Photovoltaic power distribution method based on spectral clustering and multi-modal stable matching

Abstract

The application discloses a photovoltaic power distribution method based on spectral clustering and multimodal stable matching, which is applied to the technical field of photovoltaic power generation and comprises the following steps: acquiring multi-criteria information of users and extracting feature data of the multi-criteria information; establishing a similarity matrix according to the feature data and generating an adjacency matrix; classifying users by means of spectral clustering and the adjacency matrix to obtain user classification results; establishing a first preference list of user groups and a second preference list of photovoltaic power stations according to the user classification results; matching the user groups and the photovoltaic power stations based on the first preference list, the second preference list and a multimodal stable matching algorithm, outputting a matching scheme and completing optimal matching of the photovoltaic power stations and the user groups.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of photovoltaic power generation, and in particular to a photovoltaic power distribution method based on spectral clustering and multi-modal stable matching. BACKGROUND

[0002] With the wide application of distributed energy, especially photovoltaic power generation systems, how to efficiently utilize photovoltaic power and realize reasonable matching with user loads has become an important problem in power system optimization. The intermittency and volatility of photovoltaic power generation, as well as the diversity and time-varying nature of user loads, increase the complexity of power supply and demand matching.

[0003] In existing power dispatching and load matching technologies, traditional methods usually rely on a single criterion for power distribution. This method is difficult to meet the matching requirements of photovoltaic power generation fluctuations and user load imbalance under complex multi-criteria conditions. Especially in the distributed photovoltaic power generation scenario, due to factors such as geographical dispersion and transmission loss, the matching mode of a single criterion often cannot effectively optimize the overall performance of the power system.

[0004] In order to overcome these defects, the application provides a photovoltaic power distribution method based on spectral clustering and multi-modal stable matching, which ensures efficient consumption of photovoltaic power, load balance and stability of the power system, and improves the overall energy utilization efficiency. SUMMARY

[0005] The purpose of the present application is to provide a photovoltaic power distribution method based on spectral clustering and multi-modal stable matching, aiming to solve the problems of traditional methods in dealing with the volatility of photovoltaic power generation and the diversity of user load demand.

[0006] To achieve the above purpose, the application provides the following technical solutions:

[0007] In a first aspect, the application provides a photovoltaic power distribution method based on spectral clustering and multi-modal stable matching, comprising:

[0008] Obtaining multi-criteria information of users, extracting feature data of the multi-criteria information;

[0009] Establishing a similarity matrix according to the feature data, and generating an adjacency matrix;

[0010] Classifying users through spectral clustering and the adjacency matrix to obtain user classification results;

[0011] According to the user classification results, establishing a first preference list of user groups and a second preference list of photovoltaic power stations;

[0012] Based on the first preference list, the second preference list and adopting a multi-modal stable matching algorithm, the user group and the photovoltaic power station are matched, a matching scheme is output, and optimal matching of the photovoltaic power station and the user group is completed.

[0013] In a second aspect, the present application provides a photovoltaic power distribution system based on spectral clustering and multi-modal stable matching, comprising:

[0014] The feature extraction module acquires multi-criteria information of the user and extracts feature data of the multi-criteria information.

[0015] The matrix establishment module establishes a similarity matrix according to the feature data and generates an adjacency matrix.

[0016] The clustering module classifies the user by spectral clustering and the adjacency matrix to obtain a user classification result.

[0017] The matching and output module establishes a first preference list of the user group and a second preference list of the photovoltaic power station according to the user classification result, matches the user group and the photovoltaic power station based on the first preference list, the second preference list and a multi-modal stable matching algorithm, outputs a matching scheme, and completes optimal matching of the photovoltaic power station and the user group.

[0018] In a third aspect, the present application provides a device, comprising a processor and a memory coupled to the processor, wherein the memory stores program instructions for implementing a photovoltaic power distribution method based on spectral clustering and multi-modal stable matching; and the processor is configured to execute the program instructions stored in the memory to implement photovoltaic power distribution based on spectral clustering and multi-modal stable matching.

