Photovoltaic power station site selection method based on graph characteristics considering complex geographical factors
By combining graph neural networks and capsule neural networks and integrating geographic information features, the accuracy problem of traditional photovoltaic power plant site selection methods under complex geographical factors is solved, and more efficient photovoltaic power plant site selection is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA SOUTHERN POWER GRID COMPANY
- Filing Date
- 2023-06-28
- Publication Date
- 2026-07-21
AI Technical Summary
Traditional photovoltaic power plant site selection methods are not very accurate when dealing with complex geographical factors, making it difficult to accurately predict power generation efficiency and unable to cope with site selection problems under complex geographical factors.
A graph neural network is used to model the map. By combining a feature network model and a capsule neural network, geographic information features are integrated, and the target location of the photovoltaic power station is determined through a graph feature network site selection method.
It improves the accuracy and efficiency of photovoltaic power plant site selection, comprehensively considers geographical location and complex geographical factors, and provides detailed site selection reports to support decision-making.
Smart Images
Figure CN116777069B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of power technology, and in particular to a graph feature network site selection method for photovoltaic power plants that takes into account complex geographical factors. Background Technology
[0002] With the rapid development of new energy sources, photovoltaic power plants, as an important representative of renewable energy, are gradually gaining prominence in the energy sector. Site selection is particularly important during the construction of photovoltaic power plants. It is necessary to consider not only the impact of geographical factors (such as topography and climate) on the power generation efficiency of photovoltaic power plants, but also whether the photovoltaic power plants meet the requirements of proximity to the power grid and efficient maintenance, as well as surrounding environmental factors (such as ecological environment) and socio-economic factors.
[0003] Traditional technologies employ both experience-based and rule-based methods and model-based methods for photovoltaic (PV) power plant site selection. Experience-based and rule-based methods tend to overlook factors such as topography, leading to a lack of accuracy and scientific rigor in site selection. Model-based methods, on the other hand, require large amounts of training samples and computational resources, struggle to handle the complexity of geographical factors, and are unable to accurately predict PV power plant efficiency, thus failing to address the site selection challenges posed by complex geographical conditions.
[0004] Therefore, current photovoltaic power plant site selection technologies suffer from low site selection accuracy. Summary of the Invention
[0005] Based on this, it is necessary to provide a graph feature network site selection method, apparatus, computer equipment, computer-readable storage medium, and computer program product for photovoltaic power plants that takes into account complex geographical factors and can improve the site selection accuracy, thereby addressing the aforementioned technical problems.
[0006] Firstly, this application provides a graph feature network site selection method for photovoltaic power plants that considers complex geographical factors. The method includes:
[0007] Obtain a map of the target area and the corresponding geographic information of the map;
[0008] The map is modeled based on a graph neural network to obtain the location features of the target area, and the geographic information is input into a pre-trained feature network model to obtain the geographic features of the target area.
[0009] The fused features of the location features and the geographic features are input into a pre-trained capsule neural network model to obtain the target features of the target area;
[0010] Based on the target characteristics, the target location of the photovoltaic power station is determined from the target area.
[0011] In one embodiment, the step of modeling the map based on a graph neural network to obtain the location features of the target area includes:
[0012] Candidate locations and their spatial relationships are determined from the map.
[0013] The candidate locations are used as nodes, and the edges between the nodes are determined according to the positional relationships to obtain the network graph corresponding to the map.
[0014] Based on the network diagram, the location features of the target region are obtained.
[0015] In one embodiment, obtaining the location features of the target region based on the network graph includes:
[0016] The target node is determined from the network graph;
[0017] The hidden state of the target node is determined based on the target node, the positional relationship of the target node, and the adjacent nodes of the target node;
[0018] The hidden state is input into a pre-determined local output model to obtain the positional features of the target node.
[0019] In one embodiment, the geographic information includes the topography, landforms, sunshine duration, and rainfall of the target area; the step of inputting the geographic information into a pre-trained feature network model to obtain the geographic features of the target area includes:
[0020] Based on the feature network model, the input terrain, landform, sunshine duration and rainfall are identified to obtain the geographical features of the target area.
[0021] In one embodiment, after modeling the map based on a graph neural network to obtain the location features of the target area, and inputting the geographic information into a pre-trained feature network model to obtain the geographic features of the target area, the method further includes:
[0022] The location features and the geographic features are fused to obtain the fused features of the target area.
[0023] In one embodiment, the step of inputting the fused features of the location features and the geographic features into a pre-trained capsule neural network model to obtain the target features of the target region includes:
[0024] Feature extraction is performed on the fused features to obtain the feature map corresponding to the fused features;
[0025] The feature map is input into the capsule layer of the capsule neural network model to obtain the capsule vector corresponding to the feature map;
[0026] The target features of the target region are determined based on the capsule vector.
[0027] In one embodiment, the target area includes at least one candidate location, each candidate location corresponding to a target feature; determining the target location of the photovoltaic power station from the target area based on the target feature includes:
[0028] If the target features of the candidate location meet the preset conditions, the candidate location is determined as the target location of the photovoltaic power station;
[0029] A site selection report for the photovoltaic power station is generated based on the target location and the target features of the target location.
