Power generation method and system for new energy stations in power grid considering spatial correlation

Abstract

The present invention discloses a method and system for generating power at a power grid renewable energy station that considers spatial correlation, comprising: calculating the correlation coefficient of any two different renewable energy stations based on a power change sequence; calculating the distance between stations using the correlation coefficient, and clustering based on the distance to form multiple station clusters; constructing a correlation coefficient matrix for each station cluster, and generating a new renewable energy station power change sequence through Cholesky decomposition. The present invention accurately quantifies the correlation characteristics between renewable energy units through an improved correlation analysis method; utilizes an optimized clustering algorithm to group highly correlated units, significantly reducing the sample space dimension; and adopts a random generation method based on an advanced statistical model to generate a renewable energy power sequence with true randomness and correlation, thereby maintaining the randomness of renewable energy power output while ensuring the correlation structure between units.

Description

Technical Field

[0001] The present invention relates to the field of new energy technologies, and more particularly, to a method and system for generating power at a new energy station in a power grid taking spatial correlation into consideration. Background Art

[0002] With the rapid transformation of the global energy structure and the booming development of new energy technologies, high-penetration renewable energy power grids are becoming a key development direction for the power industry. However, the power output of renewable energy stations is characterized by significant randomness and volatility, which poses significant challenges to the stable operation and dispatch of the power grid.

[0003] In power grids with a high proportion of renewable energy integration, the operating space for conventional generators is gradually shrinking. Simultaneously, problems such as reduced frequency regulation capabilities, insufficient reactive power support, exacerbated voltage stability issues, and increased risk of over-limit surges are also emerging. Traditional grid control strategies are no longer able to meet these complex challenges, especially with the large-scale integration of renewable energy.

[0004] In existing technologies, power forecasting and optimization methods are mainly based on time series analysis of historical data. For example, the autoregressive integrated moving average (ARIMA) model is a typical statistical method that is widely used for short-term power forecasting. This model stabilizes the non-stationary time series by performing differential processing on historical power data, and then uses autoregression and moving average to predict future power values. However, the ARIMA model mainly focuses on changes in the time dimension and does not fully consider the spatial correlation between stations in the power forecast of new energy stations.

[0005] In reality, renewable energy sites often exhibit strong correlations in geographic location and meteorological conditions. For example, multiple wind farms in the same region may be simultaneously affected by the same meteorological conditions, resulting in similar fluctuations in their power output. Existing technologies, ignoring these spatial correlations, suffer from significant inaccuracies in their predictions for clusters of renewable energy sites. Furthermore, existing technologies are relatively simplistic in data preprocessing and outlier handling, potentially making them overly sensitive to abnormal data and thus impacting the reliability of predictions.

[0006] Furthermore, with the rapid growth in the number of renewable energy stations and the scale of data, the computational complexity of existing statistical methods has increased significantly, resulting in excessive consumption of computing resources when processing large-scale renewable energy grid data, affecting the feasibility of real-time applications. Furthermore, existing methods have limited generalization capabilities for predictive models due to the high volatility of renewable energy power output and insufficient data from newly built stations, making it difficult to accurately respond to the complex and volatile power fluctuations of renewable energy stations.

[0007] Therefore, in the field of power prediction in power grids with high penetration of new energy, there is an urgent need for an intelligent power generation method for new energy sites that takes spatial correlation into account. Summary of the Invention

[0008] The present invention proposes a method and system for generating power of a new energy station in a power grid taking spatial correlation into consideration, so as to solve the problem of how to efficiently generate power data of the new energy station.

[0009] In order to solve the above problem, according to one aspect of the present invention, a method for generating power at a new energy station in a power grid considering spatial correlation is provided, the method comprising:

[0010] Calculate the correlation coefficient between any two different new energy stations based on the power change sequence;

[0011] Calculating the distance between stations based on the correlation coefficient, and clustering based on the distance to form a plurality of station clusters;

[0012] For each station group, a correlation coefficient matrix is ​​constructed, and a new power change sequence of new energy stations is generated through Cholesky decomposition.

[0013] Preferably, the calculation of the correlation coefficient between any two different new energy stations based on the power change sequence includes:

[0014]

[0015] Among them, ρ X,Y is the correlation coefficient between the power change sequences X and Y of the two new energy stations; X t and Y t is the power value at the tth time point; n is the number of time points; and is the mean of the corresponding sequence; w t is the weight coefficient based on the power change rate; σ X and σ Y are the standard deviations of series X and Y, respectively.

[0016] Preferably, the distance between stations is calculated based on the correlation coefficient, comprising:

[0017]

[0018] Among them, d ij is the distance between new energy station i and new energy station j; ρ X,Y Represents the correlation coefficient between the power change sequences X and Y corresponding to the new energy stations i and j.

[0019] Preferably, the method further comprises:

[0020] Set the initial range of the number of clusters K according to the data scale and expected clustering granularity [K min ,K max ];

[0021] For each candidate K value, perform the improved K-Means clustering algorithm to calculate the corresponding global weighted silhouette coefficient And look for The largest K value is taken as the optimal number of clusters K*; if [K min ,K max ] If the optimal value is not found within the range, then according to Expand your search scope based on the changing trends of the

[0022] Preferably, the global weighted silhouette coefficient is determined in the following manner, comprising:

[0023]

[0024] Among them, a i represents the average distance between sample i and its cluster and other samples; b i represents the average distance between sample i and the nearest samples in other clusters; w i represents the weight coefficient of the i-th sample, which is determined by the installed capacity, historical power generation and data quality of the new energy unit; s i represents the weighted silhouette coefficient of the i-th sample, s i >0 indicates that the samples are reasonably clustered, with close clusters and separation between clusters; i =0 means that the sample is located at the boundary of two clusters and the clustering effect is average; s i <0 indicates that the samples may be clustered incorrectly, the intra-cluster distance is greater than the inter-cluster distance, and one new energy station is one sample.

