Wind power generation power prediction method and system combining local and global features

Abstract

The invention relates to a wind power generation power prediction method and system combining local and global features. The method comprises the following steps: collecting wind power generation power influence factor data and external environment variable data; constructing a wind power generation power influence factor sequence and an external environment variable sequence; performing preliminary mixed feature extraction on the wind power generation power influence factor sequence and the external environment variable sequence; local features of wind power generation power are extracted from the preliminary mixed features through a time domain convolution branch and a wavelet transform branch, and fusion is carried out through two-dimensional convolution; through short-time Fourier transform and fast Fourier transform, capturing short-term abrupt change and long-term periodic information in the frequency domain features, and performing fusion through a global convolution kernel to obtain global features of wind power generation power; and merging the local features and the global features of the wind power generation power, and inputting the merged features into a feedforward neural network to obtain a prediction result of the wind power generation power.

Description

Technical Field

[0001] The present application relates to the technical field of power generation prediction, and more specifically, to a method and system for predicting wind power generation combining local and global characteristics. Background Art

[0002] As the global demand for clean energy continues to grow, wind power generation, as an important form of renewable energy utilization, is expanding in scale and proportion. However, accurate prediction of wind power generation has always been a key problem faced by this field. The output power of wind power generation mainly depends on wind energy resources, and wind energy itself has natural instability. The size and direction of wind speed change all the time, and this change shows complex nonlinear and non-stationary characteristics. At the same time, changes in wind direction will also have an important impact on wind power generation. There are differences in the power generation efficiency of wind turbines under different wind direction angles. In addition to the characteristics of wind energy itself, external environmental factors cannot be ignored. Changes in temperature will affect air density, which in turn affects the output of wind turbines; changes in humidity may lead to changes in equipment performance, such as affecting the insulation performance of electrical equipment; fluctuations in air pressure are also related to air density and affect the operating status of wind turbines. There is a complex relationship between these external environmental variables and wind power generation, which further increases the difficulty of power prediction.

[0003] Existing wind power generation prediction methods often find it difficult to fully and accurately capture various features and their interrelationships when faced with these complex factors. Traditional prediction methods have limitations when dealing with nonlinear and non-stationary data, resulting in low prediction accuracy. Some methods fail to fully consider the characteristics of different factors in the time and frequency domains during feature extraction, and are unable to effectively mine the potential information in the data. Moreover, the lack of comprehensive utilization of external environmental variables makes the model less adaptable and generalizable under different weather conditions. These problems seriously restrict the reliability and effectiveness of wind power generation prediction, and a new prediction method is urgently needed to improve the prediction performance to better meet the operation and management needs of wind power generation systems.

[0004] Prior art, such as a Chinese patent with patent number "CN118211494A", discloses a method for constructing a wind speed prediction hybrid model based on a correlation matrix, the method comprising the following steps: S1: sampling the time series wind speed data through a sliding time window to obtain a sub-data group, and constructing a wind speed correlation matrix with the sub-data group; S2: extracting wind speed features through a multi-scale feature enhanced convolutional network, and generating a down-sampled feature map containing wind speed information;

[0005] S3: Filter the optimal feature subset through the feature selection method of random forest; S4: Perform sequence analysis through the bidirectional long short-term memory network and output the wind speed prediction result. The wind speed prediction hybrid model based on the correlation matrix provided by the invention recognizes the long-term dependency and periodic pattern of wind speed data through the multi-scale feature enhanced convolution network, replaces a convolution layer in the traditional convolution network with a selective convolution kernel network, and can better adapt to the dynamic characteristics and complexity of wind speed data through its unique selective kernel network structure.

[0006] The problem with the above prior art is that the method uses a bidirectional long short-term memory network to perform sequence analysis to output wind speed prediction results. Although the bidirectional long short-term memory network can process sequence data, it may be affected by the gradient vanishing or gradient exploding problem. In the wind speed prediction of long time series, as the number of time steps increases, this problem may cause the model to be unable to effectively learn long-term dependencies, thereby affecting the prediction accuracy. Summary of the invention

[0007] In order to solve the above technical problems, the present invention proposes a wind power prediction method and system combining local and global characteristics.

