Deep learning-based wind power generation power prediction method and electronic equipment

Abstract

The invention provides a wind power generation power prediction method based on deep learning and electronic equipment, and belongs to the technical field of wind power generation. Reconstructing the time sequence meteorological data based on an encoder-decoder module in the trained wind power generation power prediction model to obtain reconstructed data; based on a Legendre polynomial projection module in the trained wind power generation power prediction model, projecting the reconstructed data to a Legendre polynomial space for feature extraction to obtain a feature vector; performing spectrum convolution processing on the feature vector based on a Fourier transform module in the trained wind power generation power prediction model to obtain data after spectrum convolution; and processing the data after frequency spectrum convolution based on a multi-layer perceptron module in the trained wind power generation power prediction model to obtain wind power generation power. According to the scheme, the prediction precision and accuracy of the wind power generation power can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of wind power generation, and specifically to a wind power generation power prediction method and electronic equipment based on deep learning. Background Art

[0002] At present, wind power generation prediction is mainly based on meteorological data and historical power data. Among them, meteorological data includes wind force, wind speed, wind direction, temperature and humidity, and historical power data is used to establish prediction models. Commonly used prediction models include artificial neural networks, support vector machines, and regression analysis models. These prediction models can associate meteorological data with historical power data, and obtain the prediction results of wind power generation through the trained prediction model.

[0003] In the prior art, the wind power generation prediction methods mainly include: wind power generation prediction based on statistical methods and wind power generation prediction based on physical models. However, the prediction accuracy and accuracy of these two prediction methods are low. Therefore, how to improve the prediction accuracy and accuracy of wind power generation has become a problem to be solved. Summary of the invention

[0004] The purpose of this application is to provide a wind power prediction method based on deep learning, which can solve the problems of low wind power prediction accuracy and poor accuracy in the prior art.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting wind power generation based on deep learning, the method comprising:

[0006] Obtain time series meteorological data;

[0007] Reconstructing the time series meteorological data based on the encoder-decoder module in the trained wind power prediction model to obtain reconstructed data;

[0008] Based on the Legendre polynomial projection module in the trained wind power prediction model, the reconstructed data is projected into the Legendre polynomial space for feature extraction to obtain a feature vector;

[0009] Performing spectrum convolution processing on the feature vector based on the Fourier transform module in the trained wind power prediction model to obtain spectrum convolved data;

[0010] The multi-layer perceptron module in the trained wind power prediction model processes the data after spectrum convolution to obtain the wind power.

[0011] In a possible implementation manner of the first aspect, after acquiring the time series meteorological data, the method further includes:

[0012] The time series meteorological data is averaged and downsampled to obtain the downsampled data. The expression of the downsampled data is as follows:

[0013]

[0014] in, represents the downsampled data, t represents time, k represents the pooling window size, i represents the ordinal number, x(t) represents the time series meteorological data, and x(i) represents the time series meteorological data with t in x(t) replaced by i.

[0015] In a possible implementation manner of the first aspect, reconstructing time series meteorological data based on an encoder-decoder module in a trained wind power prediction model to obtain reconstructed data includes:

[0016] Based on the encoder, the downsampled data is encoded into an implicit representation. The specific expression is as follows:

[0017]

[0018] Among them, h(t) represents implicit representation, Encoder represents encoder, Represents the downsampled data;

[0019] Based on the decoder, the implicit representation is decoded into reconstructed data. The specific expression is as follows:

[0020]

[0021] in, Represents reconstructed data, and Decoder represents a decoder.

[0022] In a possible implementation manner of the first aspect, the encoder includes several layers of neural networks, and the decoder includes several layers of neural networks; encoding the downsampled data into an implicit representation based on the encoder includes:

[0023] The downsampled data is encoded into implicit representations based on several layers of neural networks. The output of each layer of the encoder neural network is expressed as follows:

[0024] h l =f(W l h l-1 +b l )

[0025] Among them, h l represents the output of the lth layer of the encoder neural network, l represents the index of each layer of the neural network, f represents the activation function of the encoder, W lrepresents the weight matrix of the encoder's layer l, h l-1 represents the output of the encoder's l-1th layer neural network, b l represents the bias vector of the lth layer of the encoder;

[0026] The expression of the implicit representation of the encoder's final output is as follows:

[0027] h(t)=h L

[0028] Among them, h(t) represents the implicit representation, L represents the total number of layers of the encoder's neural network, and h L Represents the output of the last neural network layer of the encoder.

[0029] In a possible implementation manner of the first aspect, decoding the implicit representation into reconstructed data based on a decoder includes:

[0030] Based on several layers of neural networks, the implicit representation is converted into reconstructed data. The output of each layer of the decoder neural network is expressed as follows:

[0031] d l =g(W l 'd l-1 +b l ′)

[0032] Among them, d l represents the output of the lth layer neural network of the decoder, g represents the activation function of the decoder, and W l ′ represents the weight matrix of the lth layer of the decoder, d l-1 represents the output of the decoder's l-1th layer neural network, b l ′ represents the bias vector of the lth layer of the decoder;

[0033] The expression of the reconstructed data finally output by the encoder is as follows:

[0034]

[0035] in, represents the reconstructed data, L′ represents the total number of layers of the decoder’s neural network, and d L′ Represents the output of the last neural network layer of the decoder.

[0036] In a possible implementation manner of the first aspect, before projecting the reconstructed data into the Legendre polynomial space for feature extraction to obtain a feature vector based on the Legendre polynomial projection module in the trained wind power prediction model, the method further includes:

[0037] The reconstructed data is normalized based on the data normalization module in the trained wind power prediction model to obtain the reconstructed data after data normalization. The expression of the reconstructed data after data normalization is as follows:

[0038]

[0039] in, represents the reconstructed data after data normalization. represents the reconstructed data, μ1 represents the mean of the reconstructed data, and σ1 represents the standard deviation of the reconstructed data.

