Photovoltaic power station equivalent modeling method, device and equipment based on C-AGNES and medium

Abstract

The invention discloses a photovoltaic power station equivalent modeling method and device based on C-AGNES, equipment and a medium, and relates to the field of photovoltaic power generation. The method aims at solving the problems that an existing equivalent modeling method neglects the influence of irradiance and a current collection system on a simulation result, and the dynamic characteristics of an equivalent model and a detailed model are different. According to the method, each sample in a sample set is scanned based on a density peak strategy, then initial cluster division is performed on each sample in the sample set, and the samples are irradiance and line equivalent impedance of each photovoltaic unit of which dimension difference is eliminated in a photovoltaic power station; clustering the result of the initial cluster division by using an AGNES algorithm; and performing equivalent calculation on parameters of the photovoltaic units in each cluster after clustering by adopting a weighting method, and connecting equivalent power generation units in the same cluster in parallel to form an equivalent model of the cluster.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of photovoltaic power generation, and particularly relates to equivalent modeling of photovoltaic power stations. BACKGROUND

[0002] With the policy of developing clean and renewable energy sources, the amount of photovoltaic power generation has increased rapidly, and unlike the modeling of traditional thermal power generating units, the photovoltaic components, inverters, unit transformers and other elements in photovoltaic power stations are numerous in quantity and various in type, and the photovoltaic output is obviously affected by natural conditions such as irradiance and temperature. The detailed model simulation time is long and the calculation efficiency is not high while retaining the dynamic characteristics. Through equivalent modeling, the model size can be significantly reduced while retaining the key dynamic characteristics, and the simulation efficiency can be improved to meet the timeliness requirements of power grid planning, operation and fault analysis.

[0003] In the field of equivalent modeling of photovoltaic power generation systems, there have been some related researches at home and abroad. At present, the more common equivalent method for photovoltaic power generation systems is the single-machine equivalent method, that is, a number of photovoltaic power generation units are equivalent to one photovoltaic power generation unit, but this method ignores the differences in internal parameters, topological structures and other factors when simplifying the equivalent, which may cause differences in dynamic characteristics between the equivalent model and the detailed model.

[0004] With the large-scale access of distributed photovoltaic to distribution networks, the dynamic response characteristics of power systems are affected, and the establishment of an equivalent aggregation model of photovoltaic power stations is the key to accurately studying new power systems. In photovoltaic grid-connected analysis, photovoltaic power stations are usually equivalent to unit models. However, such modeling, although convenient to implement, ignores the influence of irradiance, power collection system and other physical output characteristics and electrical coupling characteristics on the simulation results. SUMMARY

[0005] The present application is to solve the problem that the existing equivalent modeling method in the photovoltaic power generation system ignores the influence of irradiance and power collection system on the simulation results, and there are differences in dynamic characteristics between the equivalent model and the detailed model, and now provides a C-AGNES algorithm based on the Canopy algorithm, which innovatively proposes a dual-dimensional clustering index that combines physical output characteristics and electrical coupling characteristics, combines irradiance which directly affects the power output of photovoltaic units and line equivalent impedance which reflects the loss characteristics of the power collection system, and more comprehensively and accurately represents the actual operating state of photovoltaic units.

[0006] The first aspect of the present application provides a photovoltaic power station equivalent modeling method based on C-AGNES, comprising:

[0007] Scanning each sample in the sample set based on a density peak strategy, and then performing initial cluster division on each sample in the sample set, the sample being irradiance and line equivalent impedance of each photovoltaic unit in the photovoltaic power station after eliminating dimensional differences;

[0008] Clustering the result of the initial cluster division by using an AGNES algorithm;

[0009] Performing equivalent calculation on the parameters of the photovoltaic units in each cluster after clustering by using a weighting method, connecting the equivalent power generation units in the same cluster in parallel to form an equivalent model of the cluster.