[0019] In a fourth aspect, the present application provides a storage medium storing program instructions executable by a processor, wherein the program instructions are used to execute a photovoltaic power distribution method based on spectral clustering and multi-modal stable matching.

[0020] The present application provides a photovoltaic power distribution method based on spectral clustering and multi-modal stable matching, which has the following advantages:

[0021] (1) Improving the global stability of the matching result: by introducing a multi-modal stable matching algorithm, the matching result of the photovoltaic power station and the user group under multiple criteria is ensured to be consistent and stable, avoiding the instability problem caused by single standard matching in the prior art. Especially when facing complex photovoltaic power generation fluctuations, the matching result is more robust and less susceptible to external interference.

[0022] (2) Multi-criteria optimization, improve matching accuracy: This application can handle multiple matching criteria at the same time, such as user load pattern, photovoltaic power generation capacity, geographical distance, etc., no longer limited to single power generation capacity or distance calculation; By considering these criteria comprehensively, the matching process can more accurately allocate photovoltaic power resources, ensuring that the demand and supply between each photovoltaic power station and the user group reach the best balance, adapting to the complex scenarios in distributed energy management;

[0023] (3) Strong adaptability and high robustness: Compared with traditional matching methods based on single criteria, this application can still maintain high robustness when facing changes in load demand and fluctuations in photovoltaic power generation. The matching result is not sensitive to changes in external conditions and has high consistency; Regardless of fluctuations in user load and photovoltaic power generation, this application can adaptively adjust to ensure stable operation of the system;

[0024] (4) Reduce power transmission loss and improve energy utilization: Through multi-criteria matching optimization, while ensuring supply and demand balance, it also maximizes the reduction of long-distance transmission loss of photovoltaic power; Preferentially allocate power generation capacity to user groups with shorter geographical distance, significantly improving energy utilization and reducing the operating cost of the power system;

[0025] (5) Improve economic benefits and resource efficiency: Through accurate matching and optimization process, unnecessary power waste and resource mismatch caused by unreasonable matching are reduced; For large-scale distributed photovoltaic power stations and complex user groups, this application can achieve optimal matching in a short time, reduce scheduling cost, and improve economic benefits. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a flowchart of the photovoltaic power distribution method based on spectral clustering and multi-modal stable matching of embodiment 1 of the application;

[0027] Figure 2 is a spectral clustering effect diagram using different numbers of feature vectors for the instance of the meta-graph structure C8 of embodiment 1 of the application;

[0028] Figure 3 is a spectral clustering effect diagram using different numbers of feature vectors for the instance of the meta-graph structure P8 of embodiment 1 of the application;

[0029] Figure 4 is a flowchart of the multi-modal stable matching algorithm executed by embodiment 1 of the application;

[0030] Figure 5 is a structural diagram of the photovoltaic power distribution system based on spectral clustering and multi-modal stable matching of embodiment 2 of the application;

[0031] Figure 6This is a schematic diagram of the device structure in Embodiment 3 of this application;

[0032] Figure 7 This is a schematic diagram of the storage medium structure of Embodiment 4 of this application. Detailed Implementation

[0033] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0034] The following analysis, based on relevant technologies, examines existing solutions.

[0035] To better achieve photovoltaic (PV) power generation absorption and supply-demand balance in complex power networks, researchers have gradually explored load matching methods based on multi-criteria optimization and clustering. These methods aim to achieve more accurate and stable power matching by considering multiple key factors, such as user load patterns, geographical location, and PV power plant capacity. However, existing methods still face challenges such as unstable matching results, insufficient flexibility, and difficulty in adapting to dynamic supply and demand changes.

[0036] Existing power load clustering technologies: Currently, research on power load clustering and matching mainly focuses on single clustering methods and simple load forecasting-based scheduling. Commonly used clustering algorithms in existing technologies include k-means, hierarchical clustering, and density-based DBSCAN. Although these methods perform well in traditional load analysis, their application has many shortcomings when facing the multi-dimensional characteristics of photovoltaic power generation and load fluctuations.