[0040] In a possible implementation of the first aspect, based on a Legendre polynomial projection module in a trained wind power prediction model, reconstructed data is projected into a Legendre polynomial space to extract features to obtain a feature vector, including:

[0041] For each moment, the Legendre polynomials of each order corresponding to the reconstructed data after data normalization are calculated. The expressions of the Legendre polynomials of each order are as follows:

[0042]

[0043]

[0044] in, represents the reconstructed data after data normalization. is the 0th order Legendre polynomial, is the first-order Legendre polynomial, is the nth order Legendre polynomial, n is a natural number;

[0045] The Legendre polynomials of each order are weighted and summed according to the corresponding coefficients to obtain the eigenvector. The expression of the eigenvector is as follows:

[0046]

[0047] Among them, P(t) represents the feature vector, N represents the length of the feature vector, and a n Represents the coefficients corresponding to Legendre polynomials of various orders.

[0048] In a possible implementation manner of the first aspect, a spectral convolution process is performed on a feature vector based on a Fourier transform module in a trained wind power prediction model to obtain data after spectral convolution, including:

[0049] Perform Fourier transform on the eigenvector, the specific expression is as follows:

[0050] F[P(t)]=Φ[P(t)]

[0051] Among them, F is the intermediate variable, P(t) represents the eigenvector, and Φ represents the Fourier transform;

[0052] Perform linear transformation on the eigenvector after Fourier transformation. The specific expression is as follows:

[0053] S f =W f ·F[P(t)]

[0054] Among them, S f represents linear transformation, W f represents a linear transformation matrix;

[0055] Perform inverse Fourier transform on the eigenvector after linear transformation. The specific expression is as follows:

[0056] S(t)=Φ -1 (S f )

[0057] Among them, S(t) represents the data after spectrum convolution, Φ -1 represents the inverse Fourier transform.

[0058] In a possible implementation manner of the first aspect, performing Fourier transform on the feature vector includes:

[0059]

[0060] Where F[P(t)](k) represents the Fourier transform of P(t), P(t) represents the eigenvector, k represents the frequency index, n represents the time index, N represents the length of the eigenvector, e is a natural constant, and i is an imaginary unit;

[0061] Perform linear transformation on the eigenvector after Fourier transformation, including:

[0062]

[0063] Among them, S f (k) represents linear transformation, W f (k, j) is the element in the kth row and jth column of the linear transformation matrix, F[P(t)](j) represents the Fourier transform of P(t), k represents the row of the linear transformation matrix, and j represents the column of the linear transformation matrix;

[0064] Perform inverse Fourier transform on the linearly transformed eigenvector, including:

[0065]

[0066] Among them, S(t) n Represents the data after spectral convolution.

[0067] In a possible implementation manner of the first aspect, the multilayer perceptron module includes an input layer, a hidden layer, and an output layer; and processing the data after the spectrum convolution based on the trained multilayer perceptron module to obtain the wind power generation power includes:

[0068] The data S(t) after spectrum convolution at the input layer n As input vector x;

[0069] In the hidden layer, the input of the lth layer is x l-1 , the output of the lth layer is x l , the expression of the l-th layer operation is as follows:

[0070] z l =W l x l-1 +b l

[0071] Among them, z l represents the operation of layer l, W l represents the weight matrix of the lth layer, b l represents the bias vector of the lth layer, l represents the hidden layer number;

[0072] Use activation function to transform z l The specific expression is as follows:

[0073] x l =f(z l )

[0074] Among them, x l represents the output of the lth layer, and f represents the activation function;

[0075] In the output layer, a fully connected layer is used. The input of the fully connected layer is the output of the last hidden layer, and the output of the fully connected layer is the prediction result. The specific expression is as follows:

[0076]

[0077] in, represents the prediction result, W l represents the weight matrix of the lth layer, b l Represents the bias vector of the lth layer;

[0078] The prediction results are processed by data denormalization to obtain wind power:

[0079]

[0080] Among them, y(t) represents the wind power generation, σ2 represents the standard deviation of the prediction results, and μ2 represents the mean of the prediction results.

[0081] In a possible implementation manner of the first aspect, the method further includes:

[0082] A wind power prediction model is trained according to historical meteorological data and historical wind power data; wherein the wind power prediction model includes: an encoder-decoder module, a data normalization module, a Legendre polynomial projection module, a Fourier transform module and a multi-layer perceptron module; in the process of predicting wind power based on the trained wind power prediction model, the weight parameters of the trained wind power prediction model are updated according to the latest meteorological data and the latest wind power data.

[0083] In a second aspect, an embodiment of the present application provides a wind power prediction device based on deep learning, the device comprising:

[0084] An acquisition unit, used for acquiring time series meteorological data;

[0085] A reconstruction unit, used for reconstructing the time series meteorological data based on the encoder-decoder module in the trained wind power generation prediction model to obtain reconstructed data;

[0086] A feature extraction unit is used to project the reconstructed data into the Legendre polynomial space for feature extraction based on the Legendre polynomial projection module in the trained wind power prediction model to obtain a feature vector;

[0087] A spectrum convolution unit, used for performing spectrum convolution processing on the feature vector based on the Fourier transform module in the trained wind power prediction model to obtain spectrum convolved data;

[0088] The output unit is used to process the data after the spectrum convolution based on the multi-layer perceptron module in the trained wind power prediction model to obtain the wind power.

[0089] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, any one of the wind power prediction methods based on deep learning in the first aspect is implemented.

[0090] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements any wind power prediction method based on deep learning in the above-mentioned first aspect.

[0091] In a fifth aspect, an embodiment of the present application provides a computer program product. When the computer program product runs on an electronic device, the electronic device executes any one of the wind power prediction methods based on deep learning in the above-mentioned first aspect.

[0092] The present application first reconstructs the time series meteorological data through the encoder-decoder module, then extracts the features of the reconstructed data through the Legendre polynomial projection module to obtain the feature vector, then performs spectrum convolution processing on the feature vector through the Fourier transform module, and finally processes the spectrum convolved data through the multi-layer perceptron module to obtain the wind power generation. The solution of the present application can improve the prediction accuracy and precision of wind power generation, and has strong ease of use and practicality.

