A distributed photovoltaic cluster rapid clustering and grouping method

Abstract

A distributed photovoltaic cluster rapid clustering method, comprising: S1, extracting the historical meteorological data of the grid-connected point of the distributed photovoltaic power station and the operating characteristic parameter value of the distributed photovoltaic power station, and normalizing the same; S2, performing rapid filtering on the distributed photovoltaic power station based on the approximate k-means vector quantization technology, and obtaining representative typical distributed photovoltaic power stations; S3, screening out the distributed photovoltaic power stations with similar relations between each other through the KNN algorithm; S4, obtaining the adjacent matrix and degree matrix information of the typical distributed photovoltaic power stations by means of the Gaussian kernel calculation method, and constructing a Laplace graph; S5, adopting the "average cut" graph cutting method, obtaining the clustering result of the typical distributed photovoltaic power stations; S6, calculating the distance from each distributed photovoltaic power station to all typical distributed photovoltaic power stations, and distributing to a similar class according to the distance size, and completing the clustering. The scheme realizes rapid and accurate clustering of the distributed photovoltaic cluster.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power system data clustering analysis, and particularly relates to a distributed photovoltaic cluster rapid clustering and grouping method. BACKGROUND

[0002] The emergence of renewable energy effectively solves the problems of traditional energy shortage and environmental pollution, and various countries in the world have successively introduced industrial policies to promote photovoltaic power generation. China has also deployed key and major demonstration projects related to the development of photovoltaic power generation technology, and distributed grid-connected photovoltaic power stations have developed rapidly in China. The installed capacity of photovoltaic power generation is continuously increasing every year, especially the distributed photovoltaic installed capacity is growing rapidly, and large-scale access of distributed photovoltaic to the distribution network has become an inevitable trend.

[0003] While the distributed photovoltaic is developing rapidly, the technical requirements of power grid operators for the state operation monitoring and equivalent modeling of distributed photovoltaic power stations are increasingly urgent. In the face of a large number of distributed photovoltaic power stations, whether it is the observation and monitoring of the operation state or the equivalent modeling of the distributed photovoltaic cluster, the essential requirement is to simplify, and the main technical key is to realize the clustering and grouping of the distributed photovoltaic cluster.

[0004] At present, the existing clustering and grouping method of distributed photovoltaic power stations has problems such as large clustering calculation amount, slow clustering speed, poor model adaptability, high data requirement, and insignificant clustering effect. Therefore, it is necessary to provide a distributed photovoltaic cluster rapid clustering and grouping method to improve the clustering analysis effect.

[0005] The above information disclosed in the background section is only intended to strengthen the understanding of the background of the present application, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide a distributed photovoltaic cluster rapid clustering and grouping method. In view of the deficiencies of the existing clustering method, based on the data rapid processing capability of the fast vector quantization technology and the good clustering effect of the spectral clustering analysis method, the advantages of the two are combined to realize the rapid and accurate clustering and grouping of the distributed photovoltaic cluster.

[0007] A technical scheme provided in an embodiment of the present application is a distributed photovoltaic cluster rapid clustering and grouping method, including the following steps:

[0008] S1, extracting historical meteorological data of a grid-connected point of a distributed photovoltaic power station and operation characteristic parameter values of the distributed photovoltaic power station, and normalizing the same;

[0009] S2, the approximate k-means based vector quantization technology is used to quickly filter the distributed photovoltaic power station, and a representative typical distributed photovoltaic power station is obtained;

[0010] S3, the K-Nearest Neihbor (KNN) algorithm is used to screen out distributed photovoltaic power stations with similar relationships;

[0011] S4, the Gaussian kernel calculation method is used to obtain the adjacent matrix and degree matrix information of the typical distributed photovoltaic power station, and a Laplace graph is constructed;

[0012] S5, the "average cut" cut method is used to obtain the clustering result of the typical distributed photovoltaic power station;

[0013] S6, the distance from each distributed photovoltaic power station to all typical distributed photovoltaic power stations is calculated, and according to the distance size, it is allocated to a similar class in the vicinity, and the clustering grouping is completed.

