Ultra-short-term wind power forecasting method and system

Abstract

Disclosed are an ultra-short-term wind power forecasting method and system, relating to the technical field of artificial intelligence. The method comprises: obtaining an original dataset of a wind farm, processing the original dataset, and performing training on the basis of processed original data; decomposing wind speed data in the trained original data, calculating each decomposition component, and constructing a feature matrix on the basis of the calculation results; and introducing a residual attention mechanism to reconstruct the feature matrix, using the reconstructed result to establish a network model, performing secondary training, and forecasting ultra-short-term wind power. The present invention improves the accuracy and reliability of ultra-short-term wind power forecasting and achieves significant advances in algorithm optimization, thereby providing effective support for the stable power supply of renewable energy sources such as wind farms and for power grid operation.

Description

A method and system for predicting ultra-short-term wind power Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for predicting ultra-short-term wind power. Background Technology

[0002] In today's society, with the rapid development of data science and artificial intelligence, various industries are applying big data and intelligent algorithms to solve complex problems. This is especially true in the renewable energy sector, such as wind farms, where accurate prediction of wind speed and power is becoming increasingly crucial. This trend stems from the growing reliance on renewable energy and the continuous pursuit of reliable and stable energy supply. Against this backdrop, precise ultra-short-term wind power forecasting technology has become one of the important means to improve grid operating efficiency and optimize energy utilization. Therefore, the application of data science and artificial intelligence technologies in the renewable energy field is gradually gaining attention and importance, providing new opportunities and possibilities for solving the challenges of renewable energy sources such as wind farms.

[0003] Existing methods for ultra-short-term wind power prediction often have some shortcomings. Traditional prediction models may ignore outliers and missing values ​​in the wind speed data, resulting in inaccurate prediction results. In addition, existing methods may lack effective processing of complex data structures, leading to low prediction accuracy. This invention adopts improved adaptive noise complete ensemble empirical mode decomposition model, residual attention mechanism and gold mining optimization algorithm to solve the above problems and improve the accuracy and reliability of ultra-short-term wind power prediction. Summary of the Invention

[0004] In view of the problems existing in the current ultra-short-term wind power prediction methods and systems, this invention is proposed.

[0005] Therefore, the purpose of this invention is to provide an ultra-short-term wind power prediction method and system. In view of the problems of existing technologies ignoring outliers and missing values ​​when processing wind speed data, and insufficient processing of complex data structures, this invention adopts quartile filtering and improved adaptive noise complete integration empirical mode decomposition model and other technical means to solve these problems.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, embodiments of the present invention provide an ultra-short-term wind power prediction method, which includes acquiring and processing the original dataset of a wind farm, and training based on the processed original data.

[0008] An improved adaptive noise-complete ensemble empirical mode decomposition model is adopted to decompose the wind speed data in the original data after training, calculate each decomposition component, and construct a feature matrix based on the calculation results.

[0009] A residual attention mechanism is introduced to reconstruct the feature matrix. The reconstructed result is used to build a network model and perform secondary training to predict ultra-short-term wind power.

[0010] As a preferred embodiment of the ultra-short-term wind power prediction method of the present invention, the processing includes filtering the original dataset using the quartile method, wherein the filtering includes removing outliers and missing values ​​from the original dataset, and the specific steps are as follows:

[0011] Calculate the data sequence X = {x1, x2, ..., x} in the original dataset. n The second quartile Q2 of the sequence is located in the middle if n = 2k + 1, and in the middle if n = 2k. The specific formulas for calculating the first quartile Q1, the second quartile Q2, and the third quartile Q3 are as follows:

[0012] Where k = 0, 1, 2, ..., then n = 2k + 1, and Q2 is in the middle position; if k = 1, 2, ..., then n = 2k, and the average of the two middle numbers is Q2.

[0013] The removal of outliers and missing values ​​from the original dataset includes defining I based on the difference between the third quartile Q3 and the first quartile Q1. QR Based on the definition result I QR The upper and lower limits of the difference are calculated using the following formula: I QR =Q3-Q1

[0014] Among them, the first quartile Q1 and the third quartile Q3 are used to remove outliers from the data by using the upper limit W1 and the lower limit W2;

[0015] The specific formula for calculating the upper and lower limits of the difference is: [W1,W2]=[Q1-1.5I QR Q3+1.5I QR ]

[0016] Where Q3 represents the third quartile, Q1 represents the first quartile, and Q2 represents the second quartile.