[0030] Secondly, this application also provides a graph feature network site selection device for photovoltaic power plants that considers complex geographical factors. The device includes:
[0031] The information acquisition module is used to acquire a map of the target area and the corresponding geographic information of the map;
[0032] The feature extraction module is used to model the map based on a graph neural network to obtain the location features of the target area, and to input the geographic information into a pre-trained feature network model to obtain the geographic features of the target area.
[0033] The feature determination module is used to input the fused features of the location features and the geographic features into a pre-trained capsule neural network model to obtain the target features of the target area;
[0034] The location determination module is used to determine the target location of the photovoltaic power station from the target area based on the target characteristics.
[0035] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:
[0036] Obtain a map of the target area and the corresponding geographic information of the map;
[0037] The map is modeled based on a graph neural network to obtain the location features of the target area, and the geographic information is input into a pre-trained feature network model to obtain the geographic features of the target area.
[0038] The fused features of the location features and the geographic features are input into a pre-trained capsule neural network model to obtain the target features of the target area;
[0039] Based on the target characteristics, the target location of the photovoltaic power station is determined from the target area.
[0040] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:
[0041] Obtain a map of the target area and the corresponding geographic information of the map;
[0042] The map is modeled based on a graph neural network to obtain the location features of the target area, and the geographic information is input into a pre-trained feature network model to obtain the geographic features of the target area.
[0043] The fused features of the location features and the geographic features are input into a pre-trained capsule neural network model to obtain the target features of the target area;
[0044] Based on the target characteristics, the target location of the photovoltaic power station is determined from the target area.
[0045] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:
[0046] Obtain a map of the target area and the corresponding geographic information of the map;
[0047] The map is modeled based on a graph neural network to obtain the location features of the target area, and the geographic information is input into a pre-trained feature network model to obtain the geographic features of the target area.
[0048] The fused features of the location features and the geographic features are input into a pre-trained capsule neural network model to obtain the target features of the target area;
[0049] Based on the target characteristics, the target location of the photovoltaic power station is determined from the target area.
[0050] The aforementioned graph feature network-based site selection method, apparatus, computer equipment, storage medium, and computer program product for photovoltaic power plants considering complex geographical factors acquires a map of the target area and its corresponding geographical information. It then models the map based on a graph neural network to obtain the location features of the target area. The geographical information is input into a pre-trained feature network model to obtain the geographical features of the target area. The fused features of the location and geographical features are input into a pre-trained capsule neural network model to obtain the target features of the target area. Based on these target features, the target location of the photovoltaic power plant is determined. The method extracts the location features of the target area from the map and the geographical features from the geographical information. The capsule neural network model is used to extract the fused features of the location and geographical features, comprehensively considering the geographical location and complex geographical factors of the target area for photovoltaic power plant site selection, thus improving the accuracy of photovoltaic power plant site selection. Attached Figure Description
[0051] Figure 1 This is a flowchart illustrating a graph feature network site selection method for a photovoltaic power plant that considers complex geographical factors in one embodiment.
[0052] Figure 2 This is a schematic diagram of a deep neural network model in one embodiment;
[0053] Figure 3 This is a schematic diagram of a capsule neural network model in one embodiment;
[0054] Figure 4 This is a flowchart illustrating a graph feature network location method for a photovoltaic power station in one embodiment.
[0055] Figure 5 This is a structural block diagram of a graph feature network site selection device for a photovoltaic power plant that considers complex geographical factors in one embodiment.
[0056] Figure 6 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0058] In one embodiment, such as Figure 1As shown, a graph feature network site selection method for photovoltaic power plants considering complex geographical factors is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:
[0059] Step S110: Obtain a map of the target area and the corresponding geographic information of the map.
[0060] The target area can be an area that contains candidate sites for photovoltaic power plants.
[0061] Geographic information may include, but is not limited to, topographic data, geomorphological data, the spatiotemporal distribution of sunshine duration, and the spatiotemporal distribution of rainfall.
[0062] In practice, a map of the target area can be obtained, and corresponding terrain data, landform data, and geographic information such as the spatiotemporal distribution of sunshine duration and rainfall can be collected and input into the terminal, so that the terminal can obtain the map of the target area and the corresponding geographic information.
[0063] In practical applications, topographic and geomorphic data of the target area can be obtained. At least one candidate location can be identified in the target area, and the sunshine duration of each candidate location can be collected at different times to obtain the spatiotemporal distribution information of sunshine duration. Rainfall at each candidate location can be collected at different times to obtain the spatiotemporal distribution information of rainfall. The topographic and geomorphic data, as well as the spatiotemporal distribution information of sunshine duration and rainfall, can be used as the geographic information of the target area.
[0064] Step S120: Model the map based on a graph neural network to obtain the location features of the target area, and input the geographic information into a pre-trained feature network model to obtain the geographic features of the target area.
[0065] Among them, location features can be the geographical features of the target area, which can be represented as a network graph.