[0037] (1) k-means clustering algorithm: k-means is a common clustering algorithm based on Euclidean distance, which classifies data by minimizing the distance between the cluster center and the data point. However, k-means only clusters based on a single criterion and lacks support for multiple criteria. This means that in complex load scenarios, the clustering results may not be able to effectively reflect the diversity of user load patterns, especially when photovoltaic power generation fluctuates greatly, and it cannot adapt to the changes in demand of different user loads in real time. In addition, the k-means algorithm is sensitive to the initial cluster center and is prone to getting trapped in local optima, making it unsuitable for processing load data with irregular distribution [1].

[0038] (2) Hierarchical clustering algorithm: Hierarchical clustering recursively divides the dataset into multiple subsets to form a tree-structured clustering result. Although hierarchical clustering can adapt to load data of different scales and is not sensitive to initial conditions, its computational complexity is high when dealing with large-scale high-dimensional data, making it difficult to adapt to the real-time and flexibility required for photovoltaic power generation and load matching [2]. At the same time, hierarchical clustering performs poorly in multi-criteria environments and cannot balance the constraints of multiple different criteria.

[0039] (3) DBSCAN clustering algorithm: DBSCAN is a density-based clustering algorithm that can discover clusters of arbitrary shapes and distinguish noisy data points. It has advantages in processing spatial data and can identify high-density areas. However, DBSCAN is highly dependent on parameters, especially under multi-criteria matching conditions, it is difficult to simultaneously consider the power generation characteristics, geographical location and user load demand of photovoltaic power plants, resulting in limited application effect in photovoltaic power generation scenarios [3].

[0040] Challenges of multi-criteria load matching in existing technologies: Some studies have attempted to achieve load matching through weighted multi-criteria optimization methods, such as weighting power generation capacity, load demand and transmission distance to derive matching schemes. However, these weighting methods often rely on manually set weights and lack global adaptive capabilities. In the actual scenario of photovoltaic power generation, this approach cannot ensure the globally optimal matching result under multi-dimensional conditions. Especially when facing changes in power generation and load, it is easy to cause unstable matching results, affecting the stability and economy of the power system [4].

[0041] Introduction of Multimodal Stable Matching: In recent years, the multimodal stable matching problem has evolved from the classic marriage matching problem (Gale-Shapley algorithm) and has been widely used in resource allocation, task scheduling and other fields. The core advantage of the multimodal stable matching model is that it can achieve a globally stable matching result under multiple preferences or multiple criteria, avoiding the occurrence of local optima [5]. By combining this algorithm with photovoltaic power load matching, a globally optimal and stable matching scheme can be found based on multiple dimensions such as power generation, load demand, and geographical location. For example, in the power dispatching system, photovoltaic power stations can be matched with different user groups in multiple dimensions to ensure that each photovoltaic power station can supply power to the most suitable load area, reduce transmission loss, and optimize the utilization rate of photovoltaic power generation. This application classifies the load by spectral clustering and combines multimodal stable matching technology to realize supply and demand matching under multiple criteria, so as to improve the global stability and energy utilization efficiency of the system.

[0042] [1]Hartigan, JA, & Wong, MA (1979). Algorithm AS136: A K-meansclustering algorithm. Journal of the Royal Statistical Society: Series C (Applied Statistics), 28 (1), 100-108.

[0043] [2]Ester, M., Kriegel, HP, Sander, J., & Xu,

[0044] [3]Kaufman,L.,&Rousseeuw,PJ(2009).Finding Groups in Data: An Introduction to Cluster Analysis.John Wiley&Sons.

[0045] [4]Gale, D., & Shapley, LS (1962). College admissions and the stability of marriage. The American Mathematical Monthly, 69 (1), 9-15.

[0046] [5]Chen, Jiehua, Rolf Niedermeier, and Piotr Skowron. "Stable marriage with multi-modal preferences." Proceedings of the 2018ACM Conference on Economics and Computation. 2018.

