A new energy power sequence sample generation method and system

Abstract

The application provides a new energy power sequence sample generation method and system. The method calculates the correlation coefficient between new energy stations in a new energy station group according to the power generation data of the new energy stations, groups the new energy stations through the correlation coefficient, generates a correlation coefficient sub-matrix through the correlation coefficient between the new energy stations in a new energy station group with a reducible dimension, determines the power generation sequence of the new energy station group with a reducible dimension, combines the power generation sequence with the power generation sequence of other new energy station groups without a reducible dimension, and generates the power generation sequence of the new energy station group after dimension reduction. The method and system analyze the correlation first, group and cluster next, and then randomly generate new energy power sequence samples. The randomness of the new energy power is maintained, the correlation between the new energy stations is unchanged, the dimension of the entire function sequence sample is greatly reduced, the difficulty of spatial sampling and sample generation is reduced, and the generation efficiency is high.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of machine learning, and more particularly, to a new energy power sequence sample generation method and system. BACKGROUND

[0002] As one of the important measures to realize energy transformation, the proportion of new energy in the energy supply structure is becoming higher and higher, and the fluctuation and randomness of its output have brought great challenges to traditional power planning. New energy large-scale access occupies the space of conventional unit start-up, or causes problems such as system operation frequency modulation capability decline, insufficient reactive power support, prominent voltage stability, and increased risk of exceeding the limit. Because the number of new energy units is large, if each unit is considered separately, it will definitely cause the dimension of the sample space to be too high, increasing the difficulty of sample generation; and there is usually strong correlation between new energy units, for example, wind power units in the same period and the same region may be under similar weather conditions, and their power curves show the same trend, and the sample data generated by using this correlation will also be closer to the actual situation. Therefore, how to use the correlation between new energy units to improve the efficiency of new energy unit power data generation has become a technical problem that needs to be solved in view of the difficulty of generating samples caused by the large number of new energy units. SUMMARY

[0003] In order to solve the problem of large proportion of new energy units in the power grid and difficulty of sample generation caused by considering each unit separately in the prior art, the present application provides a new energy power sequence sample generation method and system.

[0004] According to an aspect of the present application, the present application provides a new energy power sequence sample generation method, which comprises:

[0005] Obtaining the same number of power generation history data of a plurality of new energy stations in a new energy station group in the same historical period;

[0006] According to the power generation history data, calculating the correlation coefficient between any two new energy stations in the new energy station group;

[0007] According to the correlation coefficient, grouping the new energy stations in the new energy station group to determine K new energy station groups;

[0008] According to the pre-set correlation coefficient threshold, determining the dimension-reducible new energy station group in the K new energy station groups;

[0009] According to the correlation coefficient between any two new energy stations in each dimension-reducible new energy station group, generating a correlation coefficient sub-matrix, and randomly generating the power generation sequence of each dimension-reducible new energy station group based on the correlation coefficient sub-matrix;

[0010] The power generation sequence of each dimension-reducible new energy station group is combined with the power generation sequence of other non-dimension-reducible new energy station groups to generate a dimension-reduced power generation sequence of the new energy station group.

[0011] Optionally, according to the historical power generation data, a correlation coefficient between any two new energy stations in the new energy station group is calculated, wherein the calculation formula of the correlation coefficient is:

[0012]

[0013] In the formula, x ik and x jk respectively represent the kth power generation of the new energy station i and the new energy station j, and respectively represent the average of n power generations of the new energy station i and the new energy station j, ρ ij represents the correlation coefficient of the new energy station i and the new energy station j, 1≤i, j≤m, and m represents the total number of new energy stations in the new energy station group.

[0014] Optionally, the new energy stations in the new energy station group are grouped according to the correlation coefficient, and K new energy station groups are determined, including:

[0015] According to the correlation coefficient between the new energy station i and the new energy station j in the new energy station group, the distance d ij between the new energy station i and the new energy station j is determined, wherein 1≤i, j≤m, and m represents the total number of new energy stations in the new energy station group.

[0016] According to the distance d ij , a clustering algorithm is used to determine K new energy station groups.

[0017] Optionally, according to the correlation coefficient between the new energy station i and the new energy station j in the new energy station group, the distance d i,j between the new energy station i and the new energy station j is determined, wherein the calculation formula of the distance d i,j is:

[0018]

[0019] In the formula, ρ ij represents the correlation coefficient of the new energy station i and the new energy station j.

[0020] Optionally, according to the distance d ij , a clustering algorithm is used to determine K new energy station groups, including:

[0021] A parameter K is set, wherein K is a natural number greater than 1.

[0022] The distance d ij is generated as a sample point, and a sample set D = {d ij} is generated;

[0023] The sample set D is divided into K clusters according to the principle of minimum distance within a cluster and maximum distance between clusters by using a K-Means algorithm, wherein the centroid of each cluster is a sample point with the minimum average distance to other sample points in the cluster;

[0024] The set of all new energy stations corresponding to the sample points in each cluster is taken as one new energy station group.