[0008] The technical solution of the present invention is as follows:

[0009] The present invention proposes a wind power prediction method combining local and global characteristics, comprising the following steps:

[0010] Step S1, collecting data on factors affecting wind power generation and external environmental variables, the data on factors affecting wind power generation including historical wind direction, wind speed and wind power generation; the external environmental variables including temperature, humidity and air pressure;

[0011] Step S2, constructing a wind power generation power influencing factor sequence and an external environment variable sequence based on the wind power generation power influencing factor data and the external environment variable data;

[0012] Step S3, performing preliminary mixed feature extraction on the sequence of factors affecting wind power generation and the sequence of external environmental variables;

[0013] Step S4, extracting local features of wind power from the preliminary mixed features through the time domain convolution branch and the wavelet transform branch respectively, and merging the local features extracted by the time domain convolution branch and the wavelet transform branch into a unified local feature through two-dimensional convolution;

[0014] Step S5, extracting long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on the local features, and fusing the long-term periodic information and short-term mutation information through a global convolution kernel to obtain the global features of wind power generation;

[0015] Step S6, merging the local features and the global features of the wind power generation, and inputting the merged features into a feedforward neural network to obtain a prediction result of the wind power generation.

[0016] As a preferred implementation, in the step of constructing a sequence of factors affecting wind power generation and a sequence of external environmental variables:

[0017] The expression of the wind power generation factor sequence X is as follows:

[0018]

[0019] Where: v n is the wind speed at time n; θ n is the wind direction at time n; p n is the wind power generation power at time n;

[0020] The expression of the external environment variable sequence E is as follows:

[0021]

[0022] Where: w n is the temperature at time n; h n is the humidity at time n; q n is the air pressure at time n.

[0023] As a preferred implementation, the preliminary mixed feature extraction is performed on the sequence of factors affecting wind power generation and the sequence of external environmental variables. The preliminary mixed feature extraction formula is as follows:

[0024] F input =VE(X)+PE(X)+Attn(E);

[0025] Where: F input is the preliminary mixed feature; VE() is the value embedding function; PE() is the position embedding function; Attn() is the weighted attention mechanism.

[0026] As a preferred embodiment, the local features of wind power generation are extracted from the preliminary mixed features respectively through the time domain convolution branch and the wavelet transform branch, wherein:

[0027] The local time series feature expression of wind power generation extracted by time domain convolution branch is as follows:

[0028] F conv,l =Conv1D(k l ,s l ),l∈{1,2,…,m};

[0029] Where: Fconv,l is the local time series feature extracted by the lth time domain convolution branch; Conv1D() is a one-dimensional convolution operation; k l is the convolution kernel size of the lth time domain convolution branch; s l is the step size of the lth time domain convolution branch; m is the number of time domain convolution branches;

[0030] The instantaneous variation characteristic expression of wind power extracted by wavelet transform branch is as follows:

[0031] F wave,b = WaveletTransform(F input ,w b ), b ∈{1,2,…,r};

[0032] Where: F wave,b is the instantaneous change feature extracted by the bth wavelet transform branch;

[0033] WaveletTransform() is a wavelet transform operation; b is the wavelet function of the bth wavelet transform branch; r is the number of wavelet decomposition layers.

[0034] As a preferred implementation, the local features extracted by the time domain convolution branch and the wavelet transform branch are combined into a unified local feature by two-dimensional convolution, and the local feature is obtained by the following formula:

[0035] F local =Conv2D(Concat(F up,1 ,F up,2 ,…,F up,m+r ));

[0036] F local is the merged local feature; Conv2D() is a two-dimensional convolution operation; Concat() is a concatenation operation; F up,m+r The local features are obtained by linearly interpolating the local features extracted by the time domain convolution branch and the wavelet transform branch.