[0014] As preferred, step S1 comprises the following steps:

[0015] S11, the output voltage and current of the photovoltaic array in the distributed photovoltaic power station to be grouped, the output voltage and current of the photovoltaic grid-connected inverter in the photovoltaic power station, and the historical meteorological data including solar irradiance, temperature, atmospheric humidity, etc. are extracted;

[0016] S12, considering that the dimension number of each parameter is inconsistent, normalization processing is performed:

[0017] (1)

[0018] In the formula, x represents the actual measured value of a certain type of parameter, x represents the minimum value of a certain type of parameter, x represents the maximum value of a certain type of parameter.

[0019] As preferred, step S2 comprises the following steps:

[0020] S21, the solar irradiance, temperature, photovoltaic power station grid-connected point output voltage, current and power data of the distributed photovoltaic power station at T sampling time points in a period are extracted, and the corresponding photovoltaic power station parameter matrix is established:

[0021] (2)

[0022] In the formula, x represents the parameter matrix of the i-th photovoltaic power station at T sampling time points, x represents the solar irradiance of the i-th photovoltaic power station at the j-th sampling time point, Tij represents the temperature of the jth sampling time of the ith photovoltaic power station, Iij represents the current of the jth sampling time of the ith photovoltaic power station, Vij represents the voltage of the jth sampling time of the ith photovoltaic power station, Pij represents the power of the jth sampling time of the ith photovoltaic power station.

[0023] S22, assuming that the distributed photovoltaic power stations to be clustered have a total of N, a random non-replacement sampling method is adopted to extract n (0 ) power stations from the N power stations to generate a photovoltaic power station subset S which approximates the distribution characteristics of the whole set of photovoltaic power stations;

[0024] S23, the number of centers is set to k, and the representative typical photovoltaic power stations are quickly screened based on the fast vector quantization technology of the k-means algorithm.

[0025] As a preferred, in step S3, the KNN algorithm is used to select the first k distributed photovoltaic power stations most similar to each distributed photovoltaic power station.

[0026] As a preferred, step S4 includes the following steps:

[0027] S41, for the measurement of the proximity between the distributed photovoltaic power stations, only the proximity distance with the first k similar distributed power stations is considered, and the Gaussian kernel algorithm is used for calculation:

[0028] (3)

[0029] In the formula, d represents the proximity between the distributed photovoltaic power stations and the distributed photovoltaic power stations , σ represents a bandwidth parameter, and exp represents a natural exponential function.

[0030] S42, considering that in practice, the proximity between photovoltaic power stations is symmetrical, for the first k nearest neighbor photovoltaic power stations of two distributed photovoltaic power stations, if they are mutual neighbors, the proximity between them is calculated using formula (3), otherwise the proximity between them is 0, and all photovoltaic power stations are traversed to obtain a proximity matrix W.

[0031] S43, according to the calculated proximity matrix W, a degree matrix D of the typical distributed photovoltaic power station is obtained, D is a diagonal matrix, only the elements of the main diagonal are meaningful, and the rest of the elements are 0:

[0032] (4)

[0033] (5) ​​

[0034] wherein denotes the i-th diagonal element of the degree matrix D, denotes the value of the i-th row and j-th column element of the adjacency matrix W;

[0035] S44, after obtaining the adjacency matrix W and the degree matrix D, a symmetric Laplacian matrix L is constructed by equation (6) :

[0036] (6)

[0037] wherein denotes the symmetric Laplacian matrix, is the identity matrix.

