A wind power forecast correction method considering power curtailment

Abstract

The present invention relates to a wind power prediction correction method considering power restrictions, which comprises the following steps: step 1. constructing a reference database for wind power prediction based on K-means clustering; step 2. calculating the correlation coefficient between the wind power prediction data vector p to be corrected and all data vectors in the corresponding database; step 3. dividing the correlation coefficient according to subcategories, calculating the average value set of the correlation coefficients in the subcategories, and determining the subcategory prediction data vector set most relevant to the vector p to be corrected; step 4. judging whether p needs to be corrected by comparing the number of power restriction days and the number of power unrestricted days in the smallest category most relevant to p; step 5. if p needs to be corrected, constructing a correction matrix Q, and obtaining the corrected wind power prediction data vector p′ through calculation. The present invention solves the problem of inaccurate wind power prediction due to power restrictions in a certain wind farm.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind power prediction, and in particular relates to a wind power prediction correction method taking power restriction into consideration. Background Art

[0002] During operation, a wind farm needs to report its output data for the next seven days to the power grid, i.e., short-term wind power forecast data, so as to meet the power grid's power supply dispatch plan for a certain area. Inaccurate wind power forecast data will increase the difficulty of the actual dispatch operation of the power grid and affect the stability of the power grid. According to the power grid's "double rules" policy, if the wind power fails to meet the accuracy required by the power grid, it will have a serious impact on the wind farm.

[0003] As photovoltaic power plants and wind farms continue to be deployed in a given region, the capacity of new energy sources connected to the power grid continues to increase. However, the region's load and grid structure are relatively fixed, making it difficult to fully absorb new energy sources locally. This has led to power curtailment at wind farms. Currently, wind power prediction devices on the market estimate wind farm output by predicting wind speed conditions in a given region. Chinese patent application number 201210252084.9 discloses a short-term wind speed and wind power prediction method, while Chinese patent number 201410768784.2 discloses a wind power prediction method based on wind speed patterns in wind zones and wind belts. However, power curtailment conditions vary across regions, and therefore the wind power prediction devices used by wind farms do not take these conditions into account. Consequently, under power curtailment conditions, the accuracy of short-term wind power prediction data from wind farms fails to meet grid requirements.

[0004] Therefore, it is necessary to study a wind power prediction and correction method that can deeply explore the power curtailment law of wind farms in a certain area and improve the stability of power grid operation. Summary of the Invention

[0005] In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a wind power prediction correction method considering power curtailment, which solves the problem of inaccurate wind power prediction in a certain wind farm due to power curtailment.

[0006] The technical solution adopted by the present invention is: the present invention provides a wind power prediction correction method considering power curtailment, which comprises the following steps:

[0007] Step 1. Construct a reference database for wind power prediction based on K-means clustering;

[0008] Step 2. Calculate the correlation coefficient between the wind power forecast data vector p to be corrected and all data vectors in the corresponding reference database;

[0009] Step 3. Divide the correlation coefficient into subcategories, calculate the maximum value of the set of average correlation coefficients in the subcategories, and determine the set of subcategory prediction data vectors that are most correlated with the wind power prediction data vector p to be corrected;

[0010] Step 4. Determine whether p needs to be corrected by comparing the number of power-restricted days and the number of unrestricted days in the smallest category most related to p;

[0011] Step 5. If p needs to be corrected, a correction matrix Q is constructed and the corrected wind power prediction data vector p' is obtained through calculation.

[0012] Furthermore, in step 1, the construction process includes the following steps,

[0013] S11. Classify the short-term power forecast data for the entire year in previous years into four categories: A, B, C, and D according to the four seasons;

[0014] S12. Using K-means cluster analysis and silhouette coefficient as the optimization criterion, a reference database for short-term wind power prediction of A, B, C and D is constructed respectively.

[0015] Furthermore, in S11, the short-term power forecast data is daily wind power forecast data at intervals of 15 minutes, and one day's data constitutes a 1×96 column vector, and each of the four categories of data A, B, C and D contains three months of wind power forecast data.

[0016] Furthermore, in S12, the K-means cluster analysis method is to set different cluster numbers for any one of the A, B, C and D categories, namely 2, 3, 4...10, and perform K-means cluster analysis on the wind power prediction data vectors in the category under different cluster number settings, and use the silhouette coefficient as the evaluation index for each clustering result. The cluster corresponding to the maximum value of the silhouette coefficient is the optimal clustering of the data of this category. The cluster numbers of categories A, B, C and D are represented by u, v, x and y respectively.