[0010] In one possible design, the scanning each sample in the sample set based on a density peak strategy, and then performing initial cluster division on each sample in the sample set includes:

[0011] Selecting a sample in the sample set as a clustering center, and calculating a standardized Euclidean distance between the sample and a sample in the sample set , , , , , , ;

[0012] Determining a relationship between the standardized Euclidean distance and a loose threshold value and a close threshold value , if , marking the sample as "close" and dividing the sample into a cluster with the sample as the clustering center, if , marking the sample as "loose", and if , taking the sample as a new clustering center and constructing a new cluster for the new clustering center; ,

[0013] Traversing each sample in the sample set until all samples are marked as "close", and completing the initial cluster division.

[0014] In one possible design, the clustering the result of the initial cluster division by using an AGNES algorithm includes:

[0015] Calculating a physical information merging error between two clusters in the result of the initial cluster division, merging the two clusters with the minimum physical information merging error value, and repeating the merging process until the total number of clusters reaches a preset value.

[0016] ​​​​​​In one possible design, the computing the physical information merge error between each pair of clusters in the result of the initial cluster partition includes:

[0017] The physical information merge error between each pair of clusters is computed by:

[0018] ,

[0019] wherein, denotes the physical information merge error between cluster and cluster , is the sum of squared differences distance between cluster and cluster , is the electrical distance constraint term between cluster and cluster , is the topological connectivity constraint term between cluster and cluster , , and are the weights of , and respectively.

[0020] In one possible design, the sum of squared differences distance between cluster and cluster is expressed as:

[0021] ,

[0022] wherein, and denote the centers of cluster and cluster respectively, denotes the squared Euclidean distance between the two cluster centers, and denote the sample numbers of cluster and cluster respectively.

[0023] In one possible design, the electrical distance constraint term between cluster and cluster is expressed as:

[0024] ,

[0025] wherein, is the electrical distance penalty coefficient, ​​is a compactness weight coefficient, is an electrical distance threshold, is a cluster is an electrical separation between clusters , is an electrical compactness between clusters , ,

[0026] ,

[0027] ,

[0028] and are the number of samples of clusters and , is the electrical distance between photovoltaic units and : , and are the self-impedances of photovoltaic units and , is the mutual impedance between photovoltaic units and , and are the reference resistance and reactance, respectively.

[0029] In one possible design, the expression of the topological connectivity constraint term between clusters and is:

[0030] ,

[0031] where is a topological connectivity weight coefficient, is a topological connectivity threshold, is the topological diameter of the merged cluster, is the average length of the collection line of the power plant, is the topological connectivity score between clusters and ,

[0032] , is the number of samples of clusters , is the connectivity matrix between clusters and :

[0033] ​ .

[0034] The second aspect of the present application provides a C-AGNES-based photovoltaic power station equivalent modeling device, comprising:

[0035] An initial cluster division unit: based on a density peak value strategy, each sample in a sample set is scanned, and then each sample in the sample set is subjected to initial cluster division, the sample being irradiance and line equivalent impedance of each photovoltaic unit in the photovoltaic power station after eliminating dimensional differences;

[0036] A clustering unit: the AGNES algorithm is used to cluster the result of the initial cluster division;

[0037] A modeling unit: a weighted method is used to perform equivalent calculation on the parameters of the photovoltaic units in each cluster after clustering, the equivalent power generation units in the same cluster are connected in parallel, and the equivalent model of the cluster is constructed.

[0038] The third aspect of the present application provides a C-AGNES-based photovoltaic power station equivalent modeling device, comprising a processor and a memory, and at least one instruction is stored in the memory, the at least one instruction is loaded and executed by the processor to realize the C-AGNES-based photovoltaic power station equivalent modeling method as described above.

[0039] The fourth aspect of the present application provides a computer storage medium, and at least one instruction is stored in the computer storage medium, the at least one instruction is loaded and executed by the processor to realize the C-AGNES-based photovoltaic power station equivalent modeling method as described above.