[0038] As a preferred embodiment, S5 comprises the following steps:

[0039] S51, assuming that the Laplacian graph constructed based on the distributed photovoltaic power station to be clustered is cut into k sub-graphs, respectively denoted as , the intersection between these sub-graphs is empty, and the union of all sub-graphs is the whole set;

[0040] S52, the cut graph weight between two sub-graphs is defined according to equation (7):

[0041] (7)

[0042] wherein denotes the adjacency between the distributed photovoltaic power station and the distributed photovoltaic power station , i.e. the value of the i-th row and j-th column element of the adjacency matrix W;

[0043] S53, further, the cut graph weight between all sub-graphs is defined according to equation (8):

[0044] (8)

[0045] wherein denotes the complement of the sub-graph ;

[0046] S54, considering the connection between each sub-graph and the number of photovoltaic power stations in the graph, the average cut method shown in equation (9) is used for sub-graph cutting:

[0047] (9)

[0048] wherein denotes the number of distributed photovoltaic power stations in ;

[0049] S55, when the Laplacian graph is cut into k sub-graphs, The indicator matrix H is defined according to formula (10):

[0050] (10)

[0051] In the formula, H(i,j) represents the element of the i-th row and the j-th column of the indicator matrix H;

[0052] S56, the above subgraph partitioning problem is converted into an optimization problem as shown in formula (11):

[0053] (11)

[0054] In the formula, Tr represents the trace of a matrix;

[0055] S57, the trace optimization problem represented by formula (11) is solved, and the row vectors of the H matrix are input into a k-means clustering model to obtain the clustering result of the typical distributed photovoltaic power station.

[0056] As a preferred, the step S6 comprises the following steps:

[0057] S61, the distance between all distributed photovoltaic power stations and the typical distributed photovoltaic power station obtained in step S5 is calculated by the Gaussian kernel function of formula (3);

[0058] S62, according to the calculation result, the distributed photovoltaic power stations to be grouped are divided into the typical distributed photovoltaic power station categories according to the distance, and the classification of the entire distributed photovoltaic cluster is completed.

[0059] The present application has the following beneficial effects: in view of the deficiencies of the existing clustering method, the data processing speed based on the fast vector quantization technology and the good clustering effect of the spectral clustering analysis method are combined to realize the fast and accurate clustering of the distributed photovoltaic cluster; the clustering result provided by the present application can be used as the technical basis for the state monitoring and equivalent modeling of the distributed photovoltaic cluster, and can provide technical support for the state monitoring and equivalent modeling of the photovoltaic power station in the process of power system transformation.

[0060] The above summary of the application is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application, the application can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the specific embodiments of the application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0061] ​Other features, objects, and advantages of the application will become more apparent from the following detailed description when read in connection with the following drawings, in which like numerals designate like elements in the several figures. The drawings are intended to be illustrative, and not limiting of the application. Like numerals designate like elements in the several views.

[0062] Figure 1 A flow chart of a distributed photovoltaic cluster rapid clustering method of the application. DETAILED DESCRIPTION

[0063] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are merely intended to explain the present application, and do not limit the protection scope of the present application, and all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present application.

[0064] Before the example embodiments are discussed in more detail, it should be mentioned that some of the example embodiments are described as processes or methods depicted as flow charts. Although the flow charts depict the operations (or steps) as sequential processes, many of the operations (or steps) can be performed in parallel, concurrently or simultaneously. In addition, the order of the operations can be rearranged. The processes can be terminated when their operations are completed, but can also have additional steps not included in the figures; the processes can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0065] The present embodiment provides a distributed photovoltaic cluster rapid clustering method, as shown in Figure 1 comprising the following steps:

[0066] S1, extracting historical meteorological data of a grid-connected point of a distributed photovoltaic power station and operating characteristic parameter values of the distributed photovoltaic power station, and normalizing the same;

[0067] S2, performing rapid filtering on the distributed photovoltaic power station based on an approximate k-means vector quantization technology, and obtaining representative typical distributed photovoltaic power stations;

[0068] S3, screening out distributed photovoltaic power stations having a similar relationship therebetween through a KNN algorithm;

[0069] S4, obtaining neighbor matrix and degree matrix information of the typical distributed photovoltaic power stations by means of a Gaussian kernel calculation method, and constructing a Laplacian graph;

[0070] S5, obtaining clustering results of the typical distributed photovoltaic power stations by means of a "average cut" cut graph method.

