Data processing method and device of photovoltaic power station, electronic equipment and storage medium

Abstract

The application relates to the technical field of data processing, in particular to a data processing method and device of a photovoltaic power station, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring time series data of the photovoltaic power station; performing multi-dimensional data clustering on the time series data to obtain multi-dimensional similarity measurement of the photovoltaic power station; and generating a target photovoltaic scene of the photovoltaic power station according to the multi-dimensional similarity measurement of the photovoltaic power station. Thus, the application considers multiple dimensions of the power dimension, the energy dimension and the fluctuation of the photovoltaic power station, effectively extracts representative photovoltaic scenes, reduces the data volume, improves the efficiency of data analysis and processing, and evaluates the photovoltaic scenes by using indexes such as photovoltaic grid-connected random flow, annual power generation, photovoltaic daily power generation curve shape similarity, photovoltaic power generation peak period frequency value and the like, so that the accuracy, reliability and practicability of the photovoltaic scene extraction are verified from the clustering algorithm and the power.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to a data processing method and device for a photovoltaic power station, an electronic device and a storage medium. BACKGROUND

[0002] With the development of solar power generation technology and the rapid growth of photovoltaic scale, effective use of photovoltaic historical data is crucial for improving photovoltaic grid efficiency, formulating reasonable inverter control strategies, implementing uncertainty scheduling and planning, and accurately simulating and predicting photovoltaic power generation.

[0003] However, in the face of massive historical photovoltaic data, its processing faces two major problems: first, the high repeatability and regularity of the data lead to a huge waste of storage and computing resources; second, the existing technology for extracting effective data from the scene cannot truly reflect the actual photovoltaic power generation, thereby affecting the accuracy and reliability of power system planning. SUMMARY

[0004] The present application provides a data processing method and device for a photovoltaic power station, an electronic device and a storage medium to solve the problems that related technologies cannot truly reflect the actual photovoltaic power generation, affect the accuracy and reliability of photovoltaic power station planning, and have a large amount of data processing.

[0005] The first aspect embodiment of the present application provides a data processing method for a photovoltaic power station, comprising the following steps: obtaining time series data of the photovoltaic power station; performing multi-dimensional data clustering on the time series data to obtain multi-dimensional similarity metrics of the photovoltaic power station, wherein the multi-dimensions are a plurality of power dimensions, energy dimensions and volatility dimensions of the photovoltaic power station; and generating a target photovoltaic scene of the photovoltaic power station according to the multi-dimensional similarity metrics of the photovoltaic power station.

[0006] Optionally, the multi-dimensional data clustering on the time series data to obtain the multi-dimensional similarity metrics of the photovoltaic power station comprises: performing pre-clustering on the time series data to obtain a target cluster number; inputting data of the target cluster number into a clustering model, and the clustering model outputs the multi-dimensional similarity metrics of the photovoltaic power station, wherein the clustering model calculates similarity metrics, energy size similarity, cross-correlation sequence length and fluctuation position similarity of any two power time series in the data of the target cluster number, and generates the multi-dimensional similarity metrics according to the similarity metrics, the energy size similarity, the cross-correlation sequence length and the fluctuation position similarity.

[0007] Optionally, the clustering model is:

[0008] LESD(x i ,x j ,W)=w1L(x i ,x j)+w2E(x i ,x j )+w3S(x i ,x j )+w4δ(x i ,x j )

[0009] where w1, w2, w3 and w4 represent weight coefficients, which constitute a vector W∈R 4×1 , x i and x j represent a power time sequence with a length of L, x it , x jt represent the characteristics of the power time sequence x i and x j at the t-th moment, respectively;

[0010] E(x i ,x j ) represents the electrical quantity similarity of x i and x j ;

[0011] CC ω (x i ,x j )=R ω-L (x i ,x j ),ω∈{1,2,K,2L-1},

[0012] CC w (x i ,x j ) represents the cross-correlation sequence of x i and x j , s represents the moving step of the power time sequence x j , s∈[-L,L]; w is the position subscript of the cross-correlation sequence CC w (x i ,x j );

[0013] S(x i ,x j )=1-Max(NCC c (x i ,x j )), S(x i ,x j ) represents the morphological trend similarity distance of x i and x j , with a value range of [0,2];

[0014] representing sequence x j sliding to and x i displacement of the most similar morphological trend, δ(x i ,x j represents the similarity of fluctuations of the two power time series.

[0015] Optionally, before inputting the data of the target cluster number into the clustering model, further comprising: obtaining neighborhood probability of the power time series and other sequences; determining neighborhood sequences of the power time series according to the neighborhood probability.

[0016] Optionally, the target photovoltaic scene of the photovoltaic power station is generated according to the multi-dimensional similarity measure of the photovoltaic power station, comprising: extracting morphological representative scenes from the multi-dimensional similarity measure of the photovoltaic power station; and magnifying the morphological representative scenes to obtain the target photovoltaic scene of the photovoltaic power station.

[0017] The target photovoltaic scene is:

[0018]

[0019] wherein, N g is the number of the gth type of power time series; x g,i,t is the feature value of the ith power time series in the gth type at time t; is the gth type of morphological representative scene, μ g,t * is the feature value of the gth type of morphological representative scene at time t, P g represents the original sequence set of the cluster where the representative scene to be solved is located.

[0020] Optionally, before the multi-dimensional data clustering of the time series data to obtain the multi-dimensional similarity measure of the photovoltaic power station, further comprising: performing standardization processing on the time series data.

[0021] Optionally, after the target photovoltaic scene of the photovoltaic power station is generated according to the multi-dimensional similarity measure of the photovoltaic power station, further comprising: obtaining multiple index data of photovoltaic grid-connected random power flow, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value; and evaluating the accuracy level of the target photovoltaic scene based on the multiple index data of photovoltaic grid-connected random power flow, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value.