[0040] The beneficial effects of the present application are:

[0041] 1. Irradiance (reflecting power characteristics) and line equivalent impedance (reflecting electrical loss and coupling characteristics) are combined as a two-dimensional clustering index, which more comprehensively represents the real operating state of the photovoltaic unit, and makes the equivalent model more close to the detailed model in dynamic response.

[0042] 2. The Canopy algorithm quickly scans the samples through the loose / tight double threshold, generates the initial cluster center and boundary, and significantly reduces the initial dependence and calculation complexity of the AGNES algorithm. The AGNES algorithm adopts a bottom-up merging strategy, avoids the problems of sensitivity to initial center and hard division of algorithms such as K-Means, and can adaptively identify clusters of arbitrary shape, which is suitable for complex data distribution of photovoltaic power stations. The combination of the two improves the clustering efficiency and stability.

[0043] 3、The application proposes a physical information merging error, which fuses the sum of squared distances, the electrical distance constraint term and the topological connectivity constraint term, so that the clustering results not only conform to the data distribution, but also meet the operation rules of the power system, and the physical rationality and simulation reliability of the equivalent model are improved.

[0044] In summary, through algorithm fusion, index innovation and physical constraint introduction, the application realizes the leap of photovoltaic power station equivalent modeling from "mathematical simplification" to "physical equivalence", which not only improves the accuracy and rationality of the model in theory, but also significantly improves the simulation efficiency and practicality in engineering application, and is suitable for the analysis, planning and control needs of the power system under the background of high proportion of new energy access. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 The flow chart of the photovoltaic power station equivalent modeling method based on C-AGNES is shown in the figure.

[0046] Figure 2 The electrical structure diagram of the photovoltaic power station is shown in the figure.

[0047] Figure 3 The equivalent circuit diagram of the inverter is shown in the figure.

[0048] Figure 4 The equivalent circuit diagram of the photovoltaic power station is shown in the figure. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0050] In order to ensure the dynamic response characteristics of the photovoltaic power station model, a clustering algorithm based on similarity is usually used to perform equivalent modeling on the photovoltaic power station. The clustering index of the photovoltaic power generation unit needs to reflect the power output characteristics of the photovoltaic power generation unit, and also needs to be easy to obtain. The commonly used clustering indexes at present include the access point impedance, the control parameters of the inverter, the temperature and the solar irradiance, or the impedance matrix reflecting the connection degree between photovoltaic units. The photovoltaic array in the photovoltaic power station covers a wide area, and the irradiance shows a relatively large span distribution. The temperature index has a significant influence on the dynamic characteristics of the model due to the sudden change caused by weather changes or local shading. In addition, the equivalent impedance of the collection line also significantly affects the output characteristics of the photovoltaic power station.

[0051] Common clustering algorithms include K-Means clustering, but K-Means clustering has the disadvantages of being sensitive to noise and initial clustering centers, being able to only process hard division (i.e., each sample can only belong to one cluster), and being unable to express the uncertainty of a sample belonging to multiple clusters; hierarchical clustering has high computational complexity, and the merging or splitting operation is irreversible, making it difficult to handle noise interference; and the clustering indicators of the current common methods depend on the internal control parameters of the inverter, which are difficult to obtain, and the single-dimensional dynamic characteristics cannot take into account the physical and electrical coupling characteristics of irradiance and line impedance, resulting in limited precision of the equivalent model.