[0037] As a preferred implementation, the method extracts long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on local features, and fuses the long-term periodic information and short-term mutation information through a global convolution kernel to obtain the global features of wind power generation, which specifically includes the following steps:

[0038] Perform short-time Fourier transform and fast Fourier transform on local features, as shown in the following formula:

[0039]

[0040] Where: F stft is the frequency domain result after short-time Fourier transform; F Ftt is the frequency domain result after fast Fourier transform;

[0041] The convolution theorem captures the long-term dependencies in frequency domain features as follows:

[0042] F stft_freq =F stft ⊙K stft ;

[0043] F ftt_freq =F fft ⊙K fft ;

[0044] Where: F stft_freq is the convolution result of short-time Fourier transform frequency domain features; F ftt_freq is the convolution result of the fast Fourier transform frequency domain feature; ⊙ is the Hadamard product; K stft is the short-time Fourier transform global convolution kernel; K fft is the fast Fourier transform global convolution kernel;

[0045] The frequency domain features are restored to the time domain through inverse Fourier transform, as shown in the following formula:

[0046] F global =IFFT(F ftt_freq +F stft_freq );

[0047] Where: F global is the global characteristic of wind power generation; IFFT() is the inverse Fourier transform operation.

[0048] As a preferred implementation, the local features and global features of wind power generation are combined, and the combined features are input into a feedforward neural network to obtain a prediction result of wind power generation; wherein, the specific expression for combining the local features and global features of wind power generation is as follows:

[0049] F fuslon =F local +F global ;

[0050] Where: F fusion Merge results for local features and global features;

[0051] Inputting the combined features into the feedforward neural network to obtain the wind power prediction result specifically includes the following steps:

[0052] The merged features are input into the feedforward neural network, and the specific expression is as follows:

[0053] F proj =ReLU(FC1(F fusion ))+F fusion ;

[0054] Where: F proj is the output result of the first fully connected layer; ReLU() is the activation function; FC1() is the operation of the first fully connected layer;

[0055] The final prediction result is obtained by dimensional projection through the output layer. The specific expression is as follows:

[0056] Y=FC2(F proj );

[0057] Where: Y is the predicted output result; FC2() is the second fully connected layer operation.

[0058] On the other hand, the present invention also provides a wind power generation prediction system combining local and global features, comprising:

[0059] The data acquisition module is used to collect data on factors affecting wind power generation and external environmental variables. The data on factors affecting wind power generation include historical wind direction, wind speed and wind power generation; the environmental variables include temperature, humidity and air pressure.

[0060] A sequence construction module is used to construct a sequence of factors affecting wind power generation and a sequence of external environmental variables;

[0061] A preliminary mixed feature extraction module is used to perform preliminary mixed feature extraction on a sequence of factors affecting wind power generation and a sequence of external environmental variables;

[0062] A local feature extraction module extracts local features of wind power from preliminary mixed features through a time domain convolution branch and a wavelet transform branch, respectively, and merges the local features extracted by the time domain convolution branch and the wavelet transform branch into a unified local feature through two-dimensional convolution;

[0063] The global feature extraction module extracts long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on local features, and fuses the long-term periodic information and short-term mutation information through the global convolution kernel to obtain the global features of wind power generation;

[0064] The feature merging and prediction module is used to merge the local features and global features of wind power generation, and input the merged features into the feedforward neural network to obtain the prediction result of wind power generation.

[0065] On the other hand, the present invention further provides an electronic device having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for predicting wind power generation combining local and global features as described in any embodiment of the present invention is implemented.

[0066] On the other hand, the present invention also provides a computer-readable medium for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement a wind power prediction method combining local and global characteristics as described in any embodiment of the present invention.

[0067] The present invention has the following beneficial effects:

[0068] 1. A multi-branch mechanism is used to extract fine-grained local features and global long-term trends of wind power data from different time scales and feature spaces.