[0022] The second aspect embodiment of the application provides a data processing device of a photovoltaic power station, comprising: an acquisition module configured to acquire time series data of the photovoltaic power station; a clustering module configured to perform multi-dimensional data clustering on the time series data to obtain multi-dimensional similarity metrics of the photovoltaic power station, wherein the multi-dimensions are a plurality of power dimensions, energy dimensions and volatility dimensions of the photovoltaic power station; and a generation module configured to generate a target photovoltaic scenario of the photovoltaic power station according to the multi-dimensional similarity metrics of the photovoltaic power station.

[0023] Optionally, the clustering module is further configured to perform pre-clustering on the time series data to obtain a target cluster number; input data of the target cluster number into a clustering model, and the clustering model outputs the multi-dimensional similarity metrics of the photovoltaic power station, wherein the clustering model calculates similarity metrics, energy size similarity, cross-correlation sequence length and fluctuation position similarity of any two power time series in the data of the target cluster number, and generates the multi-dimensional similarity metrics according to the similarity metrics, the energy size similarity, the cross-correlation sequence length and the fluctuation position similarity.

[0024] Optionally, the clustering model is:

[0025] LESD(x i ,x j ,W)=w1L(x i ,x j )+w2E(x i ,x j )+w3S(x i ,x j )+w4δ(x i ,x j )

[0026] wherein w1, w2, w3 and w4 represent weight coefficients, and constitute a vector W∈R 4×1 , x i and x j represent a power time series with a length of L, x it , x jt represent characteristics of the power time series x i and x j at the t-th moment, respectively;

[0027] E(x i ,x j ) represents energy size similarity of x i and x j ;

[0028] CC ω (x i ,x j )=R ω-L(x i ,x j ),ω∈{1,2,K,2L-1},

[0029] CC w (x i ,x j ) represents x i and x j The cross-correlation sequence, s represents the power time series x j The moving step size, s∈[-L,L]; w is the cross-correlation sequence CC w (x i ,x j The positional subscript of );

[0030] S(x i ,x j ) = 1 - Max(NCC) c (x i ,x j )),S(x i ,x j ) represents x i With x j The similarity distance in terms of morphological trends, with a value range of [0,2];

[0031] Represents sequence x j Slide to x i The displacement when the pattern trend is most similar, δ(x) i ,x j The ) indicates the similarity of fluctuations between two power time series.

[0032] Optionally, the data processing device for the photovoltaic power station further includes: a determination module, used to obtain the neighborhood probability of the power time series and other series before inputting the data of the target cluster number into the clustering model; and to determine the neighborhood sequence of the power time series based on the neighborhood probability.

[0033] Optionally, the generation module is further used to extract morphologically representative scenes from the multidimensional similarity measure of the photovoltaic power station; the morphologically representative scenes are then magnified to obtain the target photovoltaic scene of the photovoltaic power station, wherein the target photovoltaic scene is:

[0034]

[0035] Where, N g x is the number of power time series entries of type g; g,i,t Let be the feature value of the i-th power time series in class g at time t; For the g-th type of morphology, μ g,t *a feature value of a representative scenario of a g-th mode at a t-th time point, P g representing a raw sequence set of a cluster in which the representative scenario to be solved is located.

[0036] Optionally, the data processing device of the photovoltaic power station further comprises a processing module configured to perform standardization processing on the time series data before performing multi-dimensional data clustering on the time series data to obtain the multi-dimensional similarity measure of the photovoltaic power station.

[0037] Optionally, the data processing device of the photovoltaic power station further comprises an evaluation module configured to, after generating the target photovoltaic scenario of the photovoltaic power station according to the multi-dimensional similarity measure of the photovoltaic power station, acquire a plurality of index data of random power flow of the photovoltaic grid, annual power generation, photovoltaic daily power generation curve shape similarity, and photovoltaic power generation peak period frequency value; and evaluate the accuracy level of the target photovoltaic scenario based on the plurality of index data of random power flow of the photovoltaic grid, annual power generation, photovoltaic daily power generation curve shape similarity, and photovoltaic power generation peak period frequency value.

[0038] The third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the program to implement the data processing method of the photovoltaic power station as described in the above embodiments.

[0039] The fourth aspect of the present application provides a computer readable storage medium having a computer program stored thereon, and the program is executed by a processor to implement the data processing method of the photovoltaic power station as described in the above embodiments.

[0040] Therefore, the present application includes the following beneficial effects:

[0041] The embodiments of the present application can more comprehensively reflect the actual characteristics of photovoltaic power generation by performing multi-dimensional data clustering on the time series data to obtain the multi-dimensional similarity measure of the photovoltaic power station, effectively extract representative photovoltaic scenarios based on the multi-dimensional similarity measure of the photovoltaic power station, reduce the amount of data, improve the efficiency of data analysis and processing, and provide reliable basis for the scheduling, control strategy formulation, and future scenario simulation of the power system by evaluating the target photovoltaic scenario. Therefore, the problems of the related art that the actual photovoltaic power generation cannot be truly reflected, the accuracy and reliability of photovoltaic power station planning are affected, and the amount of data processing is large are solved.

[0042] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0043] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description, taken in conjunction with the following drawings of which:

[0044] Figure 1 A flow chart of a data processing method of a photovoltaic power station according to an embodiment of the present application;

[0045] Figure 2 A flow chart of obtaining representative photovoltaic scenarios by an optimization algorithm according to an embodiment of the present application;

[0046] Figure 3 A result chart of representative photovoltaic scenarios according to an embodiment of the present application;

[0047] Figure 4 A comparison chart of probabilistic power flow results of an IEEE 69 bus power distribution system according to an embodiment of the present application;

[0048] Figure 5 A comparison chart of total power of a photovoltaic power station and power area of original data points replaced by power of representative scenarios according to an embodiment of the present application;

[0049] Figure 6 A distribution chart of relative frequency of peak time of representative scenarios and relative frequency of peak time of original scenarios according to an embodiment of the present application;

[0050] Figure 7 A structure chart of a data processing apparatus of a photovoltaic power station according to an embodiment of the present application;

[0051] Figure 8 A structure chart of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0052] Embodiments of the present application are described in detail below with reference to the attached drawing figures, wherein the same or like component have the same or similar designations. The embodiments described below are intended to serve as examples of the present application and are not intended to limit the present application in any way.