[0052] Therefore, in view of the fact that the single-machine equivalent method in the existing photovoltaic power station equivalent modeling ignores the difference between irradiance and line impedance, resulting in distortion of dynamic characteristics, and the traditional clustering algorithm (such as K-Means) has defects such as hard division, sensitivity to initial center, inability to handle clusters of arbitrary shape, and inability to reveal the hierarchical structure of data, the embodiments of the present application provide a photovoltaic power station equivalent modeling method based on C-AGNES to solve the above problems. The scheme of the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0053] Specific implementation method one: the photovoltaic power station equivalent modeling method based on C-AGNES provided in the embodiments includes:

[0054] scanning each sample in the sample set based on a density peak value strategy, and then performing initial cluster division on each sample in the sample set, the sample being the irradiance and line equivalent impedance of each photovoltaic unit in the photovoltaic power station after eliminating dimensional differences;

[0055] performing clustering on the result of the initial cluster division by using an AGNES algorithm;

[0056] performing equivalent calculation on the parameters of the photovoltaic units in each cluster after clustering by using a weighted method, connecting the equivalent power generation units in the same cluster in parallel to form an equivalent model of the cluster.

[0057] In one embodiment, the scanning of each sample in the sample set based on the density peak value strategy and the initial cluster division of each sample in the sample set include:

[0058] selecting a sample in the sample set as a clustering center, and calculating the standardized Euclidean distance between the sample and the sample , , , , , ;

[0059] judging whether the standardized Euclidean distance is less than a loose threshold ​​With close threshold Relationship, if Then the sample Marked as "tight" and included in the sample In a cluster that serves as the cluster center, if Then the sample Marked as "loose", if Then the sample As a new cluster center, and to construct a new cluster for that cluster center;

[0060] Traverse the sample set Each sample in the sample is processed until all samples are labeled as "tight", thus completing the initial cluster partitioning.

[0061] In one implementation, clustering the results of the initial cluster partitioning using the AGNES algorithm includes:

[0062] Calculate the physical information merging error between each pair of clusters in the initial cluster division result, merge the two clusters with the smallest physical information merging error value, and repeat the above merging process until the total number of clusters reaches the preset value.

[0063] In one implementation, the calculation of the physical information merging error between pairs of clusters in the initial cluster partitioning result includes:

[0064] The physical information merging error between the pairs of clusters is calculated using the following formula:

[0065] ,

[0066] in, Cluster with cluster Error in merging physical information between them For clusters with cluster The sum of squared deviations and distances between them For clusters with cluster Electrical distance constraints between them For clusters with cluster Topological connectivity constraints between them , and They are respectively , and The weight.

[0067] In one implementation, the cluster with cluster The sum of squared deviations and distances between them The expression is:

[0068] ,

[0069] in, and Representing clusters and cluster The center This represents the squared Euclidean distance between the centers of two clusters. and Representing clusters and cluster The number of samples.

[0070] In one implementation, the cluster with cluster Electrical distance constraints between The expression is:

[0071] ,

[0072] in, This is the electrical distance penalty factor. This is the compactness weighting coefficient. The electrical distance threshold. For clusters with cluster Electrical separation between them For clusters with cluster Electrical compactness between them;

[0073] ,

[0074] ,

[0075] and Representing clusters and cluster The number of samples, Photovoltaic unit and Electrical distance: , and Photovoltaic units and Self-impedance, Photovoltaic unit and mutual impedance between and These are the reference resistor and the reactance, respectively.

[0076] In one implementation, the cluster with cluster Topological connectivity constraints between The expression is:

[0077] ,

[0078] in, These are the topological connectivity weighting coefficients. This is the topological connectivity threshold. The topological diameter of the merged cluster. This represents the average length of the power station's collector lines. For clusters with cluster Topological connectivity score between them

[0079] , For clusters The number of samples, For clusters with cluster The connectivity matrix between them:

[0080] .