[0069] 2. An improved multi-scale feature fusion method combined with deep learning is used to comprehensively model the dynamic characteristics of wind power generation sequences from the time domain and frequency domain.

[0070] 3. Add external environmental variables (such as air pressure, temperature, and humidity) as auxiliary inputs, optimize the prediction effect through joint learning, and improve the model's adaptability to wind power changes under different weather conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0072] Figure 1 This is a schematic diagram of the method flow of Example 1; DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe 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 creative work are within the scope of protection of the present invention.

[0074] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0075] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0076] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0077] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0078] Embodiment 1:

[0079] In order to make the purpose, technical solutions and advantages of the present invention more clear, the following will be combined with the specific embodiments of the present application and refer to the attached Figure 1 , clearly and completely describe the technical solution of the present invention.

[0080] To solve the problems in the prior art, the present invention provides a method for predicting wind power generation by combining local and global features, comprising the following steps:

[0081] Step S1, collecting data on factors affecting wind power generation and external environmental variables, the data on factors affecting wind power generation including historical wind direction, wind speed and wind power generation; the external environmental variables including temperature, humidity and air pressure;

[0082] Step S2, constructing a wind power generation power influencing factor sequence and an external environment variable sequence based on the wind power generation power influencing factor data and the external environment variable data;

[0083] The expression of the wind power generation factor sequence X is as follows:

[0084]

[0085] Where: v n is the wind speed at time n; θ n is the wind direction at time n; p n is the wind power generation power at time n;

[0086] The expression of the external environment variable sequence E is as follows:

[0087]

[0088] Where: w n is the temperature at time n; hn is the humidity at time n; q n is the air pressure at time n.

[0089] Step S3, performing preliminary mixed feature extraction on the sequence of factors affecting wind power generation and the sequence of external environmental variables;

[0090] In this embodiment, the attention mechanism is used to dynamically weight the influence weights of different variables to perform preliminary mixed feature extraction. The specific formula is as follows:

[0091] F input =VE(X)+PE(X)+Attn(E);

[0092] Where: F input is the preliminary mixed feature; VE() is the value embedding function; PE() is the position embedding function; Attn() is the weighted attention mechanism.

[0093] Step S4, extracting local features of wind power from the preliminary mixed features through the time domain convolution branch and the wavelet transform branch respectively, and merging the local features extracted by the time domain convolution branch and the wavelet transform branch into a unified local feature through two-dimensional convolution;

[0094] Step S401: extract local time series features

[0095] Each branch of the time domain convolution branch extracts local time series features through one-dimensional convolution. The convolution kernel is convolved along the time step direction, and the kernel size and step length are set to be the same to achieve the function of compressing short sequences. The formula is as follows:

[0096] F conv,l =Conv1D(k l ,s l ),l∈{1,2,…,m};

[0097] Where: F conv,l is the local time series feature extracted by the lth time domain convolution branch; Conv1D() is a one-dimensional convolution operation; k l is the convolution kernel size of the lth time domain convolution branch; s l is the step size of the lth time domain convolution branch; m is the number of time domain convolution branches;

[0098] Step S402: Capturing the instantaneous change characteristics of wind power generation

[0099] The instantaneous variation characteristics of wind power generation are extracted through the wavelet transform branch. The formula is as follows:

[0100] F wave,b = WaveletTransform(F input ,w b),b∈{1,2,…,r};

[0101] Where: F wave,b is the instantaneous change feature extracted by the bth wavelet transform branch; WaveletTransform() is the wavelet transform operation; w b is the wavelet function of the bth wavelet transform branch; r is the number of wavelet decomposition layers.