[0025] Preferably, for each station group, constructing a correlation coefficient matrix and generating a new new energy station power change sequence through Cholesky decomposition includes:

[0026] For each station group, the corresponding correlation coefficient submatrix R is extracted from the correlation coefficient matrix constructed based on the correlation coefficient;

[0027] The correlation coefficient matrix is ​​decomposed into R = L·L based on the advanced statistical model of Cholesky decomposition. T ; Where L is a lower triangular matrix over the real field;

[0028] Generate a random matrix Z=randn(k,m) that conforms to the standard normal distribution, Where k is the length of data to be generated, m is the number of stations in the station group, and randn represents sampling from the standard normal distribution N(0,1);

[0029] The random matrix Z is linearly transformed by the lower triangular matrix L to introduce the specified correlation structure, and S = Z·L is obtained T ;

[0030] Perform inverse normalization on each column of the matrix S to obtain Each column of the generated random data matrix S is used as a new power change sequence of the new energy station that satisfies the correlation coefficient submatrix R;

[0031] in,

[0032]

[0033] Among them, S ij is a correlation-mapped matrix, each column of which is a standardized random sequence with a specified correlation; S ij ′ is the power change value; and Var(X i ) are the mean and variance of the power series of the i-th new energy station; n is the length of the time series; A T represents the transpose of A.

[0034] According to another aspect of the present invention, a power generation system for a new energy station in a power grid taking spatial correlation into consideration is provided, the system comprising:

[0035] A correlation coefficient calculation unit, used to calculate the correlation coefficient of any two different new energy stations based on the power change sequence;

[0036] A clustering unit, configured to calculate distances between stations based on correlation coefficients, and cluster the stations based on the distances to form a plurality of station clusters;

[0037] The data generation unit is used to construct a correlation coefficient matrix for each station group and generate a new power change sequence of new energy stations through Cholesky decomposition.

[0038] Preferably, the correlation coefficient calculation unit calculates the correlation coefficient of any two different new energy stations based on the power change sequence, including:

[0039]

[0040] Among them, ρ X,Y is the correlation coefficient between the power change sequences X and Y of the two new energy stations; X t and Y t is the power value at the tth time point; n is the number of time points; and is the mean of the corresponding sequence; w t is the weight coefficient based on the power change rate; σ X and σ Y are the standard deviations of series X and Y, respectively.

[0041] Preferably, the clustering unit calculates the distance between stations based on the correlation coefficient, comprising:

[0042]

[0043] Among them, d ij is the distance between new energy station i and new energy station j; ρ X,Y Represents the correlation coefficient between the power change sequences X and Y corresponding to the new energy stations i and j.

[0044] Preferably, the clustering unit further comprises:

[0045] Set the initial range of the number of clusters K according to the data scale and expected clustering granularity [K min ,K max ];

[0046] For each candidate K value, perform the improved K-Means clustering algorithm to calculate the corresponding global weighted silhouette coefficient And look for The largest K value is taken as the optimal number of clusters K*; if [K min ,K max ] If the optimal value is not found within the range, then according to Expand your search scope based on the changing trends of the

[0047] Preferably, the clustering unit determines the global weighted silhouette coefficient in the following manner, including:

[0048]

[0049] Among them, a i represents the average distance between sample i and its cluster and other samples; b i represents the average distance between sample i and the nearest samples in other clusters; w i represents the weight coefficient of the i-th sample, which is determined by the installed capacity, historical power generation and data quality of the new energy unit; s i represents the weighted silhouette coefficient of the i-th sample, s i >0 indicates that the samples are reasonably clustered, with close clusters and separation between clusters; i =0 means that the sample is located at the boundary of two clusters and the clustering effect is average; s i<0 indicates that the samples may be clustered incorrectly, the intra-cluster distance is greater than the inter-cluster distance, and one new energy station is one sample.

[0050] Preferably, the data generation unit constructs a correlation coefficient matrix for each station group and generates a new new energy station power change sequence through Cholesky decomposition, including:

[0051] For each station group, the corresponding correlation coefficient submatrix R is extracted from the correlation coefficient matrix constructed based on the correlation coefficient;

[0052] The correlation coefficient matrix is ​​decomposed into R = L·L based on the advanced statistical model of Cholesky decomposition. T ; Where L is a lower triangular matrix over the real field;

[0053] Generate a random matrix Z=randn(k,m) that conforms to the standard normal distribution, Where k is the length of data to be generated, m is the number of stations in the station group, and randn represents sampling from the standard normal distribution N(0,1);

[0054] The random matrix Z is linearly transformed by the lower triangular matrix L to introduce the specified correlation structure, and S = Z·L is obtained T ;

[0055] Perform inverse normalization on each column of the matrix S to obtain Each column of the generated random data matrix S is used as a new power change sequence of the new energy station that satisfies the correlation coefficient submatrix R;

[0056] in,

[0057]

[0058] Among them, S ij is a correlation-mapped matrix, each column of which is a standardized random sequence with a specified correlation; S ij ′ is the power change value; and Var(X i ) are the mean and variance of the power series of the i-th new energy station; n is the length of the time series; A T represents the transpose of A.