[0053] A data processing method, device, electronic equipment and storage medium of a photovoltaic power station are described below with reference to the accompanying drawings. To address the problems mentioned in the background, the present application provides a data processing method of a photovoltaic power station, which more fully considers power characteristics and power time series characteristics, so that the results are more accurate and have practical significance. Meanwhile, a method for comprehensively evaluating the accuracy of photovoltaic scene extraction is proposed, which combines probabilistic power flow analysis, power characteristics, morphological similarity verification, and relative frequency values of peak times, and evaluates through double verification indexes, namely mathematical theory verification and actual business requirements of the power system, to ensure the accuracy and practicality of the results. In this method, multi-dimensional data clustering is performed on time series data to obtain multi-dimensional similarity metrics of the photovoltaic power station, which can more comprehensively reflect the actual characteristics of photovoltaic power generation, effectively extract representative photovoltaic scenes, reduce data volume, and improve the efficiency of data analysis and processing.

[0054] Specifically, Figure 1 A flowchart of a data processing method of a photovoltaic power station provided by an embodiment of the present application is shown.

[0055] As Figure 1 shown, the data processing method of the photovoltaic power station includes the following steps:

[0056] In step S101, time series data of the photovoltaic power station is obtained.

[0057] It can be understood that the photovoltaic power generation time series data to be clustered can be obtained by data acquisition, for example, the photovoltaic power generation time series data collection range can be set to 1 year, and the collection frequency can be set to 1 h / time. In actual execution, a series of photovoltaic power generation data collected in time sequence forms a photovoltaic power generation time series data X=(x1, x2, …, x d ).

[0058] In step S102, multi-dimensional data clustering is performed on the time series data to obtain multi-dimensional similarity metrics of the photovoltaic power station.

[0059] The multi-dimension is a plurality of power dimension, energy dimension and volatility dimension of the photovoltaic power station.

[0060] In the present application, before the multi-dimensional data clustering of the time series data is performed to obtain the multi-dimensional similarity metrics of the photovoltaic power station, the time series data can be subjected to Max-Min standardization processing, which can eliminate the dimensional difference between different dimensions (such as different photovoltaic power stations or different power generation power levels), and obtain the time series data after standardization processing of different dimensions.

[0061] In an embodiment of the present application, the multi-dimensional data clustering of the time series data obtains the multi-dimensional similarity measure of the photovoltaic power station, comprising: obtaining a target cluster number by pre-clustering the time series data; inputting the data of the target cluster number into a clustering model, and the clustering model outputs the multi-dimensional similarity measure of the photovoltaic power station, wherein the clustering model calculates the similarity measure, the power size similarity, the cross-correlation sequence length and the fluctuation position similarity of any two power time series in the data of the target cluster number, and generates the multi-dimensional similarity measure according to the similarity measure, the power size similarity, the cross-correlation sequence length and the fluctuation position similarity.

[0062] It can be understood that the embodiments of the present application can pre-cluster (such as using K-means clustering) the normalized time series data obtained in the above steps, and the cluster number can be 2-100 classes. In order to avoid the randomness of initialization, the embodiments of the present application can run 100 times, and select the optimal result as the final clustering result, so as to ensure that the final clustering result is relatively stable and representative. In actual execution process, the embodiments of the present application can select the cluster number corresponding to the fastest growth speed, and can select a smaller cluster number to represent the overall information of most data, and at the same time, the calculation cost is saved to a certain extent.

[0063] Further, the embodiments of the present application can establish a clustering model considering the multi-dimensional similarity measure of power, energy and fluctuation, input the data of the target cluster number into the clustering model, and the clustering model outputs the multi-dimensional similarity measure of the photovoltaic power station, wherein the clustering model can measure the similarity of the photovoltaic power data from four aspects of distance size, power size, shape trend and fluctuation position, and improve the accuracy of photovoltaic scene extraction.

[0064] Suppose two power time series x i =(x i1 ,x i2 ,…,x iL ) T and x j =(x j1 ,x j2 ,…,x jL ) T .

[0065] The distance size similarity measure of the embodiments of the present application can measure the similarity between the two power time series x i and x j using the Euclidean distance, and the smaller the Euclidean distance, the more similar the two power time series. The similarity measure L(x i ,x j ) is defined as follows:

[0066]

[0067] Where, x it x jt x represents the power time series at time t. i and x j Features, L(x) i ,x j The squared form of the Euclidean distance is chosen, as this form is easy to derive and optimize. L(x) i ,x j The smaller the value, the more similar the changes in the characteristic values ​​of the power time series; the larger the value, the greater the changes in the characteristic values ​​of the power time series.

[0068] In this embodiment, the difference in power quantity represents the area of ​​the shaded region enclosed by the two power time series and the coordinate axis. If the two power time series curves are similar, they must satisfy the condition that their power quantities are similar, and the curve x... i and x j Similarity in battery size E(x) i ,x j The definition is as follows:

[0069]

[0070] Where, x i and x j x represents the power time series at time t. i and x j Features and These represent the sum of the eigenvalues ​​of each power time series, and the standardized metric is always greater than or equal to 1. E(x) i ,x j The smaller the value, the more similar the power levels of the two batteries; the larger the value, the greater the difference in power levels between the two batteries.