[0102] Step S403: Merge branches

[0103] Scale alignment is performed through dynamic weighted linear interpolation, the formula is as follows:

[0104] F up,l =LinearInterp(F conv,l ),l∈{1,2,…,m};

[0105] F up,b =LinearInterp(F wave,b ),b∈{1,2,…,r};

[0106] Where: F up,l It is the linear interpolation result of the local time series features extracted by the time domain convolution branch; F up,b It is the linear interpolation result of the instantaneous change characteristics extracted by the wavelet transform branch; LinearInterp() is a linear interpolation operation.

[0107] The equal-length sequences are merged into unified local features through two-dimensional convolution. The formula is as follows:

[0108] F local =Conv2D(Concat(F up,1 ,F up,2 ,…,F up,m+r ));

[0109] F local is the merged local feature; Conv2D() is a two-dimensional convolution operation; Concat() is a concatenation operation; F up,m+r The local features are obtained by linearly interpolating the local features extracted by the time domain convolution branch and the wavelet transform branch.

[0110] Step S5, extracting long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on the local features, and fusing the long-term periodic information and short-term mutation information through a global convolution kernel to obtain the global features of wind power generation;

[0111] Step S501, perform short-time Fourier transform and fast Fourier transform on the local features to extract short-term mutation and long-term periodicity information from the frequency domain. The formula is as follows:

[0112]

[0113] Where: F stft is the frequency domain result after short-time Fourier transform; F ftt is the frequency domain result after fast Fourier transform;

[0114] The convolution theorem captures the long-term dependencies in frequency domain features as follows:

[0115] F stft_freq =F stft ⊙K stft ;

[0116] F ftt_freq =F fft ⊙K fft ;

[0117] Where: F stft_freq is the convolution result of short-time Fourier transform frequency domain features; F ftt_freq is the convolution result of the fast Fourier transform frequency domain feature; ⊙ is the Hadamard product; K stft is the short-time Fourier transform global convolution kernel; K fft is the fast Fourier transform global convolution kernel;

[0118] Step S502, restore the frequency domain features to the time domain through inverse Fourier transform to obtain the global features of wind power generation, the formula is as follows:

[0119] F global =IFFT(F ftt_freq +F stft_freq );

[0120] Where: F global is the global characteristic of wind power generation; IFFT() is the inverse Fourier transform operation.

[0121] Step S6, merging the local features and the global features of the wind power generation, and inputting the merged features into a feedforward neural network to obtain a prediction result of the wind power generation.

[0122] Step S601, merging the local features and global features of wind power generation, the formula is as follows:

[0123] F fusion =F local +F global ;

[0124] Where: Ffusion Merge results for local features and global features;

[0125] Step S602: The merged features are input into the first fully connected layer of the feedforward neural network to obtain the output result of the first fully connected layer. The formula is as follows:

[0126] F proj =ReLU(FC1(F fusion ))+F fusion ;

[0127] Where: F proj is the output result of the first fully connected layer; ReLU() is the activation function; FC1() is the operation of the first fully connected layer;

[0128] Step S603: Input the output result of the first fully connected layer into the output layer for dimensional projection to obtain the final prediction result. The formula is as follows:

[0129] Y=FC2(F proj );

[0130] Where: Y is the prediction output result; FC2() is the second fully connected layer operation, which maps the features to the target dimension of wind power prediction.

[0131] Embodiment 2:

[0132] This embodiment provides a wind power generation prediction system combining local and global features, including:

[0133] The data acquisition module is used to collect data on factors affecting wind power generation and external environmental variables. The data on factors affecting wind power generation include historical wind direction, wind speed and wind power generation; the environmental variables include temperature, humidity and air pressure.

[0134] A sequence construction module is used to construct a sequence of factors affecting wind power generation and a sequence of external environmental variables;

[0135] A preliminary mixed feature extraction module is used to perform preliminary mixed feature extraction on a sequence of factors affecting wind power generation and a sequence of external environmental variables;

[0136] A local feature extraction module extracts local features of wind power from preliminary mixed features through a time domain convolution branch and a wavelet transform branch, respectively, and merges the local features extracted by the time domain convolution branch and the wavelet transform branch into a unified local feature through two-dimensional convolution;

[0137] The global feature extraction module extracts long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on local features, and fuses the long-term periodic information and short-term mutation information through the global convolution kernel to obtain the global features of wind power generation;

[0138] The feature merging and prediction module is used to merge the local features and global features of wind power generation, and input the merged features into the feedforward neural network to obtain the prediction result of wind power generation.