[0066] Among them, geographical features can be the characteristics of geographical information of the target area, and can be represented as feature vectors.
[0067] In practice, a graph neural network model can be used to model the map, with each candidate location in the target area as a node and the positional relationship between the candidate locations as an edge, to establish a network graph of positional features. Alternatively, a feature network model can be pre-trained, and geographic information can be input into the feature network model to extract geographic features from the geographic information and obtain feature vectors of geographic features.
[0068] In practical applications, during graph neural network modeling, nodes can represent candidate locations for site selection schemes, and edges can represent the relative positions and distances between candidate locations. Graph neural networks capture the dependencies in the graph by establishing a message-passing mechanism between graph nodes, thereby obtaining the corresponding features on the graph. The goal of graph neural networks is to learn the hidden state h of each candidate location. v This hidden state is used to generate the output o of candidate node locations. v Define f v Let g be the information aggregation function for candidate locations of nodes. It operates on candidate locations of all location schemes and updates the information of candidate locations based on the input neighbor features. Define g as the local output function, then the hidden state h of each candidate location is... v and the output of candidate locations v The expression is
[0069] h v =f(x) v x co[v] h ne[v] x ne[v] ),
[0070] Where, x v x co[v] h ne[v] x ne[v] These are the characteristics of candidate locations, the characteristics of their relative positions and distances, and the status and characteristics of their v-neighbors.
[0071] To find a unique solution, we apply Banach's fixed-point theorem to the above h. v =f(x) v x co[v] h ne[v] x ne[v] Rewritten as an iterative update process, the expression can be...
[0072] H t+1 =F(H t ,X),
[0073] Among them, H t H represents the hidden state of H in the t-th iteration. t+1 Let H represent the hidden state of H in the (t+1)th iteration, and X represent the features of the candidate location.
[0074] By passing features of the hidden state and candidate locations to the local output function g, the output o of the graph neural network can be computed. v The specific expression is
[0075] o v =g(h v xv ).
[0076] Feature network models can be implemented using deep neural networks. A deep neural network model is pre-trained, and topographic data, landform data, and spatiotemporal distribution information of sunshine time and rainfall are input into the trained deep neural network model to extract features and obtain the geographical features of the target area.
[0077] Figure 2 A schematic diagram of a deep neural network model is provided. According to... Figure 2 Geographic information such as topography, landforms, sunshine duration, and rainfall can be represented as X1, X2, ..., X... n The input is fed into the input layer of the deep neural network and then passed from the input layer to the first hidden layer (hidden layer 1). After a nonlinear transformation, the output is shown below. The output of the first hidden layer can be expressed as...
[0078] R1 = h(a1X + B1)
[0079] Where X represents a column vector of input data, for example, [X1, X2, ..., X...]. n ] T a1 represents the connection weight between the input layer neuron and the first hidden layer neuron, B1 represents the threshold between the input layer neuron and the first hidden layer, and h represents the nonlinear transformation function of the first hidden layer.
[0080] Each element in R1 is obtained by nonlinearly transforming the input through the excitation function f, and the transformation formula is as follows:
[0081]
[0082] Where, x i r represents the element in the i-th row of the input column vector X. 1,p This represents the p-th element in the output matrix of the first hidden layer, a. 1,p,i This represents the i-th element in the p-th row of the weight matrix between the input layer and the first hidden layer, b. 1,p This represents the p-th threshold in the threshold vector from the input layer to the first hidden layer.
[0083] After the input data is passed from the input layer to the first hidden layer, subsequent hidden layers (hidden layer 2, ..., hidden layer n) continuously learn and train while passing the data. Finally, the transformed data is sent to the output layer to obtain the final feature result y, which can be represented as...
[0084] y = c(a) n+1 R n +B n+1 (where a) n+1R represents the weight matrix between the nth hidden layer and the output layer. n B represents the vector of the nth hidden layer. n+1 Let represent the threshold vector between the nth hidden layer and the output layer, and c represent the non-linear activation function of the output layer of the deep neural network.
[0085] Step S130: Input the fused features of the location features and the geographic features into a pre-trained capsule neural network model to obtain the target features of the target area.
[0086] The target feature can be the feature output by the capsule neural network model after the fused features are input into it.
[0087] In practice, a capsule neural network model can be pre-trained to fuse location features and geographic features to obtain fused features. The fused features are then input into the trained capsule neural network model to obtain the target features of the target area output by the capsule neural network model.
[0088] Figure 3 A schematic diagram of a capsule neural network model is provided. According to... Figure 3 The capsule neural network consists of convolutional layers, a main capsule layer, and digital capsule layers. The convolutional layer uses a single 9×9 kernel to generate 256 feature maps for subsequent capsule analysis. The main capsule layer transforms the feature maps extracted from the convolutional layer into vector capsules. It uses eight sets of 9×9 kernels with a depth of 32 and a stride of 2 to perform eight convolution operations on the feature maps obtained from the convolutional layer, resulting in eight sets of 6×6×32 feature maps. These feature maps are then flattened to obtain 1152 vector neurons (1152×8, or 1152 capsules), each consisting of an 8-dimensional vector. The capsules between the main capsule layer and the digital capsule layer are iteratively updated using a dynamic routing method. The digital capsule layer outputs ten 16-dimensional vector capsules, which are obtained from the capsules generated by the convolutional operation in the main capsule layer through a dynamic routing method.