[0093] Other features and advantages of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0095] Figure 1 It is a schematic diagram of the steps of a wind power generation prediction method based on deep learning provided in an embodiment of the present application;

[0096] Figure 2 It is a schematic diagram of the overall steps of the wind power generation prediction method based on deep learning provided in an embodiment of the present application;

[0097] Figure 3 It is a structural schematic diagram of a wind power generation prediction device based on deep learning provided in an embodiment of the present application;

[0098] Figure 4 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0099] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0100] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0101] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0102] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0103] As used in this specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0104] In addition, in the description of the present application, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0105] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0106] With the intensification of the global energy crisis and the improvement of environmental awareness, renewable energy (such as wind energy and solar energy) has received widespread attention and utilization. As one of the main forms of wind energy utilization, the prediction of wind power generation is of great significance for the stable operation of the power system, the rational dispatch of energy and the effective transaction of the market.

[0107] The power generation of a wind power station is affected by many factors, such as wind force, wind speed, wind direction, temperature, and humidity. Wind power generation prediction is to input the forecast conclusions of meteorological elements into the power generation model, calculate the power generation capacity of the wind power station, connect to the grid for power generation and transmission, and finally complete the transaction in the power market. For distributed power generation systems, accurate prediction of power generation capacity is very important. How to accurately predict power output, better plan and schedule power generation behavior, and maximize the use of green electricity is one of the key technologies to ensure the safe and stable operation of the power system.

[0108] At present, wind power generation prediction is mainly based on meteorological data and historical power data. Among them, meteorological data includes wind force, wind speed, wind direction, temperature and humidity, and historical power data is used to establish prediction models. Commonly used prediction models include artificial neural networks, support vector machines, and regression analysis models. These prediction models can associate meteorological data with historical power data, and obtain the prediction results of wind power generation through the trained prediction model.

[0109] In the prior art, the wind power generation prediction methods mainly include: wind power generation prediction methods based on statistical methods and wind power generation prediction methods based on physical models. Wind power generation prediction methods based on statistical methods mainly rely on historical data and statistical models to predict future wind farm power generation, and establish prediction models by analyzing patterns, trends and correlations in historical data. Commonly used statistical methods include: regression analysis, time series analysis, and gray prediction methods. These models can capture the impact of factors such as weather, season, and time on wind farm power generation, and make predictions based on this.

[0110] The wind power prediction method based on physical models predicts the power generation of wind farms based on the physical characteristics and operating principles of wind farms, combined with meteorological data and environmental factors. This method usually requires an in-depth understanding of the components, structure and operation mode of wind farms. Commonly used physical models include: wind models, temperature models, efficiency models, etc. These models can simulate the operation of wind farms under different weather and environmental conditions, and predict their power generation accordingly.

[0111] The problems with existing wind power prediction methods mainly stem from: the volatility, intermittency and randomness of wind energy itself, as well as the limitations of prediction technology and the complexity of practical applications. The main problems are as follows: First, the wind speed prediction accuracy is insufficient: Existing prediction technology often predicts based on the average wind speed of the wind farm, without fully considering the complex impact of the topography of the wind farm on the wind speed. This simplified prediction method leads to a deviation between the predicted wind speed and the actual wind speed at the wind turbine, which in turn affects the accuracy of power prediction.

[0112] Second, the power conversion relationship is simple: In the process of converting the predicted wind speed into wind power, a simple conversion relationship is generally used, without fully considering the influence of actual factors such as towers and wind shear differences. This simplified conversion method leads to insufficient prediction accuracy of wind power. Third, insufficient combination algorithm: When using the combination algorithm for wind power generation prediction, it is often only a simple weight combination of the prediction results of each algorithm, and the combination weight is not fully optimized in combination with the shortcomings of the algorithm itself.

[0113] In view of the above defects, the embodiment of the present application provides a method for predicting wind power generation based on deep learning. First, the time series meteorological data is reconstructed through the encoder-decoder module, and then the features of the reconstructed data are extracted through the Legendre polynomial projection module to obtain feature vectors. Then, the feature vectors are subjected to spectrum convolution processing through the Fourier transform module, and finally, the data after spectrum convolution is processed through the multi-layer perceptron module to obtain wind power generation. The solution of the present application can improve the prediction accuracy and precision of wind power generation, and has strong ease of use and practicality.

[0114] The specific process implemented by this application is introduced below through specific embodiments.

[0115] See also Figure 1 , Figure 1 Schematic diagram of the steps of the wind power prediction method based on deep learning provided in the embodiment of the present application. Figure 1 As shown, the method may include the following steps:

[0116] S101, obtaining time series meteorological data.

[0117] In one embodiment, meteorological data such as wind force, wind speed, wind direction, temperature and humidity of the wind power station within a preset time period may be collected over time. The preset time period is a period of time in the future, which may be determined according to specific circumstances in actual application scenarios and is not specifically limited here.

[0118] According to an implementation of the present application, after acquiring the time series meteorological data, the method may further include the following steps:

[0119] The time series meteorological data is averaged and downsampled to obtain the downsampled data. The expression of the downsampled data is as follows:

[0120]

[0121] in, represents the downsampled data, t represents time, k represents the pooling window size, i represents the ordinal number, x(t) represents the time series meteorological data, and x(i) represents the time series meteorological data with t in x(t) replaced by i.

[0122] It should be noted that wind power generation time series meteorological data usually have significant long-term and short-term dependencies, and traditional prediction models are difficult to capture both long-term and short-term dependencies at the same time.

[0123] S102, reconstructing the time series meteorological data based on the encoder-decoder module in the trained wind power prediction model to obtain reconstructed data.

[0124] According to an embodiment of the present application, reconstructing the time series meteorological data based on the encoder-decoder module in the trained wind power prediction model to obtain the reconstructed data may include the following steps:

[0125] Based on the encoder, the downsampled data is encoded into an implicit representation. The specific expression is as follows:

[0126]

[0127] Among them, h(t) represents implicit representation, Encoder represents encoder, Represents the downsampled data.