[0139] Embodiment three:

[0140] This embodiment provides an electronic device having a computer program stored thereon. When the computer program is executed by a processor, a method for predicting wind power generation combining local and global features as described in any embodiment of the present invention is implemented.

[0141] Embodiment 4:

[0142] This embodiment provides a computer-readable medium for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement a wind power prediction method combining local and global features as described in any embodiment of the present invention.

[0143] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0144] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0145] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0146] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0147] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A wind power prediction method combining local and global features, characterized in that: The following steps are involved: Step S1, collecting data on factors affecting wind power generation and external environmental variables, the data on factors affecting wind power generation including historical wind direction, wind speed and wind power generation; the external environmental variables including temperature, humidity and air pressure; Step S2, constructing a wind power generation power influencing factor sequence and an external environment variable sequence based on the wind power generation power influencing factor data and the external environment variable data; Step S3, performing preliminary mixed feature extraction on the sequence of factors affecting wind power generation and the sequence of external environmental variables; Step S4, extracting local features of wind power from the preliminary mixed features through the time domain convolution branch and the wavelet transform branch respectively, and merging the local features extracted by the time domain convolution branch and the wavelet transform branch into a unified local feature through two-dimensional convolution; Step S5, extracting long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on the local features, and fusing the long-term periodic information and short-term mutation information through a global convolution kernel to obtain the global features of wind power generation; Step S6, merging the local features and the global features of the wind power generation, and inputting the merged features into a feedforward neural network to obtain a prediction result of the wind power generation.

2. A wind power generation prediction method combining local and global features according to claim 1, characterized in that: In the step of constructing a sequence of factors affecting wind power generation and a sequence of external environmental variables: The expression of the wind power generation factor sequence X is as follows: Where: v n is the wind speed at time n; θ n is the wind direction at time n; p n is the wind power generation power at time n; The expression of the external environment variable sequence E is as follows: Where: w n is the temperature at time n; h n is the humidity at time n; q n is the air pressure at time n.

3. The method for predicting wind power generation combining local and global features according to claim 1, characterized in that: The preliminary mixed feature extraction is performed on the sequence of factors affecting wind power generation and the sequence of external environmental variables. The preliminary mixed feature extraction formula is as follows: F input =VE(X)+PE(X)+Attn(E); Where: F input is the preliminary mixed feature; VE() is the value embedding function; PE() is the position embedding function; Attn() is the weighted attention mechanism.

4. The method for predicting wind power generation combining local and global features according to claim 1, characterized in that: The local features of wind power generation are extracted from the preliminary mixed features through the time domain convolution branch and the wavelet transform branch, wherein: The local time series feature expression of wind power generation extracted by time domain convolution branch is as follows: F conv,l =Conv1D(k l ,s l ),l∈{1,2,…,m}; Where: F conv,l is the local time series feature extracted by the lth time domain convolution branch; Conv1D() is a one-dimensional convolution operation; k l is the convolution kernel size of the lth time domain convolution branch; s l is the step size of the lth time domain convolution branch; m is the number of time domain convolution branches; The instantaneous variation characteristic expression of wind power extracted by wavelet transform branch is as follows: F wave,b =WaveletTransform(F input ,w b ), b ∈{1,2,…,r}; Where: F wave,b is the instantaneous change feature extracted by the bth wavelet transform branch; WaveletTransform() is a wavelet transform operation; b is the wavelet function of the bth wavelet transform branch; r is the number of wavelet decomposition layers.