[0047] In summary, existing photovoltaic power plant and user load matching and clustering technologies primarily rely on single criteria for load classification and matching, such as load forecasting-based scheduling methods, k-means clustering, and geographical distance-based matching methods. However, these methods suffer from several shortcomings when considering the volatility of distributed photovoltaic power generation, the diversity of user loads, and the complexity of geographical distribution. Especially under multi-criteria conditions, they struggle to guarantee the global stability, supply-demand balance, and efficient absorption of clustering and matching results. Specific drawbacks include:

[0048] (1) Limitations of a single criterion: Traditional clustering and matching algorithms are mostly based on a single criterion, which cannot handle complex multi-dimensional and multi-criteria data. For example, user load is not only affected by electricity demand, but also closely related to changes in geographical location and load patterns. Clustering and matching under a single criterion are prone to ignoring other key factors, leading to imbalances in electricity supply and demand and unreasonable matching.

[0049] (2) Lack of multimodal support: Faced with distributed photovoltaic power generation and diverse user loads, existing clustering and matching methods lack effective mechanisms to integrate information from multiple dimensions or modes. The fluctuations in photovoltaic power generation, geographical distribution, and user load patterns are complex and varied. Traditional methods cannot fully utilize the correlation between various criteria, resulting in unstable clustering and matching results and difficulty in simultaneously optimizing the performance of multiple dimensions.

[0050] (3) Unstable clustering and matching results: Existing clustering and matching algorithms are sensitive to initial conditions and are prone to getting trapped in local optima, leading to inconsistent clustering or matching results. The volatility of photovoltaic power generation further exacerbates the problem, making it difficult for the system to guarantee the stability and robustness of clustering and matching under multiple criteria.

[0051] (4) Shortcomings of multi-criteria optimization: Existing technologies address the matching and clustering of photovoltaic power generation and user loads through weighted or parameter-based methods, but these rely on manually set weights and lack adaptability and global optimality. As the scale and complexity of the power grid increase, the efficiency, stability, and accuracy of clustering and matching under multiple criteria of existing technologies decrease significantly.

[0052] To address the shortcomings of existing technologies, this application introduces spectral clustering and multimodal stable matching techniques to optimize the clustering and matching process between photovoltaic power generation and user areas under multiple criteria, ensuring the global consistency of clustering results and the stability and supply-demand balance of matching.

[0053] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0054] Example 1

[0055] Please see Figure 1 This is a flowchart illustrating the photovoltaic power distribution method based on spectral clustering and multimodal stable matching according to Embodiment 1 of this application; the steps include:

[0056] S1: Obtain the user's multi-criteria information and extract the feature data of the multi-criteria information.

[0057] In this embodiment, the feature data includes, but is not limited to, user demand, load fluctuations, and geographical location.

[0058] S2: Establish a similarity matrix based on the feature data and generate an adjacency matrix.

[0059] In this embodiment, the feature data is used to calculate similarity values ​​using distance metrics such as Euclidean distance and cosine similarity, to establish an N×N similarity matrix A; where N is the number of users, and the elements a of the similarity matrix are... ij Let be the similarity between user i and user j.

[0060] The similarity matrix A is converted into an adjacency matrix, where users with the highest similarity are connected by edges to form a graph structure.

[0061] S3: Classify users using spectral clustering and the adjacency matrix to obtain user classification results.

[0062] In this embodiment, spectral clustering is used to classify users. Spectral clustering is a graph theory-based method that identifies the global structure in the data by decomposing the eigenvalues ​​of the similarity matrix. Specific steps include:

[0063] Construct the graph Laplacian matrix L based on the adjacency matrix.

[0064] The graph Laplacian matrix L is subjected to eigenvalue decomposition, and the top k smallest eigenvectors are extracted to form a user embedding in a low-dimensional space.

[0065] The feature vectors are clustered using a clustering algorithm, dividing users into several categories. Each category includes user groups with similar load characteristics, thus generating user classification results. Each user group contains users with x similar load characteristics, laying the foundation for subsequent matching with photovoltaic power plants.

[0066] Please see Figure 2 This is a schematic diagram of the spectral clustering effect using different numbers of feature vectors for an instance of metagraph structure C8 in Embodiment 1 of this application; Figure 3 This is a schematic diagram of the spectral clustering effect using different numbers of feature vectors for an instance of metagraph structure P8 in Embodiment 1 of this application.

