A short-term wind power prediction method, device, equipment and storage medium

Abstract

The present application discloses a short-term wind power prediction method, device, equipment and storage medium, including: collecting real-time wind power data corresponding to the current wind farm, and preprocessing the real-time wind power data to obtain first preprocessed data; inputting the first preprocessed data into a pre-trained target wind power prediction model to obtain a wind power prediction result corresponding to the first historical wind power data output by the target wind power prediction model; wherein the target wind power prediction model is a pre-trained multi-branch neural network model, and in the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized. The present application optimizes the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component, thereby realizing the prediction of wind power by using the trained target wind power prediction model, thereby improving the prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power prediction, and in particular to a short-term wind power prediction method, device, equipment and storage medium. Background Art

[0002] At present, the prediction methods of wind power can be divided into indirect prediction methods and direct prediction methods. Among them, the indirect prediction method mainly refers to predicting wind power with other features, such as wind speed, temperature, air density, etc., but due to the superposition of errors, this method is usually inaccurate. However, the direct prediction method mainly uses the historical data of wind power itself to predict future wind power, which can continue the time series, fit the nonlinear relationship more accurately, and improve the prediction accuracy. However, due to the volatility and non-periodicity of wind power itself, it is difficult to directly extract the trend of the time series. Therefore, the prediction model combining signal decomposition technology with deep learning methods has begun to appear. Among them, the VMD (Variational mode decomposition) algorithm can well solve the problem of mode aliasing in the time series and will not add additional noise, but the number of decompositions of VMD and the center frequency of each component are not optimized. At the same time, most deep learning prediction methods only consider the global trend of long-term trend characteristics, and do not consider the local trend of the recent time period. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a short-term wind power prediction method, device, equipment and storage medium, which can enhance the autocorrelation of the wind power prediction model and improve the prediction accuracy of short-term wind power. The specific scheme is as follows:

[0004] In a first aspect, the present application discloses a short-term wind power prediction method, comprising:

[0005] Collecting real-time wind power data corresponding to the current wind farm, and preprocessing the real-time wind power data to obtain first preprocessed data;

[0006] Inputting the first preprocessed data into a pre-trained target wind power prediction model to obtain a wind power prediction result output by the target wind power prediction model and corresponding to the first historical wind power data;

[0007] The target wind power prediction model is a pre-trained multi-branch neural network model, and during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized.

[0008] Optionally, before inputting the first preprocessed data into a pre-trained target wind power prediction model, the method further includes:

[0009] Collecting historical wind power data of the current wind farm, and preprocessing the historical wind power data to obtain second preprocessed data;

[0010] The preset optimization algorithm is used to optimize the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm for multiple times, and the corresponding target number of decompositions and target center frequency after each optimization are obtained;

[0011] Decomposing the second preprocessed data according to the target decomposition number and the target center frequency corresponding to each optimization to obtain multiple groups of decomposition components corresponding to the target decomposition number and the center frequency;

[0012] Reconstructing the second preprocessed data with each group of the decomposed components respectively to obtain multiple groups of reconstructed data;

[0013] Using the reconstructed data, an initial wind power prediction model constructed based on a convolutional neural network model and a gated recurrent unit is trained to obtain a plurality of trained wind power prediction models corresponding to the target decomposition number and the center frequency;

[0014] A target wind power prediction model is determined from the plurality of trained wind power prediction models.

[0015] Optionally, preprocessing the historical wind power data to obtain second preprocessed data includes:

[0016] The historical wind power data is subjected to denoising processing to obtain denoised data, and the denoised data is subjected to normalization processing to obtain second pre-processed data.

[0017] Optionally, determining a target wind power prediction model from the plurality of trained wind power prediction models includes:

[0018] Determine a model evaluation index corresponding to each of the trained wind power prediction models in the plurality of trained wind power prediction models to obtain a corresponding evaluation index set;

[0019] The trained wind power prediction model corresponding to the model evaluation index with the smallest value in the evaluation index set is determined as the target wind power prediction model.

[0020] Optionally, determining the model evaluation index corresponding to each of the trained wind power prediction models in the plurality of trained wind power prediction models to obtain a corresponding evaluation index set includes:

[0021] Performing feature extraction on the reconstructed data corresponding to each of the trained wind power prediction models among the plurality of trained wind power prediction models to obtain local trend features and global trend features corresponding to the reconstructed data;

[0022] Fusing the local trend feature and the global trend feature to obtain a corresponding fused feature;

[0023] Inputting the fused features into the corresponding trained wind power prediction models respectively to obtain the wind power prediction value output by each trained wind power prediction model;

[0024] Based on the wind power prediction value and the wind power actual value corresponding to the wind power prediction value, a model evaluation index corresponding to each of the trained wind power prediction models is determined to obtain a corresponding evaluation index set.