[0071] The morphological trend in this embodiment represents the characteristic situation in different time periods. Therefore, under similar weather conditions, the morphological trends of two power time series should be similar. The cross-correlation calculation method is as follows: the displacement invariance ensures that the translation of the power time series on the time axis will not affect the results of its correlation analysis: when calculating the cross-correlation of two power time series, one series is fixed and the other series is slid in a specified direction and step size. i and x j The cross-correlation sequence is CC w (x i ,x j )=(c1,…,c w The length of this sequence is (2L-1), defined as follows:

[0072] CC ω(x i ,x j ) = R ω-L (x i ,x j ), ω ∈ {1,2,K,2L-1},

[0073]

[0074] where s represents the moving step of the power time series x j , s ∈ [-L, L]; w is the position index of the sequence CC w (x i ,x j ). The position w w corresponding to the maximum cross-correlation value can be obtained through the sequence CC i (x j ,x max ), and according to the position, the alignment sequence x j of the power time series x i relative to the power time series x j ’ can be obtained.

[0075] It can be understood that, in order to improve the comparability of the cross-correlation sequence data, the cross-correlation sequence is processed by using a coefficient normalization method, so that the zero-lag autocorrelation coefficient is equal to 1, and the coefficient normalization formula is as follows:

[0076]

[0077] where NCC c (x i ,x j ) is the cross-correlation sequence after coefficient normalization, and the value range thereof is [-1, 1]; R0(x i ,x i ) and R0(x j ,x j ) are autocorrelations of the power time series x i and x j , respectively. When NCC c (x i ,x j ) is 1 or -1, the two sequences are positively or negatively correlated, and therefore, the greater the value of NCC c (x i ,x j ), the more similar the two sequences are. For the convenience of calculation, the morphological trend similarity distance of the sequences x i and x j is defined as:

[0078] S(x i ,x j) = 1 - Max(NCC c (x i ,x j )),

[0079] where S(x i ,x j ) is in the range of [0, 2], where 0 represents the sequence x i and x j morphological trend is exactly the same, and 2 represents the complete opposite.

[0080] The fluctuation position similarity of the embodiments of the present application is an important factor of the power time series curve similarity. Since the cross-correlation is calculated under the condition of considering the shift invariance, the morphological trend distance of two power time series is calculated, i.e., the morphological trend distance of the sequence x i and the sequence x j is calculated from the sequence x i and the sequence x j ', so that the fluctuation position deviation of the two is ignored. Therefore, the fluctuation position similarity δ of two sequences is defined to reflect the difference of the fluctuation occurrence time of the two, and the formula is as follows:

[0081]

[0082] wherein, represents the shift of the sequence x j sliding to and x i morphological trend is most similar, and the smaller the value of δ(x i ,x j ) represents the more similar the fluctuation position of the two power time series; the larger the value represents the greater the difference of the fluctuation position of the two power time series.

[0083] Finally, in order to reflect the multi-dimensional similarity between the power time series, the multi-dimensional similarity measure LESD (i.e., the clustering model) is obtained according to the above four different unit type distances:

[0084] LESD(x i ,x j ,W) = w1L(x i ,x j ) + w2E(x i ,x j ) + w3S(x i ,x j ) + w4δ(x i ,x j )

[0085] wherein, w1, w2, w3 and w4 represent weight coefficients, which constitute a vector W ∈ R 4×1 , and the definition constraint is 0 < w1, w2, w3, w4 < 1, W T1=1.

[0086] In the embodiment of the present application, before inputting the data of the target cluster number into the cluster model, the embodiment further comprises: obtaining the neighborhood probability of the power time sequence and other sequences; and determining the neighborhood sequence of the power time sequence according to the neighborhood probability.

[0087] In order to fully explore the local connectivity between the power time sequences, the embodiment introduces a probability neighborhood for each power time sequence. For the power time sequence x i , the neighborhood probability between x i and any other data x j is measured by a ij , where the smaller the multi-dimensional similarity measure is, the greater the similarity a ij is. Therefore, the similarity a ij can be solved by the following formula for each power time sequence x i :

[0088]

[0089] where a ij is the connection probability between the power time sequence x i and x j , 0≤a ij ≤1 and the jth element of the vector a i ∈R d×1 is a ij , a i T 1=1, 1 is a column vector with all elements being 1, and a ij constitutes A∈R d×d , d represents days, and L represents the power values of the 24 time points of the five photovoltaic stations. The smaller the multi-dimensional similarity measure LESD(x i , x j , W) is, the greater the neighborhood probability a ij is.

[0090] The embodiment introduces a to obtain a ij , and the neighborhood of x i may be any power time sequence other than itself.

[0091] In step S103, the target photovoltaic scene of the photovoltaic power station is generated according to the multi-dimensional similarity measure of the photovoltaic power station.

[0092] In an embodiment of the present application, the target photovoltaic scene of the photovoltaic power station is generated according to the multi-dimensional similarity metric of the photovoltaic power station, comprising: extracting a morphological representative scene from the multi-dimensional similarity metric of the photovoltaic power station; and magnifying the morphological representative scene to obtain the target photovoltaic scene of the photovoltaic power station.

[0093] The embodiment of the present application can introduce a regularization parameter γ, and for any two power time series, a ij The above formula can be derived from the following problem:

[0094]

[0095] The second term in the above formula is regularization, and γ is a regularization parameter. There is only one connected component for solving the above formula, while the ideal neighborhood assignment of the algorithm has c connected branches, and therefore the following property in graph theory is used:

[0096] Theorem 1 (K connected branches): The multiplicity c of the eigenvalue 0 of the Laplacian matrix L A equals the number of connected components in the graph with the similarity matrix A.

[0097]

[0098] where L A =D A -(A T +A) / 2, the i-th diagonal element of the degree matrix D A ∈R d×d is Σ j (a ij +a ji ) / 2, and {α1,α2,…,α d} are the eigenvalues of L A .

[0099] It can be understood that the Laplacian matrix rank constraint of the similarity matrix A is imposed on the above formula according to Theorem 1 to ensure that the neighborhood assignment of A is ideal, so that the photovoltaic power data set can be divided into c clusters, and the representative scene extraction model after adding the constraint condition can be expressed in the following form:

[0100]

[0101] Theorem 2 (property of Laplacian matrix): The Laplacian matrix L A has a semi-positive definite property, that is, all eigenvalues of L A are non-negative.