[0128] Based on the decoder, the implicit representation is decoded into reconstructed data. The specific expression is as follows:

[0129]

[0130] in, Represents reconstructed data, and Decoder represents a decoder.

[0131] According to an embodiment of the present application, the encoder includes several layers of neural networks, and the decoder includes several layers of neural networks; encoding the downsampled data into an implicit representation based on the encoder may include the following steps:

[0132] The downsampled data is encoded into implicit representations based on several layers of neural networks. The output of each layer of the encoder neural network is expressed as follows:

[0133] h l =f(Wl h l-1 +b l )

[0134] Among them, h l represents the output of the lth layer of the encoder neural network, l represents the index of each layer of the neural network, f represents the activation function of the encoder, W l represents the weight matrix of the encoder's layer l, h l-1 represents the output of the encoder's l-1th layer neural network, b l Represents the bias vector of the lth layer of the encoder. The weight matrix and the bias vector can be obtained through learning.

[0135] The expression of the implicit representation of the encoder's final output is as follows:

[0136] h(t)=h L

[0137] Among them, h(t) represents the implicit representation, L represents the total number of layers of the encoder's neural network, and h L Represents the output of the last neural network layer of the encoder.

[0138] According to an embodiment of the present application, decoding the implicit representation into reconstructed data based on a decoder may include the following steps:

[0139] Based on several layers of neural networks, the implicit representation is converted into reconstructed data. The output of each layer of the decoder neural network is expressed as follows:

[0140] d l =g(W l 'd l-1 +b l ′)

[0141] Among them, d l represents the output of the lth layer neural network of the decoder, g represents the activation function of the decoder, and W l ′ represents the weight matrix of the lth layer of the decoder, s l-1 represents the output of the decoder's l-1th layer neural network, b l ′ represents the bias vector of the lth layer of the decoder. The weight matrix and the bias vector can be obtained by learning.

[0142] The expression of the reconstructed data finally output by the encoder is as follows:

[0143]

[0144] in, represents the reconstructed data, L′ represents the total number of layers of the decoder’s neural network, and d L′Represents the output of the last neural network layer of the decoder.

[0145] In this embodiment, the encoder-decoder module can effectively capture the long-term and short-term dependencies of time series meteorological data, thereby improving the accuracy of wind power generation prediction.

[0146] It should be noted that the original time series meteorological data has a high dimension and a lot of redundant information. Direct modeling may lead to high model complexity or overfitting.

[0147] S103, based on the Legendre polynomial projection module in the trained wind power prediction model, project the reconstructed data into the Legendre polynomial space for feature extraction to obtain a feature vector.

[0148] It should be noted that wind power generation data may have different dimensions and ranges, and directly using these data for modeling may result in slow convergence of the model, low prediction accuracy, or even numerical instability. According to one embodiment of the present application, before projecting the reconstructed data into the Legendre polynomial space for feature extraction to obtain a feature vector based on the Legendre polynomial projection module in the trained wind power generation prediction model, the method may also include the following steps:

[0149] The reconstructed data is normalized based on the data normalization module in the trained wind power prediction model to obtain the reconstructed data after data normalization. The expression of the reconstructed data after data normalization is as follows:

[0150]

[0151] in, represents the reconstructed data after data normalization. represents the reconstructed data, μ1 represents the mean of the reconstructed data, and σ1 represents the standard deviation of the reconstructed data.

[0152] In this embodiment, a data normalization module is used to standardize the reconstructed data so that the dimension and range of the data are the same, which can improve the numerical stability and training efficiency of the model.

[0153] According to an implementation of the present application, projecting the reconstructed data into the Legendre polynomial space to extract features to obtain feature vectors based on the Legendre polynomial projection module in the trained wind power prediction model may include the following steps:

[0154] For each moment, the Legendre polynomials of each order corresponding to the reconstructed data after data normalization are calculated. The expressions of the Legendre polynomials of each order are as follows:

[0155]

[0156] in, represents the reconstructed data after data normalization. is the 0th order Legendre polynomial, is the first-order Legendre polynomial, is the nth order Legendre polynomial, where n is a natural number.

[0157] The Legendre polynomials of each order are weighted and summed according to the corresponding coefficients to obtain the eigenvector. The expression of the eigenvector is as follows:

[0158]

[0159] Among them, P(t) represents the feature vector, N represents the length of the feature vector, and a n Represents the coefficients corresponding to Legendre polynomials of various orders, which can be obtained through learning.

[0160] In this embodiment, a Legendre polynomial projection module is used to project the reconstructed data after data normalization into the Legendre polynomial space to extract key features, thereby reducing data redundancy and improving the generalization ability and prediction performance of the model.

[0161] It should be noted that wind power generation data usually has significant periodicity and frequency characteristics, and traditional time domain models are difficult to effectively capture these characteristics.

[0162] S104, performing spectrum convolution processing on the feature vector based on the Fourier transform module in the trained wind power prediction model to obtain spectrum convolved data.

[0163] According to an embodiment of the present application, performing spectrum convolution processing on the feature vector based on the Fourier transform module in the trained wind power prediction model to obtain spectrum convolved data may include the following steps:

[0164] Perform Fourier transform on the eigenvector, the specific expression is as follows:

[0165] F[P(t)]=Φ[P(t)]

[0166] Among them, F is an intermediate variable, P(t) represents the eigenvector, Φ represents Fourier transform, and the Fourier transform can be FFT (fast Fourier transform), which has a faster calculation speed; Fourier transform is used to transform a time domain signal into a frequency domain signal, P(t) is a time domain signal, and F[P(t)] is a frequency domain signal.

[0167] Perform linear transformation on the eigenvector after Fourier transformation. The specific expression is as follows:

[0168] S f =W f ·F[P(t)]

[0169] Among them, S f represents linear transformation, W f It represents a linear transformation matrix, which can be obtained through learning, and the parameters of the linear transformation matrix can be optimized by back propagation according to historical meteorological data and historical wind power generation data.