[0025] Optionally, the fusing the local trend feature and the global trend feature to obtain a corresponding fused feature includes:

[0026] The local trend feature and the global trend feature are fused using a preset feature fusion algorithm to obtain corresponding fused features.

[0027] Optionally, the method of optimizing the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm multiple times by using a preset optimization algorithm includes:

[0028] The whale algorithm is used to optimize the number of decompositions and the center frequency of each decomposition component in the VMD algorithm multiple times.

[0029] In a second aspect, the present application discloses a wind power prediction device, comprising:

[0030] Data acquisition module, used to collect real-time wind power data corresponding to the current wind farm;

[0031] A first data preprocessing module, used for preprocessing the real-time wind power data to obtain first preprocessed data;

[0032] A wind power prediction module, configured to input the first preprocessed data into a pre-trained target wind power prediction model, and obtain a wind power prediction result output by the target wind power prediction model corresponding to the first historical wind power data;

[0033] The target wind power prediction model is a pre-trained multi-branch neural network model, and during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized.

[0034] In a third aspect, the present application discloses an electronic device, including:

[0035] Memory, used to store computer programs;

[0036] The processor is used to execute the computer program to implement the steps of the short-term wind power prediction method disclosed above.

[0037] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the short-term wind power prediction method disclosed above are implemented.

[0038] It can be seen that the present application provides a short-term wind power prediction method, including: collecting real-time wind power data corresponding to the current wind farm, and preprocessing the real-time wind power data to obtain first preprocessed data; inputting the first preprocessed data into a pre-trained target wind power prediction model to obtain a wind power prediction result corresponding to the first historical wind power data output by the target wind power prediction model; wherein the target wind power prediction model is a pre-trained multi-branch neural network model, and during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized. It can be seen that during the training process of the multi-branch neural network model, the present application optimizes the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component, thereby determining a trained target wind power prediction model, and then preprocessing the collected real-time wind power data, inputting the first preprocessed data into the target wind power prediction model, and obtaining the corresponding wind power prediction result output by the target wind power prediction model. It can be seen that the technical solution of the present application can enhance the autocorrelation of the wind power prediction model and improve the prediction accuracy of short-term wind power. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, 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 embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0040] Figure 1A flow chart of a short-term wind power prediction method disclosed in this application;

[0041] Figure 2 A flow chart of a specific short-term wind power prediction method disclosed in this application;

[0042] Figure 3 A schematic diagram of a wind power data time series corresponding to a wind farm disclosed in this application;

[0043] Figure 4 A schematic diagram of VMD decomposition of wind power data disclosed in this application;

[0044] Figure 5 A flow chart of a method for determining a model evaluation index set disclosed in this application;

[0045] Figure 6 A schematic diagram of comparing a wind power prediction value and an actual wind power value disclosed in the present application;

[0046] Figure 7 A specific wind power prediction flow chart disclosed in this application;

[0047] Figure 8 This is a schematic diagram of the structure of a wind power prediction device disclosed in this application;

[0048] Fig. 9 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0049] 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.

[0050] Currently, in the prior art, the VMD algorithm can solve the problem of mode aliasing in time series well without adding additional noise, but the number of VMD decompositions and the center frequency of each decomposition component are not optimized. At the same time, most deep learning prediction methods only consider the global trend of long-term trend characteristics, without considering the local trend of the recent time period. To this end, the present application provides a short-term wind power prediction solution that can enhance the autocorrelation of the prediction model and greatly improve the prediction accuracy.

[0051] The embodiment of the present invention discloses a short-term wind power prediction method, see Figure 1 As shown, the method includes:

[0052] Step S11: collecting real-time wind power data corresponding to the current wind farm, and preprocessing the real-time wind power data to obtain first preprocessed data.

[0053] In this embodiment, if the wind power data of the current wind farm in the future is to be predicted, the real-time wind power data corresponding to the current wind farm can be collected, and then the real-time wind power data is preprocessed to obtain the first preprocessed data. It should be noted that the collected real-time wind power data is the real wind power data with time series characteristics in the historical time of the current wind farm, not the predicted wind power data. When the real-time wind power data is preprocessed, the preprocessing operation may include but is not limited to data filling and other operations.