[0017] Furthermore, the optimal clustering is: Class A data is refined into u-class data, a1, a2...au are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the uth class; Class B data is refined into v-class data, b1, b2...bv are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the vth class; Class C data is refined into x-class data, c1, c2...cx are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the xth class; Class D data is refined into y-class data, d1, d2...dy are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the yth class.

[0018] Furthermore, in step 3, the correlation coefficient is divided into subcategories as follows:

[0019] If the class corresponding to the vector p to be corrected is A, the calculated correlation coefficient is

[0020] P A ={ρ a1_1 ,ρ a1_2 ,…,ρ a1_a1}∪{ρ a2_1 ,ρ a2_2 ,…,ρ a2_a2}…∪…{ρ au_1 ,ρ au_2 ,…,ρ au_au}(4)

[0021] ρ a1_1 is the correlation coefficient between the first vector in class a1 in A and the vector to be corrected p according to formula (3), ρ a1_2 is the correlation coefficient between the second vector in class a1 in A and the vector to be corrected p according to formula (3), and so on, ρ a1_a1 is the correlation coefficient between the a1th vector in the a1 class in A and the vector to be corrected p according to formula (3), ρ au_au is the correlation coefficient between the au-th vector in the au class in A and the vector to be corrected p according to formula (3);

[0022] If the class corresponding to the vector p to be corrected is B, the calculated correlation coefficient is:

[0023] P B ={ρ b1_1 ,ρ b1_2 ,…,ρ b1_b1}∪{ρ b2_1 ,ρ b2_2 ,…,ρ b2_b2}…∪…{ρ bv_1 ,ρ bv_2 ,…,ρ bv_bv}(5)

[0024] ρ b1_1 is the correlation coefficient between the first vector in class b1 in B and the vector to be corrected p according to formula (3), ρ b1_2 is the correlation coefficient between the second vector in class b1 in B and the vector to be corrected p according to formula (3), and so on, ρ b1_b1 is the correlation coefficient between the b1th vector in the b1 class in B and the vector to be corrected p according to formula (3), ρ bv_bv is the correlation coefficient between the bvth vector in the bv class in B and the vector to be corrected p according to formula (3);

[0025] If the class corresponding to the vector p to be corrected is C, the calculated correlation coefficient is:

[0026] P C ={ρ c1_1 ,ρ c1_2 ,…,ρ c1_c1}∪{ρ c2_1 ,ρ c2_2 ,…,ρ c2_c2}…∪…{ρ cx_1 ,ρ cx_2 ,…,ρ cx_cx}(6)

[0027] ρ c1_1 is the correlation coefficient between the first vector in the c1 class in C and the vector to be corrected p according to formula (3), ρ c1_2 is the correlation coefficient between the second vector in the c1 class in C and the vector to be corrected p according to formula (3), and so on, ρ c1_c1 is the correlation coefficient between the c1th vector in the c1th class in C and the vector to be corrected p according to formula (3), ρ cx_cx is the correlation coefficient between the cx-th vector in the cx class in C and the vector to be corrected p according to formula (3);

[0028] If the class corresponding to the vector p to be corrected is D, the calculated correlation coefficient is:

[0029] P D ={ρ d1_1 ,ρ d1_2 ,…,ρ d1_d1}∪{ρ d2_1 ,ρ d2_2 ,…,ρ d2_d2}…∪…{ρ dy_1 ,ρ dy_2 ,…,ρ dy_dy}(7)

[0030] ρ d1_1 is the correlation coefficient between the first vector in class d1 in D and the vector to be corrected p according to formula (3), ρ d1_2 is the correlation coefficient between the second vector in class d1 in D and the vector to be corrected p according to formula (3), and so on, ρ d1_d1 is the correlation coefficient between the c1th vector in the d1 class in D and the vector to be corrected p according to formula (3), ρ dy_dy is the correlation coefficient between the dyth vector in the dy class in D and the vector to be corrected p according to formula (3).

