Wind power prediction method, device, equipment and medium based on WTC-Informer algorithm

Abstract

A wind power prediction method, device, equipment and medium based on the WTC-Informer algorithm relate to the technical field of wind power prediction. The wind power prediction method includes: obtaining the operating data of the wind turbine. The operating data is input into a pre-trained wind power prediction model based on the WTC-Informer algorithm, and the following data processing steps are performed to obtain a prediction sequence of the active power of the wind turbine. Based on the operating data, high-frequency components and low-frequency components are extracted by wavelet transform and spliced ​​to obtain multi-scale features. Through the ECA channel attention mechanism, weights are dynamically assigned to each feature channel. Then, the computational complexity of attention is reduced by stacking multiple sparse self-attention encoding layers into an informer encoder. Finally, through a generative decoder, sparse attention is used to extract the temporal pattern, and combined with the connection feature map output by the cross-attention fusion encoder, the future wind power prediction sequence is output.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power prediction, and in particular to a wind power prediction method, device, equipment and medium based on a WTC-Informer algorithm. Background Art

[0002] Wind power forecasting is the process of predicting wind power generation at a specific moment or over a period of time in the future by analyzing and modeling historical wind farm power data. Accurate wind power forecasting can effectively alleviate the immense pressure and instability that large-scale wind power grid integration imposes on wind power companies' power systems. It can also effectively reduce power system operating costs and optimize wind farm operation and maintenance plans, thereby improving wind energy utilization, competitiveness, and market value in the electricity market.

[0003] Accurate wind power forecasting requires a large amount of historical data. However, in practice, data is often missing, anomaly, or inconsistent. These problematic data can negatively impact the accuracy of the forecasting model. Furthermore, wind power is often affected by a combination of long-term and short-term cyclical variations. For example, hourly changes in wind speed and temperature can have a direct impact on short-term wind power output, while seasonal variations can affect the overall power generation capacity of a wind farm. To improve forecasting accuracy, the model needs to be able to simultaneously process these multi-level, multi-scale cyclical characteristics. Therefore, a wind power forecasting model must be able to capture changes at different time scales, thereby more comprehensively reflecting the fluctuation patterns of wind power.

[0004] Deep learning has been widely used in wind power forecasting in recent years due to its powerful nonlinear modeling and feature extraction capabilities. However, existing deep learning models still face challenges in wind power forecasting, such as high computational complexity and insufficient multi-scale feature extraction. For example, the Transformer-based time series forecasting model introduces a multi-head self-attention mechanism, which enables the model to simultaneously focus on different parts of the input sequence and learn global features. While it excels at capturing global dependencies, its excessive focus on long-range dependencies can neglect short-term or local dependency information, thus affecting the accuracy and efficiency of the forecast.

[0005] In addition, the computational complexity of its dot product attention is ,As the length of the input sequence increases, the computation and ,memory consumption will rise dramatically, resulting in poor performance in ,long-term prediction. Summary of the Invention

[0006] The present invention provides a wind power prediction method, device, equipment and medium based on the WTC-Informer algorithm to improve at least one of the above technical problems.

[0007] In a first aspect, the present invention provides a wind power prediction method based on the WTC-Informer algorithm, which comprises:

[0008] Obtain wind turbine operating data.

[0009] The operating data is input into a pre-trained wind power prediction model based on the WTC-Informer algorithm, and the following data processing steps are performed to obtain a prediction sequence of the active power of the wind turbine.

[0010] According to the operating data, high-frequency components and low-frequency components are extracted by wavelet transform and spliced ​​to obtain multi-scale features.

[0011] Through the ECA channel attention mechanism, weights are dynamically assigned to each feature channel to obtain weighted multi-scale features.

[0012] The Informer encoder is composed of multiple sparse self-attention encoding layers, which reduces the computational complexity of attention from the standard Transformer to down to , obtain the connection feature graph; where, Indicates the time complexity, is the length of the input sequence.

[0013] Through the generative decoder, sparse attention is used to extract the temporal pattern, combined with the connection feature map output by the cross-attention fusion encoder to output the future wind power prediction sequence.