[0054] Step S12: Input the first preprocessed data into a pre-trained target wind power prediction model to obtain a wind power prediction result corresponding to the first historical wind power data output by the target wind power prediction model. The target wind power prediction model is a pre-trained multi-branch neural network model, and during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized.

[0055] In this embodiment, before using the first preprocessed data obtained after preprocessing the above-mentioned real-time wind power data for power prediction, it is necessary to pre-train a target wind power prediction model, that is, pre-construct an initial multi-branch neural network model, configure corresponding initial parameters, and then in the training process of the above-mentioned initial multi-branch neural network model, by optimizing the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component, the final target wind power prediction model for wind power prediction is determined, and then the above-mentioned first preprocessed data is input into the above-mentioned target wind power prediction model, so as to obtain the wind power prediction result corresponding to the first historical wind power data output by the target wind power prediction model.

[0056] It can be seen that in the training process of the multi-branch neural network model in the embodiment of the present application, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized to determine the trained target wind power prediction model, and then the collected real-time wind power data is pre-processed and input into the target wind power prediction model to obtain the corresponding wind power prediction result. It can be seen that the technical solution of the present application can enhance the autocorrelation of the prediction model and greatly improve the prediction accuracy.

[0057] See also Figure 2As shown, the embodiment of the present invention discloses a specific short-term wind power prediction method. Compared with the previous embodiment, this embodiment further illustrates and optimizes the technical solution.

[0058] Step S21: collecting historical wind power data of the current wind farm, and preprocessing the historical wind power data to obtain second preprocessed data.

[0059] In this embodiment, by collecting historical wind power data of the current wind farm, such as Figure 3 As shown, historical wind power data of a preset time period is collected, and then the historical wind power data is preprocessed to obtain second preprocessed data. A training set and a test set can be divided based on the second preprocessed data. The training set is used for model training, and the test set is used to verify the trained target model.

[0060] In this embodiment, the above-mentioned preprocessing of the historical wind power data to obtain the second preprocessed data may include: denoising the historical wind power data to obtain denoised data, and normalizing the denoised data to obtain the second preprocessed data. It can be understood that after the collected original historical wind power data is denoised, the denoised data is normalized. For example, after the above-mentioned historical wind power data is denoised, the denoised data is normalized using the Min-Max method to obtain normalized data, and the above-mentioned normalization can eliminate the dimensional differences between different data. The calculation formula of the Min-Max method is:

[0061]

[0062] in, represents the minimum value in the denoised data, represents the maximum value in the denoised data, p i represents the normalized data, x i Represents the raw wind power.

[0063] Step S22: Optimize the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm multiple times using a preset optimization algorithm to obtain the target number of decompositions and the target center frequency corresponding to each optimization.

[0064] In this embodiment, the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm are optimized multiple times to obtain the target number of decompositions and the target center frequency corresponding to each optimization. It should be noted that the termination conditions for multiple optimizations of the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm are determined based on the model evaluation index corresponding to the trained model. Each optimization corresponds to a new model training, thereby obtaining the corresponding trained model. During the optimization process, the optimization and training are stopped until the above termination conditions are met. In addition, the preset optimization algorithm may include but is not limited to WOA (Whale Optimization Algorithm), genetic algorithm, gray wolf algorithm, etc. For example, the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm are optimized multiple times using the whale algorithm. That is to say, the fitness of each whale is calculated by the whale optimization algorithm, and the position of each whale is updated until the optimal result is obtained. The above optimal results are the target number of decompositions and the target center frequency of each decomposition component in the above VMD algorithm. Among them, the calculation formula for updating the position of the above whale algorithm is:

[0065]

[0066] Among them, t represents the number of generations of the current iteration; X * (t) represents the optimal position closest to the determined prey when iterating to the tth generation; D represents the intermediate parameter of the distance between the leader whale and the prey; A represents the convergence factor; D′ represents the distance between the ith whale and the prey; b represents a constant that defines the shape of the spiral; l represents an element-by-element multiplication, usually a random number between [-1, 1]; p represents a probability value, which can be any value between [0, 1] in practical applications; X(t+1) represents the updated position.

[0067] Step S23: performing data decomposition on the second preprocessed data according to the target decomposition number and the target center frequency corresponding to each optimization, to obtain multiple groups of decomposition components corresponding to the target decomposition number and the center frequency.