[0071] S6, calculate the distance from each distributed photovoltaic power station to all typical distributed photovoltaic power stations, and allocate to similar categories according to the distance size to complete clustering grouping.

[0072] Specifically, the output voltage and current of the photovoltaic array in all the distributed photovoltaic power stations to be grouped in a region are extracted; the output voltage and current of the photovoltaic grid-connected inverter in all the distributed photovoltaic power stations to be grouped are extracted; and the historical meteorological data of the grid-connected point of all the distributed photovoltaic power stations to be grouped are extracted, including but not limited to solar irradiance, temperature, atmospheric humidity, etc.

[0073] Further, considering that the dimensions and orders of magnitude of the parameters such as the distributed photovoltaic power station and the meteorological data are inconsistent, the extracted external characteristic data and meteorological data are normalized, and the processing process is shown in formula (1):

[0074] (1)

[0075] In the above formula, x represents the actual measured value of a certain type of parameter, x represents the minimum value of a certain type of parameter, x represents the maximum value of a certain type of parameter.

[0076] Further, the solar irradiance, temperature, output voltage, current and power data of the distributed photovoltaic power station at T sampling time points in a period are extracted, and the parameter matrix of all the distributed photovoltaic power stations to be grouped is established. Without loss of generality, the parameter matrix of the ith photovoltaic power station is shown in formula (2):

[0077] (2)

[0078] In the above formula, x represents the parameter matrix of the ith photovoltaic power station at T sampling time points, x represents the solar irradiance of the ith photovoltaic power station at the jth sampling time point, x represents the temperature of the ith photovoltaic power station at the jth sampling time point, x represents the current of the ith photovoltaic power station at the jth sampling time point, x represents the voltage of the ith photovoltaic power station at the jth sampling time point, x represents the power of the ith photovoltaic power station at the jth sampling time point.

[0079] Further, assuming that there are N distributed photovoltaic power stations to be grouped, in order to reduce the calculation amount of clustering analysis and improve the calculation speed of clustering analysis, a random non-replacement sampling method is adopted to extract n ) to generate a subset S of photovoltaic power stations that approximately preserves the distribution characteristics of the original set of photovoltaic power stations.

[0080] Further, for the subset S of photovoltaic power stations obtained by random sampling, the number of centers is set to k, and the representative typical photovoltaic power stations are quickly screened based on the fast vector quantization technology of the k-means algorithm. The typical photovoltaic power stations obtained by the k-means algorithm are an approximation of the original photovoltaic power stations, which largely preserves the data characteristics of the original power stations.

[0081] Further, after the initial screening of the distributed photovoltaic power stations is completed, spectral clustering algorithm is further used for clustering. The spectral clustering algorithm includes two parts: one is the construction of the Laplacian graph, and the other is the subgraph segmentation.

[0082] For the construction of the Laplacian graph, the KNN algorithm is used to select the first k most similar distributed photovoltaic power stations for each distributed photovoltaic power station, the proximity between the distributed photovoltaic power stations is calculated by the Gaussian kernel function, the proximity matrix W is constructed, and then the degree matrix D is calculated by transforming the proximity matrix, and then the Laplacian matrix L is obtained by transformation. The specific steps are as follows:

[0083] For the measurement of the proximity between the distributed photovoltaic power stations, only the proximity distances to the first k similar distributed power stations are considered, and the Gaussian kernel algorithm is used for calculation:

[0084] (3)

[0085] In the above formula, represents the proximity between the distributed photovoltaic power station and the distributed photovoltaic power station , represents the bandwidth parameter, and exp represents the natural exponential function.

[0086] Considering that the proximity between photovoltaic power stations is symmetric in practice, for the first k nearest photovoltaic power stations of two distributed photovoltaic power stations, if they are mutual neighbors, the proximity between them is calculated using formula (3), otherwise the proximity between them is 0. All photovoltaic power stations are traversed to obtain the proximity matrix W.