[0031] Furthermore, if the class corresponding to p is A, the maximum value of the correlation coefficient average value set is calculated as:

[0032] ρ a =mamx{ρ'1,ρ'2…ρ' au} (8)

[0033] In formula (8), ρ'1, ρ'2…ρ' au The formula is:

[0034]

[0035] If the category corresponding to p is B, the maximum value of the correlation coefficient mean value set is calculated as follows:

[0036] ρ b =max{ρ'1,ρ'2…ρ' bv} (10)

[0037] In formula (10), ρ'1, ρ'2…ρ' bv The formula is:

[0038]

[0039] If the category corresponding to p is C, the maximum value of the correlation coefficient mean value set is calculated as follows:

[0040] ρ c =max{ρ'1,ρ'2…ρ' cx} (12)

[0041] In formula (12), ρ'1, ρ'2…ρ' cx The formula is:

[0042]

[0043] If the category corresponding to p is D, the maximum value of the correlation coefficient average value set is calculated as follows:

[0044] ρ d =max{ρ'1,ρ'2…ρ' dy} (14)

[0045] In formula (14), ρ'1, ρ'2…ρ' dy The formula is:

[0046]

[0047] Furthermore, in step 4, the condition for whether p is to be corrected is: counting the number of days with power restrictions and the number of days without power restrictions in the small category most relevant to p. If the number of days without power restrictions is greater than the number of days with power restrictions, the wind power prediction data vector p to be corrected is not corrected. Otherwise, p needs to be corrected, and a correction matrix Q needs to be constructed.

[0048] Furthermore, in step 5, the correction matrix Q is constructed as follows: the number of power-limited days in the small category is defined as n, and the prediction data vector of the power-limited days is [P 1测 , P 2测 ,…,P n测 ], the corresponding actual power data vector is [P 1实 , P 2实 ,…,P n实 ], then the calculation formula of the correction matrix Q is:

[0049]

[0050] The calculation described in step 5 is:

[0051] p'=pQ (17)

[0052] In formula (17), p' is the corrected power prediction data vector.

[0053] The beneficial effects of the present invention are:

[0054] The present invention improves the accuracy of wind power prediction under power restriction by deeply exploring the power restriction rules of wind farms in a certain area, and provides a certain reference value for reducing the assessment of wind farms caused by "double regulations". BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0056] Figure 1 This is a flow chart of a wind power prediction and correction method considering power curtailment disclosed in the present invention;

[0057] Figure 2 A sub-process for building a wind power prediction reference database based on correlation;

[0058] Figure 3 This is the structure diagram of the reference database for wind power prediction. DETAILED DESCRIPTION

[0059] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0060] Example 1

[0061] The present invention provides a wind power prediction correction method considering power curtailment, such as Figure 1 As shown, it includes the following steps,

[0062] Step 1. Construct a reference database for wind power prediction based on K-means clustering;

[0063] Step 2. Calculate the correlation coefficient between the wind power forecast data vector p to be corrected and all data vectors in the corresponding reference database;

[0064] The wind power prediction data vector p to be corrected is directly obtained from the wind power prediction system of the wind farm and is a 1*96 column vector. The correlation coefficient calculation method is derived from formula (1):

[0065]

[0066] Step 3. Divide the correlation coefficient into subcategories, calculate the maximum value of the set of average correlation coefficients in the subcategories, and determine the set of subcategory prediction data vectors that are most correlated with the wind power prediction data vector p to be corrected;

[0067] Step 4. Determine whether p needs to be corrected by comparing the number of power-restricted days and the number of unrestricted days in the smallest category most related to p;

[0068] Step 5. If p needs to be corrected, a correction matrix Q is constructed and the corrected wind power prediction data vector p' is obtained through calculation.

[0069] Example 2

[0070] This embodiment is further limited based on the embodiment 1.

[0071] In step 1 of this embodiment, a reference database for wind power prediction is constructed based on correlation. The method for constructing the database is as follows: Figure 2 As shown. The construction process includes the following steps,

[0072] S11: Classify the short-term power forecast data for the entire year into four categories: A, B, C, and D, based on the four seasons. In step S11, the short-term power forecast data is the daily wind power forecast data, sampled at 15-minute intervals. One day's data forms a 1×96 column vector. Each of the four categories A, B, C, and D contains three months of wind power forecast data.

[0073] S12, K-means cluster analysis was used, with silhouette coefficient as the optimization criterion, to construct reference databases for short-term wind power prediction for A, B, C, and D, respectively.