[0068] In this embodiment, the second preprocessed data is subjected to an adaptive VMD decomposition operation according to the target decomposition number and the target center frequency corresponding to each optimization, and then a plurality of groups of decomposition components corresponding to the target decomposition number and the center frequency are obtained. That is, each pair of the target decomposition number and the target center frequency corresponding to the optimization respectively performs a VMD decomposition operation on the second preprocessed data, and a plurality of IMFs (Intrinsic Mode Functions, intrinsic mode components) and residual signals corresponding to each pair of the target decomposition number and the target center frequency corresponding to the optimization are respectively obtained.

[0069] For example, Figure 4 As shown, the collected pre-processed historical wind power signal is subjected to VMD decomposition operation according to the target decomposition number and the target center frequency corresponding to the optimization, and n IMFs and 1 residual component m(t) are obtained. Assuming that the historical wind power signal is f, the historical wind power signal can be decomposed into k modal components, and it is necessary to ensure that each modal component of the decomposition has a center frequency. Under the constraint that the sum of all modal components is equal to the pre-processed historical wind power signal, the sum of the estimated bandwidths of each modal component is minimized. Therefore, the variational expression constructed for the pre-processed historical wind power signal can be:

[0070]

[0071] Where k represents the number of modal components to be decomposed, k is generally a positive integer, {u k} represents the kth modal component after decomposition, {ω k} represents the center frequency corresponding to the kth modal component, δ(t) represents the Dirac function, * represents the convolution operator, Represents the sum of the decomposed components.

[0072] Step S24: reconstructing the second preprocessed data with each group of the decomposed components respectively to obtain multiple groups of reconstructed data.

[0073] In this embodiment, since there is a certain correlation between the decomposed components obtained by decomposing the preprocessed historical wind power signal and the original historical wind power, the preset feature splicing method can be used to reconstruct the decomposed components obtained after the decomposition and the preprocessed historical wind power data. If only one aspect of the data is used when predicting wind power, some features will be lost.

[0074] Step S25: using the reconstructed data to train an initial wind power prediction model constructed based on a convolutional neural network model and a gated recurrent unit, to obtain a plurality of trained wind power prediction models corresponding to the target decomposition number and the center frequency.

[0075] In this embodiment, the reconstructed data is used as a training sample for training the model, that is, the reconstructed data is used to train the initial wind power prediction model constructed based on the convolutional neural network model and the gated cyclic unit, and a plurality of trained wind power prediction models corresponding to the target number of decompositions and the center frequency are obtained. It can be understood that, by using each group of reconstructed data to train the initial wind power prediction model constructed based on the convolutional neural network model and the gated cyclic unit, a plurality of trained wind power prediction models corresponding to the target number of decompositions and the center frequency can be obtained. In other words, each group of reconstructed data is input into the initial convolutional neural network and the initial gated cyclic unit for training, and a plurality of trained convolutional neural networks and trained gated cyclic units corresponding to the target number of decompositions and the center frequency are obtained.

[0076] Step S26: determining a target wind power prediction model from the plurality of trained wind power prediction models.

[0077] In this embodiment, the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm are optimized multiple times to obtain the target number of decompositions and the target center frequency corresponding to each optimization. During the model training process, the pre-processed historical wind power data is processed based on the target number of decompositions and the target center frequency corresponding to each optimization, so as to obtain multiple trained wind power prediction models corresponding to the target number of decompositions and the target center frequency, and then the target wind power prediction model is determined from the multiple trained wind power prediction models.

[0078] In this implementation, the determination of the target wind power prediction model from the multiple trained wind power prediction models may include: determining the model evaluation index corresponding to each of the multiple trained wind power prediction models to obtain a corresponding evaluation index set; determining the trained wind power prediction model corresponding to the model evaluation index with the smallest value in the evaluation index set as the target wind power prediction model. It can be understood that the target wind power prediction model is determined based on the model evaluation index corresponding to the trained wind power prediction model, that is, the trained wind power prediction model corresponding to the model evaluation index with the smallest value is determined as the target wind power prediction model. Among them, the determination of the model evaluation index corresponding to each of the multiple trained wind power prediction models to obtain a corresponding evaluation index set, such as Figure 5 As shown, it may specifically include:

[0079] Step S31: performing feature extraction on the reconstructed data corresponding to each of the trained wind power prediction models in the plurality of trained wind power prediction models to obtain local trend features and global trend features corresponding to the reconstructed data.