[0081] To further illustrate the implementation scheme of this application, Figure 1 A photovoltaic power plant equivalent modeling method based on C-AGNES is provided, including steps 1 to 4. The numbering of these steps does not necessarily restrict their execution order. Each step is described in detail below:

[0082] I. The Canopy clustering algorithm, as a typical approximate clustering method, demonstrates good engineering applicability and computational efficiency in large-scale data preprocessing and clustering scenarios. It achieves this by introducing a relaxed threshold. With close threshold The dual threshold mechanism ( Canopy can quickly generate initial cluster boundaries and candidate centers without requiring precise distance calculations, thus significantly reducing the initialization dependency and computational complexity of subsequent FCM clustering algorithms. Canopy uses a relaxed threshold... and tight threshold Perform a fast scan of the dataset, including:

[0083] S1: Randomly select a sample as the Canopy center;

[0084] S2: Calculate the standardized Euclidean distance from the remaining samples to the center. ;

[0085] S3: Determine the standardized Euclidean distance for each sample. The scope, if If the sample is "close", it will be marked as "close" and added to the current Canopy; if If so, then only mark it as "loose"; if If the sample is used as the new center, a new Canopy will be generated.

[0086] Ultimately, each Canopy provides a potential initial cluster center and a corresponding "compact" sample set, achieving data compression and initial center estimation.

[0087] II. The AGNES algorithm is a classic clustering algorithm. Its core idea is to adopt a "bottom-up" strategy, where each sample point initially forms its own cluster, and then the two most similar clusters are gradually merged until a preset number of clusters or a distance threshold is met.

[0088] The AGNES algorithm aims to minimize the intra-cluster information loss caused by each merge, and its core lies in defining the inter-cluster distance. This embodiment uses the Ward method (sum of squared deviations) as the merging criterion, and its mathematical expression is:

[0089] ,

[0090] In the formula, and Representing clusters and cluster The number of samples in the sample; and Representing clusters and cluster The center; This represents the squared Euclidean distance between the centers of two clusters.

[0091] III. Based on the above ideas, a method for equivalent modeling of photovoltaic power plants is proposed:

[0092] To balance the spatial distribution and electrical coupling characteristics of photovoltaic power plant output, irradiance and line equivalent impedance are used together as clustering indicators. First, the Canopy algorithm is used to perform density pre-segmentation on the original dataset to automatically obtain the initial number of clusters. The initial cluster division is then used as the input for AGNES. The refined division is completed through hierarchical merging. Finally, the weighted method is used to perform equivalent calculations on the parameters of the collector lines, inverters and other components in each cluster, and a multi-machine equivalent model that can retain the differences in internal losses is established.

[0093] The specific steps are as follows:

[0094] Step 1: Data Preprocessing

[0095] The irradiance and equivalent impedance of each photovoltaic unit in the photovoltaic power station are standardized to eliminate dimensional differences and form a sample set.

[0096] Step 2: Canopy pre-segmentation:

[0097] Scan each sample in the sample set based on the density peak strategy.

[0098] If the standardized Euclidean distance corresponding to a certain sample satisfies: If the sample is "close", it will be marked as "close" and added to the current Canopy.

[0099] If the standardized Euclidean distance corresponding to a certain sample satisfies: If the sample is "loose", it means that the sample may belong to the current Canopy, but the density is low.

[0100] If the standardized Euclidean distance corresponding to a certain sample satisfies: If the sample cannot be covered by the current Canopy, then the sample will become a new Canopy center and a new cluster will be created for it.

[0101] The algorithm continues scanning samples that have not yet been labeled "tight" (including "loose" samples and newly discovered centers) until all samples are covered by at least one Canopy, meaning there are no more samples more than a loose threshold away from any existing center. The sample.

[0102] The final Canopy, represented by its respective center, is determined by the Canopy algorithm. There are two initial clusters. Meanwhile, the set of samples marked as "tight" within each Canopy constitutes the initial cluster partitioning result.

[0103] Step 3: AGNES hierarchical merging:

[0104] Starting from the initial clusters in step 2, calculate the physical information merging error between each pair of clusters, merge the two clusters with the smallest physical information merging error, and update the center of the new cluster and the physical information merging error with other clusters. Repeat this process until the number of clusters reaches a preset value. Output the final cluster labels and hierarchical structure tree diagram. Specifically:

[0105] The traditional AGNES algorithm only uses the sum of squared deviations and distances between clusters for merging, which may produce results that are "mathematically optimal but physically unreasonable," such as merging cells that are electrically distant but have similar irradiance, forcibly aggregating topologically disconnected cells, or violating the basic operating rules of the power system.