[0074] In step S12, the K-means cluster analysis method is to set different cluster numbers for any one of the categories A, B, C and D, i.e., 2, 3, 4...10, and perform K-means cluster analysis under different cluster number settings on the wind power prediction data vector in the category, i.e., the three-month wind power prediction data contained in each of the four categories A, B, C and D obtained in step S11, and use the silhouette coefficient as the evaluation indicator for each clustering result. The cluster corresponding to the maximum value of the silhouette coefficient is the optimal cluster for the data of this category. The cluster numbers of categories A, B, C and D are represented by u, v, x and y respectively.

[0075] The optimal clustering is as follows: Class A data is refined into u-class data, a1, a2...au are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the uth class; Class B data is refined into v-class data, b1, b2...bv are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the vth class; Class C data is refined into x-class data, c1, c2...cx are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the xth class; Class D data is refined into y-class data, d1, d2...dy are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the yth class.

[0076] Taking the data in category A as an example, different cluster numbers are set, namely 2, 3, 4...10, and the wind power prediction data vectors in category A are subjected to K_means cluster analysis under different cluster number settings. The silhouette coefficient is used as the evaluation indicator of each clustering result. The cluster corresponding to the maximum silhouette coefficient is the optimal clustering of category A data. The number of clusters at this time is represented by u. At this time, category A data is refined into u category data, and a1, a2...au are the numbers of wind power prediction data vectors in category 1, category 2...u respectively. The calculation formula of the silhouette coefficient is as follows:

[0077]

[0078] In formula (2), a(i) is the distance from vector i to all other points in the cluster to which it belongs, and b(i) is the average distance from vector i to all points in a cluster that does not contain it. The average of the silhouette coefficients of all points is the total silhouette coefficient of the clustering result, which ranges from [-1, 1].

[0079] The method of constructing the wind power prediction reference database of B, C, and D is the same as that of A. After performing K-means clustering and optimization, they can be divided into v, x, and y categories respectively. The number of categories divided by A, B, C, and D may not be equal. The actual power data of the wind farm on the corresponding day is added to the daily data, and whether there is power restriction on that day is marked. At this point, the reference database for wind power prediction is established. The characteristics of the database structure are shown in the attached figure. Figure 3 shown.

[0080] Example 3

[0081] This embodiment is further limited based on the embodiment 2.

[0082] Calculate the correlation coefficient between the wind power forecast data vector p to be corrected and all data vectors in the corresponding database. The correlation coefficient calculation formula is:

[0083]

[0084] In formula (3), X and Y are two column vectors, n=96, and the correlation coefficient set between p and the data vectors in the corresponding class can be obtained according to the above formula.

[0085] In step 3, the correlation coefficient is divided into subcategories in the following manner:

[0086] If the class corresponding to the vector p to be corrected is A, the calculated correlation coefficient is:

[0087] P A ={ρ a1_1 ,ρ a1_2 ,…,ρ a1_a1}∪{ρ a2_1 ,ρ a2_2 ,…,ρ a2_a2}…∪…{ρ au_1 ,ρ au_2 ,…,ρ au_au} (4)

[0089] ρ a1_1 is the correlation coefficient between the first vector in class a1 in A and the vector to be corrected p according to formula (3), ρ a1_2 is the correlation coefficient between the second vector in class a1 in A and the vector to be corrected p according to formula (3), and so on, ρ a1_a1 is the correlation coefficient between the a1th vector in the a1 class in A and the vector to be corrected p according to formula (3), ρ au_au is the correlation coefficient between the au-th vector in the au class in A and the vector to be corrected p according to formula (3).

[0090] If the class corresponding to the vector p to be corrected is B, the calculated correlation coefficient is:

[0091] P B ={ρ b1_1 ,ρ b1_2 ,…,ρ b1_b1}∪{ρ b2_1 ,ρ b2_2 ,…,ρ b2_b2}…∪…{ρ bv_1 ,ρ bv_2 ,…,ρ bv_bv} (5)

[0093] ρ b1_1 is the correlation coefficient between the first vector in class b1 in B and the vector to be corrected p according to formula (3), ρ b1_2 is the correlation coefficient between the second vector in class b1 in B and the vector to be corrected p according to formula (3), and so on, ρ b1_b1 is the correlation coefficient between the b1th vector in class b1 in B and the vector to be corrected p according to formula (3), ρ bv_bv is the correlation coefficient between the bvth vector in the bv class in B and the vector to be corrected p according to formula (3).