[0080] Step S32: Fusing the local trend feature and the global trend feature to obtain a corresponding fused feature.

[0081] Step S33: inputting the fused features into the corresponding trained wind power prediction models respectively to obtain the wind power prediction value output by each trained wind power prediction model.

[0082] Step S34: determining a model evaluation index corresponding to each of the trained wind power prediction models based on the wind power prediction value and the wind power actual value corresponding to the wind power prediction value, and obtaining a corresponding evaluation index set.

[0083] That is, after obtaining the trained wind power prediction model, this embodiment performs multi-branch feature extraction on the reconstructed data corresponding to each trained wind power prediction model to obtain local trend features and global trend features corresponding to each group of reconstructed data, and then uses a preset feature fusion algorithm to fuse the local trend features and the global trend features corresponding to each group of reconstructed data to obtain corresponding fused features, and then inputs the fused features into the trained wind power prediction model accordingly, thereby obtaining wind power prediction values ​​output by different trained wind power prediction models, wherein the wind power prediction value output by each trained wind power prediction model is expressed as:

[0084] P=F(C(yt )⊙G(y t ));

[0085] Among them, y t represents the reconstructed data, C(y t ) represents the local trend characteristics corresponding to the reconstructed data, G(y t ) is the global trend feature corresponding to the reconstructed data, ⊙ represents the feature fusion operation, F() represents the trained wind power prediction model, and P represents the wind power prediction value output by the trained wind power prediction model.

[0086] After obtaining the wind power prediction value output by the trained wind power prediction model, the model evaluation index corresponding to each of the trained wind power prediction models is determined based on the wind power prediction value and the wind power actual value corresponding to the wind power prediction value to obtain a corresponding evaluation index set. Figure 6 As shown above Figure 6 The comparison between the wind power prediction value and the wind power actual value is shown, and the average absolute value error between the wind power prediction value and the wind power actual value corresponding to the wind power prediction value is calculated, wherein the calculation formula of the average absolute value error is:

[0087]

[0088] in, Indicates the actual value of wind power, p i is the wind power prediction value, MAE represents the mean absolute error, and n represents the total number of training samples.

[0089] Step S27: collecting real-time wind power data corresponding to the current wind farm, and preprocessing the real-time wind power data to obtain first preprocessed data.

[0090] Step S28: inputting the first preprocessed data into the target wind power prediction model to obtain a wind power prediction result output by the target wind power prediction model and corresponding to the first historical wind power data.

[0091] For the specific contents of the above steps S27 to S28, reference may be made to the corresponding contents disclosed in the above embodiments, which will not be repeated here.

[0092] like Figure 7As shown, the number of decompositions and the center frequency of each decomposition component in the VMD algorithm are optimized to obtain an optimized VMD algorithm, and then the optimized VMD algorithm is used to decompose the wind power data. At the same time, the decomposition components and the wind power data are reconstructed, and the reconstructed data are respectively input into CNN (Convolutional Neural Networks) and GRU (Gated Recurrent Unit) for modeling and prediction, and the prediction results output by CNN and GRU are fused to obtain the final prediction result. Using the VMD algorithm to decompose wind power data can enhance the autocorrelation of the prediction model and greatly improve the prediction accuracy. That is to say, since the input and output of the prediction model are both wind power, the prediction is made according to the changing trend of wind power, so the stronger the autocorrelation of the model, the better the prediction effect. Optimizing and adjusting the relevant parameters in VMD can realize the automatic optimization of the VMD algorithm in signal decomposition, and reconstruct the historical wind power data after decomposition components and preprocessing, and use the reconstructed data as new input features. Feature extraction of the reconstructed data based on CNN and GRU not only enables the constructed multi-branch neural network to realize the learning of global trend features, but also takes into account the influence of local trend features on the final prediction results, thereby improving the accuracy of wind power prediction.

[0093] It can be seen that in the embodiment of the present application, during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized to determine the trained target wind power prediction model, and then the collected real-time wind power data is preprocessed, and the first preprocessed data is input into the target wind power prediction model to obtain the corresponding wind power prediction result output by the target wind power prediction model. It can be seen that the technical solution of the present application can enhance the autocorrelation of the wind power prediction model and improve the prediction accuracy of short-term wind power.