[0067] For instances with special inter-class structures in optimal clustering, using a smaller number of feature vectors can not only reduce computational overhead but also improve clustering accuracy. Figure 2 and Figure 3This section showcases spectral clustering results for C8 and P8 metagraph structures, randomly generated based on a stochastic block model, using different numbers of feature vectors. Here, p represents the probability of adding edges within clusters, q represents the probability of adding edges between clusters, and t represents the number of feature vectors used. It can be seen that, within a specific metagraph structure, spectral clustering using fewer feature vectors yields better clustering results. Figure 2 In the above, the probability of adding an edge within a cluster is lowest when t=3; the probability of adding an edge within a cluster is highest when t=8. Figure 3 In the above, the probability of adding an edge within a cluster is lowest when t=5, and highest when t=8.

[0068] S4: Based on the user classification results, establish a first preference list for user groups and a second preference list for photovoltaic power plants.

[0069] In this embodiment, after obtaining the user classification results, preference lists based on different criteria are constructed for user groups and photovoltaic power plants respectively; specifically including:

[0070] Establish a primary preference list for the user group: based on the user group's total load demand, including daily peak electricity consumption and fluctuations; the distance between the user group and the photovoltaic power station, which directly affects power transmission losses and efficiency; and the power generation of the photovoltaic power station.

[0071] Establish a second preference list for photovoltaic power plants: select user groups whose power generation capacity matches their load demand, prioritizing groups with a high degree of matching with their own power generation capacity; prioritize the nearest user groups based on geographical distance to reduce power transmission losses; and prioritize user groups with the smallest load fluctuations to improve power supply stability.

[0072] Based on the above criteria, a first preference list for user groups and a second preference list for photovoltaic power plants are generated for use in the subsequent matching process.

[0073] S5: Based on the first preference list, the second preference list, and the multimodal stable matching algorithm, match the user group with the photovoltaic power station, output the matching scheme, and complete the optimal matching between the photovoltaic power station and the user group.

[0074] In this embodiment, after generating the preference list, a multimodal stable matching algorithm is used to optimize the matching process under multiple criteria, achieving global matching between user groups and photovoltaic power plants. Specific steps include:

[0075] Construct constraints for the multimodal stable matching problem that satisfy:

[0076]

[0077] Where, x ij This indicates whether participant i matches participant j, where i and j both belong to participant set A; if i matches j, then x ij =1; otherwise x ij =0; C is the set of preference criteria; L(j|i,k) is the set of options in the preference list of i that are ranked lower than j under criterion k; where participants refer to users who participate in the matching.

[0078] Construct the corresponding set of bad events, that is, the events that are opposite to the events represented by the constraints of the original problem;

[0079] Perform random paired sampling on the constraint variables and determine whether a bad event has occurred; if a bad event has occurred, resample.

[0080] The probability of correct sampling is analyzed using Lovász's local lemma; by analyzing the probability of bad events, the number of sampling rounds required for the algorithm to obtain the optimal solution after random sampling of variables is derived.

[0081] Please see Figure 4 This is a flowchart illustrating the execution of the multimodal stable matching algorithm in Embodiment 1 of this application. The multimodal stable matching algorithm guarantees that the desired match will be obtained with a certain probability within polynomial time. If no variable assignment satisfies the preset conditions within a preset time threshold, the variable assigned a value of 1 and appearing most frequently among all violated constraints is assigned a value of 0, until no violated constraints remain; the final variable assignment is then converted into the corresponding match.

[0082] Furthermore, for unmatched nodes, each node regenerates a new preference list based on the existing matching situation, and then uses the Gale-Shapley algorithm to match the unmatched nodes. Specifically, this involves: the user group sending matching requests to the photovoltaic power station sequentially according to the first preference list, prioritizing the photovoltaic power station that best meets their needs; the photovoltaic power station then selects from the user group's requests according to the second preference list, choosing the user group whose load demand best matches their power generation capacity and geographical distance. Through the stable matching mechanism of the Gale-Shapley algorithm, the matching between the user group and the photovoltaic power station is optimized until a preset stable matching state is reached, thus achieving the optimal matching between the photovoltaic power station and the user group.