[0017] As a preferred embodiment of the ultra-short-term wind power prediction method of the present invention, the wind speed data decomposition includes using the ICEEMDAN algorithm to decompose the wind speed data into n IMF components, calculating the entropy value of each IMF component, and constructing the original meteorological data. The specific calculation steps are as follows:

[0018] The data sequence in the original dataset is used to construct an m-dimensional vector, specifically using the formula: Y(t)=[y(t),y(t+1),…,y(t-m+1)]

[0019] Where m represents the embedding dimension, t represents the number of times, and Y(t) represents the vector;

[0020] Define vectors Y(j) and Y(k) based on the constructed vector Y(t), and calculate the distance d[Y(t),Y(j)] between vectors Y(t) and Y(j). The specific calculation formula is as follows:

[0021] Where Y(j) is the j-th dimension vector, z represents the parameter, and d[Y(t),Y(j)] represents the distance between vector Y(t) and vector Y(j);

[0022] The ratio of the number of values ​​where the distance between Y(t) and Y(k) is less than r to the total number of values ​​Nm. Where r = [0.1*std + 0.25*std], the ratio The specific calculation formula is as follows:

[0023] in, This represents the ratio between Y(t) and Y(k), where m and N represent variables;

[0024] ratio Perform a loop calculation to obtain B. m+1 (t), according to and B m+1 (t) Calculate the entropy value of each IMF component. The specific calculation formula is as follows:

[0025] Where m represents the embedding dimension, and r = [0.1*std 0.25*std].

[0026] As a preferred embodiment of the ultra-short-term wind power prediction method of the present invention, the residual attention mechanism includes allocating weights to the original meteorological data, calculating residual connections based on the allocation results, and introducing residual connections to optimize the feature matrix.

[0027] The specific formula for calculating the residual connection is as follows:

[0028] Where X represents the original meteorological data matrix, and This indicates that the weight matrix can be updated. and The bias term corresponding to the updatable weight matrix is ​​δ, which represents the ReLU activation function and σ represents the Sigmoid function.

[0029] The specific calculation formula for optimizing the feature matrix by introducing residual connections is: X xin =A⊙XX att =X+X xin

[0030] Where ⊙ represents the Hadamard product, X xin This represents the feature matrix after adding attention weights.

[0031] As a preferred embodiment of the ultra-short-term wind power prediction method of the present invention, the reconstruction includes constructing matrix M using a gold mining optimization algorithm. GP Determine the initial search agent location for the gold prospector and migrate to a new location based on the current search agent location;

[0032] The matrix M GP The specific calculation formula is as follows:

[0033] Where, x ij Let d represent the position of prospector i in the j-th dimension, d represent the dimension size, and n represent the number of gold prospectors.

[0034] The migration includes the initial gold prospectors based on matrix M. GP The specific formula for calculating and obtaining the new migration location is as follows:

[0035] in, This indicates the location of the gold mine at iteration number t. This indicates the location of the gold prospector at iteration number t.

[0036] As a preferred embodiment of the ultra-short-term wind power prediction method of the present invention, the establishment of the network model includes introducing Tent chaotic mapping and nonlinear inertia weighting factors to arrange the population positions of the gold mining optimization algorithm. The specific calculation formula of the Tent chaotic mapping is as follows:

[0037] Where, a∈(0,1), x n Representation matrix M GP Search agent location within.

[0038] As a preferred embodiment of the ultra-short-term wind power prediction method of the present invention, the nonlinear inertia weighting factor includes iterating over the arranged positions, and the specific calculation formula is as follows:

[0039] As the value of w gradually increases, the ability of the optimization individual to be influenced by the current gold miner's position and its local search ability become stronger.

[0040] The secondary training includes calculating the iterative convergence rate using the iterative results to complete the ultra-short-term wind power prediction. The specific formula for calculating the iterative convergence rate is as follows:

[0041] in, Let r represent the position of the gold prospector, and w represent the nonlinear inertia weighting factor. -- This indicates the optimal gold mine location at iteration number t, and the ultra-short-term wind power is predicted based on different iteration locations.

[0042] Secondly, embodiments of the present invention provide an ultra-short-term wind power prediction method and system, which includes: a training module, which acquires the original dataset of a wind farm, processes the original dataset, and performs training based on the processed original data;

[0043] The computation module decomposes the wind speed data in the original training data by using an improved adaptive noise complete ensemble empirical mode decomposition model.

[0044] The prediction module uses the reconstructed results to build a network model and perform secondary training to predict ultra-short-term wind power.