[0094] If the class corresponding to the vector p to be corrected is C, the calculated correlation coefficient is:

[0095] P C ={ρ c1_1 ,ρ c1_2 ,…,ρ c1_c1}∪{ρ c2_1 ,ρ c2_2 ,…,ρ c2_c2}…∪…{ρ cx_1 ,ρ cx_2 ,…,ρ cx_cx}(6)

[0096] ρ c1_1 is the correlation coefficient between the first vector in the c1 class in C and the vector to be corrected p according to formula (3), ρ c1_2 is the correlation coefficient between the second vector in the c1 class in C and the vector to be corrected p according to formula (3), and so on, ρ c1_c1 is the correlation coefficient between the c1th vector in the c1th class in C and the vector to be corrected p according to formula (3), ρ cx_cx is the correlation coefficient between the cx-th vector in the cx class in C and the vector to be corrected p according to formula (3).

[0097] If the class corresponding to the vector p to be corrected is D, the calculated correlation coefficient is:

[0098] P D ={ρ d1_1 ,ρ d1_2 ,…,ρ d1_d1}∪{ρ d2_1 ,ρ d2_2 ,…,ρ d2_d2}…∪…{ρ dy_1 ,ρ dy_2 ,…,ρ dy_dy}(7)

[0099] ρ d1_1 is the correlation coefficient between the first vector in class d1 in D and the vector to be corrected p according to formula (3), ρ d1_2 is the correlation coefficient between the second vector in class d1 in D and the vector to be corrected p according to formula (3), and so on, ρ d1_d1 is the correlation coefficient between the c1th vector in the d1 class in D and the vector to be corrected p according to formula (3), ρ dy_dy is the correlation coefficient between the dyth vector in the dy class in D and the vector to be corrected p according to formula (3).

[0100] For example: Take P A For example, the class corresponding to the vector p to be corrected is A, which includes data of 90 days in three months. A In the equation, a1 is 10 data, a2 is 20 data, a3 is 30 data, a4 is 30 data, and u is 4.

[0101] In step 3, the method for determining the small category prediction data vector set most relevant to the vector p to be corrected is:

[0102] If the class corresponding to p is A, the maximum value of the correlation coefficient average value set is calculated as follows:

[0103] ρ a =max{ρ'1,ρ'2…ρ' au} (8)

[0104] In formula (8), ρ'1, ρ'2…ρ' au The formula is:

[0105]

[0106] In the above example, according to formula (9), ρ'1 is 0.8, ρ'2 is 0.9, ρ'3 is 0.7, and ρ'4 is 0.3. ρ'2 is the maximum value, so ρ a The value is 0.2, and the minimum category is 20 days. There are 15 days of power restrictions and 5 days of no power restrictions. Get 15 days of power restriction data.

[0107] If the class corresponding to p is B, the maximum value of the correlation coefficient average value set is calculated as follows:

[0108] ρ b =max{ρ'1,ρ'2…ρ' bv} (10)

[0109] In formula (10), ρ'1, ρ'2…ρ' bv The formula is:

[0110]

[0111] If the category corresponding to p is C, the maximum value of the correlation coefficient mean value set is calculated as follows:

[0112] ρ c =max{ρ'1,ρ'2…ρ' cx} (12)

[0113] In formula (12), ρ'1, ρ'2…ρ' cx The formula is:

[0114]

[0115] If the category corresponding to p is D, the maximum value of the correlation coefficient average value set is calculated as follows:

[0116] ρ d =max{ρ'1,ρ'2…ρ' dy} (14)

[0117] In formula (14), ρ'1, ρ'2…ρ' dy The formula is:

[0118]

[0119] This determines the set of small-category prediction data vectors most relevant to the vector p to be corrected. In step 4, the condition for whether p should be corrected is to count the number of days with and without power restrictions in the small-category most relevant to p. If the number of days with and without power restrictions is greater than the number of days with power restrictions, it indicates a high probability of power restrictions for the wind farm, and the vector p to be corrected is not corrected. Conversely, the probability of power restrictions is high, and the power prediction data vector p must be corrected. This requires constructing a correction matrix Q.