[0102] It can be understood that in the embodiment of the present application, it is assumed that the g-th smallest eigenvalue of the Laplacian matrix L A is σ g (LA )≥0, according to theorem 2, σ g (L A )≥0, on this basis, a large enough parameter λ is introduced, combined with theorem 1, the constraint condition rank(L A )=d-c can be met, so the above formula is converted into the following form:

[0103]

[0104] Theorem 3 (Ky Fan theorem): for matrix F∈R d×c , there is the following minimization problem:

[0105]

[0106] Where F∈R d×c , the i-th row of F is the function value assigned to each data point Tr(·) is the trace of the matrix.

[0107] It can be understood that according to theorem 3, the sum of the first c minimum eigenvalues of any Laplacian matrix L A and L A can be obtained as follows:

[0108]

[0109] Finally, the following representative scene extraction model can be obtained:

[0110]

[0111] Further, the representative photovoltaic scene can be obtained by the alternating optimization algorithm and the two-stage optimization extraction algorithm.

[0112] In actual execution process, the representative scene is divided into two stages for solving, the first stage uses optimization idea to extract intra-cluster morphological representative scene μ g * , and the second stage uses the same ratio amplification method to amplify it into the actual representative scene μ g . The first stage uses the intra-cluster morphological representative scene which is the power time sequence curve closest to the multi-dimensional similarity distance of other curves in the class, and the morphological representative scene μ g * is expressed as follows:

[0113]

[0114] Where P g represents the original sequence set of the cluster where the representative scene to be solved is located.

[0115] The second stage uses the same ratio amplification method to amplify it into the actual representative scene μ g , set to the original cluster power time series mean and shape representative scene μ g * The ratio of the mean value, the actual representative scene μ g is expressed as follows:

[0116]

[0117] Where N g is the number of power time series of the gth type; x g,i,t is the feature value of the ith power time series in the gth type at time t; μ g,t * is the feature value of the gth type at time t; P g represents the original sequence set of the cluster where the representative scene to be solved is located.

[0118] In summary, the embodiments of the present application measure the adjacency probability of data points based on multi-dimensional similarity distance, and obtain the actual representative scene of each photovoltaic power station through two-stage optimization extraction method, thereby completing the division and extraction of photovoltaic power output representative scene.

[0119] The embodiments of the present application can be described in detail as follows. As shown in Figure 2 , the method comprises the following steps:

[0120] Step S201: acquiring photovoltaic power generation time series data X=(x1, x2, …, x d ) to be clustered, the time series length being L.

[0121] Specifically, the photovoltaic power generation time series data X to be clustered is acquired, with the whole year of B years as the collection range, the collection frequency being 1h / time, 24 sampling data points per day, and a total of 365 data.

[0122] Step S202: Max-Min standardization processing is performed on X.

[0123] S203: K-means pre-clustering is performed on the standardized X, the number of clusters being 2-100, and in order to avoid the randomness of initialization, 100 times are run to obtain stable clustering results, and the cluster number corresponding to the fastest growth rate is selected.

[0124] S204: a clustering model is established and optimized to obtain a representative photovoltaic scene.

[0125] Specifically, the proposed clustering model is optimized by an alternating optimization algorithm including linear optimization to solve W, Lagrange multiplier method to solve A and eigenvalue decomposition F until the model converges; and

[0126] To make the proposed method more intuitive and clear, a visual method for representative photovoltaic scenes is provided, as shown in Figure 3 Fig. 6 is a result diagram of a photovoltaic representative scene according to an embodiment of the present application, and the proportion of each scene is shown in Table 1.

[0127] Table 1

[0128]

[0129] Further, after generating the target photovoltaic scene of the photovoltaic power station according to the multi-dimensional similarity measure of the photovoltaic power station, the method further includes: obtaining multiple index data of photovoltaic grid-connected random power flow, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value; and evaluating the accuracy level of the target photovoltaic scene based on the multiple index data of photovoltaic grid-connected random power flow, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value.

[0130] Specifically, as shown in Figure 2 Step S205, the embodiments of the present application use power flow analysis, annual power generation, peak time statistics and shape similarity analysis as evaluation indexes to compare the effectiveness of the embodiments of the present application and other related technologies. By using the index of photovoltaic grid-connected random power flow, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value, the accuracy, reliability and practicability of the representative photovoltaic scene extraction are verified from the angles of clustering algorithm and power.

[0131] Specifically, the peak time relative frequency is calculated as follows:

[0132] Calculate the peak time frequency of each category after standardization: for each category in the photovoltaic power generation data, calculate the peak time frequency of each hour, that is, count the time when the maximum value of the power of the category per day appears:

[0133]

[0134] where H t,g is the number of peaks of category g in the tth hour, and N g is the number of data points in category g.

[0135] For example, in order to verify the correctness of the representative photovoltaic power generation scene in the power system, the MATPOWER software package is used. Four comparison algorithms include sparse subspace clustering SSC (Sparse Subspace Clustering), constrained Laplacian rank based upon the L2-norm CLR-L2 (Constrained Laplacian Rank Based Upon the L2-norm), constrained Laplacian rank based upon the L1-norm CLR-L1 (Constrained Laplacian Rank Based Upon the L1-norm), and shared weight graph clustering SWGC (Shared Weight Graph Clustering), and the results of the SSC are the average values obtained by running 100 times in order to avoid randomness. Specifically, the photovoltaic power station is integrated into nodes 2, 22, 33, 18, and 25 of the case69 test case in turn, as shown in Figure 4 , a schematic diagram of the probability power flow result of the IEEE 69 bus power distribution system of an embodiment of the present application.

[0136] As can be seen from Figure 4 , in terms of the mean value, the model proposed in the embodiment of the present application is closest to the true value compared with the other five related technologies; in terms of the standard deviation, it is also very close to the original true scene, and the comparison result of the PPF result shows that the proposed model has good performance, and therefore the algorithm of the embodiment of the present application is superior to the comparison algorithm.