[0014] Preferably, obtaining the operating data of the wind turbine specifically includes:

[0015] Obtain the original operating data of the wind turbine.

[0016] The raw operating data is preprocessed to obtain preprocessed operating data; wherein the preprocessing includes outlier screening, missing value filling, data standardization, and feature selection.

[0017] Preferably, according to the operating data, high-frequency components and low-frequency components are extracted by wavelet transform and spliced ​​to obtain a multi-scale feature tensor, which specifically includes:

[0018] Initialize the filter bank containing the multi-layer wavelet decomposition Wavelet transform module. Among them, represents the low-frequency wavelet filter, represents a high-frequency wavelet filter.

[0019] The operating data is input into the wavelet transform module, and the high-frequency components are extracted by the high-frequency wavelet filters of the wavelet decomposition filter bank, and the low-frequency components are extracted by the low-frequency wavelet filters of the wavelet decomposition filter bank. The coefficients of the current layer are calculated through a convolution operation. The low-frequency components extracted by the wavelet decomposition filter bank of the previous layer are used as the input of the wavelet decomposition filter bank of the next layer and are further decomposed into high-frequency components and low-frequency components.

[0020] According to the high-frequency components of all levels obtained by the wavelet transform module, noise suppression is performed on each high-frequency component through a depth-wise separable convolution and point-by-point convolution to capture local changes, and a residual connection is made with the original high-frequency component to obtain the high-frequency features after reducing the high-frequency noise.

[0021] The low-frequency components obtained by the last layer of wavelet transform are interpolated to make their size consistent with the original input sequence. Then, the first-order difference is used to capture the change pattern of the sequence. Finally, the low-frequency components are enhanced through residual connection to obtain the enhanced low-frequency features.

[0022] The high-frequency features and low-frequency features of all levels are concatenated together to obtain the multi-scale feature tensor.

[0023] Preferably, the ECA channel attention mechanism is used to dynamically assign weights to each feature channel to obtain weighted multi-scale features, specifically including:

[0024] Adaptive average pooling is performed on each channel of the multi-scale feature tensor to compress the spatial dimension; wherein, Where, For the Adaptive average pooling results for channels, is the input sequence length, is the sequence number of the moment, For multi-scale features, The sequence number of the channel.

[0025] The local channel relationship is captured by one-dimensional convolution to obtain the weight. Where, For weight, represents convolution, is the multi-scale feature after adaptive average pooling, Indicates transpose.

[0026] Use the Sigmoid function to normalize the weights, then broadcast the weights to the original spatiotemporal dimensions, perform channel-level scaling, and obtain weighted multi-scale features. Where, is the weighted multi-scale feature tensor, For multi-scale features, represents the channel-by-channel product, Represents the Sigmoid function, For weight, Indicates transpose.

[0027] Sin-cosine position encoding is used to convert the weighted multi-scale feature timestamps into position information.

[0028] Preferably, the encoder model is:

[0029] .

[0030] in, For the layer Samples of time, represents the maximum pooling, is the activation function, represents one-dimensional convolution, Indicates the layer Samples of time, Represents a multi-head sparse attention block.

[0031] Preferably, the generative decoder uses sparse attention to extract the temporal pattern, combines the connection feature map output by the cross-attention fusion encoder, and outputs the future wind power prediction sequence, specifically including:

[0032] The sparse self-attention module inputs a data sequence, and uses a multi-head sparse self-attention mechanism to capture the internal correlation features of the data sequence. The data sequence includes the running data sequence and the sequence to be predicted.

[0033] The connection feature map is input into the cross-attention module, and the associated features and the connection feature map are interacted by cross-attention. The key features captured by the encoder are integrated into the prediction generation process of the decoder through the attention mechanism to obtain the decoding vector of the hidden representation.

[0034] The decoded vector is mapped to a wind power prediction result through a fully connected layer to obtain a target sequence.

[0035] Preferably, the wind power prediction model based on the WTC-Informer algorithm is trained by the following steps:

[0036] The historical operation data of the wind turbine is obtained, wherein the historical operation data includes characteristic data of a plurality of wind turbines.

