the spatial correlation characteristics of the terrain. In combination of the two, and with the aid of various parameters of the box-type transformer, the position determination process has both high efficiency of data processing and full consideration of the particularity of the sand mountain environment. Compared with the general computing method in the prior art, this targeted design can more accurately deal with the complex scene of the sand mountain photovoltaic power station, and technically guarantees the reliability and rationality of the position determination result.

[0104] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "arrangement", "installation", "connection", "link", "fixing" should be understood in a broad sense, for example, can be fixedly connected, can also be detachably connected, or integrally connected; can be mechanically connected, can also be electrically connected; can be directly connected, can also be indirectly connected through an intermediate medium, can be internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0105] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various equivalent changes, modifications, replacements and variations of the embodiments can be made without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalent scope.

Claims

1. A method for determining the position of a box-type transformer of a sand dune photovoltaic power station, characterized in that, Comprise: Step S1: feature extraction of the sand mountain photovoltaic power station site terrain data is carried out through the graph embedding model, the spatial characteristics of terrain elevation, slope and slope direction are converted into high-dimensional vector representation, the terrain feature atlas containing photovoltaic module layout area is constructed, the potential photovoltaic module layout point in the atlas is divided into grid by using the plane point matrix model arrangement algorithm, and the initial point matrix is formed; Step S2: based on the rated capacity, voltage level and no-load loss parameters of the box-type transformer, combined with the estimated installed capacity of the photovoltaic module in the initial point matrix, the correlation weight of different points and the surrounding terrain characteristics is calculated by the graph embedding model, the initial point matrix is screened, and the effective point meeting the capacity matching condition is reserved; Step S3: the effective point is analyzed by using the plane point matrix model arrangement algorithm, according to the maximum power supply radius parameter of the box-type transformer, the clustering result is divided into several power supply partitions, and several candidate box transformer layout points are generated in each partition; Step S4: the electrical connection feature vector of different candidate points and the effective point of photovoltaic module in the partition is calculated by the graph embedding model, combined with the short-circuit impedance and load loss parameters of the box-type transformer, the electrical performance evaluation matrix of the candidate point is constructed; Step S5: the electrical performance evaluation matrix is optimized and solved by using the plane point matrix model arrangement algorithm, considering the stability parameters of sand mountain terrain, the candidate point of box transformer with optimal electrical performance in different partitions is screened out; Step S6: based on the traffic accessibility feature atlas of the whole site generated by the graph embedding model, combined with the transportation weight and installation size parameters of the box-type transformer, the optimal candidate point in different partitions is finally verified, and the layout position of the box-type transformer is determined.

2. The method for determining the position of a box-type transformer of a sand dune photovoltaic power station according to claim 1, characterized in that, In step S2, the graph embedding model calculates the correlation weight by the following formula: wherein W i，j is the correlation weight of the ith point and the jth topographic feature point, f(v i ) is the feature vector of the ith point output by the graph embedding model, f(v j ) is the feature vector of the jth topographic feature point, n is the total number of topographic feature points, S i is the estimated installed capacity of the photovoltaic module corresponding to the ith point, S rated is the rated capacity parameter of the box-type transformer.

3. The method for determining the position of a box-type transformer of a sand dune photovoltaic power station according to claim 1, characterized in that, In step S3, when the plane point matrix model arrangement algorithm is used for clustering analysis, the following formula is used to divide the power supply partition: Wherein, C m is the mth power supply partition, p is the effective point of photovoltaic module, P is the effective point set, dist(p, c m ) is the distance from point p to the center of the partition c m , R max is the maximum power supply radius parameter of the box transformer, S p is the photovoltaic module capacity of point p, S rated is the rated capacity parameter of the box transformer.

4. The method of claim 1, wherein, In step S4, when the electrical performance evaluation matrix is constructed, the evaluation value is calculated by the following formula: wherein E k is the electrical performance evaluation value of the kth candidate point, Z k is the short-circuit impedance calculation value of the box transformer corresponding to the point, Z rated is the rated short-circuit impedance parameter of the box transformer, P k is the load loss calculation value of the box transformer corresponding to the point, P rated is the rated load loss parameter, and α, β, and γ are weight coefficients, f elec (p, k) is the length of the electrical connection feature vector of the point p and the candidate point k output by the graph embedding model.

5. The method for determining the position of a box-type transformer of a sand dune photovoltaic power station according to claim 1, characterized in that, In step S5, when the plane point matrix model arrangement algorithm is used for optimization and solution, the following formula is used to screen the optimal candidate point: wherein O k is the comprehensive optimization value of the kth candidate point, E k is the electrical performance evaluation value, δ is the terrain stability influence coefficient, S slope，k is the slope parameter at the candidate point k, which is extracted from the terrain feature map by the graph embedding model.