[0089] Step S140: Determine the target location of the photovoltaic power station from the target area based on the target characteristics.
[0090] The target location can be the site of a photovoltaic power station.
[0091] In practice, the target features correspond to the candidate locations in the target area. If the target features meet the preset conditions, the candidate locations corresponding to the target features can be determined as the target locations, and the target locations can be used as the sites for photovoltaic power plants.
[0092] For example, if 100 candidate locations are set in the target area, the capsule neural network model outputs the capsule vector corresponding to each candidate location as the target feature. The capsule vector with the largest modulus is selected, and the candidate location corresponding to the capsule vector is determined as the target location. It is recommended to build a photovoltaic power station at the target location.
[0093] The aforementioned graph feature network-based site selection method for photovoltaic power plants, which considers complex geographical factors, obtains a map of the target area and its corresponding geographical information. It then models the map using a graph neural network to obtain the location features of the target area. The geographical information is input into a pre-trained feature network model to obtain the geographical features of the target area. The fused features of the location and geographical features are input into a pre-trained capsule neural network model to obtain the target features of the target area. Based on these target features, the target location of the photovoltaic power plant is determined. This method can extract the location features of the target area from the map and the geographical features from the geographical information. The capsule neural network model is used to extract the fused features of the location and geographical features, comprehensively considering the geographical location and complex geographical factors of the target area for photovoltaic power plant site selection, thus improving the accuracy of photovoltaic power plant site selection.
[0094] In one embodiment, step S120 may specifically include: determining candidate locations and the positional relationships between candidate locations from the map; using candidate locations as nodes and determining the edges between nodes according to the positional relationships to obtain a network graph corresponding to the map; and obtaining the positional features of the target area based on the network graph.
[0095] Among them, the candidate location can be the site of a candidate photovoltaic power station.
[0096] The positional relationship can be the relative position and distance between candidate positions.
[0097] A network graph can be a graph that includes nodes and edges between nodes.
[0098] In practice, candidate locations for photovoltaic power stations are determined from the map, and the positional relationship between any two candidate locations is determined. The candidate locations are used as nodes, and the edges between the nodes are marked with positional relationships to obtain a network graph containing all candidate locations. The positional features of each candidate location in the target area can be extracted from this network graph.
[0099] In practical applications, the k candidate locations of a photovoltaic power station are represented as s1, s2, ..., s k Using candidate positions as nodes and the lines connecting them as edges, we can also calculate the positional relationship between any two candidate positions. This relationship includes relative position and distance. For example, we can calculate the distance between s1 and s2, obtaining l. 12It can also calculate the azimuth angle of s2 relative to s1, and obtain θ. 12 Then the positional relationship between s1 and s2 is (l 12 θ 12 ), and the positional relationship (l 12 θ 12 Mark the lines between the corresponding nodes s1 and s2, and perform the above processing on all nodes to obtain the network graph corresponding to the k candidate positions.
[0100] In this embodiment, candidate locations and their spatial relationships are determined from the map; candidate locations are used as nodes, and edges between nodes are determined based on their spatial relationships to obtain a network graph corresponding to the map; based on the network graph, the location features of the target area are obtained, and the candidate locations of the target area can be modeled using a graph feature network to extract the spatial information of the candidate locations, form a network graph, and then the network graph is processed, reducing the complexity of spatial information processing.
[0101] In one embodiment, the step of obtaining the location features of the target region based on the network graph may specifically include: determining the target node from the network graph; determining the hidden state of the target node based on the target node, the corresponding positional relationship of the target node, and the adjacent nodes of the target node; and inputting the hidden state into a predetermined local output model to obtain the location features of the target node.
[0102] In practice, any node in the network graph can be designated as the target node, denoted by x. v The characteristics of the target node (e.g., the geographic coordinates of the target node) are represented by x. co[v] The azimuth angle, represented by h, is the angle marked on all edges connected to the target node. ne[v] The distance, represented by x, is the distance along all edges connected to the target node. ne[v] Let f represent the characteristics of all neighboring nodes of the target node (e.g., the geographic coordinates of neighboring nodes), where the distance between neighboring nodes and the target node does not exceed a preset threshold. Then, the hidden state of the target node can be represented as follows:
[0103] h v =f(x) v x co[v] h ne[v] x ne[v] ).
[0104] You can also set a local output function g to output the hidden state h. v and the feature x of the target node v Input the local output function g to obtain the position features o of the target node. v The specific formula is as follows
[0105] o v =g(h v x v ).
[0106] The information aggregation function f and the local output function g can each be expressed by a neural network.