[0025] Optionally, the K new energy station groups are determined according to a preset correlation coefficient threshold value, including:

[0026] For any one of the K new energy station groups, when the correlation coefficients between all new energy stations in the group are greater than the set correlation coefficient threshold value, the new energy station group is determined as a new energy station group that can be reduced in dimension.

[0027] Optionally, a correlation coefficient sub-matrix is generated according to the correlation coefficients between any two new energy stations in each new energy station group that can be reduced in dimension, and a power generation sequence of each new energy station group is randomly generated based on the correlation coefficient sub-matrix, including:

[0028] A correlation coefficient sub-matrix R l×l is generated according to the correlation coefficients between any two new energy stations in each new energy station group that can be reduced in dimension, wherein l is the total number of new energy stations in the new energy station group that can be reduced in dimension.

[0029] The correlation coefficient sub-matrix R l×l is decomposed by using Cholesky decomposition to determine a lower triangular matrix L, wherein the correlation coefficient sub-matrix R l×l and the lower triangular matrix L satisfy the following relationship:

[0030] R = L·L T

[0031] In the formula, L T is the transpose of the lower triangular matrix L.

[0032] According to the lower triangular matrix L and the power generation data of any one new energy station in each new energy station group that can be reduced in dimension within a time period in which the new energy station group is to generate a power generation sequence sample, a power generation sequence S of each new energy station group that can be reduced in dimension is generated.

[0033] Optionally, according to the lower triangular matrix L and the acquired power generation data of any new energy power station in each new energy power station group that can be reduced in dimension within a time period in which the new energy power station group is to generate a power generation sequence sample, a power generation sequence S of each new energy power station group is generated, including:

[0034] The power generation data of any new energy power station in each new energy power station group that can be reduced in dimension acquired within a time period in which the new energy power station group is to generate a power generation sequence sample is input into a random sequence generator rand conforming to a uniform distribution;

[0035] The random sequence generator rand generates the power generation sequence S based on the lower triangular matrix L and the total number l of new energy power stations in each reduced new energy power station group, wherein the expression of the power generation sequence S is:

[0036] S = rand(num, l) · L

[0037] In the formula, num is the number of power generation data of each new energy power station in the set power generation sequence sample to be generated.

[0038] Optionally, the power generation sequence of each new energy power station group that can be reduced in dimension is combined with the power generation sequence of other new energy power station groups that cannot be reduced in dimension to generate a power generation sequence of the new energy power station group after reduction in dimension, including:

[0039] According to the number of power generation data of each new energy power station in the set power generation sequence sample to be generated and the time period, the power generation data of each new energy power station in the new energy power station group that cannot be reduced in dimension is acquired to generate a power generation sequence of each new energy power station group that cannot be reduced in dimension;

[0040] The power generation sequences of all new energy power station groups that can be reduced in dimension and the power generation sequences of all new energy power station groups that cannot be reduced in dimension are spliced to generate a power generation sequence sample Y of the new energy power station group num×m , wherein m represents the total number of new energy power stations in the new energy power station group.

[0041] According to another aspect of the present application, the present application provides a new energy power sequence sample generation system, the system comprising:

[0042] A data acquisition unit is configured to acquire the same number of power generation history data of multiple new energy power stations in a new energy power station group in a same historical period;

[0043] A correlation coefficient unit is configured to calculate a correlation coefficient between any two new energy power stations in the new energy power station group according to the power generation history data;

[0044] The first grouping unit is configured to group new energy power stations in the new energy power station group according to the correlation coefficients, and determine K new energy power station groups;

[0045] The second grouping unit is configured to determine a new energy power station group that can be reduced in dimension in the K new energy power station groups according to a preset correlation coefficient threshold value.

[0046] The dimension reduction sample unit is configured to generate a correlation coefficient sub-matrix according to the correlation coefficient between any two new energy power stations in each new energy power station group that can be reduced in dimension, and randomly generate a power generation sequence of each new energy power station group based on the correlation coefficient sub-matrix.

[0047] The power sample unit is configured to combine the power generation sequence of each new energy power station group that can be reduced in dimension with the power generation sequence of other new energy power station groups that cannot be reduced in dimension, and generate a power generation sequence of the new energy power station group after dimension reduction.