[0170] Perform inverse Fourier transform on the eigenvector after linear transformation. The specific expression is as follows:

[0171] S(t)=φ -1 (S f )

[0172] Among them, S(t) represents the data after spectrum convolution, φ -1 represents an inverse Fourier transform, which may be an IFFT (Inverse Fast Fourier Transform) used to convert a frequency domain signal back to a time domain signal. f is the frequency domain signal, and S(t) is the time domain signal.

[0173] According to an implementation of the present application, the specific formula for performing Fourier transform on the feature vector is as follows:

[0174]

[0175] Where F[P(t)](k) represents the Fourier transform of P(t), P(t) represents the eigenvector, k represents the frequency index, n represents the time index, N represents the length of the eigenvector, e is a natural constant, and i is an imaginary unit.

[0176] The specific formula for linear transformation of the eigenvector after Fourier transformation is as follows:

[0177]

[0178] Among them, S f(k) represents linear transformation, W f (k,j) is the element in the kth row and jth column of the linear transformation matrix, F[P(t)](j) represents the Fourier transform of P(t), k represents the row of the linear transformation matrix, and j represents the column of the linear transformation matrix.

[0179] The specific formula for inverse Fourier transform of the eigenvector after linear transformation is as follows:

[0180]

[0181] Among them, S(t) n Represents the data after spectral convolution.

[0182] In this embodiment, the Fourier transform module is used to capture the periodicity and frequency characteristics in the data through frequency domain feature extraction, thereby enhancing the model's ability to handle periodic changes and improving the accuracy of prediction.

[0183] It should be noted that there are complex nonlinear relationships in wind power generation data, and simple linear models are difficult to effectively model these relationships.

[0184] S105, processing the data after the spectrum convolution based on the multi-layer perceptron module in the trained wind power prediction model to obtain the wind power.

[0185] In one embodiment, the data after spectrum convolution can be input into the MLP (Multilayer Perceptron) module for final prediction to obtain the final prediction result, that is, the wind power generation power. The specific expression is as follows:

[0186]

[0187] in, represents wind power, MLP represents data processing by the multi-layer perceptron module, and S(t) represents data after spectrum convolution.

[0188] According to an embodiment of the present application, the multi-layer perceptron module includes an input layer, a hidden layer, and an output layer. Processing the data after spectrum convolution based on the trained multi-layer perceptron module to obtain wind power generation may include the following steps:

[0189] At the input layer, the data S(t) after spectrum convolution n As input vector x. In the hidden layer, the input of the lth layer is x l-1 , the output of the lth layer is x l , the expression of the l-th layer operation is as follows:

[0190] z l =Wl x l-1 +b l

[0191] Among them, z l represents the operation of layer l, W l represents the weight matrix of the lth layer, b l Represents the bias vector of the lth layer, l represents the order of the hidden layer; the weight matrix and bias vector can be obtained through learning.

[0192] Use activation function to transform z l The specific expression is as follows:

[0193] x l =f(z l )

[0194] Among them, x l represents the output of the lth layer, and f represents the activation function.

[0195] In the output layer, a fully connected layer is used. The input of the fully connected layer is the output of the last hidden layer, and the output of the fully connected layer is the prediction result. The specific expression is as follows:

[0196]

[0197] in, represents the prediction result, W l represents the weight matrix of the lth layer, b l Represents the bias vector of the lth layer; the weight matrix and bias vector can be obtained through learning.

[0198] It should be noted that the prediction model needs to maintain the consistency of the preprocessed data and the output results to ensure that there is no deviation in the data transmission and processing in the model. In one embodiment, the prediction results are denormalized to obtain the wind power:

[0199]

[0200] Among them, y(t) represents wind power, σ2 represents the standard deviation of the prediction results, and μ2 represents the mean of the prediction results. The data normalization module and the data denormalization module ensure the consistency of data in the input and output stages, so that the data processed by the model is consistent in dimension and range, ensuring the accuracy of the prediction results.

[0201] In this embodiment, the multi-layer perceptron module performs nonlinear modeling on the extracted features through multi-layer nonlinear mapping, captures complex nonlinear relationships, and improves the prediction performance of the model.

[0202] According to one embodiment of the present application, the method may further include the following steps:

[0203] The wind power prediction model is trained according to the historical meteorological data and the historical wind power data; wherein the wind power prediction model includes: an encoder-decoder module, a data normalization module, a Legendre polynomial projection module, a Fourier transform module and a multi-layer perceptron module. In the process of predicting the wind power based on the trained wind power prediction model, the weight parameters of the trained wind power prediction model are updated according to the latest meteorological data and the latest wind power data.

[0204] Among them, the weight parameters of the wind power prediction model include: the weight matrix and bias vector in the encoder-decoder module, the corresponding coefficients of each order Legendre polynomial in the Legendre polynomial projection module, the parameters of the linear transformation matrix in the Fourier transform module, and the weight matrix and bias vector in the multilayer perceptron module.

[0205] It should be noted that since the data normalization module does not contain weight parameters that need to be obtained through training, the data normalization module does not need to be trained, and other modules can be trained. The training goal is to minimize the mean square error of wind power generation, minimize the training loss, and minimize the mean square error. The training loss is calculated according to the loss function, and the expression of the loss function is as follows:

[0206]

[0207] Among them, L represents the loss function, N represents the length of the feature vector, i represents the ordinal number, and y i Indicates the actual value of wind power generation. Indicates the predicted value of wind power generation.

[0208] It should be noted that wind power generation forecasting may require the combination of multi-source data (such as meteorological data, historical wind power generation data, etc.). How to effectively fuse these data and enhance feature representation is a challenge. Through modular design, each module (such as encoder-decoder module, Legendre polynomial projection module, Fourier transform module) processes different types of feature information respectively, and finally performs comprehensive modeling in the multi-layer perceptron module to achieve effective fusion and feature enhancement of multi-source data.