[0059] The present invention provides a method and system for generating power at renewable energy stations in a power grid that takes spatial correlation into account, comprising: calculating the correlation coefficient of any two different renewable energy stations based on a power change sequence; calculating the distance between stations based on the correlation coefficient, and clustering based on the distance to form multiple station clusters; constructing a correlation coefficient matrix for each station cluster, and generating a new renewable energy station power change sequence through Cholesky decomposition. The present invention accurately quantifies the correlation characteristics between renewable energy units through an improved correlation analysis method; utilizes an optimized clustering algorithm to group highly correlated units together, significantly reducing the sample space dimension; and adopts a random generation method based on an advanced statistical model to generate a renewable energy power sequence with true randomness and correlation, thereby maintaining the randomness of renewable energy power output while ensuring the correlation structure between units, and having high efficiency and applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0061] Figure 1 Flowchart of a method 100 for generating power at a new energy station in a power grid considering spatial correlation according to an embodiment of the present invention;

[0062] Figure 2 is an example diagram of a distribution histogram of all correlation coefficients according to an embodiment of the present invention;

[0063] Figure 3 is an example diagram of a distribution histogram of maximum correlation coefficients of all stations according to an embodiment of the present invention;

[0064] Figure 4 4 is a schematic structural diagram of a power generation system 400 for a power grid new energy station considering spatial correlation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0065] Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to provide a thorough and complete disclosure of the present invention and to fully convey the scope of the present invention to those skilled in the art. The terminology used in the exemplary embodiments shown in the accompanying drawings is not intended to limit the present invention. In the accompanying drawings, identical elements are denoted by the same reference numerals.

[0066] Unless otherwise specified, the terms used herein (including technical terms) have the meanings commonly understood by those skilled in the art. In addition, it is understood that terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0067] The present invention focuses on optimizing the intelligent generation of power at new energy stations through spatial correlation analysis, further improving the efficiency and stability of the power grid. The core of this method is to use spatial correlation analysis to deeply explore the interdependence of power generation between new energy stations, thereby achieving high-precision prediction and control of power fluctuations. Through refined statistical analysis, the inherent connections between different new energy stations can be revealed, providing a solid data foundation for the effective integration and utilization of new energy. In particular, the method of the present invention processes the power data of wind farms and photovoltaic power stations separately, ensuring the accuracy and reliability of the analysis.

[0068] Figure 1 FIG. 1 is a flow chart of a method 100 for generating power at a new energy station in a power grid considering spatial correlation according to an embodiment of the present invention. Figure 1 As shown, the method for generating power at a new energy station in a power grid that takes spatial correlation into account, provided by an embodiment of the present invention, accurately quantifies the correlation characteristics between new energy units through an improved correlation analysis method; utilizes an optimized clustering algorithm to group highly correlated units together, significantly reducing the sample space dimension; and adopts a random generation method based on an advanced statistical model to generate a new energy power sequence with true randomness and correlation, thereby maintaining the randomness of the new energy power output while ensuring the correlation structure between units, and having high efficiency and applicability. The method 100 for generating power at a new energy station in a power grid that takes spatial correlation into account, provided by an embodiment of the present invention, starts at step 101. In step 101, the correlation coefficient of any two different new energy stations is calculated based on the power change sequence.

[0069] Preferably, the calculation of the correlation coefficient between any two different new energy stations based on the power change sequence includes:

[0070]

[0071]

[0072] Among them, ρ X,Y is the correlation coefficient between the power change sequences X and Y of the two new energy stations; X t and Y t is the power value at the tth time point; n is the number of time points; and is the mean of the corresponding sequence; w tis the weight coefficient based on the power change rate; σ X and σ Y are the standard deviations of series X and Y, respectively.

[0073] In the present invention, spatial correlation analysis is the core link, which aims to deeply explore the power output correlation between different new energy stations and provide a solid foundation for subsequent data generation and model optimization. Traditional statistical analysis methods usually use the standard Pearson correlation coefficient to measure the degree of linear correlation between two variables. However, when processing the power data of new energy stations, the standard Pearson correlation coefficient may not be able to fully capture the complex characteristics of the data, especially when facing nonlinear relationships and outliers. Therefore, in the present invention, spatial correlation analysis is performed in the following way, as follows:

[0074]

[0075] In the formula, X and Y represent the power change sequences of two different new energy stations, X t and Y t is the power value at the tth time point; n is the number of time points; and is the mean of the corresponding sequence; w t is a weight coefficient based on the power change rate, which is used to reduce the impact of outliers on the correlation calculation and is defined as follows:

[0076]

[0077] Among them, σ X and σ Y are the standard deviations of sequences X and Y, respectively. This weight coefficient utilizes the idea of ​​the Gaussian kernel function. When the power change rate is large (possibly an abnormal fluctuation), the weight will be reduced, thereby reducing its impact on the correlation calculation.

[0078] In step 102, the distances between stations are calculated based on the correlation coefficient, and clustering is performed based on the distances to form a plurality of station groups.