[0048] The new energy power sequence sample generation method and system provided by the technical scheme of the present application calculate the correlation coefficients between new energy power stations in a new energy power station group according to the power generation data of the new energy power stations, group the new energy power stations according to the correlation coefficients, generate a correlation coefficient sub-matrix according to the correlation coefficients between the new energy power stations in a new energy power station group that can be reduced in dimension, determine a power generation sequence of the new energy power station group that can be reduced in dimension, combine the power generation sequence with the power generation sequence of other new energy power station groups that cannot be reduced in dimension, and generate a power generation sequence of the new energy power station group after dimension reduction. The method and system first analyze the correlation, then group and cluster, and finally randomly generate new energy power sequence samples, which not only maintains the randomness of new energy power, but also ensures that the correlation between new energy power stations remains unchanged, and greatly reduces the dimension of the entire function sequence sample, reduces the difficulty of spatial sampling and sample generation, and has high generation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0049] The exemplary embodiments of the present application can be more completely understood in reference to the following drawings:

[0050] Figure 1 A flowchart of the new energy power sequence sample generation method according to the preferred embodiment of the present application;

[0051] Figure 2 A distribution histogram of wind farm correlation coefficient values and correlation coefficient samples according to the preferred embodiment of the present application;

[0052] Figure 3 A distribution histogram of wind farm correlation coefficient values and wind farm numbers according to the preferred embodiment of the present application;

[0053] Figure 4A schematic diagram for determining a centroid when grouping new energy station groups according to a preferred embodiment of the present application;

[0054] Figure 5 A structural schematic diagram of a new energy power sequence sample generation system according to a preferred embodiment of the present application. DETAILED DESCRIPTION

[0055] Reference will now be made to the drawings to describe the exemplary embodiments of the present application in detail. The present application may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art. The terminology used in the description of the exemplary embodiments presented herein is not intended to be limiting in scope, but is intended to be illustrative of examples of products of the present application. Identical elements are numbered alike in the various figures of the drawings.

[0056] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and the present disclosure and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0057] Exemplary method

[0058] Figure 1 A flowchart of a new energy power sequence sample generation method according to a preferred embodiment of the present application. As shown in Figure 1 the new energy power sequence sample generation method according to the preferred embodiment starts from step 101.

[0059] In step 101, a same number of power generation history data of a plurality of new energy stations in a new energy station group in a same historical period is obtained.

[0060] In one embodiment, since there are a large number of new energy unit data, and the modeling level of different regions may be different, the power statistics method with station as a unit is uniformly used to count the wind farm power in the online data of Northeast Power Grid from September to December 2019, wherein there are 138 wind farms, and a same number of power generation history data is collected for each wind farm. It should be noted that, due to different characteristics of new energy stations, when collecting power generation history data to generate power sample sequences, wind farms and photovoltaic power stations need to be counted separately.

[0061] In step 102, a correlation coefficient between any two new energy stations in the new energy station group is calculated according to the power generation history data.

[0062] Preferably, according to the historical power generation data, a correlation coefficient between any two new energy stations in the new energy station group is calculated, wherein the calculation formula of the correlation coefficient is:

[0063]

[0064] wherein x ik and x jk respectively represent the kth power generation of the new energy station i and the new energy station j, and respectively represent the average of the n power generations of the new energy station i and the new energy station j, p ij represents the correlation coefficient of the new energy station i and the new energy station j, 1≤i, j≤m, and m represents the total number of new energy stations in the new energy station group.

[0065] In the preferred embodiment, the Pearson correlation coefficient is used as the basis for calculating the correlation coefficient between stations, but other correlation coefficient calculation methods are also within the protection scope of the present application. The Pearson correlation coefficient method is used to calculate the correlation coefficient of the historical power generation data of the wind power stations in the Northeast Power Grid, and a 138*138-dimensional correlation coefficient matrix is obtained. Except for the part on the diagonal line with a correlation coefficient of 1.0, there are a total of 18,906 correlation coefficient results, of which 18,140 are positively correlated, accounting for 95.95%. When the correlation coefficient threshold is set to 0.5, and the correlation coefficient is greater than 0.5, it means that the two wind power sounds are strongly correlated, and there are 3,010 strong correlation results, accounting for 15.92%. Figure 2 The figure is a distribution histogram of the correlation coefficient values and correlation coefficient samples of the wind power stations according to the preferred embodiment of the present application. As shown in Figure 2 of the 18,906 correlation coefficient samples, except for a few negative correlations, about 1,000 are unrelated, and the vast majority are positively correlated. Among them, the number of samples with a correlation coefficient of about 0.2 is the largest, exceeding 4,000, and the number of samples with a correlation coefficient of about 0.3 is the second, nearly 4,000.