[0106] To address the aforementioned issues, this embodiment proposes a method for resolving inter-cluster physical information merging errors. , is represented as:

[0107] ,

[0108] in, For clusters in the traditional AGNES algorithm with cluster The sum of squared deviations between them:

[0109] ,

[0110] in, and Representing clusters and cluster The center This represents the squared Euclidean distance between the centers of two clusters.

[0111] For clusters with cluster Electrical distance constraints between:

[0112] ,

[0113] in, The electrical distance penalty factor is 0.3 in this embodiment; The compactness weighting coefficient is set to 1.2 in this embodiment; This is the electrical distance threshold. A penalty is applied when the average electrical distance between clusters exceeds this value. In this embodiment, it is set to 2.

[0114] For clusters with cluster Electrical compactness:

[0115] , Cluster The number of samples, Photovoltaic unit and Electrical distance: , and Photovoltaic units and Self-impedance, Photovoltaic unit and mutual impedance between and These are the reference resistor and the reactance, respectively.

[0116] For clusters with cluster Electrical separation between:

[0117] , Cluster The number of samples.

[0118] For clusters with cluster Topological connectivity constraints between them:

[0119]

[0120] in, This is the topological connectivity weighting coefficient, which is set to 0.2 in this embodiment; This is the topological connectivity threshold; in this embodiment, it is set to 0.6. The topological diameter of the merged cluster. This represents the average length of the power station's collector lines.

[0121] For clusters with cluster Topological connectivity score between:

[0122] ,

[0123] in, For clusters with cluster The connectivity matrix between them:

[0124] .

[0125] , and These are the weights of each item.

[0126] The cluster used in this embodiment with cluster Physical information merging error By integrating electrical distance constraints and topological connectivity constraints, a fundamental shift from "mathematically optimal" to "physically reasonable" has been achieved, resulting in a comprehensive improvement in accuracy, efficiency, robustness, and usability. This method, which combines physical laws with data-driven approaches, significantly enhances the technical performance of equivalent modeling for photovoltaic power plants.

[0127] Step 4: Use a weighted average method to perform equivalent calculations on the parameters of the photovoltaic units within each cluster. Connect the equivalent power generation units within the same cluster in parallel to form the equivalent model of that cluster (i.e., one equivalent collector system connected to multiple equivalent inverters). In this way, the entire photovoltaic power station is equivalent to multiple equivalent collector systems (one for each cluster).

[0128] This embodiment does not require pre-defined cluster geometry and can adaptively discover clusters of arbitrary shapes, scientifically handling complex data distributions caused by factors such as weather and terrain. By selecting appropriate linking criteria, it effectively suppresses noise from damaging the overall structure, enhancing the algorithm's robustness and providing a novel solution for the hierarchical management and refined modeling of photovoltaic power plants.

[0129] Specific Implementation Method Two: The photovoltaic power plant equivalent modeling device based on C-AGNES described in this implementation method includes:

[0130] Initial cluster partitioning unit: Based on the density peak strategy, each sample in the sample set is scanned, and then each sample in the sample set is partitioned into initial clusters. The samples are the irradiance and line equivalent impedance of each photovoltaic unit in the photovoltaic power station after eliminating the difference in dimensions.

[0131] Clustering unit: The initial cluster partitioning results are clustered using the AGNES algorithm;

[0132] Modeling Unit: The parameters of photovoltaic units within each cluster are calculated using a weighted method. The equivalent power generation units within the same cluster are connected in parallel to form the equivalent model of the cluster.