[0107] In this embodiment, the target node is determined from the network graph; the hidden state of the target node is determined based on the target node, the positional relationship of the target node, and the adjacent nodes of the target node; the hidden state is input into a pre-determined local output model to obtain the positional features of the target node. This can determine the positional features of each node in the network graph and improve the efficiency of positional feature extraction.
[0108] In one embodiment, the geographic information includes the topography, landforms, sunshine duration, and rainfall of the target area; step S120 may further include: identifying the input topography, landforms, sunshine duration, and rainfall based on a feature network model to obtain the geographic features of the target area.
[0109] Among them, the feature network model can be a deep neural network model.
[0110] In practice, the terrain data, landform data, and spatiotemporal distribution information of sunshine duration and rainfall corresponding to the candidate locations in the target area can be input in parallel into a pre-trained deep neural network model, and the output of the deep neural network model can be used as the geographical features of the candidate location.
[0111] For example, according to Figure 2 , will candidate position s i The terrain data X1, landform data X2, the spatiotemporal distribution of sunshine duration X3, and the spatiotemporal distribution of rainfall X4 are input into the input layer of a deep neural network, passed through n hidden layers, and the final output y can be used as the candidate location s. i Geographical features of the location.
[0112] It should be noted that other information besides terrain, landform, sunshine duration and rainfall can also be input into the feature network model, and this application does not impose any restrictions on this.
[0113] In this embodiment, the geographical features of the target area are obtained by identifying the input terrain, landforms, sunshine duration and rainfall based on a feature network model. This can comprehensively determine geographical features by integrating information from various aspects such as terrain, landforms, sunshine duration and rainfall, thereby improving the accuracy of geographical feature determination.
[0114] In one embodiment, after step S120, the method may further include: fusing location features and geographic features to obtain fused features of the target area.
[0115] In practice, the location features and geographic features of the candidate location can be fused to obtain the fused features of the candidate location. This process can be applied to all candidate locations in the target area to obtain the fused features of the target area.
[0116] In practical applications, the geographical features of candidate locations can be labeled on the network graph of the target area to obtain the fused features of the target area.
[0117] In this embodiment, by fusing location features and geographical features, the fused features of the target area are obtained. Geographical location can be integrated with information such as topography, landform, sunshine duration and rainfall. The fused information is used to comprehensively consider the site selection of photovoltaic power plants, thereby improving the accuracy of photovoltaic power plant site selection.
[0118] In one embodiment, step S130 may specifically include: extracting features from the fused features to obtain a feature map corresponding to the fused features; inputting the feature map into the capsule layer of the capsule neural network model to obtain a capsule vector corresponding to the feature map; and determining the target features of the target region based on the capsule vector.
[0119] In practice, the fused features of the candidate locations in the target region can be input into the convolutional layer of the capsule neural network. The convolutional layer extracts features from the fused features to obtain a feature map corresponding to the fused features. Then, the feature map is input into the main capsule layer of the capsule neural network to convert the feature map into a capsule vector. The digital capsule layer of the capsule neural network iteratively updates the capsule vector to obtain a capsule vector corresponding to the feature map. This capsule vector is then determined as the target feature of the candidate location in the target region.
[0120] In this embodiment, feature extraction is performed on the fused features to obtain the feature map corresponding to the fused features; the feature map is input into the capsule layer of the capsule neural network model to obtain the capsule vector corresponding to the feature map; based on the capsule vector, the target features of the target area are determined. The features of the candidate positions corresponding to the fused features can be determined by the capsule neural network. Since the fused features comprehensively consider information such as geographical location, terrain, landform, sunshine duration and rainfall, the accuracy of the candidate position feature determination is improved.
[0121] In one embodiment, the target area includes at least one candidate location, and each candidate location corresponds to a target feature; step S140 may specifically include: determining the candidate location as the target location of the photovoltaic power station if the target feature of the candidate location meets the preset conditions; and generating a site selection report for the photovoltaic power station based on the target location and the target feature of the target location.
[0122] The preset condition can be that the vector magnitude of the target feature is the largest.
[0123] The target location can be a recommended site for a photovoltaic power plant.
[0124] In practice, target features can be represented as vectors. For each candidate location in the target area, the amplitude of the corresponding target feature is calculated. The target feature with the largest amplitude is selected, and the candidate location corresponding to that target feature is determined as the target location. A photovoltaic power station is recommended to be built at this target location. A site selection report can also be generated based on the target location and its corresponding target features and displayed on the terminal display to explain why the photovoltaic power station is selected at the target location.
[0125] For example, a site selection report can be generated based on the target features corresponding to the target location. Since the target features take into account information such as geographical location, topography, landform, sunshine duration and rainfall, a detailed site selection report can be generated to explain the reasons for recommending the construction of a photovoltaic power station at the target location.
[0126] In this embodiment, if the target features of the candidate location meet the preset conditions, the candidate location is determined as the target location of the photovoltaic power station; based on the target location and the target features of the target location, a site selection report for the photovoltaic power station is generated. Since the site selection report can display the target location and explain the reasons for the site selection, the efficiency of the photovoltaic power station site selection can be improved.