[0045] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of the above-described ultra-short-term wind power prediction method.

[0046] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the above-described ultra-short-term wind power prediction method.

[0047] The beneficial effects of this invention are as follows: By employing the quartile method to filter outliers and missing values, and introducing a residual attention mechanism and a gold mining optimization algorithm, this invention optimizes the feature matrix reconstruction and network model establishment process, enabling the model to better capture the characteristics of wind speed data and effectively reduce prediction errors. Furthermore, by using Tent chaotic mapping and nonlinear inertia weighting factors to arrange the population positions of the optimization algorithm, the search efficiency and convergence speed of the algorithm are further improved. This invention not only improves the accuracy and reliability of ultra-short-term wind power prediction but also achieves significant progress in optimization algorithms, providing effective support for the stable power supply of wind farms and other renewable energy sources and the operation of the power grid. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0049] Figure 1 is a flowchart of an ultra-short-term wind power prediction method and system provided in an embodiment of the present invention.

[0050] Figure 2 is a flowchart of a method and system for predicting ultra-short-term wind power according to an embodiment of the present invention. Detailed Implementation

[0051] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0052] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0054] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0055] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0056] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0057] Example 1

[0058] Referring to Figures 1 and 2, the first embodiment of the present invention provides a method for predicting ultra-short-term wind power, including:

[0059] S1: Obtain the raw dataset of the wind farm and process it, then train the system based on the processed raw data.

[0060] The processing includes filtering the original dataset using the quartile method. This filtering involves removing outliers and missing values ​​from the original dataset. The specific steps are as follows:

[0061] Calculate the data sequence X = {x1, x2, ..., x} in the original dataset. n The second quartile Q2 of the sequence is located in the middle if n = 2k + 1, and in the middle if n = 2k. The specific formulas for calculating the first quartile Q1, the second quartile Q2, and the third quartile Q3 are as follows:

[0062] Where k = 0, 1, 2, ..., then n = 2k + 1, and Q2 is in the middle position; if k = 1, 2, ..., then n = 2k, and the average of the two middle numbers is Q2.

[0063] Removing outliers and missing values ​​from the original dataset involves defining I based on the difference between the third quartile Q3 and the first quartile Q1. QR Based on the definition result I QR The upper and lower limits of the difference are calculated using the following formula: I QR =Q3-Q1

[0064] Among them, the first quartile Q1 and the third quartile Q3 are used to remove outliers from the data by using the upper limit W1 and the lower limit W2;

[0065] The specific formula for calculating the upper and lower limits of the difference is: [W1,W2]=[Q1-1.5I QR Q3+1.5I QR ]

[0066] Where Q3 represents the third quartile, Q1 represents the first quartile, and Q2 represents the second quartile.

[0067] S2: An improved adaptive noise-complete ensemble empirical mode decomposition model is adopted to decompose the wind speed data in the original data after training, calculate each decomposition component, and construct a feature matrix based on the calculation results.

[0068] The decomposition of wind speed data includes using the ICEEMDAN algorithm to decompose the wind speed data into n IMF components, calculating the entropy value of each IMF component, and constructing the original meteorological data. The specific calculation steps are as follows:

[0069] The data sequence in the original dataset is used to construct an m-dimensional vector, specifically using the formula: Y(t)=[y(t),y(t+1),…,y(t-m+1)]

[0070] Where m represents the embedding dimension, t represents the number of times, and Y(t) represents the vector;

[0071] Define vectors Y(j) and Y(k) based on the constructed vector Y(t), and calculate the distance d[Y(t),Y(j)] between vectors Y(t) and Y(j). The specific calculation formula is as follows:

[0072] Where Y(j) is the j-th dimension vector, z represents the parameter, and d[Y(t),Y(j)] represents the distance between vector Y(t) and vector Y(j);

[0073] The ratio of the number of values ​​where the distance between Y(t) and Y(k) is less than r to the total number of values ​​Nm. Where r = [0.1*std + 0.25*std], the ratio The specific calculation formula is as follows:

[0074] in, This represents the ratio between Y(t) and Y(k), where m and N represent variables;

[0075] ratio Perform a loop calculation to obtain B. m+1 (t), according to and B m+1 (t) Calculate the entropy value of each IMF component. The specific calculation formula is as follows:

[0076] Where m represents the embedding dimension, and r = [0.1*std 0.25*std].