[0120] The smallest category most relevant to p in step 4 is not ABCD, but a category that is further subdivided than them. See the attached Figure 3 .

[0121] In step 5, the correction matrix Q is constructed as follows: the number of power-limited days in the small category is defined as n, and the prediction data vector of the power-limited days is [P 1测, P 2测 ,…,P n测 ], the corresponding actual power data vector is [P 1实 , P 2实 ,…,P n实 ], then the calculation formula of the correction matrix Q is:

[0122]

[0123] Among them, [P 1测 , P 2测 ,…,P n测 ] is the prediction vector of the number of power-restriction days in the small category database corresponding to vector p, [P 1实 , P 2实 ,…,P n实 ] is the actual power of the power-restriction days in the small category database corresponding to the vector p. This power can be directly obtained by the wind farm and is the historical record data.

[0124] The calculation method of p' in step 5 is:

[0125] p'=pQ (17)

[0126] In formula (17), p' is the corrected power prediction data vector.

[0127] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0128] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A wind power forecast correction method considering power curtailment, characterized in that: It includes the following steps, Step 1. Construct a reference database for wind power prediction based on K-means clustering; Step 2. Calculate the correlation coefficient between the wind power forecast data vector p to be corrected and all data vectors in the corresponding reference database; Step 3. Divide the correlation coefficient into subcategories, calculate the maximum value of the set of average correlation coefficients in the subcategories, and determine the set of subcategory prediction data vectors that are most correlated with the wind power prediction data vector p to be corrected; Step 4. Determine whether p needs to be corrected by comparing the number of power-restricted days and the number of unrestricted days in the smallest category most related to p; Step 5. If p needs to be corrected, a correction matrix Q is constructed and the corrected wind power prediction data vector p' is obtained through calculation.

2. A wind power prediction correction method considering power curtailment according to claim 1, characterized in that: In step 1, the construction process includes the following steps, S11. Classify the short-term power forecast data for the entire year in previous years into four categories: A, B, C, and D according to the four seasons; S12. Using K-means cluster analysis and silhouette coefficient as the optimization criterion, a reference database for short-term wind power prediction of A, B, C and D is constructed respectively.

3. The method for correcting wind power prediction considering power curtailment according to claim 2, characterized in that: In S11, the short-term power forecast data is daily wind power forecast data at intervals of 15 minutes. One day's data constitutes a 1×96 column vector. Each of the four types of data, A, B, C, and D, contains three months of wind power forecast data.

4. The method for correcting wind power prediction considering power curtailment according to claim 2, characterized in that: In S12, the K-means cluster analysis method is to set different cluster numbers for any one of the categories A, B, C and D, i.e., 2, 3, 4...10, and perform K-means cluster analysis on the wind power prediction data vectors in the category under different cluster number settings, and use the silhouette coefficient as the evaluation index for each clustering result. The cluster corresponding to the maximum value of the silhouette coefficient is the optimal clustering of the data of this category. The cluster numbers of categories A, B, C and D are represented by u, v, x and y, respectively.

5. The method for correcting wind power prediction considering power curtailment according to claim 4, characterized in that: The optimal clustering is as follows: Class A data is refined into u-class data, a1, a2...au are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the uth class; Class B data is refined into v-class data, b1, b2...bv are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the vth class; Class C data is refined into x-class data, c1, c2...cx are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the xth class; Class D data is refined into y-class data, d1, d2...dy are respectively the number of stroke power prediction data vectors of the 1st class, the 2nd class...the yth class.