[0094] Accordingly, the present application also discloses a wind power prediction device, see Figure 8 As shown, the device comprises:

[0095] The data acquisition module 11 is used to collect the real-time wind power data corresponding to the current wind farm;

[0096] A first data preprocessing module 12, used for preprocessing the real-time wind power data to obtain first preprocessed data;

[0097] A wind power prediction module 13, configured to input the first preprocessed data into a pre-trained target wind power prediction model, and obtain a wind power prediction result output by the target wind power prediction model corresponding to the first historical wind power data;

[0098] The target wind power prediction model is a pre-trained multi-branch neural network model, and during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized.

[0099] As can be seen from the above, in the training process of the multi-branch neural network model in this embodiment, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized to determine the trained target wind power prediction model, and then the collected real-time wind power data is preprocessed, and the first preprocessed data is input into the target wind power prediction model to obtain the corresponding wind power prediction result output by the target wind power prediction model. It can be seen that the technical solution of the present application can enhance the autocorrelation of the wind power prediction model and improve the prediction accuracy of short-term wind power.

[0100] In some specific embodiments, the wind power prediction device may further include:

[0101] A data collection module, used for the data collection unit to collect historical wind power data of the current wind farm;

[0102] A second data preprocessing module, used for preprocessing the historical wind power data to obtain second preprocessed data;

[0103] An optimization module is used to optimize the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm multiple times using a preset optimization algorithm to obtain the target number of decompositions and the target center frequency corresponding to each optimization;

[0104] A second data preprocessing module, used for performing data decomposition on the second preprocessed data according to the target decomposition number and the target center frequency corresponding to each optimization, to obtain multiple groups of decomposition components corresponding to the target decomposition number and the center frequency;

[0105] A data reconstruction module, used to reconstruct the second preprocessed data with each group of the decomposed components to obtain multiple groups of reconstructed data;

[0106] A model training module, used to train an initial wind power prediction model constructed based on a convolutional neural network model and a gated recurrent unit using the reconstructed data, to obtain a plurality of trained wind power prediction models corresponding to the target decomposition number and the center frequency;

[0107] The model determination module is used to determine a target wind power prediction model from the plurality of trained wind power prediction models.

[0108] In some specific embodiments, the second data preprocessing module may specifically include:

[0109] A data denoising submodule, used for denoising the historical wind power data to obtain denoised data;

[0110] The data normalization submodule is used to perform normalization processing on the denoised data to obtain second preprocessed data.

[0111] In some specific embodiments, the model determination module may specifically include:

[0112] An evaluation index determination submodule, used to determine a model evaluation index corresponding to each of the trained wind power prediction models in the plurality of trained wind power prediction models to obtain a corresponding evaluation index set;

[0113] The target model determination submodule is used to determine the trained wind power prediction model corresponding to the model evaluation index with the smallest value in the evaluation index set as the target wind power prediction model.

[0114] In some specific embodiments, the evaluation index determination submodule may specifically include:

[0115] A feature extraction unit, used for performing feature extraction on the reconstructed data corresponding to each of the trained wind power prediction models in the plurality of trained wind power prediction models, to obtain local trend features and global trend features corresponding to the reconstructed data;

[0116] A feature fusion unit, used for fusing the local trend feature and the global trend feature to obtain a corresponding fused feature;

[0117] A feature input unit, used to input the fused features into the corresponding trained wind power prediction models respectively, to obtain the wind power prediction value output by each trained wind power prediction model;

[0118] An evaluation index determination unit is used to determine the model evaluation index corresponding to each of the trained wind power prediction models based on the wind power prediction value and the wind power actual value corresponding to the wind power prediction value, so as to obtain a corresponding evaluation index set.

[0119] In some specific embodiments, the feature fusion submodule may specifically include:

[0120] The feature fusion unit is used to fuse the local trend feature and the global trend feature using a preset feature fusion algorithm to obtain a corresponding fused feature.

[0121] In some specific embodiments, the optimization module may specifically include:

[0122] The optimization submodule is used to use the whale algorithm to optimize the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component multiple times.

[0123] Furthermore, an embodiment of the present application also provides an electronic device. Fig. 9 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

[0124] Fig. 9 A schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the short-term wind power prediction method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0125] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0126] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0127] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, and may be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the short-term wind power prediction method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 may further include computer programs that can be used to complete other specific tasks.

[0128] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium, in which a computer program is stored. When the computer program is loaded and executed by a processor, the steps of the short-term wind power prediction method disclosed in any of the aforementioned embodiments are implemented.