[0037] The historical operation data is preprocessed, wherein the preprocessing includes detecting the historical operation data according to a preset abnormal value determination condition, deleting all abnormal data, and filling in missing data with the value of the previous moment.

[0038] The preprocessed historical running data is divided into training set, validation set and test set, and standardized.

[0039] A correlation analysis is performed on the various features of the standardized historical operating data to screen out features with a correlation with wind power higher than a preset value, and obtain the final training set, validation set, and test set.

[0040] A wind power prediction model based on the WTC-Informer algorithm was constructed and trained using the final training set and a pre-defined loss function to establish a mapping relationship between input features and wind power. The model's performance was evaluated after each training cycle using a validation set. An early stopping strategy was used to terminate training early if performance on the validation set did not improve over several consecutive training cycles to ensure the model's generalization ability.

[0041] The trained model is evaluated by inputting the final test set. When the model meets the preset standards, the model parameters are saved and the pre-trained wind power prediction model based on the WTC-Informer algorithm is obtained.

[0042] In a second aspect, the present invention provides a wind power prediction device based on the WTC-Informer algorithm, which is used to implement a wind power prediction method based on the WTC-Informer algorithm as described in any paragraph of the first aspect.

[0043] The wind power prediction device comprises:

[0044] The operation data acquisition module is used to obtain the operation data of the wind turbine.

[0045] The prediction module is used to input the operating data into a pre-trained wind power prediction model based on the WTC-Informer algorithm, perform data processing through the following units, and obtain a prediction sequence of the active power of the wind turbine.

[0046] The wavelet transform subunit is used to extract high-frequency components and low-frequency components through wavelet transform according to the operating data and to splice them to obtain multi-scale features.

[0047] The ECA channel attention subunit is used to dynamically assign weights to each feature channel through the ECA channel attention mechanism to obtain weighted multi-scale features.

[0048] The encoder unit is used to stack multiple sparse self-attention encoding layers into an Informer encoder, which reduces the computational complexity of attention from the standard Transformer down to , obtain the connection feature graph; where, Indicates the time complexity, is the length of the input sequence.

[0049] The decoder unit is used to extract the temporal pattern using sparse attention through a generative decoder, and combine it with the connection feature map output by the cross-attention fusion encoder to output the future wind power prediction sequence.

[0050] In a third aspect, the present invention provides a wind power forecasting device based on the WTC-Informer algorithm, comprising a processor, a memory, and a computer program stored in the memory. The computer program can be executed by the processor to implement the wind power forecasting method based on the WTC-Informer algorithm as described in any paragraph of the first aspect.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a wind power prediction method based on the WTC-Informer algorithm as described in any paragraph of the first aspect.

[0052] By adopting the above technical solution, the present invention can achieve the following technical effects:

[0053] The present invention's wind power forecasting method, based on the WTC-Informer algorithm, uses a multi-scale feature extraction module to perform multi-level wavelet decomposition on the input sequence. Using a wavelet filter bank, the time series data is decomposed layer by layer into multiple high-frequency components and deep low-frequency components. Based on different frequency characteristics, the module employs layered processing to perform deep separable convolution on the high-frequency components followed by linear interpolation reconstruction. The low-frequency components are differentially enhanced and residually joined with the original data. Finally, a channel attention mechanism dynamically integrates multi-scale features, enabling cross-scale feature interaction from short-term fluctuations to long-term trends, significantly improving the modeling capabilities of wind power series.

[0054] Using the Informer as a predictor, the sparse attention mechanism only calculates the small number of point pairs that contribute most to the output, thereby reducing the complexity of attention calculations. The encoder reduces the sequence length through layer-by-layer downsampling, retaining the most important feature points at each layer to further improve computational efficiency. A multi-head sparse self-attention mechanism is also used to capture correlated features within the input sequence. Furthermore, the Informer's cross-attention mechanism is used to interact the decoder input with the global contextual features of the encoder output, incorporating key features captured by the encoder into the decoder's prediction generation process. The decoder output uses a fully connected layer to map the hidden representation to the final prediction result, generating the target sequence. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the specific implementation methods of the present invention. It should be understood that the following drawings only show certain specific implementation methods of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 It is a flow chart of the wind power prediction method.