[0127] To facilitate a deeper understanding of the embodiments of this application by those skilled in the art, a specific example will be used for illustration below.
[0128] Figure 4 A flowchart illustrating a graph feature network-based site selection method for photovoltaic power plants is provided. Figure 4 Combining graph neural networks, feature networks, and capsule neural networks for photovoltaic power plant site selection can improve the model's robustness and generalization ability, thereby enabling better site selection. Specifically, the graph feature network site selection method for photovoltaic power plants may include the following steps:
[0129] Step S210: Collect data from high-precision satellite maps and topographic data sources; at the same time, considering the complexity of geographical location factors, it is necessary to collect information on the spatial and temporal distribution characteristics of sunshine duration and rainfall.
[0130] Step S220: The collected data is modeled using a graph neural network. The geographical location information and feature information in the input data are represented as graphs, with nodes representing geographical locations and edges representing the relationships between geographical locations, thereby realizing feature extraction and abstraction of map data.
[0131] The graph is a data structure composed of nodes and edges. Nodes represent candidate locations for site selection, and edges represent the relative positions and distances between candidate locations. Graph neural networks capture the dependencies in the graph by establishing a message-passing mechanism between nodes, thereby obtaining the corresponding features on the graph. The goal of the graph neural network is to learn the hidden state h of each candidate location. v This hidden state is used to generate the output o of candidate node locations. v Define f v Let g be the information aggregation function for candidate node addresses. It operates on candidate locations of all location schemes and updates the information of candidate locations based on the input neighbor features. Define g as the local output function, then the hidden state h of each candidate location is... v and the output of candidate locations v The expression is:
[0132] h v =f(x) v x co[v] h ne[v] x ne[v] ),
[0133] In the formula, x v x co[v] h ne[v] x ne[v] These include the characteristics of candidate locations, the characteristics of the relative positions and distances between candidate locations, and the status and characteristics of candidate location v-neighbors;
[0134] Since we are looking for a unique solution, applying Banach's fixed-point theorem, we can rewrite the above equation as an iterative update process. The update expression is:
[0135] H t+1 =F(H t ,X),
[0136] In the formula, H t H is used to represent the result of H in the t-th iteration. t+1 Used to represent the result of H in the (t+1)th iteration;
[0137] The output o of the graph neural network is computed by passing the state and features to a local output function. v Output o v The expression is:
[0138] o v =g(h v x v ),
[0139] In the formula, h v For each hidden candidate location, x is the hidden state.v denoted by , where v is a feature of the candidate location; g is a local output function.
[0140] Step S230: To better process the feature information in the input data, topographic features and spatial-temporal feature distribution information are input into the feature network model for further processing and feature extraction. Key feature information is extracted, and the extracted features are fused with the features extracted from the graph neural network. All processed and feature-extracted data are used as the initial input to the deep neural network. The input is fed into the neural network, passed from the input layer to the first hidden layer, and output from the first hidden layer after a nonlinear transformation. The output of the first hidden layer is expressed as:
[0141] R1 = h(a1X + B1),
[0142] In the formula, X represents the column vector of input feature data, R1 refers to the output after the data reaches the first hidden layer from the input layer, a1 refers to the connection weight between the neurons in the input layer and the neurons in the first hidden layer, and B1 refers to the threshold between the neurons in the input layer and the first hidden layer.
[0143] Each element in R1 is obtained by nonlinearly transforming the input through the excitation function f, and the transformation formula is as follows:
[0144]
[0145] In the formula, x i This represents the element in the i-th row of the input feature data column vector X, r 1,p This refers to the p-th element in the output matrix of the first hidden layer, a 1,p,i This represents the i-th element in the p-th row of the weight matrix between the input layer and the first hidden layer, b 1,p This refers to the p-th threshold in the threshold vector from the input layer to the first hidden layer;
[0146] After the input data is passed from the input layer to the first hidden layer, subsequent hidden layers continuously learn and train while passing the data. Finally, the transformed data is transmitted to the output layer to obtain the final feature result y, which is represented as:
[0147] y = c(a) n+1 R n +B n+1 ),
[0148] In the formula, a n+1 This refers to the weight matrix between the nth hidden layer and the output layer, R. n B refers to the vector of the nth hidden layer. n+1...
[0149] Figure 2 A structural diagram of a deep neural network is provided. Figure 2 All processed and feature-extracted data are used as the initial inputs X1, X2, ..., X... of the deep neural network. n The feature is extracted and output as y after passing through hidden layer 1, hidden layer 2, ..., hidden layer n.
[0150] Step S240: The capsule neural network is used to further process the features output by the feature network, identify and represent the location of the photovoltaic power station, encode the node features, extract spatial distribution features, retain information at different levels and scales, aggregate and process the features to obtain more representative node features, improve the robustness and generalization ability of the model, and thus make better site selection.