[0133] In one implementation, the step of scanning each sample in the sample set based on the density peak strategy, and then performing initial clustering on each sample in the sample set, includes:

[0134] Select the sample set Samples in As cluster centers, and calculate the samples With sample Standardized Euclidean distance between , , , ;

[0135] Determine the standardized Euclidean distance With a relaxed threshold With close threshold Relationship, if Then the sample Marked as "tight" and included in the sample In a cluster that serves as the cluster center, if Then the sample Marked as "loose", if Then the sample As a new cluster center, and to construct a new cluster for that cluster center;

[0136] Traverse the sample set Each sample in the sample is processed until all samples are labeled as "tight", thus completing the initial cluster partitioning.

[0137] In one implementation, clustering the results of the initial cluster partitioning using the AGNES algorithm includes:

[0138] Calculate the physical information merging error between each pair of clusters in the initial cluster division result, merge the two clusters with the smallest physical information merging error value, and repeat the above merging process until the total number of clusters reaches the preset value.

[0139] In one implementation, the calculation of the physical information merging error between pairs of clusters in the initial cluster partitioning result includes:

[0140] The physical information merging error between the pairs of clusters is calculated using the following formula:

[0141] ,

[0142] in, Cluster with cluster Error in merging physical information between them For clusters with cluster The sum of squared deviations and distances between them For clusters with cluster Electrical distance constraints between them For clusters with cluster Topological connectivity constraints between them , and They are respectively , and The weight.

[0143] In one implementation, the cluster with cluster The sum of squared deviations and distances between them The expression is:

[0144] ,

[0145] in, and Representing clusters and cluster The center This represents the squared Euclidean distance between the centers of two clusters. and Representing clusters and cluster The number of samples.

[0146] In one implementation, the cluster with cluster Electrical distance constraints between The expression is:

[0147] ,

[0148] in, This is the electrical distance penalty factor. This is the compactness weighting coefficient. The electrical distance threshold. For clusters with cluster Electrical separation between them For clusters with cluster Electrical compactness between them;

[0149] ,

[0150] ,

[0151] and Representing clusters and cluster The number of samples, Photovoltaic unit and Electrical distance: , and Photovoltaic units and Self-impedance, Photovoltaic unit and mutual impedance between and These are the reference resistor and the reactance, respectively.

[0152] In one implementation, the cluster with cluster Topological connectivity constraints between The expression is:

[0153] ,

[0154] in, These are the topological connectivity weighting coefficients. This is the topological connectivity threshold. The topological diameter of the merged cluster. This represents the average length of the power station's collector lines. For clusters with cluster Topological connectivity score between them

[0155] , For clusters The number of samples, For clusters with cluster The connectivity matrix between them:

[0156] .

[0157] Specific Implementation Method 3: The photovoltaic power plant equivalent modeling device based on C-AGNES described in this implementation method includes a processor and a memory. The memory stores at least one instruction, which is loaded and executed by the processor to implement the photovoltaic power plant equivalent modeling method based on C-AGNES as described in Specific Implementation Method 1.

[0158] Specific Implementation Method Four: A computer storage medium according to this implementation method is characterized in that the computer storage medium stores at least one instruction, which is loaded and executed by a processor to implement the equivalent modeling method for photovoltaic power plants based on C-AGNES as described in Specific Implementation Method One.

[0159] While specific embodiments of this application have been described herein with reference to them, it should be understood that these embodiments are merely examples of the principles and applications of this application. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of this application as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A photovoltaic power plant equivalent modeling method based on C-AGNES, characterized in that, include: Based on the density peak strategy, each sample in the sample set is scanned, and then the initial clustering of each sample in the sample set is performed. The sample is the irradiance and line equivalent impedance of each photovoltaic unit in the photovoltaic power station after eliminating the difference in dimensions. The initial clustering results were clustered using the AGNES algorithm. The parameters of photovoltaic units within each cluster are calculated using a weighted method. The equivalent power generation units within the same cluster are then connected in parallel to form the equivalent model of that cluster.