[0057] Figure 2 This is a flowchart of the training process of the wind power prediction model.

[0058] Figure 3 This is the architecture diagram of the wind power prediction model based on the WTC-Informer algorithm.

[0059] Figure 4 It is a schematic diagram of wavelet transform. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0061] Example 1, please refer to Figures 1 to 4A first embodiment of the present invention provides a wind power forecasting method based on the WTC-Informer algorithm, which can be performed by a wind power forecasting device based on the WTC-Informer algorithm (hereinafter referred to as the wind power forecasting device). Specifically, steps A1 to A2 are performed by one or more processors in the wind power forecasting device. It is understood that the wind power forecasting device can be an electronic device with computing capabilities, such as a portable notebook computer, desktop computer, server, smartphone, or tablet computer.

[0062] A1. Obtaining the operating data of the wind turbine. Preferably, step A1 specifically includes steps A11 to A12.

[0063] A11. Obtain the original operating data of the wind turbine.

[0064] A12. Preprocess the original operating data to obtain preprocessed operating data; wherein the preprocessing includes outlier screening, missing value filling, data standardization, and feature selection.

[0065] A2. Input the operating data into a pre-trained wind power prediction model based on the WTC-Informer algorithm, and execute steps B1 to B4 to obtain a prediction sequence of the active power of the wind turbine for a period of time in the future.

[0066] Specifically, the wind power forecasting model based on the WTC-Informer algorithm integrates wavelet transforms with the Informer architecture to construct a nonlinear mapping relationship from historical input features to future active power. The model takes preprocessed time series data as input and, through multi-scale feature extraction and long-sequence dependency modeling, directly outputs a series of wind turbine active power forecasts for a period of time in the future. This model can provide wind power operators with accurate decision-making support, supporting them in optimizing wind farm power scheduling, grid stability control, and reserve capacity allocation, thereby promoting the widespread application and sustainable development of wind power generation.

[0067] B1. Extract high-frequency components and low-frequency components from the operating data through wavelet transform and combine them to obtain multi-scale features. Preferably, B1 specifically includes steps B11 to B15.

[0068] The process of wavelet transform in this embodiment can be understood as using high-frequency wavelet filter and low-frequency wavelet filter to perform convolution operation on the input sequence respectively to obtain high-frequency component and low-frequency component. Figure 4 As shown, the low-frequency components represent global characteristics and trends, while the high-frequency components represent rapidly changing detail characteristics.

[0069] B11. Initialize the filter bank containing multi-layer wavelet decomposition Wavelet transform module. Among them, represents the low-frequency wavelet filter, Represents a high-frequency wavelet filter. For the input sequence, the filter is used to perform low-frequency wavelet decomposition on each variable step by step.

[0070] B12, input the operating data into the wavelet transform module, extract the high-frequency component through the high-frequency wavelet filter of the wavelet decomposition filter bank, extract the low-frequency component through the low-frequency wavelet filter of the wavelet decomposition filter bank, and calculate the coefficient of the current layer through a convolution operation. Figure 3 As shown in FIG, the low-frequency component extracted by the previous wavelet decomposition filter bank is further decomposed into high-frequency component and low-frequency component as the input of the next wavelet decomposition filter bank.

[0071] Specifically, the input signal (i.e., running data) is decomposed layer by layer into low-frequency and high-frequency components using a multi-layer wavelet decomposition filter bank. For each layer of wavelet decomposition, a convolution operation is used to calculate the coefficients of the current layer, and the low-frequency component of the signal is used as the input for the next layer. The high-frequency component of each layer is saved and used as the subsequent high-frequency features.

[0072] The calculation model of the wavelet transform module is:

[0073] .

[0074] in, Indicates the level of wavelet transform, is the sequence number of the variable, Indicates low frequency, Indicates high frequency, Indicates the The input sequence of variables, Indicates the The variable The low-frequency components extracted by the wavelet transform, Indicates the The variable The high frequency components extracted by wavelet transform, represents convolution, represents the low-frequency wavelet filter, represents a high-frequency wavelet filter, Indicates the The variable The low-frequency components extracted by the wavelet transform. hour, Use the running data preprocessed in step A1.