[0079] Preferably, the distance between stations is calculated based on the correlation coefficient, comprising:

[0080]

[0081] Among them, d ij is the distance between new energy station i and new energy station j; ρ X,Y Represents the correlation coefficient between the power change sequences X and Y corresponding to the new energy stations i and j.

[0082] Preferably, the method further comprises:

[0083] Set the initial range of the number of clusters K according to the data scale and expected clustering granularity [K min ,K max ];

[0084] For each candidate K value, perform the improved K-Means clustering algorithm to calculate the corresponding global weighted silhouette coefficient And look for The largest K value is taken as the optimal number of clusters K*; if [K min ,K max ] If the optimal value is not found within the range, then according to Expand your search scope based on the changing trends of the

[0085] Preferably, the global weighted silhouette coefficient is determined in the following manner, comprising:

[0086]

[0087] Among them, a i represents the average distance between sample i and its cluster and other samples; b i represents the average distance between sample i and the nearest samples in other clusters; w i represents the weight coefficient of the i-th sample, which is determined by the installed capacity, historical power generation and data quality of the new energy unit; s i represents the weighted silhouette coefficient of the i-th sample, s i >0 indicates that the samples are reasonably clustered, with close clusters and separation between clusters; i =0 means that the sample is located at the boundary of two clusters and the clustering effect is average; s i <0 indicates that the samples may be clustered incorrectly, the intra-cluster distance is greater than the inter-cluster distance, and one new energy station is one sample.

[0088] In this paper, to effectively group renewable energy stations, a K-means algorithm was adopted, and innovative optimization was performed based on the characteristics of renewable energy station power data. The traditional K-means algorithm typically uses Euclidean distance as a similarity metric, but this approach has limitations when dealing with negative or zero values ​​in renewable energy station power data that violate the non-negativity principle of distance metric and cannot fully reflect the spatial correlation between stations and the similarity of power output.

[0089] To solve this problem, this paper proposes a distance measurement method based on correlation coefficient, which directly uses the correlation between stations for distance calculation. Specifically, the following distance conversion formula is designed:

[0090]

[0091] Among them, d ij represents the distance between station i and station j, ρ X,Y is the correlation coefficient between the power change sequences X and Y of the two new energy stations. The correlation coefficient is converted into a distance value through the nonlinear mapping of the exponential function. ij When it is close to 1 (i.e. highly correlated), the corresponding distance d ij Close to 0; when ρ ij When it is close to 0 or negative (i.e. weak correlation or negative correlation), the distance d ij This mapping method can more keenly capture the subtle differences in the correlation between stations and improve the clustering resolution. At the same time, it ensures that the distance d ij It is always non-negative, meeting the basic requirements of distance measurement.

[0092] For the K-Means clustering algorithm, one of the key parameters is the number of clusters K, the selection of which directly affects the effectiveness of the clustering results and the accuracy of subsequent analysis. In the clustering of power data of new energy stations, the traditional fixed K value method may not fully capture the complex structure of the data, resulting in poor clustering results. To this end, this application proposes an adaptive K value selection method that combines the silhouette coefficient (Silhouette Coefficient) and data characteristics to accurately determine the optimal number of clusters.

[0093] The traditional silhouette coefficient is defined as follows:

[0094]

[0095] Where a is the average distance between a sample and other samples in the same cluster, reflecting the compactness within the cluster; b is the average distance between a sample and the samples in the nearest cluster, reflecting the separation between clusters.

[0096] However, in the power data of new energy stations, the correlation between stations may be affected by multiple factors such as geographical location and meteorological conditions, resulting in uneven data distribution. To this end, the present invention improves the silhouette coefficient and adds a weight factor w i To reflect the importance or representativeness of each station.

[0097]

[0098] Weight factor w i It can be determined based on the installed capacity, historical power generation or data quality of the site. The global weighted silhouette coefficient is obtained by averaging the weighted silhouette coefficients of all samples.

[0099]

[0100] Among them, ai represents the average distance between sample i and its cluster and other samples (intra-cluster distance); b i represents the average distance between sample i and the nearest samples in other clusters (the nearest inter-cluster distance); w i The weight coefficient of the i-th sample reflects the importance or representativeness of the sample, which is determined by the installed capacity, historical power generation and data quality of the new energy unit in this patent; s i represents the weighted silhouette coefficient of the i-th sample, s i >0 indicates that the samples are reasonably clustered, with close clusters and separation between clusters; i =0 means that the sample is located at the boundary of two clusters and the clustering effect is average; s i <0 indicates that the samples may be clustered incorrectly, and the intra-cluster distance is greater than the inter-cluster distance.

[0101] In order to find the optimal K value, the present invention designs an adaptive search algorithm: First, the initial range of K value [K min ,K max ], for each candidate K value, perform the improved K-Means clustering algorithm and calculate the corresponding global weighted silhouette coefficient And look for The largest K value is taken as the optimal number of clusters K*. If [K min ,K max ] No obvious optimal value is found within the range, so According to the changing trend of the search results, expand the search scope appropriately.

[0102] In step 103, for each station group, a correlation coefficient matrix is ​​constructed, and a new new energy station power change sequence is generated through Cholesky decomposition.