[0151] The capsule neural network consists of a feature extraction part, a main capsule layer, and a digital capsule layer. The feature extraction part uses a single 9×9 convolutional layer to obtain 256 feature maps, which are used for subsequent capsule analysis. The main capsule layer transforms the feature maps extracted by the convolutional layer into vector capsules. The main capsule layer uses eight sets of 9×9 convolutional kernels with a depth of 32 and a stride of 2 to perform eight convolution operations on the feature maps obtained after the first convolution, resulting in eight sets of 6×6×32 feature maps. The feature maps are then flattened to obtain vector neurons of size 1152×8, or 1152 capsules, each consisting of an 8-dimensional vector. The capsules between the main capsule layer and the digital capsule layer are iteratively updated using a dynamic routing method. The third fully connected layer outputs ten 16-dimensional vector capsules, which are obtained from the capsules obtained after the convolution operation of the second main capsule layer through a dynamic routing method.
[0152] Figure 3 A structural diagram of a capsule neural network is provided. Figure 3 The input features are extracted through convolutional layers, main capsule layers, and digital capsule layers. The node features are encoded, spatial distribution features are extracted, information at different levels and scales is preserved, and the features are aggregated and processed to obtain more representative node features for better site selection.
[0153] Step S250: Use the trained model to select a site for the photovoltaic power station, provide the optimal site selection scheme, output the results to the user, and provide a detailed site selection report. The report includes the specific location of the site selection scheme and the reasons for the site selection, which can be used for reference and decision-making.
[0154] Traditional photovoltaic (PV) power plant site selection methods often only consider basic factors such as sunlight intensity and land utilization, neglecting the complexity of geographical location. The graph feature network (GNN) site selection method for PV power plants described above considers more geographical factors, such as topography, geographical location, and meteorological data, providing a more accurate basis for PV power plant site selection. Specifically, this method first collects relevant data required for PV power plant site selection, including high-precision satellite maps, topography, and spatial-temporal feature distribution information. Then, it preprocesses the collected data and uses a graph neural network to model the preprocessed data, representing the geographical location and feature information in the input data as a graph. Nodes represent geographical locations, and edges represent the relationships between geographical locations. Based on the graph neural network, a feature network is constructed to further extract and represent topography and spatial-temporal feature distribution information. Finally, based on the feature network, a capsule neural network is used to further process the features output by the feature network, identifying and representing the location of the PV power plant, retaining information at different levels and scales, obtaining more representative node features, further optimizing the results, improving the accuracy of the site selection scheme, and ultimately providing the optimal site selection scheme.
[0155] Moreover, by using graph feature networks to model geographic location factors, spatial information and relationships can be extracted, effectively solving the difficulties in geographic information processing in traditional methods.
[0156] Furthermore, by using capsule neural networks to extract image features, it has stronger recognition and classification capabilities and can better process image data.
[0157] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0158] Based on the same inventive concept, this application also provides a graph feature network site selection device for photovoltaic power plants considering complex geographical factors, which is used to implement the graph feature network site selection method for photovoltaic power plants considering complex geographical factors described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the graph feature network site selection device for photovoltaic power plants considering complex geographical factors provided below can be found in the limitations of the graph feature network site selection method for photovoltaic power plants considering complex geographical factors described above, and will not be repeated here.
[0159] In one embodiment, such as Figure 5 As shown, a graph feature network site selection device for photovoltaic power plants considering complex geographical factors is provided, including: an information acquisition module 310, a feature extraction module 320, a feature determination module 330, and a location determination module 340, wherein:
[0160] Information acquisition module 310 is used to acquire a map of the target area and the geographic information corresponding to the map;
[0161] The feature extraction module 320 is used to model the map based on a graph neural network to obtain the location features of the target area, and to input the geographic information into a pre-trained feature network model to obtain the geographic features of the target area.
[0162] The feature determination module 330 is used to input the fused features of the location features and the geographic features into a pre-trained capsule neural network model to obtain the target features of the target area;
[0163] The location determination module 340 is used to determine the target location of the photovoltaic power station from the target area based on the target characteristics.
[0164] In one embodiment, the feature extraction module 320 further includes:
[0165] A location determination module is used to determine candidate locations and the positional relationships between the candidate locations from the map;
[0166] A network graph generation module is used to take the candidate positions as nodes and determine the edges between the nodes according to the positional relationships to obtain the network graph corresponding to the map.
[0167] The graph feature extraction module is used to obtain the location features of the target region based on the network graph.
[0168] In one embodiment, the graph feature extraction module is further configured to determine a target node from the network graph; determine the hidden state of the target node based on the target node, the positional relationship corresponding to the target node, and the neighboring nodes of the target node; and input the hidden state into a predetermined local output model to obtain the positional features of the target node.
[0169] In one embodiment, the feature extraction module 320 is further configured to identify the input terrain, landform, sunshine duration and rainfall based on the feature network model to obtain the geographical features of the target area.
[0170] In one embodiment, the above-mentioned graph feature network site selection device for photovoltaic power plants that considers complex geographical factors further includes:
[0171] The feature fusion module is used to fuse the location features and the geographic features to obtain the fused features of the target area.