[0077] Furthermore, the above text describes the specific steps for wind speed data decomposition based on the improved adaptive noise-complete ensemble empirical mode decomposition model. First, the wind speed data sequences in the original dataset are constructed into 2^n vectors, where the embedding dimension is n, representing the length of the time series. Based on the constructed vectors, a vector sum is defined, and the distance between each vector is calculated. The number of vector pairs with a distance less than a threshold is counted and compared with the total number of vector pairs to obtain the ratio. Based on the ratio, the entropy values ​​of the n IMF components are calculated iteratively. The entropy value is calculated using the formula: where n represents the embedding dimension.

[0078] Specifically, if the wind speed data sequence in the original dataset is 100 units long and has an embedding dimension of 5, then firstly, a vector of dimension 5 is constructed. Next, the distance between each vector is calculated, and the number of vector pairs with a distance less than a certain threshold is counted. For example, if 60 vector pairs have a distance less than the threshold, the ratio is 60 / 100 = 0.6. Based on this ratio, a loop is performed to obtain the entropy values ​​of the n IMF components. Assuming the calculated entropy values ​​are 0.3, 0.4, 0.5, 0.6, and 0.7, these five values ​​are the entropy values ​​of each IMF component. The wind speed data decomposition is shown in Table 1 below.

[0079] Table 1. Decomposition of Wind Speed ​​Data

[0080] The table above summarizes the key numerical values ​​in the wind speed data decomposition process. First, the wind speed data sequence has a length of 100 and an embedding dimension of 5, representing the time length of the original data sequence and the dimension of the vectors. Second, based on the distance between vectors, 60 vector pairs with a distance less than a threshold were identified, out of a total of 100 vector pairs. The calculated ratio is 0.6, reflecting the degree of similarity between vectors. Finally, the comparison values ​​are iteratively calculated to obtain the entropy values ​​of the five IMF components: 0.3, 0.4, 0.5, 0.6, and 0.7. These entropy values ​​reflect the information content and complexity of each IMF component. The entire process, through numerical calculation, provides a reliable quantitative basis for the decomposition of wind speed data.

[0081] S3: Introduce a residual attention mechanism to reconstruct the feature matrix, use the reconstructed result to build a network model and perform secondary training to predict ultra-short-term wind power.

[0082] The residual attention mechanism includes assigning weights to the original meteorological data, calculating residual connections based on the assignment results, and introducing residual connections to optimize the feature matrix.

[0083] The specific formula for calculating residual join is:

[0084] Where X represents the original meteorological data matrix, and This indicates that the weight matrix can be updated. and The bias term corresponding to the updatable weight matrix is ​​δ, which represents the ReLU activation function and σ represents the Sigmoid function.

[0085] The specific calculation formula for optimizing the feature matrix by introducing residual connectivity is: X xin =A⊙XX att =X+X xin

[0086] Where ⊙ represents the Hadamard product, X xin This represents the feature matrix after adding attention weights.

[0087] S3.1: Reconstruction includes constructing matrix M using a gold mining optimization algorithm. GP Determine the initial search agent location for the gold prospector and migrate to a new location based on the current search agent location;

[0088] Matrix M GP The specific calculation formula is as follows:

[0089] Where, x ijLet d represent the position of prospector i in the j-th dimension, d represent the dimension size, and n represent the number of gold prospectors.

[0090] Migration includes the initial gold prospectors based on matrix M GP The specific formula for calculating and obtaining the new migration location is as follows:

[0091] in, This indicates the location of the gold mine at iteration number t. This indicates the location of the gold prospector at iteration number t.

[0092] Furthermore, establishing the network model includes introducing the Tent chaotic map and a nonlinear inertia weight factor to arrange the population positions of the gold mining optimization algorithm. The specific calculation formula for the Tent chaotic map is as follows:

[0093] Where, a∈(0,1), x n Representation matrix M GP Search agent location within.

[0094] Furthermore, the nonlinear inertia weighting factor includes iterating over the rearranged positions, and the specific calculation formula is as follows:

[0095] As the value of w gradually increases, the ability of the optimization individual to be influenced by the current gold miner's position and its local search ability become stronger.

[0096] Secondary training involves calculating the iterative convergence rate using the iterative results to complete the ultra-short-term wind power prediction. The specific formula for calculating the iterative convergence rate is as follows:

[0097] in, Let r represent the position of the gold prospector, and w represent the nonlinear inertia weighting factor. -- This indicates the optimal gold mine location at iteration number t, and the ultra-short-term wind power is predicted based on different iteration locations.