[0103] Preferably, for each station group, constructing a correlation coefficient matrix and generating a new new energy station power change sequence through Cholesky decomposition includes:

[0104] For each station group, the corresponding correlation coefficient submatrix R is extracted from the correlation coefficient matrix constructed based on the correlation coefficient;

[0105] The correlation coefficient matrix is ​​decomposed into R = L·L based on the advanced statistical model of Cholesky decomposition. T ; Where L is a lower triangular matrix over the real field;

[0106] Generate a random matrix Z=randn(k,m) that conforms to the standard normal distribution, Where k is the length of data to be generated, m is the number of stations in the station group, and randn represents sampling from the standard normal distribution N(0,1);

[0107] The random matrix Z is linearly transformed by the lower triangular matrix L to introduce the specified correlation structure, and S = Z·L is obtained T ;

[0108] Perform inverse normalization on each column of the matrix S to obtain Each column of the generated random data matrix S is used as a new power change sequence of the new energy station that satisfies the correlation coefficient submatrix R;

[0109] in,

[0110]

[0111] Among them, S ij is a correlation-mapped matrix, each column of which is a standardized random sequence with a specified correlation; S ij ′ is the power change value; and Var(X i ) are the mean and variance of the power series of the i-th new energy station; n is the length of the time series; A T represents the transpose of A.

[0112] The random generation process of the present invention is based on an advanced statistical model, aiming to maintain the correlation structure between each subsequence in the generated sample data consistent with the correlation coefficient matrix R of the actual observed data. To this end, the present invention systematically optimizes the extraction and construction of the correlation coefficient submatrix and the random data generation process.

[0113] First, for each station cluster, the corresponding correlation coefficient submatrix R is extracted from the global correlation coefficient matrix. This submatrix R is a symmetric matrix, and its element values ​​usually range from 0.5 to 1.0, reflecting the strong correlation between stations within the station cluster. This provides a reliable statistical basis for subsequent data generation. In order to construct a matrix A that can accurately reflect the correlation between stations within the cluster, the original data is standardized. The constructed matrix A is as follows:

[0114]

[0115] Among them, d ij represents the power change value of the i-th station at the j-th time point, and Var(X i ) are the mean and variance of the power sequence of the ith station, m is the number of stations in the group, and n is the length of the time series.

[0116] Through the above standardization process, each column of matrix A is standard normalized data with a mean of zero and a variance of one. Next, verify the properties of the correlation coefficient submatrix R constructed from matrix A. Specifically, matrix R can be expressed as:

[0117]

[0118] To ensure the accuracy and stability of R, it is proved that R is a semi-positive definite matrix. When the column vectors of matrix A are linearly independent (i.e., the column of A is full rank), R is a positive definite matrix. The proof process is as follows: For any non-zero vector There is b T Rb=b T A T Ab=(Ab) T ·Ab=||Ab|| 2 ≥0. Therefore, R is a semi-positive definite matrix. Since in practical applications, the data of matrix A comes from real measurement data and usually satisfies n>>m, the possibility of linear correlation between the column vectors of A is extremely low. It can be considered that the columns of A are full rank, and thus R is a positive definite matrix. After confirming that R is a positive definite matrix, we use the Cholesky decomposition technique to decompose it into R = L·L T Where L is a lower triangular matrix over the real field. The existence and uniqueness of the Cholesky decomposition ensure the stability and efficiency of the decomposition process. This decomposition provides a solid mathematical foundation for subsequent random data generation. Next, in order to generate a random data sequence with a specified correlation structure, the following steps are taken:

[0119] Generate a random matrix that follows a standard normal distribution Where k is the length of data to be generated (i.e., the number of samples), m is the number of stations in the station group, Z = randn(k,m), and randn represents sampling from the standard normal distribution N(0,1).

[0120] The random matrix Z is linearly transformed by the lower triangular matrix L to introduce the specified correlation structure: S = Z·L T , after this linear transformation, the column vectors of matrix S have a correlation structure consistent with R.

[0121] To restore the actual scale of the data, we denormalize each column of the matrix S using the previously calculated mean and standard deviation Var(X i ):

[0122]

[0123] Through the above steps, each column of the generated random data matrix S can be obtained as a station power change sequence that satisfies the preset correlation coefficient matrix R.

[0124] The present invention focuses on the spatial correlation between new energy stations, and adopts an improved Pearson correlation coefficient, a distance conversion formula based on the correlation coefficient, and an optimized K-means clustering algorithm to effectively capture the mutual influence and dependency between new energy stations. In addition, by utilizing advanced statistical models and algorithms, it is possible to accurately generate new energy power data based on the correlation between stations, providing an efficient and reliable method for intelligent power generation for new power grids. This technology not only improves the accuracy of new energy power prediction, but also provides strong data support for new energy integration and power grid optimization, showing obvious advancement and practical value. Through the method of the present invention, the effective reduction of the original feature space dimension is achieved, the difficulty of spatial sampling and sample preparation is simplified, and the efficiency of sequence generation is significantly improved.

[0125] To validate the effectiveness and advancement of our method, we conducted a statistical analysis of power data from renewable energy stations based on online data from a provincial power grid with a high penetration of renewable energy. The dataset includes operational data from 69 renewable energy stations (including wind farms and photovoltaic power plants) covering a total of 8,457 time points.