[0137] For example, in order to verify that the representative photovoltaic power generation scene retains the power characteristics of the original scene, the total power after clustering is calculated by replacing the original power generation data points with the representative scene extracted according to the model in this paper, and the total power of the photovoltaic power station and the power area of the representative scene after clustering are visualized and compared, as shown in Figure 5 . Wherein, the x-axis represents the output time of 365 days, the y-axis represents the power, the curve represents the daily power generation of the original data, and the bar chart represents the daily power generation of the extracted typical scene.

[0138] As can be seen from the figure, the daily power generation of the representative scene extracted by the representative scene extraction model of the embodiment of the present application mostly fits the daily power generation of the original data. As can be seen from the observation figure, the curve area is equal to the total area of the bar chart, and the cumulative power E obtained by the representative scene extraction model proposed in this paper is equal to the original total power, which shows that the photovoltaic representative scene extracted in the embodiment of the present application better reflects the total power of the original scene set.

[0139] For example, in order to verify that the representative photovoltaic power generation scene maintains the volatility characteristics of the original scene, the power generation amount of each time point of the five photovoltaic power stations is summed, the time point of the maximum power amount per day is the peak time of the day, the relative frequency of the peak time of the representative scene is compared with the relative frequency of the peak time of the original scene, and the results of the present technology are shown in FIG. 6. Figure 6 The larger the blue circle is, the greater the probability that the original scene is the peak time at this time is. The peak time of the representative scene is connected by a line.

[0140] The peak time of the representative scene extracted by the embodiment of the present application is consistent with the position of the peak time of the original scene, and the representative scene of most clusters can be consistent with the high-probability peak time of the original scene, which indicates that the clustering result of the representative scene extraction model of the embodiment of the present application is optimal under the evaluation index.

[0141] For example, in order to verify that the representative photovoltaic power generation scene maintains the shape similarity characteristics of the original scene, the shape trend similarity distance of the photovoltaic power generation data in the cluster and the representative scene of the cluster is calculated, and the maximum value of the similarity distance of the photovoltaic power generation data in each cluster and the representative scene of the cluster is compared. The smaller the value is, the more similar the photovoltaic power generation data is to the representative scene of the cluster. The results of each comparison technology are shown in Table 2.

[0142] Table 2

[0143]

[0144] By comparing the shape similarity distance in each cluster, compared with other technologies, the similarity distance calculated by the representative scene extraction model of the present application is mostly the minimum value, which indicates that the clustering result of the representative scene extraction model of the present application is optimal under the evaluation index. By comparing the maximum values of the 13 clusters, the maximum value of the shape similarity distance calculated by the representative scene extraction model of the present application is the smallest among these comparison models, which indicates that the clustering result of the representative scene extraction model of the present application is optimal under the evaluation index.

[0145] According to the data processing method of the photovoltaic power station proposed in the embodiment of the present application, the multi-dimensional similarity measurement of the photovoltaic power station is obtained by multi-dimensional data clustering of the time series data, which can more comprehensively reflect the actual characteristics of photovoltaic power generation, effectively extract representative photovoltaic scenes, reduce the data amount, and improve the efficiency of data analysis and processing.

[0146] Secondly, the data processing device of the photovoltaic power station according to the embodiment of the present application is described with reference to the accompanying drawings.

[0147] Figure 7is a block schematic diagram of a data processing device of a photovoltaic power station according to an embodiment of the present application.

[0148] As shown in Figure 7 , the data processing device 10 of the photovoltaic power station comprises an acquisition module 100, a clustering module 200 and a generation module 300.

[0149] The acquisition module 100 is configured to acquire time series data of the photovoltaic power station; the clustering module 200 is configured to perform multi-dimensional data clustering on the time series data to obtain multi-dimensional similarity metrics of the photovoltaic power station, wherein the multi-dimensions are a plurality of dimensions of power dimension, energy dimension and volatility dimension of the photovoltaic power station; and the generation module 300 is configured to generate a target photovoltaic scene of the photovoltaic power station according to the multi-dimensional similarity metrics of the photovoltaic power station.

[0150] In an embodiment of the present application, the clustering module 200 is further configured to perform pre-clustering on the time series data to obtain a target cluster number; and input data of the target cluster number into a clustering model, and the clustering model outputs the multi-dimensional similarity metrics of the photovoltaic power station, wherein the clustering model calculates similarity metrics, energy size similarity, cross-correlation sequence length and fluctuation position similarity of any two power time series in the data of the target cluster number, and generates the multi-dimensional similarity metrics according to the similarity metrics, the energy size similarity, the cross-correlation sequence length and the fluctuation position similarity.

[0151] In an embodiment of the present application, the clustering model is:

[0152] LESD(x i ,x j ,W)=w1L(x i ,x j )+w2E(x i ,x j )+w3S(x i ,x j )+w4δ(x i ,x j )

[0153] wherein w1, w2, w3 and w4 represent weight coefficients, and constitute a vector W∈R 4×1 , x i and x j represent a power time series with a length of L, x it , x jt represent features of the power time series x i and x j at the t-th moment, respectively.

[0154] E(x i ,xj ) represents x i and x j The magnitude similarity of the electric quantity of x

[0155] CC ω (x i ,x j ) = R ω-L (x i ,x j ), ω ∈ {1, 2, K, 2L-1},

[0156] CC w (x i ,x j ) represents the cross-correlation sequence of x i and x j , s represents the moving step of the power time sequence x j , s ∈ [-L, L]; w is the position index of the cross-correlation sequence CC w (x i ,x j );

[0157] S(x i ,x j ) = 1 - Max(NCC c (x i ,x j )), S(x i ,x j ) represents the morphological trend similarity distance of x i and x j , the value range is [0, 2];

[0158] represents the displacement of the sequence x j sliding to the most similar morphological trend of x i , δ(x i ,x j ) represents the similarity of the fluctuations of the two power time sequences.