[0098] In a preferred embodiment, an ultra-short-term wind power prediction method and system are provided. The system includes a training module, which acquires the original dataset of the wind farm, processes the original dataset, and trains based on the processed original data.

[0099] The computation module decomposes the wind speed data in the original training data by using an improved adaptive noise complete ensemble empirical mode decomposition model.

[0100] The prediction module uses the reconstructed results to build a network model and perform secondary training to predict ultra-short-term wind power.

[0101] The above-mentioned unit modules can be embedded in the processor of the computer device in hardware form or independent of it, or they can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0102] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0103] In summary, this invention employs the quartile method to filter outliers and missing values, and introduces a residual attention mechanism and a gold mining optimization algorithm to optimize the feature matrix reconstruction and network model building process. This allows the model to better capture the characteristics of wind speed data and effectively reduce prediction errors. Furthermore, the use of Tent chaotic mapping and nonlinear inertia weighting factors to arrange the population positions of the optimization algorithm further improves the search efficiency and convergence speed. This invention not only improves the accuracy and reliability of ultra-short-term wind power prediction but also achieves significant progress in optimization algorithms, providing effective support for the stable power supply of wind farms and other renewable energy sources and the operation of the power grid.

[0104] Example 2

[0105] Referring to Figures 1 and 2, the second embodiment of the present invention is provided, which provides a method for predicting ultra-short-term wind power. To verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0106] Outlier handling is based on quartile calculations. Suppose we have a dataset where the first quartile is 10, the second quartile (median) is 20, and the third quartile is 30. The calculated difference is 20. Values ​​greater than 30 plus 20 or less than 10 minus 20 are considered outliers. Therefore, in this dataset, values ​​greater than 50 or less than -10 will be considered outliers. The outlier handling data is shown in Table 2 below.

[0107] Table 2. Outlier Handling Data Table

[0108] The outlier handling table displays the three quartiles of the dataset. The differences and upper and lower limits of outliers calculated based on these quartiles can be used to determine which values ​​in the dataset are considered outliers and to perform appropriate processing to ensure the accuracy and reliability of subsequent analysis. A comparison with existing techniques is shown in Table 3 below.

[0109] Table 3 Comparison with Existing Technologies

[0110] This invention uses the quartile method to define the upper and lower limits of outliers and missing values, which is more accurate than traditional statistical methods. In addition, this invention utilizes the residual attention mechanism to optimize features, which can better capture data features and improve the accuracy of the prediction model compared to traditional modeling methods. At the same time, this invention introduces the gold mining optimization algorithm to accelerate convergence, which is significantly better than traditional optimization algorithms in terms of algorithm efficiency. Overall, this invention has significant advantages in handling outliers and missing values, building prediction models, and algorithm efficiency, thereby improving the accuracy and reliability of wind power prediction.

[0111] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting ultra-short-term wind power, characterized in that: include, Obtain and process the raw dataset of the wind farm, and then train the system based on the processed raw data. An improved adaptive noise-complete ensemble empirical mode decomposition model is used to decompose the wind speed data in the original training data, calculate each decomposition component, and construct a feature matrix based on the calculation results. A residual attention mechanism is introduced to reconstruct the feature matrix. The reconstructed result is used to build a network model and perform secondary training to predict ultra-short-term wind power.

2. The ultra-short-term wind power prediction method as described in claim 1, characterized in that: The processing includes filtering the original dataset using the quartile method. This filtering includes removing outliers and missing values ​​from the original dataset. The specific steps are as follows: Calculate the data sequence X = {x1, x2, ..., x} in the original dataset. n The second quartile Q2 of the sequence is located in the middle if n = 2k + 1, and in the middle if n = 2k. The specific formulas for calculating the first quartile Q1, the second quartile Q2, and the third quartile Q3 are as follows: Where k = 0, 1, 2, ..., then n = 2k + 1, and Q2 is in the middle position; where k = 1, 2, ..., then n = 2k, and the average of the two middle numbers is Q2. The removal of outliers and missing values ​​from the original dataset includes defining I based on the difference between the third quartile Q3 and the first quartile Q1. QR Based on the definition result I QR The upper and lower limits of the difference are calculated using the following formula: I QR =Q3-Q1 Among them, the first quartile Q1 and the third quartile Q3 are used to remove outliers from the data by using the upper limit W1 and the lower limit W2; The specific formulas for calculating the upper and lower limits of the difference are as follows: [W1,W2]=[Q1-1.5I QR ,Q3+1.5I QR ] Where Q3 represents the third quartile, Q1 represents the first quartile, and Q2 represents the second quartile. quartiles.