[0126] First, using the improved Pearson correlation coefficient, we conducted a detailed spatial correlation analysis of the power data of 69 new energy sites (including wind farms and photovoltaic power stations). We constructed a 69×69 correlation coefficient matrix. Excluding the part with a correlation coefficient of 1.0 on the diagonal, the number of remaining correlation coefficients is 4692. This method introduces weight coefficients through prior knowledge, reduces the impact of outliers and data noise on the correlation calculation, and improves the accuracy and robustness of the analysis. Among the total 4692 site pairs, there are 3018 positively correlated site pairs, accounting for 64.32%. Strong positive correlation (|ρ X,Y There are 1446 station pairs with a correlation coefficient greater than 0.5|, accounting for 30.82%. All stations have a strong positive correlation with at least one other station. These results fully verify the effectiveness of our improved correlation analysis method. Strong positive correlation mainly occurs between stations with close geographical locations, which is highly consistent with the similarity of meteorological conditions in local areas, reflecting the accurate capture of spatial correlation by the method. The distribution of all correlation coefficient results is shown in the figure. Figure 2 As shown in Figure 2, since this analysis combines the data of wind farms and photovoltaic power stations, the proportion of weak negative correlation is relatively high. Further statistics on each new energy station show that all stations have at least one other station with strong correlation. The distribution of the maximum correlation coefficient of all wind farms is shown in Figure 2. Figure 3 shown.

[0127] To address the unique characteristics of power data from renewable energy stations, the previously proposed K-Means clustering algorithm, optimized using a distance metric based on correlation coefficients, was used. To determine the optimal number of clusters, K, an improved silhouette coefficient method was employed. This method, by introducing a weighting factor, comprehensively considers both intra-cluster compactness and inter-cluster separation, improving the accuracy of cluster assessment. Experiments were conducted with K values ​​ranging from 5 to 20, ultimately determining the optimal number of clusters to be K = 8, corresponding to an improved silhouette coefficient of 0.4154.

[0128] Clustering results show that the 69 stations were effectively divided into eight groups, of which six groups had minimum internal correlation coefficients greater than 0.5, indicating that the stations within the groups were highly correlated. Group 1 contained 16 photovoltaic power plants, and the remaining five groups were wind farms, confirming the low correlation between wind farms and photovoltaic power plants. Through clustering, the original 69-dimensional input data was reduced to 8 dimensions, of which 49 stations were classified into six highly correlated groups, as shown in Table 1. Reducing the input dimension from 49 to 6 dimensions significantly reduced the complexity of data processing. This optimized clustering method not only improved the accuracy of the clustering results but also effectively reduced the dimensionality of the feature space, simplifying the subsequent sample generation process.

[0129] Table 1 List of power grid station groups

[0130] serial number Number of stations Average correlation coefficient Minimum correlation coefficient 1 16 0.90 0.79 2 15 0.79 0.61 3 6 0.87 0.79 4 5 0.74 0.64 5 4 0.80 0.61 6 3 0.90 0.87

[0131] During the random data generation process, an advanced statistical model based on Cholesky decomposition was employed. By performing Cholesky decomposition on the correlation coefficient matrix R, we successfully generated a new energy power series with the same correlation structure as the original data. Taking Group 4 (containing five wind farms) as an example, we generated a data series with a length of 8457, which is consistent with the length of the original online data. Comparing the correlation coefficient matrices of the generated data with the original data revealed that the deviation between corresponding elements was generally below 0.01, with a maximum deviation of only 0.0105. Tables 2 and 3 show the correlation coefficient matrices for Group 4 of the original online data and the generated data.

[0132] Table 2 Correlation coefficient matrix of group 4 in the original online data

[0133] Station 1 Station 2 Station 3 Station 4 Station 5 Station 1 1.0000 0.6645 0.8200 0.7588 0.6484 Station 2 0.6645 1.0000 0.7225 0.6389 0.6956 Station 3 0.8200 0.7225 1.0000 0.8586 0.7686 Station 4 0.7588 0.6389 0.8586 1.0000 0.7754 Station 5 0.6484 0.6956 0.7686 0.7754 1.0000

[0134] Table 3 Correlation coefficient matrix of group 4 in the generated data

[0135]

[0136]

[0137] Comparing the generated data with the original data reveals that the correlation coefficient matrix of the generated data is highly consistent with the original data, with minimal deviation, demonstrating the effectiveness of the method in maintaining inter-station correlation. The generated power series closely matches the original data in terms of volatility and trend, demonstrating the accuracy of the method in capturing the power characteristics of renewable energy sources. The improvement in computational efficiency is attributed to clustering dimensionality reduction and an optimized random generation algorithm. This significantly reduces the time cost of data processing and generation, meeting the needs of large-scale data processing and satisfying the requirements of online data processing.

[0138] Figure 4 FIG. 4 is a schematic structural diagram of a power generation system 400 for a new energy station in a power grid considering spatial correlation according to an embodiment of the present invention. Figure 4 As shown, the power generation system 400 of a new energy station in a power grid considering spatial correlation provided by an embodiment of the present invention includes: a correlation coefficient calculation unit 401, a clustering unit 402 and a data generation unit 403.

[0139] Preferably, the correlation coefficient calculation unit 401 is used to calculate the correlation coefficient between any two different new energy stations based on the power change sequence.

[0140] Preferably, the correlation coefficient calculation unit 401 calculates the correlation coefficient of any two different new energy stations based on the power change sequence, including:

[0141]

[0142] Among them, ρ X,Y is the correlation coefficient between the power change sequences X and Y of the two new energy stations; X t and Y t is the power value at the tth time point; n is the number of time points; and is the mean of the corresponding sequence; w t is the weight coefficient based on the power change rate; σ X and σ Y are the standard deviations of series X and Y, respectively.