[0087] According to the calculated proximity matrix W, the degree matrix D of the typical distributed photovoltaic power station is obtained. D is a diagonal matrix, only the elements on the main diagonal are meaningful, and the rest are 0:

[0088] (4)

[0089] (5)

[0090] where denotes the i-th diagonal element of the degree matrix D, denotes the value of the i-th row and j-th column element of the adjacency matrix W;

[0091] After obtaining the adjacency matrix W and the degree matrix D, the symmetric Laplacian matrix L is constructed by equation (6) :

[0092] (6)

[0093] where denotes the symmetric Laplacian matrix, is the identity matrix.

[0094] Suppose the Laplacian graph constructed based on the distributed photovoltaic power stations to be clustered is cut into k sub-graphs, denoted as respectively, the intersection between these sub-graphs is empty, and the union of all sub-graphs is the whole set.

[0095] According to equation (7), the cut graph weight between two sub-graphs is defined as:

[0096] (7)

[0097] where denotes the adjacency degree between the distributed photovoltaic power station and the distributed photovoltaic power station , i.e., the value of the i-th row and j-th column element in the adjacency matrix W.

[0098] Further, according to equation (8), the cut graph weight between all sub-graphs is defined as:

[0099] (8)

[0100] where denotes the complement of the sub-graph .

[0101] Considering the connection between each sub-graph and the number of photovoltaic power stations in the graph, the average cut method shown in equation (9) is used for sub-graph cutting:

[0102] (9)

[0103] where denotes the number of distributed photovoltaic power stations in .

[0104] When the Laplacian graph is cut into k sub-graphs, , the indicator matrix H is defined according to equation (10):

[0105] (10)

[0106] In the above formula, H represents the i-th row and j-th column element of the indication matrix H.

[0107] The above subgraph partitioning problem is converted into an optimization problem as shown in formula (11):

[0108] (11)

[0109] In the above formula, Tr represents the trace of the matrix.

[0110] The trace optimization problem represented by formula (11) is solved, and the row vectors of the H matrix are input into the k-means clustering model to obtain the clustering result of the typical distributed photovoltaic power station.

[0111] Since the original photovoltaic power station is sampled by using a random non-replacement sampling method at the beginning of clustering, after completing the spectral clustering analysis, most of the power stations still have not been clustered into groups. Considering this problem, based on the results of clustering analysis, the distance between each distributed photovoltaic power station and the typical photovoltaic power station is calculated, and it is classified into the same category according to clustering. The specific steps are as follows:

[0112] The distance between all distributed photovoltaic power stations and typical distributed photovoltaic power stations is calculated by using the Gaussian kernel function of formula (3);

[0113] According to the calculation results, the distributed photovoltaic power stations to be grouped are divided into the typical distributed photovoltaic power station categories according to the distance, and the classification of the entire distributed photovoltaic cluster is completed.

[0114] The typical photovoltaic power station obtained in the clustering process can be used as the representative power station of the photovoltaic cluster, and the state monitoring of the entire distributed photovoltaic cluster can be realized by observing and supervising the operating state thereof;

[0115] The distributed photovoltaic clustering analysis result obtained by clustering analysis can be used as the basis for multi-machine equivalent modeling of the distributed photovoltaic cluster.