[0083] This stable matching process ensures the matching stability between photovoltaic power plants and user groups under multiple criteria, avoiding local optima and maximizing the absorption of photovoltaic power and supply-demand balance. The final matching scheme ensures efficient absorption of photovoltaic power generation, reduces power transmission losses, and maintains dynamic balance of user loads.

[0084] In summary, Embodiment 1 of this application combines spectral clustering and multimodal stable matching techniques to optimize the matching relationship between photovoltaic power generation and user load. Through spectral clustering, users are accurately classified to form user groups with similar load demands. Through multimodal stable matching, an optimized match between the photovoltaic power station and user groups is achieved by comprehensively considering multiple criteria such as user load, photovoltaic power station generation capacity, and geographical distance. This application not only improves the global stability and accuracy of matching but also reduces losses during power transmission, realizing precise matching and efficient scheduling in distributed photovoltaic power generation systems. It is suitable for complex load management and scheduling scenarios in distributed photovoltaic power generation.

[0085] Example 2

[0086] Please see Figure 5 This is a schematic diagram of the photovoltaic power distribution system based on spectral clustering and multimodal stable matching according to Embodiment 2 of this application; the specific content includes:

[0087] Feature extraction module: Acquires multi-criteria information of the user and extracts feature data from the multi-criteria information;

[0088] Matrix building module: Builds a similarity matrix based on the feature data and generates an adjacency matrix;

[0089] Clustering module: Classifies users using spectral clustering and the adjacency matrix to obtain user classification results;

[0090] Matching and Output Module: Based on the user classification results, a first preference list for the user group and a second preference list for the photovoltaic power station are established; based on the first preference list, the second preference list, and a multimodal stable matching algorithm, the user group and the photovoltaic power station are matched, and a matching scheme is output to achieve the optimal matching between the photovoltaic power station and the user group.

[0091] Example 3

[0092] Please see Figure 6 This is a schematic diagram of the device structure in Embodiment 3 of this application. The device 50 includes a processor 51 and a memory 52 coupled to the processor 51.

[0093] The memory 52 stores program instructions for implementing the photovoltaic power distribution method based on spectral clustering and multimodal stable matching described above.

[0094] The processor 51 is used to execute program instructions stored in the memory 52 to realize photovoltaic power distribution based on spectral clustering and multimodal stable matching.

[0095] The processor 51 can also be referred to as a CPU (Central Processing Unit).

[0096] Processor 51 may be an integrated circuit chip with signal processing capabilities. Processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor.

[0097] Example 4

[0098] Please see Figure 7 This is a schematic diagram of the storage medium in Embodiment 4 of this application. The storage medium in this embodiment stores a program file 61 capable of implementing all the above methods. This program file 61 can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or devices such as computers, servers, mobile phones, and tablets.

[0099] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0100] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

[0101] Although embodiments of this application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the appended claims and their equivalents.

[0102] Of course, the present invention may have many other embodiments. Based on this embodiment, other embodiments obtained by those skilled in the art without any creative effort are all within the scope of protection of the present invention.