[0209] It should be noted that the training efficiency and convergence speed of complex models are important considerations in practical applications. Traditional models may not be able to achieve good results within a reasonable time. Improve training stability through data normalization, reduce model complexity through modular design and feature extraction (such as Legendre polynomial projection module, Fourier transform module), and improve model training efficiency and convergence speed.

[0210] The embodiment of the present application trains a deep learning model through historical data, uses the model to predict the wind power generation in the future, and uses the latest data to update the model parameters to improve the accuracy of the prediction. First, the encoder-decoder module based on the Transformer (converter) focuses on the important parts of the input time series meteorological data; secondly, the encoder output is standardized based on the data normalization module to avoid problems such as different dimensions of wind power data and large differences in numerical ranges; thirdly, feature extraction is performed based on the Legendre polynomial projection module to solve the problem of retaining and utilizing historical information; then, the dimension is reduced by combining Fourier transform and linear transformation matrix approximation of the objective function to minimize the impact of noise signals in time series meteorological data and alleviate the problem of over-extraction; finally, the prediction is performed through the multi-layer perceptron module, and the prediction result is denormalized and output, and the wind power generation is finally predicted.

[0211] It should be noted that the key to the encoder-decoder module is that it can effectively capture the long-term and short-term dependencies of the input time series and enhance the modeling ability of complex time series data. The key to the data normalization module and the data denormalization module is to standardize data processing, ensure the numerical stability of the input data, and improve the efficiency and accuracy of model training and prediction. The key to the Legendre polynomial projection module is to project the time series data into the polynomial space for feature extraction, enhance the expression ability of the features, and improve the prediction performance of the model. The key to the Fourier transform module is to capture the periodicity and frequency characteristics in the data through frequency domain feature extraction and processing, and improve the prediction accuracy of the model. The key to the multi-layer perceptron module is to effectively model the extracted features through multi-layer nonlinear mapping, and finally output a high-precision prediction result, that is, wind power generation.

[0212] See also Figure 2 , Figure 2 Schematic diagram of the overall steps of the wind power generation prediction method based on deep learning provided in the embodiment of the present application. Figure 2 As shown, the method may include the following steps:

[0213] S201, pre-training a wind power prediction model according to historical data, and updating model parameters according to the latest data in the process of predicting wind power based on the trained wind power prediction model.

[0214] Among them, the historical data are historical meteorological data and historical wind power data, the latest data are the latest meteorological data and the latest wind power data, the updated model parameters are the weight parameters of the updated wind power prediction model, and the wind power prediction model includes: encoder-decoder module, data normalization module, Legendre polynomial projection module, Fourier transform module and multi-layer perceptron module.

[0215] S202, obtaining time series meteorological data x(t);

[0216] S203, average pooling downsamples the time series meteorological data x(t) to obtain downsampled data

[0217] S204, reconstructing the time series meteorological data x(t) based on the encoder-decoder module to obtain reconstructed data

[0218] S205, reconstructing the data based on the data normalization module Perform data normalization to obtain reconstructed data after data normalization

[0219] S206, projecting the reconstructed data into the Legendre polynomial space based on the Legendre polynomial projection module to extract features, and obtaining a feature vector P(t);

[0220] S207, performing spectrum convolution processing on the feature vector based on the Fourier transform module to obtain spectrum convolved data S(t);

[0221] S208, processing the data S(t) after the spectrum convolution based on the multi-layer perceptron (MLP) module, and outputting the wind power generation power Finally, the prediction result of wind power generation is obtained.

[0222] It should be noted that the specific implementation principles of steps S201 to S208 have been described in the above embodiments and will not be repeated here.

[0223] The wind power generation prediction method based on deep learning provided in the embodiment of the present application first reconstructs the time series meteorological data through the encoder-decoder module, then extracts the features of the reconstructed data through the Legendre polynomial projection module to obtain the feature vector, then performs spectrum convolution processing on the feature vector through the Fourier transform module, and finally processes the data after spectrum convolution through the multi-layer perceptron module to obtain the wind power generation. The solution of the embodiment of the present application can improve the prediction accuracy and precision of wind power generation, and has strong ease of use and practicality.

[0224] The deep learning model of the embodiment of the present application can comprehensively consider the influence of various factors such as wind force, wind speed, wind direction, temperature and humidity of the wind power station to predict the wind power generation in the future. The deep learning model has the following technical performances: the prediction accuracy exceeds the level of similar models and can provide accurate wind power generation prediction data; it has the characteristics of stable performance and high detection accuracy, and can meet the needs of long-term and continuous operation; it is suitable for wind power generation prediction of various wind farms and has wide applicability.

[0225] In addition, the embodiment scheme of the present application has the following advantages: First, high-precision prediction: The encoder-decoder module is used to capture the long-term and short-term dependencies of the input time series, and can effectively process the timing characteristics of wind power generation data. The data normalization module is used to standardize the input data, eliminate dimensional differences, and improve the convergence speed and prediction accuracy of the model. The Legendre polynomial projection module is used to project the data into the polynomial space for feature extraction, improve the feature representation capability, and thus enhance the prediction performance of the model. The Fourier transform module captures the periodicity and frequency characteristics in the input data through frequency domain feature extraction and processing, thereby improving the prediction accuracy. The multi-layer perceptron module is used for nonlinear mapping and final wind power generation power prediction, and can effectively model the extracted features.

[0226] Second, good generalization ability: Through the encoder-decoder module and the Legendre polynomial projection module, effective information can be extracted from different time scales and frequency domain features. It has good generalization ability and is suitable for different types of wind power generation data. Third, strong interpretability and feature extraction capabilities: The Legendre polynomial projection module provides good feature extraction and interpretation capabilities through polynomial expansion, which helps to understand the key features in the input data. The Fourier transform module introduces frequency domain features, which provides the model with the ability to interpret the periodicity and frequency characteristics of the data.