[0143] Preferably, the clustering unit 402 is configured to calculate the distance between stations based on a correlation coefficient, and perform clustering based on the distance to form a plurality of station clusters.

[0144] Preferably, the clustering unit 402 calculates the distance between stations based on the correlation coefficient, including:

[0145]

[0146] Among them, d ijis the distance between new energy station i and new energy station j; ρ X,Y Represents the correlation coefficient between the power change sequences X and Y corresponding to the new energy stations i and j.

[0147] Preferably, the clustering unit 402 further includes:

[0148] Set the initial range of the number of clusters K according to the data scale and expected clustering granularity [K min ,K max ];

[0149] For each candidate K value, perform the improved K-Means clustering algorithm to calculate the corresponding global weighted silhouette coefficient And look for The largest K value is taken as the optimal number of clusters K*; if [K min ,K max ] If the optimal value is not found within the range, then according to Expand your search scope based on the changing trends of the

[0150] Preferably, the clustering unit 402 determines the global weighted silhouette coefficient in the following manner, including:

[0151]

[0152] Among them, a i represents the average distance between sample i and its cluster and other samples; b i represents the average distance between sample i and the nearest samples in other clusters; w i represents the weight coefficient of the i-th sample, which is determined by the installed capacity, historical power generation and data quality of the new energy unit; s i represents the weighted silhouette coefficient of the i-th sample, s i >0 indicates that the samples are reasonably clustered, with close clusters and separation between clusters; i =0 means that the sample is located at the boundary of two clusters and the clustering effect is average; s i <0 indicates that the samples may be clustered incorrectly, the intra-cluster distance is greater than the inter-cluster distance, and one new energy station is one sample.

[0153] Preferably, the data generating unit 403 is configured to construct a correlation coefficient matrix for each station group, and generate a new new energy station power change sequence through Cholesky decomposition.

[0154] Preferably, the data generating unit 403 constructs a correlation coefficient matrix for each station group, and generates a new new energy station power change sequence through Cholesky decomposition, including:

[0155] For each station group, the corresponding correlation coefficient submatrix R is extracted from the correlation coefficient matrix constructed based on the correlation coefficient;

[0156] The correlation coefficient matrix is ​​decomposed into R = L·L based on the advanced statistical model of Cholesky decomposition. T ; Where L is a lower triangular matrix over the real field;

[0157] Generate a random matrix Z=randn(k,m) that conforms to the standard normal distribution, Where k is the length of data to be generated, m is the number of stations in the station group, and randn represents sampling from the standard normal distribution N(0,1);

[0158] The random matrix Z is linearly transformed by the lower triangular matrix L to introduce the specified correlation structure, and S = Z·L is obtained T ;

[0159] Perform inverse normalization on each column of the matrix S to obtain Each column of the generated random data matrix S is used as a new power change sequence of the new energy station that satisfies the correlation coefficient submatrix R;

[0160] in,

[0161]

[0162] Among them, S ij is a correlation-mapped matrix, each column of which is a standardized random sequence with a specified correlation; S ij ′ is the power change value; and Var(X i ) are the mean and variance of the power series of the i-th new energy station; n is the length of the time series; A T represents the transpose of A.

[0163] The power generation system 400 for a power grid new energy station considering spatial correlation according to an embodiment of the present invention corresponds to the power generation method 100 for a power grid new energy station considering spatial correlation according to another embodiment of the present invention, and will not be described in detail here.

[0164] The invention has been described above with reference to a few embodiments. However, it is readily apparent to a person skilled in the art that other embodiments than the ones disclosed above are equally within the scope of the invention, as defined by the appended patent claims.

[0165] Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise herein. All references to "a / the [means, component, etc.]" are to be interpreted openly as referring to at least one instance of the means, component, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not necessarily need to be performed in the exact order disclosed, unless explicitly stated otherwise.

[0166] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0167] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0168] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0169] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A method for generating power at a new energy station in a power grid considering spatial correlation, characterized in that: The method comprises: Calculate the correlation coefficient between any two different new energy stations based on the power change sequence; Calculating the distance between stations based on the correlation coefficient, and clustering based on the distance to form a plurality of station clusters; For each station group, a correlation coefficient matrix is ​​constructed, and a new power change sequence of new energy stations is generated through Cholesky decomposition; The calculation of the correlation coefficient between any two different new energy stations based on the power change sequence includes: Among them, ρ X,Y is the correlation coefficient between the power change sequences X and Y of the two new energy stations; X t and Y t is the power value at the tth time point; n is the number of time points; and is the mean of the corresponding sequence; w t is the weight coefficient based on the power change rate; σ X and σ Y are the standard deviations of sequences X and Y respectively; The method of constructing a correlation coefficient matrix for each station group and generating a new power change sequence of new energy stations through Cholesky decomposition includes: For each station group, the corresponding correlation coefficient submatrix R is extracted from the correlation coefficient matrix constructed based on the correlation coefficient; Based on the advanced statistical model of Cholesky decomposition, the correlation coefficient submatrix is ​​decomposed into R = L·L T ; Where L is a lower triangular matrix over the real field; Generate a random matrix Z=randn(k,m) that conforms to the standard normal distribution, Where k is the length of data to be generated, m is the number of stations in the station group, and randn represents sampling from the standard normal distribution N(0,1); The random matrix Z is linearly transformed by the lower triangular matrix L to introduce the specified correlation structure, and S = Z·L is obtained T ; Perform inverse normalization on each column of the matrix S to obtain Each column of the generated random data matrix S is used as a new power change sequence of the new energy station that satisfies the correlation coefficient submatrix R; in, Among them, S ij is a correlation-mapped matrix, each column of which is a standardized random sequence with a specified correlation; S ij ′ is the power change value; and Var(X i ) are the mean and variance of the power series of the i-th new energy station; n is the length of the time series; A T represents the transpose of A.