[0159] In an embodiment of the present application, the data processing device 10 of the photovoltaic power station further comprises a determination module, configured to obtain the neighborhood probability of the power time sequence and other sequences before inputting the data of the target cluster number into the cluster model; and determine the neighborhood sequence of the power time sequence according to the neighborhood probability.

[0160] In an embodiment of the present application, the generation module 300 is further configured to extract a morphological representative scene from the multi-dimensional similarity measure of the photovoltaic power station; and magnify the morphological representative scene to obtain a target photovoltaic scene of the photovoltaic power station, wherein the target photovoltaic scene is:

[0161]

[0162] wherein, N g is the number of power time series of the gth type; x g,i,t is the feature value of the ith power time series of the gth type at time t; μ g,t * is the feature value of the gth type representative scene at time t. P g represents the original sequence set of the cluster where the representative scene to be solved is located.

[0163] In an embodiment of the present application, the data processing device 10 of the photovoltaic power station further comprises a processing module, configured to perform standardization processing on the time series data before obtaining the multi-dimensional similarity measure of the photovoltaic power station by performing multi-dimensional data clustering on the time series data.

[0164] In an embodiment of the present application, the data processing device 10 of the photovoltaic power station further comprises an evaluation module, configured to, after generating the target photovoltaic scene of the photovoltaic power station according to the multi-dimensional similarity measure of the photovoltaic power station, obtain a plurality of index data of random power flow of the photovoltaic grid-connected, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value; and evaluate the accuracy level of the target photovoltaic scene based on the plurality of index data of random power flow of the photovoltaic grid-connected, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value.

[0165] It should be noted that the foregoing explanation and description of the embodiment of the data processing method of the photovoltaic power station also applies to the data processing device of the embodiment of the photovoltaic power station, which will not be described here.

[0166] The data processing device of the photovoltaic power station according to the embodiment of the present application can more comprehensively reflect the actual characteristics of photovoltaic power generation by obtaining the multi-dimensional similarity measure of the photovoltaic power station through multi-dimensional data clustering on the time series data, effectively extracts representative photovoltaic scenes, reduces the amount of data, and improves the efficiency of data analysis and processing.

[0167] Figure 8 The electronic device provided in the embodiment of the present application is shown in the structural schematic diagram. The electronic device can include:

[0168] The memory 801, the processor 802, and the computer program stored in the memory 801 and executable on the processor 802.

[0169] The processor 802 implements the data processing method of the photovoltaic power station provided in the above embodiment when executing the program.

[0170] Further, the electronic device further includes:

[0171] The communication interface 803 is configured to communicate between the memory 801 and the processor 802.

[0172] The memory 801 is configured to store a computer program executable on the processor 802.

[0173] The memory 801 can include a high-speed RAM (Random Access Memory) memory, and can further include a nonvolatile memory such as at least one disk memory.

[0174] If the memory 801, the processor 802 and the communication interface 803 are independently implemented, the communication interface 803, the memory 801 and the processor 802 can be connected through a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 8 In the figure, only one thick line is used to represent that there is only one bus or only one type of bus.

[0175] Optionally, in a specific implementation, if the memory 801, the processor 802 and the communication interface 803 are integrated on a chip, the memory 801, the processor 802 and the communication interface 803 can communicate with each other through an internal interface.

[0176] The processor 802 can be a CPU (Central Processing Unit) or an ASIC (Application Specific Integrated Circuit) or one or more integrated circuits configured to implement the embodiments of the present application.

[0177] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the data processing method of the photovoltaic power station as above.

[0178] In the description of the application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In the description of the application, the illustrative description of the above terms is not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or N embodiments or examples. In addition, different embodiments or examples described in the description of the application and the features of different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0179] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0180] Any process or method descriptions in flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for implementing specific logic functions (or steps) in the process, and that the various embodiments of the application can include additional or fewer steps or processes in alternative implementations, as will be appreciated by those skilled in the art. The various embodiments of the application can be implemented in hardware, software, firmware, or a combination thereof, as desired.

[0181] It should be understood that parts of the application can be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. As in another embodiment implemented in hardware, any one or a combination of the following technologies known in the art can be used: discrete logic circuit with logic gate circuit for implementing logic functions on data signals, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array, field programmable gate array, etc.

[0182] Those skilled in the art of the art can understand that the method of implementing the above-mentioned embodiments carries out all or part of the steps. The above-mentioned program can be stored in a computer readable storage medium, which includes one or a combination of the steps of the method embodiment when executed.

[0183] Although the embodiments of the present application have been shown and described above, it is understood that the above-described embodiments are exemplary and are not to be construed as limiting the present application, and that changes, modifications, substitutions and variations can be made by those skilled in the art without departing from the scope of the present application.