[0075] B13. Based on the high-frequency components of all levels obtained by the wavelet transform module, noise suppression is performed on each high-frequency component through a depth-wise separable convolution and point-by-point convolution to capture local changes, and a residual connection is made with the original high-frequency component to obtain the high-frequency component after reducing high-frequency noise, so as to reduce the impact of high-frequency noise on the prediction.

[0076] The processing model for high-frequency components is:

[0077] .

[0078] in, Indicates the The variable The high frequency components extracted by wavelet transform, Represents point-by-point convolution operation, Indicates splicing, Represents a depthwise separable convolution operation.

[0079] B14. Interpolate the low-frequency components obtained by the last layer of wavelet transform to make their size consistent with the original input sequence. Then use the first-order difference to capture the change pattern of the sequence. Finally, enhance the features of the low-frequency components through residual connection to obtain the enhanced low-frequency components.

[0080] Specifically, the low-frequency component obtained by the last layer of wavelet transform Represents the trend or long-term change in the input sequence. In this embodiment, the low-frequency component is first Interpolation is performed to ensure that the size is consistent with the original input sequence. First-order differences are then used to capture the changing pattern of the sequence. Finally, residual connections are used to enhance the low-frequency components.

[0081] The processing model of the low-frequency component obtained by the last layer of wavelet transform is:

[0082] .

[0083] .

[0084] in, for The first-order difference of time, Indicates splicing, Indicates the variables The low-frequency components extracted at each moment, Indicates the variables The low-frequency components extracted at each moment, Represents the low-frequency component obtained by the last layer of wavelet transform, Represents a point-wise convolution operation.

[0085] B15. Combine the high-frequency features and low-frequency features of all levels to obtain the multi-scale features. The calculation model of the multi-scale feature tensor is: .in, For multi-scale features, Indicates splicing, Indicates the The variable The high frequency components extracted by wavelet transform, Represents the low-frequency component obtained by the last layer of wavelet transform.

[0086] B2. Dynamically assign weights to each feature channel through the ECA channel attention mechanism to obtain weighted multi-scale features to further enhance the model's ability to learn features at different scales. Preferably, step B2 specifically includes steps B21 to B24.

[0087] B21. Perform adaptive average pooling on each channel of the multi-scale feature tensor to compress the spatial dimension.

[0088] .

[0089] in, For the Adaptive average pooling results for channels, is the input sequence length, is the sequence number of the moment, For multi-scale features, The sequence number of the channel. . The total number of moments.

[0090] B22. Capture local channel relationships through one-dimensional convolution and obtain weights.

[0091] .

[0092] Where, For weight, represents convolution, is the multi-scale feature after adaptive average pooling, Indicates transpose.

[0093] B23. Use the Sigmoid function to normalize the weights, then broadcast the weights to the original spatiotemporal dimensions, perform channel-level scaling, and obtain weighted multi-scale features.

[0094] .

[0095] Where, is the weighted multi-scale feature tensor, For multi-scale features, represents the channel-by-channel product, Represents the Sigmoid function, For weight, Indicates transpose.

[0096] B24. Use sine and cosine position encoding to convert the weighted multi-scale feature timestamps into position information.

[0097] Specifically, since the attention mechanism itself is insensitive to the order of sequence data, in order to enhance the model's ability to model the periodicity of the wind turbine operating status sequence, the present invention adopts sine and cosine position encoding to convert timestamps into position information, so that the model can distinguish different time steps and effectively retain the time sequence information, thereby improving the prediction accuracy.

[0098] The calculation model using sine and cosine position encoding is:

[0099] .

[0100] .

[0101] in, For position coding, Represents the position index of the current time step, Represents the encoding dimension, is the dimension of the model, is a sine function, is the cosine function.

[0102] B3, through the Informer encoder stacked with multiple sparse self-attention encoding layers, the computational complexity of attention is reduced from the standard Transformer down to , obtain the connection feature graph; where, Indicates the time complexity, is the length of the input sequence.