2. The photovoltaic power plant equivalent modeling method based on C-AGNES according to claim 1, characterized in that, The method of scanning each sample in the sample set based on the density peak strategy, and then performing initial clustering on each sample in the sample set, includes: Select the sample set Samples in As cluster centers, and calculate the samples With sample Standardized Euclidean distance between , , , ; Determine the standardized Euclidean distance With a relaxed threshold With close threshold Relationship, if Then the sample Marked as "tight" and included in the sample In a cluster that serves as the cluster center, if Then the sample Marked as "loose", if Then the sample As a new cluster center, and to construct a new cluster for that cluster center; Traverse the sample set Each sample in the sample is processed until all samples are labeled as "tight", thus completing the initial cluster partitioning.

3. The photovoltaic power plant equivalent modeling method based on C-AGNES according to claim 1 or 2, characterized in that, The clustering of the initial cluster partitioning results using the AGNES algorithm includes: Calculate the physical information merging error between each pair of clusters in the initial cluster division result, merge the two clusters with the smallest physical information merging error value, and repeat the above merging process until the total number of clusters reaches the preset value.

4. The photovoltaic power plant equivalent modeling method based on C-AGNES according to claim 3, characterized in that, The physical information merging error between pairs of clusters in the result of calculating the initial cluster partitioning includes: The physical information merging error between the pairs of clusters is calculated using the following formula: ， in, Cluster with cluster Error in merging physical information between them For clusters with cluster The sum of squared deviations and distances between them For clusters with cluster Electrical distance constraints between them For clusters with cluster Topological connectivity constraints between them , and They are respectively , and The weight.

5. The photovoltaic power plant equivalent modeling method based on C-AGNES according to claim 4, characterized in that, The cluster with cluster The sum of squared deviations and distances between them The expression is: ， in, and Representing clusters and cluster The center This represents the squared Euclidean distance between the centers of two clusters. and Representing clusters and cluster The number of samples.

6. The photovoltaic power plant equivalent modeling method based on C-AGNES according to claim 4, characterized in that, The cluster with cluster Electrical distance constraints between The expression is: ， in, This is the electrical distance penalty factor. This is the compactness weighting coefficient. The electrical distance threshold. For clusters with cluster Electrical separation between them For clusters with cluster Electrical compactness between them; ， ， and Representing clusters and cluster The number of samples, Photovoltaic unit and Electrical distance: , and Photovoltaic units and Self-impedance, Photovoltaic unit and mutual impedance between and These are the reference resistor and the reactance, respectively.

7. The photovoltaic power plant equivalent modeling method based on C-AGNES according to claim 4, characterized in that, The cluster with cluster Topological connectivity constraints between The expression is: ， in, These are the topological connectivity weighting coefficients. This is the topological connectivity threshold. The topological diameter of the merged cluster. This represents the average length of the power station's collector lines. For clusters with cluster Topological connectivity score between them , For clusters The number of samples, For clusters with cluster The connectivity matrix between them: 。 8. A photovoltaic power plant equivalent modeling device based on C-AGNES, characterized in that, include: Initial cluster partitioning unit: Based on the density peak strategy, each sample in the sample set is scanned, and then each sample in the sample set is partitioned into initial clusters. The samples are the irradiance and line equivalent impedance of each photovoltaic unit in the photovoltaic power station after eliminating the difference in dimensions. Clustering unit: The initial cluster partitioning results are clustered using the AGNES algorithm; Modeling Unit: The parameters of photovoltaic units within each cluster are calculated using a weighted method. The equivalent power generation units within the same cluster are connected in parallel to form the equivalent model of the cluster.

9. A photovoltaic power plant equivalent modeling device based on C-AGNES, characterized in that, The C-AGNES-based photovoltaic power plant equivalent modeling device includes a processor and a memory. The memory stores at least one instruction, which is loaded and executed by the processor to implement the C-AGNES-based photovoltaic power plant equivalent modeling method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The computer storage medium stores at least one instruction, which is loaded and executed by a processor to implement the C-AGNES-based photovoltaic power plant equivalent modeling method as described in any one of claims 1 to 7.

Images