[0066] Further, the wind power stations are statistically analyzed according to the calculated correlation coefficient values. Figure 3 The figure is a distribution histogram of the correlation coefficient values and the number of wind power stations according to the preferred embodiment of the present application. As shown in Figure 3 , the number of wind power stations with a correlation coefficient value in the interval (0.8, 0.9) exceeds 30, and the number of wind power stations with a correlation coefficient value in the interval (0.7, 0.8) also exceeds 30. Overall, the number of wind power stations with at least one strong correlation result reaches 137, accounting for 99.28%; only one wind power station has no strong correlation result, but the maximum correlation coefficient result of the wind power station reaches 0.46. From Figure 3It can be seen that the power generation of the wind farms mainly presents strong positive correlation (geographical location is close) or no correlation (geographical location is far), which is due to the fact that the local meteorological conditions are usually similar, and there are almost no examples of strong negative correlation; according to the correlation coefficient, the new energy stations can be grouped, so that the new energy station groups with strong correlation are uniformly generated power sequences by the sequence generator, and the new energy station groups without strong correlation are independently collected to generate power sequences, thereby achieving the purpose of reducing the dimension of the original feature space.

[0067] In step 103, the new energy stations in the new energy station group are grouped according to the correlation coefficient, and K new energy station groups are determined.

[0068] Preferably, the new energy stations in the new energy station group are grouped according to the correlation coefficient, and K new energy station groups are determined, including:

[0069] According to the correlation coefficient between the new energy station i and the new energy station j in the new energy station group, the distance d ij between the new energy station i and the new energy station j is determined, wherein 1≤i, j≤m, and m represents the total number of new energy stations in the new energy station group.

[0070] According to the distance d ij , a clustering algorithm is used to determine K new energy station groups.

[0071] Preferably, according to the correlation coefficient between the new energy station i and the new energy station j in the new energy station group, the distance d i,j between the new energy station i and the new energy station j is determined, wherein the distance d i,j is calculated according to the following formula:

[0072]

[0073] In the formula, ρ ij represents the correlation coefficient between the station i and the station j.

[0074] Preferably, according to the distance d ij , a clustering algorithm is used to determine K new energy station groups, including:

[0075] Setting a parameter K, wherein K is a natural number greater than 1;

[0076] Taking the distance d ij as a sample point, a sample set D={d ij} is generated;

[0077] The K-Means algorithm is used to divide the sample set D into K clusters according to the principle that the distance in a cluster is minimum and the distance between clusters is maximum, wherein the centroid of each cluster is the sample point with the minimum average distance to other sample points in the cluster.

[0078] The set of all new energy stations corresponding to the sample points in each cluster is taken as one new energy station group.

[0079] The similarity degree between different wind power stations can be obtained through the correlation coefficient analysis between wind power stations, and the greater the similarity coefficient, the higher the similarity degree between stations. In an embodiment, the K-Mean algorithm, i.e., the K-means algorithm, is used to cluster 138 wind power stations, wherein K is a hyperparameter representing the number of clusters and is specified in advance. Before clustering the 138 wind power stations, the 18906 correlation coefficients of the 138 wind power stations are converted into distances between stations, and the greater the distance, the lower the similarity degree. According to the distance relationship between stations, the K-Mean algorithm is used to divide the sample set composed of the 18906 distances between stations of the 138 wind power stations into K clusters, so that the distance in a cluster is minimum (high similarity) and the distance between clusters is maximum (low similarity), thereby grouping similar samples. The mean position of all sample points in each cluster is called the "centroid", and the centroid is a virtual sample point representing the center position of a cluster. However, the method described in this embodiment uses the correlation coefficient between new energy stations to represent similarity, i.e., only the distance between stations, without the actual coordinates of each sample point, so the position of the centroid cannot be calculated. Based on this, the traditional K-Means algorithm is improved in this embodiment: the sample point with the minimum average distance to other sample points in the cluster is selected as the centroid position of the cluster. Figure 4 The principle diagram for determining the centroid when grouping new energy station groups according to the preferred embodiment of the present application is shown in FIG. 1. Figure 4 As shown in FIG. 1, it is assumed that when grouping and clustering new energy station groups, there are 4 sample points (represented by solid hollow circles 1 to 4) in one cluster, and the centroid is shown as a dashed circle a. However, since the coordinates of the dashed circle a cannot be obtained, the centroid can only be represented by the actual existing solid hollow circle 1, and at this time the coverage range of the cluster is expanded from the dashed circle b to the dashed circle c. This improvement slightly reduces the stability of the K-Means clustering process, but through practice it is found that the improved K-Means algorithm can still well complete the task of grouping new energy stations. In this embodiment, when K is 20, the 138 wind power stations are divided into 20 new energy station groups.

[0080] In step 104, the new energy station groups with reducible dimensions in the K new energy station groups are determined according to a pre-set correlation coefficient threshold.

[0081] Preferably, the new energy station groups with the possibility of dimension reduction in the K new energy station groups are determined according to a preset correlation coefficient threshold, and the method comprises the following steps:

[0082] For any one of the K new energy station groups, when the correlation coefficients between all new energy stations in the group are greater than the set correlation coefficient threshold, the new energy station group is determined as a new energy station group with the possibility of dimension reduction.