[0227] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0228] Corresponding to the method of the above embodiment, Figure 3 1 is a schematic diagram of the structure of a wind power prediction device based on deep learning provided in an embodiment of the present application. For ease of explanation, only the parts related to the embodiment of the present application are shown.

[0229] Reference Figure 3 , the device comprises:

[0230] An acquisition unit 301 is used to acquire time series meteorological data;

[0231] A reconstruction unit 302 is used to reconstruct the time series meteorological data based on the encoder-decoder module in the trained wind power prediction model to obtain reconstructed data;

[0232] A feature extraction unit 303 is used to project the reconstructed data into the Legendre polynomial space for feature extraction based on the Legendre polynomial projection module in the trained wind power prediction model to obtain a feature vector;

[0233] The spectrum convolution unit 304 is used to perform spectrum convolution processing on the feature vector based on the Fourier transform module in the trained wind power generation prediction model to obtain spectrum convolved data;

[0234] The output unit 305 is used to process the data after the spectrum convolution based on the multi-layer perceptron module in the trained wind power prediction model to obtain the wind power.

[0235] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0236] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0237] Figure 4 This is a schematic diagram of the structure of the electronic device 4 provided in the embodiment of the present application. Figure 4 As shown, the electronic device 4 of this embodiment may include the following steps: at least one processor 401 ( Figure 4 Only one is shown in the figure), a memory 403, and a computer program 402 stored in the memory 403 and executable on at least one processor 401, and when the processor 401 executes the computer program 402, the steps in the above method embodiment are implemented.

[0238] The electronic device 4 may be a computing device such as a desktop computer, a notebook, a PDA, or a mobile phone. The electronic device 4 may include, but is not limited to, a processor 401 and a memory 403. Those skilled in the art will appreciate that Figure 4 It is only an example of the electronic device 4 and does not constitute a limitation on the electronic device 4. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, it may also include input and output devices, network access devices, etc.

[0239] The processor 401 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0240] In some embodiments, the memory 403 may be an internal storage unit of the electronic device 4, such as a hard disk or memory of the electronic device 4. In other embodiments, the memory 403 may also be an external storage device of the electronic device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital card (SecureDigital, SD), a flash card (Flash Card), etc. equipped on the electronic device 4. Further, the memory 403 may also include both an internal storage unit of the electronic device 4 and an external storage device. The memory 403 is used to store an operating system, an application program, a boot loader (Boot Loader), data, and other programs, such as program codes of a computer program, etc. The memory 403 may also be used to temporarily store data that has been output or is to be output.

[0241] If the above-mentioned integrated unit 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 such an understanding, when the present application implements all or part of the processes in the above-mentioned embodiment method, it can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps applied to the method embodiment can be implemented. Among them, the computer program includes a computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable storage medium may at least include: any entity or device that can carry the computer program code to the computing device / electronic device, a recording medium, a computer memory, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), an electric carrier signal, a telecommunication signal, and a software distribution medium, such as a USB flash drive, a mobile hard disk, a magnetic disk or an optical disk. In some jurisdictions, according to legislation and patent practice, a computer-readable storage medium cannot be an electric carrier signal and a telecommunication signal.

[0242] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0243] An embodiment of the present application provides a computer program product. When the computer program product is run on an electronic device, the electronic device executes the steps in the above-mentioned method embodiments.

[0244] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0245] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of 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.

[0246] In the embodiments provided in the present application, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. The device / electronic device embodiments described above are merely schematic, and the division of the above modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, and some features can be ignored and not executed. Another point is that the indirect coupling, direct coupling or communication connection between each other shown or discussed can be an indirect coupling, direct coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0247] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0248] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A wind power prediction method based on deep learning, characterized in that: The method comprises: Obtain time series meteorological data; Reconstructing the time series meteorological data based on the encoder-decoder module in the trained wind power prediction model to obtain reconstructed data; Based on the Legendre polynomial projection module in the trained wind power prediction model, the reconstructed data is projected into the Legendre polynomial space for feature extraction to obtain a feature vector; Performing spectrum convolution processing on the feature vector based on the Fourier transform module in the trained wind power prediction model to obtain spectrum convolved data; The multi-layer perceptron module in the trained wind power prediction model processes the data after spectrum convolution to obtain the wind power.

2. The wind power generation prediction method based on deep learning according to claim 1 is characterized in that: After acquiring the time series meteorological data, the method further includes: The time series meteorological data is average-pooled downsampled to obtain downsampled data. The expression of the downsampled data is as follows: in, represents the downsampled data, t represents time, k represents the pooling window size, i represents the ordinal number, x(t) represents the time series meteorological data, and x(i) represents the time series meteorological data with t in x(t) replaced by i.

3. The wind power prediction method based on deep learning according to claim 2 is characterized in that: The time series meteorological data is reconstructed based on the encoder-decoder module in the trained wind power generation prediction model to obtain reconstructed data, including: Based on the encoder, the downsampled data is encoded into an implicit representation. The specific expression is as follows: Among them, h(t) represents implicit representation, Encoder represents encoder, Represents the downsampled data; Based on the decoder, the implicit representation is decoded into reconstructed data. The specific expression is as follows: in, Represents reconstructed data, and Decoder represents a decoder.

4. The method for predicting wind power generation based on deep learning according to claim 3 is characterized in that: The encoder includes several layers of neural networks, and the decoder includes several layers of neural networks; encoding the downsampled data into implicit representations based on the encoder includes: The downsampled data is encoded into implicit representations based on several layers of neural networks. The output of each layer of the encoder neural network is expressed as follows: h l =f(W l h l-1 +b l ) Among them, h represents the output of the lth layer of the encoder neural network, l represents the index of each layer of the neural network, f represents the activation function of the encoder, and W l represents the weight matrix of the encoder's layer l, h l-1 represents the output of the encoder's l-1th layer neural network, b l represents the bias vector of the lth layer of the encoder; The expression of the implicit representation of the encoder's final output is as follows: h(t)=h L Among them, h(t) represents the implicit representation, L represents the total number of layers of the encoder's neural network, and h L Represents the output of the last neural network layer of the encoder.