Claims

1. A photovoltaic power distribution method based on spectral clustering and multimodal stable matching, characterized in that, include: Obtain multi-criteria information from the user and extract feature data from the multi-criteria information; A similarity matrix is ​​established based on the feature data, and an adjacency matrix is ​​generated. Users are classified using spectral clustering and the adjacency matrix to obtain user classification results; Based on the user classification results, establish a first preference list for user groups and a second preference list for photovoltaic power plants; Based on the first preference list, the second preference list, and a multimodal stable matching algorithm, the user group and the photovoltaic power station are matched, a matching scheme is output, and the optimal matching between the photovoltaic power station and the user group is achieved. The step of classifying users using spectral clustering and the adjacency matrix to obtain user classification results specifically includes the following steps: Construct the graph Laplacian matrix based on the adjacency matrix. L ; For the graph Laplace matrix L Perform eigenvalue decomposition and extract the eigenvalues ​​before extraction. k The smallest feature vector is used to form the user embedding in a low-dimensional space; The feature vectors are clustered using a clustering algorithm to divide users into several categories. Each category includes a group of users with similar workload characteristics, thus generating user classification results. The steps of matching the user group and the photovoltaic power station based on the first preference list, the second preference list, and using a multimodal stable matching algorithm to output a matching scheme and achieve optimal matching between the photovoltaic power station and the user group specifically include the following steps: In the multimodal stable matching algorithm, constraints are constructed for the multimodal stable matching problem to satisfy: ， in, Indicates participants i Whether with participants j Matching, where i and j All belong to the participant set A ;like i and j If matched, then ;otherwise ; C A set of preference criteria; In order to be in the standard k Down, i Ranked in the preference list j The later set of options; Construct the corresponding set of bad events, that is, the events that are opposite to the events represented by the constraints of the original problem; Perform random paired sampling on the constraint variables and determine whether a bad event has occurred; if a bad event has occurred, resample. The probability of correct sampling is analyzed using Lovász's local lemma. If no variable is assigned a value that meets the preset conditions within the preset time threshold, the variable that is assigned a value of 1 and appears most frequently among all violated constraints will be assigned a value of 0, until there are no more violated constraints; the final variable assignment will be converted into the corresponding match. For unmatched nodes, a new preference list is generated, and the Gale-Shapley algorithm is used to match the unmatched nodes.

2. The photovoltaic power distribution method based on spectral clustering and multimodal stable matching according to claim 1, characterized in that, The steps of establishing a similarity matrix and generating an adjacency matrix based on the feature data specifically include the following steps: The feature data is used to calculate similarity values ​​using distance metrics, and an N×N similarity matrix A is established; where N is the number of users, and the elements of the similarity matrix are... For users With users The similarity between them; The similarity matrix A is converted into an adjacency matrix, where users with the highest similarity are connected by edges to form a graph structure.

3. The photovoltaic power distribution method based on spectral clustering and multimodal stable matching according to claim 1, characterized in that, The steps of establishing a first preference list for user groups and a second preference list for photovoltaic power plants based on the user classification results specifically include the following steps: Based on the total load demand of the user group, the distance between the user group and the photovoltaic power station, and the power generation of the photovoltaic power station, a first preference list of the user group is established. Establish a second preference list for photovoltaic power plants: select user groups whose power generation capacity matches their load demand; prioritize the closest user groups based on their geographical distance; and prioritize user groups with the smallest load fluctuations.

4. The photovoltaic power distribution method based on spectral clustering and multimodal stable matching according to claim 1, characterized in that, For unmatched nodes, a new preference list is generated. The step of matching unmatched nodes using the Gale-Shapley algorithm specifically includes the following steps: The matching between the user group and the photovoltaic power station is optimized by the Gale-Shapley algorithm until a preset stable matching state is reached, thus achieving the optimal matching between the photovoltaic power station and the user group.

5. A system applying the photovoltaic power distribution method based on spectral clustering and multimodal stable matching as described in claim 1, characterized in that, include: Feature extraction module: Acquires multi-criteria information of the user and extracts feature data from the multi-criteria information; Matrix building module: Builds a similarity matrix based on the feature data and generates an adjacency matrix; Clustering module: Classifies users using spectral clustering and the adjacency matrix to obtain user classification results; Matching and Output Module: Based on the user classification results, establish a first preference list for user groups and a second preference list for photovoltaic power plants; Based on the first preference list, the second preference list, and a multimodal stable matching algorithm, the user group and the photovoltaic power station are matched, and a matching scheme is output to achieve the optimal matching between the photovoltaic power station and the user group.

6. A device, characterized in that, The device includes a processor and a memory coupled to the processor, wherein the memory stores program instructions for implementing the photovoltaic power distribution method based on spectral clustering and multimodal stable matching as described in any one of claims 1-4; the processor is used to execute the program instructions stored in the memory to implement photovoltaic power distribution based on spectral clustering and multimodal stable matching.

7. A storage medium, characterized in that, The device stores processor-executable program instructions for performing the photovoltaic power distribution method based on spectral clustering and multimodal stable matching as described in any one of claims 1-4.

Images