6. A wind power prediction correction method considering power curtailment according to claim 5, characterized in that: In step 3, the correlation coefficient is divided into subcategories as follows: If the class corresponding to the vector p to be corrected is A, the calculated correlation coefficient is P A ={ρ a1_1 ,r a1_2 ,…,r a1_a1 }∪{r a2_1 ,r a2_2 ,…,r a2_a2 }…∪…{ρ au_1 ,r au_2 ,…,r au_au } (4) ρ a1_1 is the correlation coefficient between the first vector in class a1 in A and the vector to be corrected p, ρ a1_2 is the correlation coefficient between the second vector in class a1 in A and the vector to be corrected p, and so on, ρ a1_a1 is the correlation coefficient between the a1th vector in the a1 class in A and the vector to be corrected p, ρ au_au is the correlation coefficient between the au-th vector in the au class in A and the vector p to be corrected; If the class corresponding to the vector p to be corrected is B, the calculated correlation coefficient is: P B ={ρ b1_1 ,r b1_2 ,…,r b1_b1 }∪{r b2_1 ,r b2_2 ,…,r b2_b2 }…∪…{ρ bv_1 ,r bv_2 ,…,r bv_bv } (5) ρ b1_1 is the correlation coefficient between the first vector in class b1 in B and the vector to be corrected p, ρ b1_2 is the correlation coefficient between the second vector in class b1 in B and the vector to be corrected p, and so on, ρ b1_b1 is the correlation coefficient between the b1th vector in class b1 in B and the vector to be corrected p, ρ bv_bv is the correlation coefficient between the bvth vector in the bv class in B and the vector to be corrected p; If the class corresponding to the vector p to be corrected is C, the calculated correlation coefficient is: P C ={ρ c1_1 ,r c1_2 ,…,r c1_c1 }∪{r c2_1 ,r c2_2 ,…,r c2_c2 }…∪…{ρ cx_1 ,r cx_2 ,…,r cx_cx } (6) ρ c1_1 is the correlation coefficient between the first vector in class c1 in C and the vector to be corrected p, ρ c1_2 is the correlation coefficient between the second vector in the c1 class in C and the vector to be corrected p, and so on, ρ c1_c1 is the correlation coefficient between the c1th vector in the c1th class in C and the vector to be corrected p, ρ cx_cx is the correlation coefficient between the cx-th vector in the cx class in C and the vector p to be corrected; If the class corresponding to the vector p to be corrected is D, the calculated correlation coefficient is: P D ={ρ d1_1 ,r d1_2 ,…,r d1_d1 }∪{r d2_1 ,r d2_2 ,…,r d2_d2 }…∪…{ρ dy_1 ,r dy_2 ,…,r dy_dy } (7) ρ d1_1 is the correlation coefficient between the first vector in class d1 in D and the vector to be corrected p, ρ d1_2 is the correlation coefficient between the second vector in class d1 in D and the vector to be corrected p, and so on, ρ d1_d1 is the correlation coefficient between the c1th vector in the d1 class in D and the vector to be corrected p, ρ dy_dy is the correlation coefficient between the dyth vector in the dy class in D and the vector p to be corrected.

7. The method for correcting wind power prediction considering power curtailment according to claim 6, characterized in that: If the class corresponding to p is A, the maximum value of the correlation coefficient average value set is calculated as follows: r a =max{ρ'1,ρ'2…ρ' au } (8) In formula (8), ρ'1, ρ'2…ρ' au The formula is: If the category corresponding to p is B, the maximum value of the correlation coefficient mean value set is calculated as follows: r b =max{ρ'1,ρ'2…ρ' bv } (10) In formula (10), ρ'1, ρ'2…ρ' bv The formula is: If the category corresponding to p is C, the maximum value of the correlation coefficient mean value set is calculated as follows: r c =max{ρ'1,ρ'2…ρ' cx } (12) In formula (12), ρ'1, ρ'2…ρ' cx The formula is: If the category corresponding to p is D, the maximum value of the correlation coefficient mean value set is calculated as follows: r d =max{ρ'1,ρ'2…ρ' dy } (14) In formula (14), ρ'1, ρ'2…ρ' dy The formula is:

8. The method for correcting wind power prediction considering power curtailment according to claim 1, characterized in that: In step 4, the condition for whether p is to be corrected is: the number of days with power restrictions and the number of days without power restrictions in the small category most relevant to p are counted. If the number of days without power restrictions is greater than the number of days with power restrictions, the wind power forecast data vector p to be corrected is not corrected. Otherwise, p needs to be corrected, and the correction matrix Q needs to be constructed.

9. The method for correcting wind power prediction considering power curtailment according to claim 1, characterized in that: In step 5, the correction matrix Q is constructed as follows: the number of power-limited days in the small category is defined as n, and the prediction data vector of power-limited days is [P 1测 , P 2测 ,…,P n测 ], the corresponding actual power data vector is [P 1实 , P 2实 ,…,P n实 ], then the calculation formula of the correction matrix Q is: The p described in step 5 ’ The calculation method is: p'=pQ (17) In formula (17), p' is the corrected power prediction data vector.

Images