5. The method for predicting wind power generation based on deep learning according to claim 3, characterized in that: Decoding the implicit representation into reconstructed data based on a decoder, comprising: The implicit representation is converted into the reconstructed data based on several layers of neural networks, and the output of each layer of the neural network of the decoder is expressed as follows: d l =g(W l ′d l-1 +b′ l ) Among them, d l represents the output of the lth layer neural network of the decoder, g represents the activation function of the decoder, and W l ′ represents the weight matrix of the lth layer of the decoder, d l-1 represents the output of the decoder's l-1th layer neural network, b′ l represents the bias vector of the lth layer of the decoder; The expression of the reconstructed data finally output by the encoder is as follows: in, represents the reconstructed data, L′ represents the total number of layers of the decoder’s neural network, and d L′ Represents the output of the last neural network layer of the decoder.

6. The method for predicting wind power generation based on deep learning according to claim 3, characterized in that: Before projecting the reconstructed data into Legendre polynomial space for feature extraction to obtain feature vectors based on the Legendre polynomial projection module in the trained wind power prediction model, the method further includes: The reconstructed data is subjected to data normalization processing based on the data normalization module in the trained wind power generation prediction model to obtain the reconstructed data after data normalization processing. The expression of the reconstructed data after data normalization processing is as follows: in, represents the reconstructed data after data normalization. represents the reconstructed data, μ1 represents the mean of the reconstructed data, and σ1 represents the standard deviation of the reconstructed data.

7. The method for predicting wind power generation based on deep learning according to claim 6, characterized in that: Based on the Legendre polynomial projection module in the trained wind power prediction model, the reconstructed data is projected into the Legendre polynomial space to extract features to obtain feature vectors, including: For each moment, the Legendre polynomials of each order corresponding to the reconstructed data after data normalization are calculated. The expressions of the Legendre polynomials of each order are as follows: in, represents the reconstructed data after data normalization. is the 0th order Legendre polynomial, is the first-order Legendre polynomial, is the nth order Legendre polynomial, n is a natural number; The Legendre polynomials of each order are weighted and summed according to the corresponding coefficients to obtain the eigenvector, and the expression of the eigenvector is as follows: Among them, P(t) represents the feature vector, N represents the length of the feature vector, and a n Represents the coefficients corresponding to Legendre polynomials of various orders.

8. The method for predicting wind power generation based on deep learning according to claim 7, characterized in that: Based on the Fourier transform module in the trained wind power prediction model, the feature vector is subjected to spectrum convolution processing to obtain data after spectrum convolution, including: The eigenvector is subjected to Fourier transform, and the specific expression is as follows: F[P(t)]=Φ[P(t)] Among them, F is the intermediate variable, P(t) represents the eigenvector, and Φ represents the Fourier transform; Perform linear transformation on the eigenvector after Fourier transformation. The specific expression is as follows: S f =W f ·F[P(t)] Among them, S f represents linear transformation, W f represents a linear transformation matrix; Perform inverse Fourier transform on the eigenvector after linear transformation. The specific expression is as follows: S(t)=Φ -1 (S f ) Among them, S(t) represents the data after spectrum convolution, Φ -1 represents the inverse Fourier transform.

9. The method for predicting wind power generation based on deep learning according to claim 8, characterized in that: Performing Fourier transform on the feature vector includes: Where F[P(t)](k) represents the Fourier transform of P(t), P(t) represents the eigenvector, k represents the frequency index, n represents the time index, N represents the length of the eigenvector, e is a natural constant, and i is an imaginary unit; Perform linear transformation on the eigenvector after Fourier transformation, including: Among them, S f (k) represents linear transformation, W f (k, j) is the element in the kth row and jth column of the linear transformation matrix, F[P(t)](j) represents the Fourier transform of P(t), k represents the row of the linear transformation matrix, and j represents the column of the linear transformation matrix; Perform inverse Fourier transform on the linearly transformed eigenvector, including: Among them, S(t) n Represents the data after spectral convolution.

10. The wind power generation prediction method based on deep learning according to claim 8, characterized in that: The multi-layer perceptron module includes an input layer, a hidden layer, and an output layer. The data after spectrum convolution is processed based on the trained multi-layer perceptron module to obtain wind power generation, including: The data S(t) after spectrum convolution at the input layer n As input vector x; In the hidden layer, the input of the lth layer is x l-1 , the output of the lth layer is x l , the expression of the l-th layer operation is as follows: z l =W l x l-1 +b l Among them, z l represents the operation of layer l, W l represents the weight matrix of the lth layer, b l represents the bias vector of the lth layer, l represents the hidden layer number; Use activation function to transform z l The specific expression is as follows: x l =f(z l ) Among them, x l represents the output of the lth layer, and f represents the activation function; In the output layer, a fully connected layer is used. The input of the fully connected layer is the output of the last hidden layer, and the output of the fully connected layer is the prediction result. The specific expression is as follows: in, represents the prediction result, W l represents the weight matrix of the lth layer, b l Represents the bias vector of the lth layer; The prediction results are processed by data denormalization to obtain wind power: Among them, y(t) represents the wind power generation, σ2 represents the standard deviation of the prediction results, and μ2 represents the mean of the prediction results.

11. The method for predicting wind power generation based on deep learning according to any one of claims 1 to 10, characterized in that: The method further comprises: The wind power generation prediction model is trained according to historical meteorological data and historical wind power generation data; wherein the wind power generation prediction model comprises: an encoder-decoder module, a data normalization module, a Legendre polynomial projection module, a Fourier transform module and a multi-layer perceptron module; In the process of predicting wind power generation based on the trained wind power generation prediction model, the weight parameters of the trained wind power generation prediction model are updated according to the latest meteorological data and the latest wind power generation data.

12. 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 computer program, the wind power prediction method based on deep learning described in any one of claims 1-11 is implemented.

13. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the deep learning-based wind power prediction method according to any one of claims 1 to 11 is implemented.

Images