[0083] In the preferred embodiment, the set correlation coefficient threshold is 0.5, and when the correlation coefficient between two stations is greater than 0.5, the two stations are considered to be strongly correlated. After dividing the 138 wind power stations into 20 new energy station groups, the correlation coefficients of all stations in 11 station groups are greater than 0.5, that is, all the stations in the group are strongly correlated, as shown in Table 1. The 11 station groups include 75 stations, and the minimum correlation coefficient between the stations is greater than 0.5, which is a new energy station group with the possibility of dimension reduction.

[0084] Table 1: List of wind power station groups with the possibility of dimension reduction in Northeast Power Grid

[0085]

[0086]

[0087] In step 105, a correlation coefficient sub-matrix is generated according to the correlation coefficient between any two new energy stations in each new energy station group with the possibility of dimension reduction, and a power generation sequence of each new energy station group is randomly generated based on the correlation coefficient sub-matrix.

[0088] Preferably, the correlation coefficient sub-matrix is generated according to the correlation coefficient between any two new energy stations in each new energy station with the possibility of dimension reduction, and a power generation sequence of each new energy station group is randomly generated based on the correlation coefficient sub-matrix, and the method comprises the following steps:

[0089] A correlation coefficient sub-matrix R is generated according to the correlation coefficient between any two new energy stations in each new energy station group with the possibility of dimension reduction l×l , wherein l is the total number of new energy stations in the new energy station group with the possibility of dimension reduction;

[0090] The correlation coefficient sub-matrix R is decomposed by Cholesky decomposition l×l to determine a lower triangular matrix L, wherein the correlation coefficient sub-matrix R and the lower triangular matrix L satisfy the following relationship: l×l

[0091] R = L·L T

[0092] In the formula, L T is the transpose of the lower triangular matrix L.​

[0093] According to the lower triangular matrix L and the acquired power generation data of any new energy power station in each new energy power station group with reducible dimension in a time period in which the new energy power station group is supposed to generate power generation sequence samples, a power generation sequence S of each new energy power station group is generated.

[0094] Preferably, according to the lower triangular matrix L and the acquired power generation data of any new energy power station in each new energy power station group with reducible dimension in a time period in which the new energy power station group is supposed to generate power generation sequence samples, a power generation sequence S of each new energy power station group is generated, including:

[0095] The acquired power generation data of any new energy power station in each new energy power station group with reducible dimension in a time period in which the new energy power station group is supposed to generate power generation sequence samples is input into a random sequence generator rand conforming to a uniform distribution;

[0096] The random sequence generator rand generates the power generation sequence S based on the lower triangular matrix L and the total number l of new energy power stations in each new energy power station group with reducible dimension, wherein the expression of the power generation sequence S is:

[0097] S=rand(num, l)·L

[0098] In the formula, num is the number of power generation data of each new energy power station in the supposed power generation sequence samples.

[0099] In step 106, the power generation sequence of each new energy power station group with reducible dimension is combined with the power generation sequence of other new energy power station groups without reducible dimension, to generate a power generation sequence of the new energy power station group after dimension reduction.

[0100] In one embodiment, after 11 dimension-reducible new energy station groups in a new energy station group are screened out, if it is needed to generate power generation sequence samples of the new energy station group in any time period, only the power generation data of any new energy station in the 11 dimension-reducible new energy station groups needs to be collected, and a random sequence generator can be used to automatically generate the power generation sequence according to the expression of the power generation sequence. Since only the power generation data of 11 new energy stations in the 11 dimension-reducible new energy station groups needs to be collected, the power generation samples of all 75 new energy stations in the 11 dimension-reducible new energy station groups can be generated by using the random sequence generator, which is equivalent to reducing the originally needed 75-dimensional data to 11-dimensional data when obtaining the power sequence samples of the new energy station group. Compared with the existing sample generation method, the difficulty of spatial sampling and sample generation is reduced, and the generation efficiency is improved. For the remaining 9 non-dimension-reducible new energy station groups, the power generation data samples of all new energy stations in each non-dimension-reducible new energy station group are collected according to the time period, the number of power samples, and the power sequence samples are generated. Finally, the power generation sequence samples of the 20 new energy station groups are combined to obtain the power generation sequence samples of the new energy station group.

[0101] Exemplary system

[0102] Figure 5 A structure diagram of a new energy power sequence sample generation system according to a preferred embodiment of the present application is shown in FIG. 1. As shown in FIG. 1, the new energy power sequence sample generation system according to the preferred embodiment of the present application includes: Figure 5

[0103] A data acquisition unit 501 is configured to acquire the same number of power generation history data of a plurality of new energy stations in a new energy station group in a same historical period.

[0104] A correlation coefficient unit 502 is configured to calculate a correlation coefficient between any two new energy stations in the new energy station group according to the power generation history data.

[0105] A first grouping unit 503 is configured to group the new energy stations in the new energy station group according to the correlation coefficient, and determine K new energy station groups.