[0129] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0130] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0131] The above is a detailed introduction to a short-term wind power prediction method, device, equipment and storage medium provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A short-term wind power prediction method, characterized in that: include: Collecting real-time wind power data corresponding to the current wind farm, and preprocessing the real-time wind power data to obtain first preprocessed data; Inputting the first preprocessed data into a pre-trained target wind power prediction model to obtain a wind power prediction result output by the target wind power prediction model and corresponding to the first historical wind power data; The target wind power prediction model is a pre-trained multi-branch neural network model, and during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized; Before inputting the first preprocessed data into a pre-trained target wind power prediction model, the method further includes: Collecting historical wind power data of the current wind farm, and preprocessing the historical wind power data to obtain second preprocessed data; The preset optimization algorithm is used to optimize the number of decompositions and the center frequency corresponding to each decomposition component in the VMD algorithm for multiple times, and the corresponding target number of decompositions and target center frequency after each optimization are obtained; Decomposing the second preprocessed data according to the target decomposition number and the target center frequency corresponding to each optimization to obtain multiple groups of decomposition components corresponding to the target decomposition number and the center frequency; Reconstructing the second preprocessed data with each group of the decomposed components respectively to obtain multiple groups of reconstructed data; Using the reconstructed data, an initial wind power prediction model constructed based on a convolutional neural network model and a gated recurrent unit is trained to obtain a plurality of trained wind power prediction models corresponding to the target decomposition number and the center frequency; Determining a target wind power prediction model from the plurality of trained wind power prediction models; Wherein, determining a target wind power prediction model from the plurality of trained wind power prediction models comprises: Determine a model evaluation index corresponding to each of the trained wind power prediction models in the plurality of trained wind power prediction models to obtain a corresponding evaluation index set; Determine the trained wind power prediction model corresponding to the model evaluation index with the smallest value in the evaluation index set as the target wind power prediction model; Wherein, the determining of the model evaluation index corresponding to each of the trained wind power prediction models in the plurality of trained wind power prediction models to obtain a corresponding evaluation index set includes: Performing feature extraction on the reconstructed data corresponding to each of the trained wind power prediction models among the plurality of trained wind power prediction models to obtain local trend features and global trend features corresponding to the reconstructed data; Fusing the local trend feature and the global trend feature to obtain a corresponding fused feature; Inputting the fused features into the corresponding trained wind power prediction models respectively to obtain the wind power prediction value output by each trained wind power prediction model; Based on the wind power prediction value and the wind power actual value corresponding to the wind power prediction value, a model evaluation index corresponding to each of the trained wind power prediction models is determined to obtain a corresponding evaluation index set.

2. The short-term wind power prediction method according to claim 1, characterized in that: The preprocessing of the historical wind power data to obtain second preprocessed data includes: The historical wind power data is subjected to denoising processing to obtain denoised data, and the denoised data is subjected to normalization processing to obtain second pre-processed data.

3. The short-term wind power prediction method according to claim 1, characterized in that: The fusing the local trend feature and the global trend feature to obtain a corresponding fused feature includes: The local trend feature and the global trend feature are fused using a preset feature fusion algorithm to obtain corresponding fused features.

4. The short-term wind power prediction method according to any one of claims 1 to 3, characterized in that: The method of using a preset optimization algorithm to optimize the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component multiple times includes: The whale algorithm is used to optimize the number of decompositions and the center frequency of each decomposition component in the VMD algorithm multiple times.

5. A wind power prediction device, characterized in that: The wind power prediction device is used to implement the short-term wind power prediction method according to any one of claims 1 to 4, and the wind power prediction device includes: Data acquisition module, used to collect real-time wind power data corresponding to the current wind farm; A first data preprocessing module, used for preprocessing the real-time wind power data to obtain first preprocessed data; A wind power prediction module, configured to input the first preprocessed data into a pre-trained target wind power prediction model to obtain a wind power prediction result output by the target wind power prediction model and corresponding to the first historical wind power data; The target wind power prediction model is a pre-trained multi-branch neural network model, and during the training process of the multi-branch neural network model, the number of decompositions in the VMD algorithm and the center frequency corresponding to each decomposition component are optimized.

6. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the steps of the short-term wind power prediction method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: Used to store computer programs; wherein, when the computer program is executed by a processor, the steps of the short-term wind power prediction method according to any one of claims 1 to 4 are implemented.

Images