3. The ultra-short-term wind power prediction method as described in claim 2, characterized in that: The decomposition of the wind speed data includes using the ICEEMDAN algorithm to decompose the wind speed data into n IMF components, calculating the entropy value of each IMF component, and constructing the original meteorological data. The specific calculation steps are as follows: The data sequence in the original dataset is used to construct an m-dimensional vector. The specific construction formula is as follows: Y(t)=[y(t),y(t+1),·.·,y(t-m+1)] Where m represents the embedding dimension, t represents the number of times, and Y(t) represents the vector; Define vectors Y(j) and Y(k) based on the constructed vector Y(t), and calculate the distance d[Y(t),Y(j)] between vectors Y(t) and Y(j). The specific calculation formula is as follows: Where Y(j) is the j-th dimension vector, z represents the parameter, and d[Y(t),Y(j)] represents the distance between vector Y(t) and vector Y(j); The ratio of the number of values ​​where the distance between Y(t) and Y(k) is less than r to the total number of values ​​Nm. Where r = [0.1*std + 0.25*std], the ratio The specific calculation formula is as follows: in, This represents the ratio between Y(t) and Y(k), where m and N represent variables; ratio Perform a loop calculation to obtain B. m+1 (t), according to and B m+1 (t) Calculate the entropy value of each IMF component. The specific calculation formula is as follows: Where m represents the embedding dimension, and r = [0.1*std 0.25*std].

4. The ultra-short-term wind power prediction method as described in claim 3, characterized in that: The residual attention mechanism includes assigning weights to the original meteorological data, calculating residual connections based on the assignment results, and introducing residual connections to optimize the feature matrix. The specific formula for calculating the residual connection is as follows: Where X represents the original meteorological data matrix, and This indicates that the weight matrix can be updated. and The bias term corresponding to the updatable weight matrix is ​​δ, which represents the ReLU activation function and σ represents the Sigmoid function. The specific calculation formula for optimizing the feature matrix by introducing residual connections is as follows: X xin =A⊙X X att =X+X xin Where ⊙ represents the Hadamard product, X xin This represents the feature matrix after adding attention weights.

5. The ultra-short-term wind power prediction method as described in claim 4, characterized in that: The reconstruction includes constructing matrix M using a gold mining optimization algorithm. GP Determine the initial search agent location for the gold prospector and migrate to a new location based on the current search agent location; The matrix M GP The specific calculation formula is as follows: Where, x ij Let d represent the position of prospector i in the j-th dimension, d represent the dimension size, and n represent the number of gold prospectors. The migration includes the initial gold prospectors based on matrix M. GP The specific formula for calculating and obtaining the new migration location is as follows: in, This indicates the location of the gold mine at iteration number t. This indicates the location of the gold prospector at iteration number t.

6. The ultra-short-term wind power prediction method as described in claim 5, characterized in that: The establishment of the network model includes introducing the Tent chaotic map and a nonlinear inertia weight factor to arrange the population positions of the gold mining optimization algorithm. The specific calculation formula of the Tent chaotic map is as follows: Where, a∈(0,1), x n Representation matrix M GP Search agent location within.

7. The ultra-short-term wind power prediction method as described in claim 6, characterized in that: The nonlinear inertia weighting factor involves iterating over the arranged positions, and the specific calculation formula is as follows: As the value of w gradually increases, the ability of the optimization individual to be influenced by the current gold prospector's position and its local search ability improve. Become stronger; The secondary training includes calculating the iterative convergence rate using the iterative results to complete the ultra-short-term wind power prediction. The specific formula for calculating the iterative convergence rate is as follows: in, Let r represent the position of the gold prospector, and w represent the nonlinear inertia weighting factor. -- This indicates the optimal gold mine location at iteration number t, and the ultra-short-term wind power is predicted based on different iteration locations.

8. A short-term wind power prediction system, based on the short-term wind power prediction method according to any one of claims 1 to 7, characterized in that: include, The training module acquires the original dataset of the wind farm, processes the original dataset, and performs training based on the processed original data. The computation module decomposes the wind speed data in the original training data by using an improved adaptive noise complete ensemble empirical mode decomposition model. The prediction module uses the reconstructed results to build a network model and perform secondary training to predict ultra-short-term wind power.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the ultra-short-term wind power prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the ultra-short-term wind power prediction method according to any one of claims 1 to 7.

Images