[0106] A second grouping unit 504 is configured to determine dimension-reducible new energy station groups in the K new energy station groups according to a pre-set correlation coefficient threshold.

[0107] ​The dimension reduction sample unit 505 is used to generate a correlation coefficient sub-matrix based on the correlation coefficient between any two new energy power stations in each dimension reduction new energy power station group, and to randomly generate the power generation sequence of each dimension reduction new energy power station group based on the correlation coefficient sub-matrix.

[0108] The power sample unit 506 is used to combine the power generation sequence of each dimensionality-reducible new energy power station group with the power generation sequence of other non-dimensionality-reducible new energy power station groups to generate the dimensionality-reduced power generation sequence of the new energy power station group.

[0109] Preferably, the correlation coefficient unit 502 calculates the correlation coefficient between any two new energy power stations in the new energy power station group based on the historical power generation data, wherein the formula for calculating the correlation coefficient is:

[0110]

[0111] In the formula, x ik and x jk Let i and j represent the k-th generating capacity of the new energy power station, respectively. and Let ρ represent the average of the n power generation values ​​of renewable energy power plants i and j, respectively. ij Let represent the correlation coefficient between new energy power station i and new energy power station j, where 1 ≤ i, j ≤ m, and m represents the total number of new energy power stations in the new energy power station group.

[0112] Preferably, the first grouping unit 503 groups the new energy power stations in the new energy power station cluster according to the correlation coefficient, and determines K new energy power station groups, including:

[0113] The distance d between new energy power station i and new energy power station j in the new energy power station cluster is determined based on the correlation coefficient between them. ij Where 1≤i, j≤m, and m represents the total number of new energy power stations in the new energy power station cluster;

[0114] According to the distance d ij K new energy power station groups were determined using a clustering algorithm.

[0115] Preferably, the first grouping unit 503 determines the distance d between new energy power station i and new energy power station j in the new energy power station group based on the correlation coefficient between new energy power station i and new energy power station j. i,j , where the distance d i,j The calculation formula is:

[0116]

[0117] In the formula, ρij a correlation coefficient between the power generation of the power station i and the power generation of the power station j.

[0118] Preferably, the first grouping unit 503 determines the distance d ij , and determines K new energy power station groups by using a clustering algorithm, including:

[0119] setting a parameter K, wherein K is a natural number greater than 1;

[0120] determining the distance d ij between each two new energy power stations in the sample set D as the sample point, and generating a sample set D = {d ij};

[0121] dividing the sample set D into K clusters by using a K-Means algorithm according to the principle of minimum intra-cluster distance and maximum inter-cluster distance, wherein the centroid of each cluster is the sample point with the minimum average distance to other sample points in the cluster;

[0122] taking the set of all new energy power stations corresponding to the sample points in each cluster as one new energy power station group.

[0123] Preferably, the second grouping unit 504 determines a new energy power station group with a reducible dimension in the K new energy power station groups according to a pre-set correlation coefficient threshold, including:

[0124] for any one new energy power station group in the K new energy power station groups, when the correlation coefficient between all new energy power stations in the group is greater than the set correlation coefficient threshold, determining the new energy power station group as a new energy power station group with a reducible dimension.

[0125] Preferably, the dimension reduction sample unit 505 generates a correlation coefficient sub-matrix according to the correlation coefficient between any two new energy power stations in each new energy power station group with a reducible dimension, and randomly generates a power generation sequence of each new energy power station group based on the correlation coefficient sub-matrix, including:

[0126] generating a correlation coefficient sub-matrix R l×l according to the correlation coefficient between any two new energy power stations in each new energy power station group with a reducible dimension, wherein l is the total number of new energy power stations in the new energy power station group with a reducible dimension;

[0127] decomposing the correlation coefficient sub-matrix R l×l by using Cholesky decomposition to determine a lower triangular matrix L, wherein the relationship between the correlation coefficient sub-matrix R l×l and the lower triangular matrix L is:

[0128] R = L·L T

[0129] wherein L Tis the transpose of the lower triangular matrix L;

[0130] According to the lower triangular matrix L and the generated power data of any new energy plant station in each dimension-reducible new energy plant station group within a time period in which the new energy plant group is to generate a power generation sequence sample, the power generation sequence S of each dimension-reducible new energy plant station group is generated.

[0131] Preferably, the dimension-reduced sample unit 505 generates the power generation sequence S of each dimension-reducible new energy plant station group according to the lower triangular matrix L and the generated power data of any new energy plant station in each dimension-reducible new energy plant station group within a time period in which the new energy plant group is to generate a power generation sequence sample, including:

[0132] The generated power data of any new energy plant station in each dimension-reducible new energy plant station group within a time period in which the new energy plant group is to generate a power generation sequence sample is input into a random sequence generator rand conforming to a uniform distribution;

[0133] The random sequence generator rand generates the power generation sequence S based on the lower triangular matrix L and the total number l of new energy plant stations in each dimension-reducible new energy plant station group, wherein the expression of the power generation sequence S is:

[0134] S=rand(num, l)·L

[0135] In the formula, num is the number of generated power data of each new energy plant station in the set power generation sequence sample.