6. The method of claim 1, wherein, In step S6, the traffic accessibility feature atlas generated by the graph embedding model is finally verified by the following formula: wherein T k is the traffic accessibility verification value of the kth candidate site, f traff (k) is the traffic feature vector of the candidate site k, f std is the standard traffic accessibility feature vector, sim() is the vector similarity calculation function, W rated is the rated transportation weight parameter of the box transformer, W k is the actual transportation weight limit value at the candidate site k, L rated is the installation size parameter of the box transformer, L k is the installation space size at the candidate site k.

7. The method of claim 1, wherein, The step S3 comprises the following steps: S3.1: the coordinate data of the effective point is imported into the plane point matrix model arrangement algorithm, the neighborhood search range of the matrix is set according to the maximum power supply radius parameter of the box-type transformer, and the local neighborhood matrix of different points is generated; S3.2: the spatial correlation characteristics of different points in the local neighborhood matrix are calculated by the graph embedding model, and the clustering initial center is determined combined with the power supply efficiency parameter of the box-type transformer; S3.3: the effective point is iteratively clustered based on the initial center, and the clustering center position is adjusted according to the load rate parameter of the box-type transformer after each iteration until the clustering result converges; S3.4: the boundary of the converged clustering result is divided to ensure that the total capacity of photovoltaic modules in each partition matches the rated capacity parameter of the box-type transformer, and the power supply partition is formed.

8. The method of claim 1, wherein, The step S4 comprises the following steps: S4.1: Collect three-dimensional coordinate data of different candidate point positions, combine the installation height parameters of the box-type transformer, and construct the spatial coordinate system of the candidate point positions through the graph embedding model; S4.2: Map the coordinates of the effective point positions of the photovoltaic components in the subarea to the coordinate system, calculate the electrical distance between different effective point positions and the candidate point positions, combine the impedance parameters of the box-type transformer, and generate a preliminary connection matrix; S4.3: Perform sparse processing on the preliminary connection matrix through the plane point position matrix model arrangement algorithm, and retain the connection paths that meet the maximum current-carrying capacity parameters of the box-type transformer; S4.4: Based on the connection paths after sparse processing, combine the loss parameters of the box-type transformer, and construct an electrical performance evaluation matrix containing path loss and voltage drop indicators.

9. The method of claim 1, wherein, The step S5 comprises the following sub-steps: S5.1: Perform eigenvalue decomposition on the electrical performance evaluation matrix using the plane point position matrix model arrangement algorithm, and extract principal component features representing the electrical performance; S5.2: Convert the stability parameters of the sand mountain terrain into feature vectors through the graph embedding model, fuse them with the principal component features, and form a comprehensive feature matrix; S5.3: Sort the comprehensive feature matrix, and select the top 30% of the candidate point positions in the comprehensive feature values as the key optimization objects; S5.4: Combine the operating environment temperature parameters of the box-type transformer, perform secondary evaluation on the key optimization objects, and determine the box-type transformer candidate point positions with the optimal electrical performance in different subareas.

10. The method for determining the position of a box-type transformer of a sandhill photovoltaic power station according to any one of claims 1-9, characterized in that, The method is implemented through different units, including: a terrain feature map construction and point position matrix generation unit, which is connected with a photovoltaic component parameter acquisition sensor device, receives the capacity parameters of the photovoltaic components, and outputs an initial point position matrix; an effective point position screening unit, which is connected with the terrain feature map construction and point position matrix generation unit and a box-type transformer parameter database, receives the initial point position matrix and the rated capacity parameters of the box-type transformer, and outputs an effective point position set; a power supply subarea division unit, which is connected with the effective point position screening unit and a plane point position calculation module, receives the effective point position set, and outputs the power supply subarea result; an electrical performance evaluation matrix construction unit, which is connected with the power supply subarea division unit and the box-type transformer parameter database, receives the power supply subarea result and the electrical parameters of the box-type transformer, and outputs the electrical performance evaluation matrix; an optimal candidate point position screening unit, which is connected with the electrical performance evaluation matrix construction unit and a terrain stability analysis module, receives the electrical performance evaluation matrix and the terrain parameters, and outputs the optimal candidate point position; a final position determination unit, which is connected with the optimal candidate point position screening unit, a traffic accessibility analysis module, and the box-type transformer parameter database, receives the optimal candidate point position, traffic data, and physical parameters of the box-type transformer, and outputs the box-type transformer layout position.

Images