5. The method for predicting wind power generation combining local and global features according to claim 1, characterized in that: The local features extracted by the time domain convolution branch and the wavelet transform branch are combined into a unified local feature through two-dimensional convolution. The local feature is obtained by the following formula: F local =Conv2D(Concat(F up,1 ,F up,2 ,…,F up,m+r )); F local is the merged local feature; Conv2D() is a two-dimensional convolution operation; Concat() is a concatenation operation; F up,m+r The local features are obtained by linearly interpolating the local features extracted by the time domain convolution branch and the wavelet transform branch.

6. A wind power generation prediction method combining local and global features according to claim 1, characterized in that: The method extracts long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on local features, and fuses the long-term periodic information and short-term mutation information through a global convolution kernel to obtain the global features of wind power generation, which specifically includes the following steps: Perform short-time Fourier transform and fast Fourier transform on local features, as shown in the following formula: F stft =STFT(F local ) F ftt =FFT(F local ); Where: F stft is the frequency domain result after short-time Fourier transform; F ftt is the frequency domain result after fast Fourier transform; The convolution theorem captures the long-term dependencies in frequency domain features as follows: F stft_freq =F stft ⊙K stft ; F ftt_freq =F fft ⊙K fft ; Where: F stft_freq It is the convolution result of short-time Fourier transform frequency domain features; F ftt_freq It is the convolution result of fast Fourier transform frequency domain features; ⊙ is the Hadamard product; K stft is the short-time Fourier transform global convolution kernel; K fft is the fast Fourier transform global convolution kernel; The frequency domain features are restored to the time domain through inverse Fourier transform, as shown in the following formula: F global =IFFT(F ftt_freq +F stft_freq ); Where: F global is the global characteristic of wind power generation; IFFT() is an inverse Fourier transform operation.

7. The method for predicting wind power generation combining local and global features according to claim 1, characterized in that: The local features and global features of wind power generation are combined, and the combined features are input into the feedforward neural network to obtain the prediction result of wind power generation; wherein, the specific expression for combining the local features and global features of wind power generation is as follows: F fusion =F local +F global ; Where: F fusion Merge results for local features and global features; Inputting the combined features into the feedforward neural network to obtain the wind power prediction result specifically includes the following steps: The merged features are input into the feedforward neural network, and the specific expression is as follows: F proj =ReLU(FC1(F fusion ))+F fusion ; Where: F proj is the output result of the first fully connected layer; ReLU() is the activation function; FC1() is the operation of the first fully connected layer; The final prediction result is obtained by dimensional projection through the output layer. The specific expression is as follows: Y=FC2(F proj ); Where: Y is the predicted output result; FC2() is the second fully connected layer operation.

8. A wind power generation prediction system combining local and global features, characterized in that: include: The data acquisition module is used to collect data on factors affecting wind power generation and external environmental variables. The data on factors affecting wind power generation include historical wind direction, wind speed and wind power generation; the environmental variables include temperature, humidity and air pressure. A sequence construction module is used to construct a sequence of factors affecting wind power generation and a sequence of external environmental variables; A preliminary mixed feature extraction module is used to extract preliminary mixed features from a sequence of factors affecting wind power generation and a sequence of external environmental variables; A local feature extraction module extracts local features of wind power from preliminary mixed features through a time domain convolution branch and a wavelet transform branch, respectively, and merges the local features extracted by the time domain convolution branch and the wavelet transform branch into a unified local feature through two-dimensional convolution; The global feature extraction module extracts long-term periodic information and short-term mutation information from the frequency domain by performing short-time Fourier transform and fast Fourier transform on local features, and fuses the long-term periodic information and short-term mutation information through the global convolution kernel to obtain the global features of wind power generation; The feature merging and prediction module is used to merge the local features and global features of wind power generation, and input the merged features into the feedforward neural network to obtain the prediction result of wind power generation.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, a wind power prediction method combining local and global features as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a method for predicting wind power generation combining local and global features as described in any one of claims 1 to 7 is implemented.

Images