[0116] The above specific embodiments are the preferred embodiments of the distributed photovoltaic cluster rapid clustering and grouping method of the present application, and are not intended to limit the specific implementation range of the present application. The scope of the present application includes but is not limited to the specific embodiments, and equivalent changes made according to the shape and structure of the present application are within the protection scope of the present application.​

Claims

1. A distributed photovoltaic cluster fast clustering method, characterized in that, The method comprises the following steps: S1, extracting historical meteorological data of a grid-connected point of a distributed photovoltaic power station and operation characteristic parameter values of the distributed photovoltaic power station, and normalizing the same; S2, rapidly filtering the distributed photovoltaic power station based on an approximate k-means vector quantization technology, and obtaining representative typical distributed photovoltaic power stations; S3, screening distributed photovoltaic power stations having a similar relationship therebetween by using a KNN algorithm; S4, obtaining neighbor matrix and degree matrix information of the typical distributed photovoltaic power stations by using a Gaussian kernel calculation method, and constructing a Laplace graph; S5, obtaining clustering results of the typical distributed photovoltaic power stations by using a "average cut" cut graph method; S6, calculating distances from each distributed photovoltaic power station to all the typical distributed photovoltaic power stations, and distributing to a similar class in proximity according to the distance size to complete clustering and grouping.

2. The method of claim 1, wherein, Step S1 comprises the following steps: S11, extracting photovoltaic array output voltage and current of a distributed photovoltaic power station to be grouped, output voltage and current of a photovoltaic grid-connected inverter in the photovoltaic power station, and historical meteorological data; S12, considering that the dimensions of various parameters are inconsistent in quantity, performing normalization processing.

3. The method of claim 2, wherein, The historical meteorological data include but are not limited to solar irradiance, temperature, and atmospheric humidity.

4. The method of claim 1, wherein, Step S2 comprises the following steps: S21, extracting solar irradiance, temperature, photovoltaic power station grid-connected point output voltage, current, and power data of a distributed photovoltaic power station at a time period and T sampling time points, and establishing a corresponding photovoltaic power station parameter matrix; S22, using a random non-replacement sampling method to extract n power stations from N distributed photovoltaic power stations to generate a photovoltaic power station subset S; S23, setting the number of clustering centers as k, and rapidly screening representative typical photovoltaic power stations based on a fast vector quantization technology of a k-means algorithm.

5. The method of claim 1, wherein, In step S3, a KNN algorithm is used to select the first k distributed photovoltaic power stations most similar to each distributed photovoltaic power station.

6. The method of claim 1, wherein, Step S4 comprises the following steps: S41, for measuring the proximity degree between distributed photovoltaic power stations, only considering the proximity distance of the first k similar distributed power stations, and calculating by using a Gaussian kernel algorithm; S42, for the first k neighboring photovoltaic power stations of two distributed photovoltaic power stations, if the two are mutual neighbors, the proximity degree between the two is calculated by using a Gaussian kernel algorithm, otherwise the proximity degree between the two is 0, and the proximity matrix W is obtained by traversing all the photovoltaic power stations; S43, obtaining the degree matrix D of the typical distributed photovoltaic power stations according to the calculated proximity matrix W; S44、After obtaining the adjacent matrix W and the degree matrix D, a symmetric Laplace matrix L is constructed sym .

7. The method of claim 1, wherein, Step S5 comprises the following steps: S51, cut the Laplacian graph constructed based on the distributed photovoltaic power station to be grouped into k subgraphs, denoted as {S1, S2, …, S k}, the intersection between the k subgraphs is empty, and the union of the k subgraphs is the universal set; S52, defining the cut graph weight between two subgraphs; S53, further, defining the cut graph weight between all subgraphs; S54, considering the connection between each subgraph and the number of photovoltaic power stations in the graph, using an average cut method to cut the subgraph; S55, split the Laplacian graph into k sub-graphs, {S1, S2, …, Sk} k} define an indicator matrix; S56, converting the subgraph segmentation problem into a trace optimization problem; S57, solving the trace optimization problem to obtain the clustering results of the typical distributed photovoltaic power stations.

8. The method of claim 1, wherein, Step S6 comprises the following steps: S61, calculating the distance between all distributed photovoltaic power stations and the typical distributed photovoltaic power stations by using a Gaussian kernel function. S62, according to the calculation result, the to-be-grouped distributed photovoltaic power station is classified into the typical distributed photovoltaic power station category in proximity according to the distance, and the classification of the entire distributed photovoltaic cluster is completed.

Images