[0136] Preferably, the power sample unit 506 combines the power generation sequence of each dimension-reducible new energy plant station group with the power generation sequence of other non-dimension-reducible new energy plant station groups to generate the power generation sequence of the dimension-reduced new energy plant group, including:

[0137] According to the number of generated power data of each new energy plant station in the set power generation sequence sample and the time period, the power generation data of each new energy plant station in the non-dimension-reducible new energy plant station group is obtained to generate the power generation sequence of each non-dimension-reducible new energy plant station group.

[0138] The power generation sequences of all dimension-reducible new energy plant station groups and the power generation sequences of all non-dimension-reducible new energy plant station groups are spliced to generate the power generation sequence sample Y of the new energy plant group num×m , wherein m represents the total number of new energy plant stations in the new energy plant group.

[0139] The new energy power sequence sample generation system of the preferred embodiment generates the power generation power sequence of the new energy field station group with the dimensionality reduction through correlation analysis, clustering grouping and random generation, and then combines it with the power generation power sequence of the new energy field station group without the dimensionality reduction to generate the power sequence sample of the new energy field station group after the dimensionality reduction. The steps of the new energy power sequence sample generation method of the preferred embodiment are the same, and the technical effects achieved are also the same. Here, no further description is given.

[0140] The application has been described by reference to a few embodiments. However, as will be apparent to those skilled in the art in light of the disclosure, some embodiments of the application differ from those explicitly described to result in an equivalent or similar result. Any feature of the described embodiments can be used in combination with or substituted for other features. Such embodiments of the application also are intended to fall within the scope of the application as defined by the appended claims.

[0141] 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" or "an" means "at least one" unless otherwise clearly indicated by the context of the only language "the" or "that". The steps of any methods disclosed herein do not have to be performed in the exact order disclosed unless explicitly stated.

[0142] Those skilled in the art will appreciate that embodiments of the application can be devised for use with any computer or processor readable medium, device or chips configured to store or execute program code, wherein the program code used implements the steps recited in the claims. Accordingly, the embodiments of the application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects all designed according to the principles of the application.

[0143] The present application is described with reference to the accompanying drawings, which are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the application and, together with the description, serve to explain the principles of the application. In the drawings: Figure 1 The device for performing the function specified in the one or more flows and / or blocks. Figure 1 The device for performing the function specified in the one or more flows and / or blocks.

[0144] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0145] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flows and / or blocks Figure 1 one or more blocks or multiple blocks.

[0146] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, rather than limiting the same. Even though the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced equivalently, and any modification or replacement without departing from the spirit and scope of the present application should be covered within the protection scope of the claims of the present application.

Claims

1. A method for generating new energy power sequence samples, characterized in that, The method comprises: Obtaining the same number of historical power generation data of a plurality of new energy stations in a new energy station group in the same historical period; According to the power generation history data, the correlation coefficient between any two new energy stations in the new energy station group is calculated; According to the correlation coefficient, the new energy stations in the new energy station group are grouped to determine K new energy station groups; According to the pre-set correlation coefficient threshold, the dimension-reducible new energy station group in the K new energy station groups is determined; According to the correlation coefficient between any two new energy stations in each dimension-reducible new energy station group, a correlation coefficient submatrix is generated, and a power generation sequence of each dimension-reducible new energy station group is randomly generated based on the correlation coefficient submatrix, comprising: A correlation coefficient sub-matrix R is generated according to the correlation coefficient between any two new energy station groups in each new energy station group l×l wherein, l is the total number of new energy stations in the new energy station group The correlation coefficient sub-matrix R l×l is decomposed to determine a lower triangular matrix L, wherein the correlation coefficient sub-matrix R l×l and the lower triangular matrix L have the following relationship: R = L • L T wherein L T is the transpose of the lower triangular matrix L; According to the lower triangular matrix L and the power generation data of any one new energy station in each dimension-reducible new energy station group obtained in the time period in which the new energy station group is to generate a power generation sequence sample, a power generation sequence S of each dimension-reducible new energy station group is generated; The power generation sequence of each dimension-reducible new energy station group is combined with the power generation sequence of other non-dimension-reducible new energy station groups to generate a power generation sequence of the new energy station group after dimension reduction.

2. The method of claim 1, wherein, According to the power generation history data, the correlation coefficient between any two new energy stations in the new energy station group is calculated, wherein the calculation formula of the correlation coefficient is: wherein x ik and x jk respectively represent the kth power generation of the new energy station i and the new energy station j, and respectively represent the average of the n power generations of the new energy station i and the new energy station j, ρ ij represents the correlation coefficient of the new energy station i and the new energy station j, 1≤i, j≤m, and m represents the total number of new energy stations in the new energy station group.