[0172] In one embodiment, the feature determination module 330 is further configured to extract features from the fused features to obtain a feature map corresponding to the fused features; input the feature map into the capsule layer of the capsule neural network model to obtain a capsule vector corresponding to the feature map; and determine the target features of the target region based on the capsule vector.
[0173] In one embodiment, the location determination module 340 is further configured to determine the candidate location as the target location of the photovoltaic power station if the target features of the candidate location meet preset conditions; and generate a site selection report for the photovoltaic power station based on the target location and the target features of the target location.
[0174] The modules in the graph feature network site selection device for photovoltaic power plants that considers complex geographical factors can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0175] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a graph feature network site selection method for photovoltaic power plants that considers complex geographical factors. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0176] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0177] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.
[0178] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.
[0179] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0180] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.
[0181] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.
[0182] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0183] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A graph feature network site selection method for photovoltaic power plants considering complex geographical factors, characterized in that, The method includes: Obtain a map of the target area and the corresponding geographic information of the map; The map is modeled using a graph neural network to obtain the location features of the target area. The geographic information is then input into a pre-trained feature network model to obtain the geographic features of the target area. The location features are the geographical features of the target area, represented as a network graph. Modeling the map using a graph neural network to obtain the location features of the target area includes: determining candidate locations and their relationships from the map; using the candidate locations as nodes; determining the edges between the nodes based on the relationships; obtaining the network graph corresponding to the map; and obtaining the location features of the target area based on the network graph. The feature network model is a deep neural network model. Topographic data, landform data, and the spatiotemporal distribution information of sunshine duration and rainfall are input into the trained deep neural network model for feature extraction to obtain the geographic features of the target area. Feature extraction is performed on the fusion features of the location features and the geographic features to obtain a feature map corresponding to the fusion features. The feature map is then input into the capsule layer of a pre-trained capsule neural network model to obtain a capsule vector corresponding to the feature map. Based on the capsule vector, the target features of the target region are determined. Based on the target characteristics, the target location of the photovoltaic power station is determined from the target area.
2. The method according to claim 1, characterized in that, The step of obtaining the location features of the target region based on the network graph includes: The target node is determined from the network graph; The hidden state of the target node is determined based on the target node, the positional relationship of the target node, and the adjacent nodes of the target node; The hidden state is input into a pre-determined local output model to obtain the positional features of the target node.
3. The method according to claim 1, characterized in that, The geographic information includes the topography, landforms, sunshine duration, and rainfall of the target area; the step of inputting the geographic information into a pre-trained feature network model to obtain the geographic features of the target area includes: Based on the feature network model, the input terrain, landform, sunshine duration and rainfall are identified to obtain the geographical features of the target area.
4. The method according to claim 1, characterized in that, After modeling the map based on a graph neural network to obtain the location features of the target area, and inputting the geographic information into a pre-trained feature network model to obtain the geographic features of the target area, the method further includes: The location features and the geographic features are fused to obtain the fused features of the target area.
5. The method according to claim 1, characterized in that, The target region includes at least one candidate location, and each candidate location corresponds to one of the target features; Determining the target location of the photovoltaic power station from the target area based on the target characteristics includes: If the target features of the candidate location meet the preset conditions, the candidate location is determined as the target location of the photovoltaic power station; A site selection report for the photovoltaic power station is generated based on the target location and the target features of the target location.
6. A graph feature network site selection device for photovoltaic power plants considering complex geographical factors, characterized in that, The device includes: The information acquisition module is used to acquire a map of the target area and the corresponding geographic information of the map; The feature extraction module is used to model the map based on a graph neural network to obtain the location features of the target area, and to input the geographic information into a pre-trained feature network model to obtain the geographic features of the target area; the location features are the features of the geographic location in the target area, represented as a network graph, and the feature network model is a deep neural network model; The feature extraction module is further configured to determine candidate locations and positional relationships between candidate locations from the map, use the candidate locations as nodes, determine the edges between the nodes according to the positional relationships, obtain a network graph corresponding to the map, and obtain the positional features of the target region according to the network graph. The feature extraction module is also used to input terrain data, landform data, and the spatiotemporal distribution information of sunshine time and rainfall into a trained deep neural network model to perform feature extraction and obtain the geographical features of the target area. The feature determination module is used to extract features from the fusion features of the location features and the geographic features to obtain a feature map corresponding to the fusion features, input the feature map into the capsule layer of a pre-trained capsule neural network model to obtain a capsule vector corresponding to the feature map, and determine the target features of the target region based on the capsule vector. The location determination module is used to determine the target location of the photovoltaic power station from the target area based on the target characteristics.
7. The apparatus according to claim 6, characterized in that, The feature extraction module is further configured to determine the target node from the network graph; determine the hidden state of the target node based on the target node, the positional relationship of the target node and the neighboring nodes of the target node; and input the hidden state into a pre-determined local output model to obtain the positional features of the target node.
8. The apparatus according to claim 6, characterized in that, The geographic information includes the topography, landforms, sunshine duration and rainfall of the target area. The feature extraction module is also used to identify the input topography, landforms, sunshine duration and rainfall based on the feature network model to obtain the geographic features of the target area.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.