2. The method according to claim 1, characterized in that The distance between stations is calculated based on the correlation coefficient, including: Among them, d ij is the distance between new energy station i and new energy station j; ρ X,Y Represents the correlation coefficient between the power change sequences X and Y corresponding to the new energy stations i and j.

3. The method according to claim 1, characterized in that The method further comprises: Set the initial range of the number of clusters K according to the data size and expected clustering granularity [K min ,K max ]; For each candidate K value, perform the improved K-Means clustering algorithm to calculate the corresponding global weighted silhouette coefficient And look for The largest K value is taken as the optimal number of clusters K*; if [K min ,K max ] If the optimal value is not found within the range, then according to Expand your search scope based on the changing trends of the 4. The method according to claim 3, characterized in that The global weighted silhouette coefficient is determined using the following methods, including: Among them, a i represents the average distance between sample i and its cluster and other samples; b i represents the average distance between sample i and the nearest samples in other clusters; w i represents the weight coefficient of the i-th sample, which is determined by the installed capacity, historical power generation and data quality of the new energy unit; s i represents the weighted silhouette coefficient of the i-th sample, s i >0 indicates that the samples are reasonably clustered, with close clusters and separation between clusters; i =0 means that the sample is located at the boundary of two clusters and the clustering effect is average; s i <0 indicates that the samples may be clustered incorrectly, the intra-cluster distance is greater than the inter-cluster distance, and one new energy station is one sample.

5. A power generation system for a new energy station in a power grid considering spatial correlation, characterized in that: The system comprises: A correlation coefficient calculation unit, used to calculate the correlation coefficient of any two different new energy stations based on the power change sequence; A clustering unit, configured to calculate distances between stations based on correlation coefficients, and cluster the stations based on the distances to form a plurality of station clusters; The data generation unit is used to construct a correlation coefficient matrix for each station group and generate a new power change sequence of new energy stations through Cholesky decomposition; The correlation coefficient calculation unit calculates the correlation coefficient of any two different new energy stations based on the power change sequence, including: Among them, ρ X,Y is the correlation coefficient between the power change sequences X and Y of the two new energy stations; X t and Y t is the power value at the tth time point; n is the number of time points; and is the mean of the corresponding sequence; w t is the weight coefficient based on the power change rate; σ X and σ Y are the standard deviations of sequences X and Y respectively; The data generation unit constructs a correlation coefficient matrix for each station group and generates a new power change sequence of new energy stations through Cholesky decomposition, including: For each station group, the corresponding correlation coefficient submatrix R is extracted from the correlation coefficient matrix constructed based on the correlation coefficient; Based on the advanced statistical model of Cholesky decomposition, the correlation coefficient submatrix is ​​decomposed into R = L·L T ; Where L is a lower triangular matrix over the real field; Generate a random matrix Z=randn(k,m) that conforms to the standard normal distribution, Where k is the length of data to be generated, m is the number of stations in the station group, and randn represents sampling from the standard normal distribution N(0,1); The random matrix Z is linearly transformed by the lower triangular matrix L to introduce the specified correlation structure, and S = Z·L is obtained T ; Perform inverse normalization on each column of the matrix S to obtain Each column of the generated random data matrix S is used as a new power change sequence of the new energy station that satisfies the correlation coefficient submatrix R; in, Among them, S ij is a correlation-mapped matrix, each column of which is a standardized random sequence with a specified correlation; S ij ′ is the power change value; and Var(X i ) are the mean and variance of the power series of the i-th new energy station; n is the length of the time series; A T represents the transpose of A.

6. The system according to claim 5, characterized in that The clustering unit calculates the distance between stations based on the correlation coefficient, including: Among them, d ij is the distance between new energy station i and new energy station j; ρ X,Y Represents the correlation coefficient between the power change sequences X and Y corresponding to the new energy stations i and j.

7. The system according to claim 5, characterized in that The clustering unit further includes: Set the initial range of the number of clusters K according to the data size and expected clustering granularity [K min ,K max ]; For each candidate K value, perform the improved K-Means clustering algorithm to calculate the corresponding global weighted silhouette coefficient And look for The largest K value is taken as the optimal number of clusters K*; if [K min ,K max ] If the optimal value is not found within the range, then according to Expand your search scope based on the changing trends of the 8. The system according to claim 7, characterized in that The clustering unit determines the global weighted silhouette coefficient in the following manner, including: Among them, a i represents the average distance between sample i and its cluster and other samples; b i represents the average distance between sample i and the nearest samples in other clusters; w i represents the weight coefficient of the i-th sample, which is determined by the installed capacity, historical power generation and data quality of the new energy unit; s i represents the weighted silhouette coefficient of the i-th sample, s i >0 indicates that the samples are reasonably clustered, with close clusters and separation between clusters; i =0 means that the sample is located at the boundary of two clusters and the clustering effect is average; s i <0 indicates that the samples may be clustered incorrectly, the intra-cluster distance is greater than the inter-cluster distance, and one new energy station is one sample.

Images