3. The method of claim 1, wherein, According to the correlation coefficient, the new energy stations in the new energy station group are grouped to determine K new energy station groups, comprising: determine the distance d between the new energy station i and the new energy station j according to the correlation coefficient between the new energy station i and the new energy station j in the new energy station group ij wherein 1≤i,j≤m, m represents the total number of new energy stations in the new energy station group; According to the distance d ij , a clustering algorithm is used to determine K new energy station groups.

4. The method of claim 3, wherein, The distance d between the new energy station i and the new energy station j is determined according to a correlation coefficient between the new energy station i and the new energy station j in the new energy station group i,j wherein the distance d i,j The calculation formula of the distance d is: wherein p ij denotes the correlation coefficient between the field station i and the field station j.

5. The method of claim 3, wherein, According to the distance d ij , a clustering algorithm is used to determine K new energy station groups, including: Setting parameter K, wherein K is a natural number greater than 1; The distance d ij As sample points, a sample set D = {d ij} is generated. Using K-Means algorithm, according to the principle of minimum intra-cluster distance and maximum inter-cluster distance, the sample set D is divided into K clusters, wherein the centroid of each cluster is the sample point with the minimum average distance to other sample points in the cluster. The set of all new energy stations corresponding to the sample points in each cluster is regarded as one new energy station group.

6. The method of claim 1, wherein, According to the pre-set correlation coefficient threshold, the dimension-reducible new energy station group in the K new energy station groups is determined, comprising: For any one new energy station group in the K new energy station groups, when the correlation coefficient between all new energy stations in the group is greater than the set correlation coefficient threshold, the new energy station group is determined as a dimension-reducible new energy station group.

7. The method of claim 1, wherein, According to the lower triangular matrix L and the power generation data of any one new energy station in each dimension-reducible new energy station group obtained in the time period in which the new energy station group is to generate a power generation sequence sample, a power generation sequence S of each dimension-reducible new energy station group is generated, comprising: The power generation data of any one new energy station in each dimension-reducible new energy station group obtained in the time period in which the new energy station group is to generate a power generation sequence sample is input into a random sequence generator rand conforming to uniform distribution; The random sequence generator rand generates the power generation sequence S based on the lower triangular matrix L and the total number l of new energy stations in each dimension-reducible new energy station group, wherein the expression of the power generation sequence S is: S = rand(num, l) · L In the formula, num is the number of power generation data of each new energy station in the set power generation sequence sample.

8. The method of claim 7, wherein, The power generation sequence of each new energy station group is combined with the power generation sequence of other non-reducible new energy station groups to generate the power generation sequence of the new energy station group after dimension reduction, including: According to the number of power generation data of each new energy station in the set power generation sequence sample and the time period, the power generation data of each new energy station in the non-reducible new energy station group is obtained to generate the power generation sequence of each non-reducible new energy station group. The power generation power sequence of all the new energy power station groups and the power generation power sequence of all the non-reducible new energy power station groups are spliced to generate the power generation power sequence sample Y of the new energy power station group num×m wherein m represents the total number of new energy power stations in the new energy power station group.

9. A new energy power sequence sample generation system, characterized in that, The system comprises: A data acquisition unit is configured to acquire the same number of power generation history data of a plurality of new energy stations in a new energy station group in a same historical period; A correlation coefficient unit is configured to calculate the correlation coefficient between any two new energy stations in the new energy station group according to the power generation history data; A first grouping unit is configured to group the new energy stations in the new energy station group according to the correlation coefficient to determine K new energy station groups; A second grouping unit is configured to determine the reducible new energy station group in the K new energy station groups according to a pre-set correlation coefficient threshold; A dimension reduction sample unit is configured to generate a correlation coefficient sub-matrix according to the correlation coefficient between any two new energy stations in each reducible new energy station group, and randomly generate the power generation sequence of each reducible new energy station group based on the correlation coefficient sub-matrix, including: A correlation coefficient sub-matrix R is generated according to correlation coefficients between any two new energy station groups in the new energy station group l×l wherein, l is the total number of new energy stations in the new energy station group The correlation coefficient sub-matrix R l×l is decomposed to determine a lower triangular matrix L, wherein the correlation coefficient sub-matrix R l×l and the lower triangular matrix L satisfy the following relationship: R = L • L T wherein L T is the transpose of the lower triangular matrix L; According to the lower triangular matrix L and the obtained power generation data of any one new energy station in each reducible new energy station group in the time period of the new energy station group set power generation sequence sample, the power generation sequence S of each reducible new energy station group is generated; A power sample unit is configured to combine the power generation sequence of each reducible new energy station group with the power generation sequence of other non-reducible new energy station groups to generate the power generation sequence of the new energy station group after dimension reduction.

Images