Claims

1. A data processing method for a photovoltaic power plant, characterized in that, The method comprises the following steps: acquiring time series data of a photovoltaic power station; performing multi-dimensional data clustering on the time series data to obtain a multi-dimensional similarity measure of the photovoltaic power station, wherein the multi-dimensional data comprises power dimension, energy dimension and fluctuation dimension of the photovoltaic power station; generating a target photovoltaic scenario of the photovoltaic power station according to the multi-dimensional similarity measure of the photovoltaic power station; the multi-dimensional data clustering on the time series data to obtain the multi-dimensional similarity measure of the photovoltaic power station comprises: performing pre-clustering on the time series data to obtain a target cluster number; inputting data of the target cluster number into a clustering model, wherein the clustering model outputs the multi-dimensional similarity measure of the photovoltaic power station, wherein the clustering model calculates similarity measure, power size similarity, cross-correlation sequence length and fluctuation position similarity of any two power time series in the data of the target cluster number, and generates a multi-dimensional similarity measure according to the similarity measure, the power size similarity, the cross-correlation sequence length and the fluctuation position similarity, wherein the clustering model is: LESD(x i ,x j ,W) = w1L(x i ,x j ) + w2E(x i ,x j ) + w3S(x i ,x j ) + w4δ(x i ,x j ) where w1, w2, w3 and w4 represent weight coefficients, which constitute a vector W e R 4×1 , x i and x j represent a power time series with length L, x it , x jt represent the features of the power time series x i and x j at the t-th moment, respectively; E(x i ,x j ) represents the similarity of the magnitude of the electric quantity of x i and x j ; CC ω (x i ,x j )=R ω-L (x i ,x j ),ω∈{1,2,K,2L-1}, CC w (x i ,x j ) represents the cross-correlation sequence of x i and x j , s represents the moving step of the power time sequence x j , s ∈ [-L, L]; w is the position index of the cross-correlation sequence CC w (x i ,x j ). S(x i ,x j ) = 1 - Max(NCC c (x i ,x j )), S(x i ,x j ) represents the morphological trend similarity distance between x i and x j , the value range is [0, 2], and Max(NCC c (x i , x j )) represents the maximum value in the coefficient normalized cross-correlation sequence; denotes the sequence x j slides to and x i displacement of the morphological trend most similar, δ(x i , x j ) denotes the similarity of the fluctuations of the two power time series; the generating of the target photovoltaic scenario of the photovoltaic power station according to the multi-dimensional similarity measure of the photovoltaic power station comprises: extracting a representative scenario of the target photovoltaic scenario from the multi-dimensional similarity measure of the photovoltaic power station; and magnifying the representative scenario to obtain the target photovoltaic scenario of the photovoltaic power station, wherein the target photovoltaic scenario is: wherein N g is the number of power time series of the gth type; x g,i,t is the feature value of the ith power time series of the gth type at time t, is the representative scene of the gth type, μ g,t * is the feature value of the representative scene of the gth type at time t, P g represents the original sequence set of the cluster where the representative scene to be solved is located, LESD(μ g , x i , W) represents the multi-dimensional similarity measure value of μ g and x i .

2. A data processing method for a photovoltaic power plant according to claim 1, characterized in that, before inputting the data of the target cluster number into the clustering model, further comprising: acquiring a neighborhood probability of a power time series and other sequences; determining a neighborhood sequence of the power time series according to the neighborhood probability.

3. A method of processing data of a photovoltaic power plant according to claim 1, characterized in that, before performing the multi-dimensional data clustering on the time series data to obtain the multi-dimensional similarity measure of the photovoltaic power station, further comprising: performing standardization processing on the time series data.

4. A method of processing data of a photovoltaic power plant according to claim 1, characterized in that, after generating the target photovoltaic scenario of the photovoltaic power station according to the multi-dimensional similarity measure of the photovoltaic power station, further comprising: acquiring a plurality of index data of random power flow, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value of photovoltaic grid connection; based on the plurality of index data of random power flow, annual power generation, photovoltaic daily power generation curve shape similarity and photovoltaic power generation peak period frequency value of photovoltaic grid connection, evaluating the accuracy level of the target photovoltaic scenario.

5. A data processing device for a photovoltaic power plant, characterized in that comprise: an acquisition module, configured to acquire time series data of a photovoltaic power station; a clustering module, configured to perform multi-dimensional data clustering on the time series data to obtain a multi-dimensional similarity measure of the photovoltaic power station, wherein the multi-dimensional data comprises power dimension, energy dimension and fluctuation dimension of the photovoltaic power station; a generation module, configured to generate a target photovoltaic scenario of the photovoltaic power station according to the multi-dimensional similarity measure of the photovoltaic power station; The clustering module is further configured to: pre-cluster the time series data to obtain a target cluster number; input data of the target cluster number into a clustering model, and the clustering model outputs a multi-dimensional similarity measure of the photovoltaic power station, wherein the clustering model calculates a similarity measure, a power size similarity, a cross-correlation sequence length, and a fluctuation position similarity of any two power time series in the data of the target cluster number, and generates a multi-dimensional similarity measure according to the similarity measure, the power size similarity, the cross-correlation sequence length, and the fluctuation position similarity, wherein the clustering model is: LESD(x i ,x j ,W) = w1L(x i ,x j ) + w2E(x i ,x j ) + w3S(x i ,x j ) + w4δ(x i ,x j ) where w1, w2, w3 and w4 represent weight coefficients, which constitute a vector W e R 4×1 , x i and x j respectively represent a power time sequence with a length of L, x it , x jt respectively represent features of the power time sequence x i and x j at the t time. E(x i ,x j ) represents the electrical magnitude similarity of x i and x j ; CC ω (x i ,x j )=R ω-L (x i ,x j ),ω∈{1,2,K,2L-1}, CC w (x i ,x j ) represents the cross-correlation sequence of x i and x j , s represents the moving step of the power time sequence x j , s ∈ [-L, L]; w is the position index of the cross-correlation sequence CC w (x i ,x j ). S(x i ,x j ) = 1 - Max(NCC c (x i ,x j )), S(x i ,x j ) represents the morphological trend similarity distance between x i and x j , the value range is [0, 2], and Max(NCC c (x i , x j )) represents the maximum value in the coefficient normalized cross-correlation sequence; denotes the sequence x j slides to and x i displacement, δ(x i , x j ) denotes the similarity of the fluctuations of the two power time series; The generation module is further configured to: extract a representative scenario of the form from the multi-dimensional similarity measure of the photovoltaic power station; and magnify the representative scenario of the form to obtain a target photovoltaic scenario of the photovoltaic power station, wherein the target photovoltaic scenario is: wherein N g is the number of power time series of the gth type; x g,i,t is the feature value of the ith power time series of the gth type at time t, is the representative scene of the gth type, μ g,t * is the feature value of the representative scene of the gth type at time t, P g denotes the original sequence set of the cluster where the representative scene to be solved is located, LESD(μ g , x i , W) denotes the multi-dimensional similarity measure value of μ g and x i .

6. An electronic device, comprising: comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the program to implement the data processing method of the photovoltaic power station according to any one of claims 1-4.

7. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions are executed to implement the data processing method of the photovoltaic power station according to any one of claims 1-4.

Images