[0103] In this embodiment, the encoder is composed of multiple stacked encoding modules, which extract global dependencies and local features from the historical sequence and compress them into an efficient contextual representation. The encoder reduces the sequence length by downsampling layer by layer. The most important feature points are retained at each layer, further improving efficiency.

[0104] The encoder model is:

[0105] .

[0106] in, For the layer Samples of time, represents the maximum pooling, is the activation function, represents one-dimensional convolution, Indicates the layer Samples of time, Represents a multi-head sparse attention block.

[0107] The encoding module uses a sparse attention mechanism to calculate only a small number of point pairs that contribute most to the output, reducing the complexity of attention calculation from the standard Transformer down to In the sparse attention mechanism, the input is converted into three parts: query ,key ,value In all Random sampling is performed indivual , whose distribution can represent all The distribution of Calculate each The attention score,

[0108] The process can be expressed as:

[0109] .

[0110] .

[0111] .

[0112] .

[0113] in, Represents the input of the encoding module, To generate a learnable matrix for query, is the learnable matrix for generating keys, is the learnable matrix that generates the values, The calculated The attention score of the query, Indicates the queries, For query The index of Indicates the keys, Key The index of Indicates transposition, represents the dimension of the input vector, It is randomly selected The number of

[0114] The sparse self-attention calculation formula is:

[0115] .

[0116] in, represents the output of the sparse self-attention mechanism, For query, For key, For value, represents the softmax function, Indicates the top selected based on the attention score "active" queries, Indicates transposition, Indicates the dimension of the input vector.

[0117] B4. Using a generative decoder, extracting temporal patterns using sparse attention, combining them with the connection feature graph output by the cross-attention fusion encoder, and outputting a future wind power prediction sequence. Preferably, step B4 specifically includes steps B41 to B43.

[0118] B41. Input a data sequence into the sparse self-attention module and capture the internal correlation features of the data sequence through the multi-head sparse self-attention mechanism. The data sequence includes the running data sequence and the sequence to be predicted.

[0119] B42. Input the connection feature map into the cross-attention module, use cross-attention to interact the associated features and the connection feature map, integrate the key features captured by the encoder into the prediction generation process of the decoder through the attention mechanism, and obtain the decoding vector of the hidden representation.

[0120] B43. Map the decoded vector to a wind power prediction result through a fully connected layer to obtain a target sequence.

[0121] In this embodiment, the decoder consists of a sparse self-attention module and a cross-attention module. Based on the context vector output by the encoder, the decoder generates a prediction sequence of the wind turbine active power for a period of time in the future. The historical input sequence and length are The sequence to be predicted is filled with 0 to generate the prediction result.

[0122] Specifically, the decoder utilizes a multi-head sparse self-attention mechanism to capture correlated features within the decoder input sequence. Cross-attention is used to interact the decoder input with the global contextual features of the encoder output. This attention mechanism incorporates key features captured by the encoder into the decoder's prediction generation process. Finally, the decoder output passes through a fully connected layer to map the hidden representation to the final prediction result, generating the target sequence.

[0123] The wind power prediction method based on the WTC-Informer algorithm of the present invention integrates wavelet transform and deep neural network architecture to realize cross-scale feature learning of multivariate time series data. The model performs multi-level wavelet transform on multi-channel inputs through the wavelet decomposition module, wherein the high-frequency components use deep separable convolution and residual gating to achieve noise robustness processing, and the low-frequency components are fused with the original signal features through first-order differences to enhance the trend characterization capability. Through dynamic weighted fusion of the channel attention network, a multi-scale feature tensor is formed. This architecture breaks through the limitations of traditional single-scale convolution and improves the modeling accuracy of non-stationary features and cross-scale dependencies in wind power sequences.

[0124] The sparse self-attention mechanism based on the Informer algorithm achieves efficient attention computation, fully capturing the internal correlation features of the input sequence. Based on the context vector output by the encoder, the decoder directly generates predictions for future sequences, thereby achieving accurate and efficient wind power forecasting. Power dispatchers can use these predictions as a basis for power demand scheduling, effectively alleviating the immense pressure and instability that large-scale wind power integration brings to the power system. This can also effectively reduce power system operating costs and optimize the operation and maintenance plans of the entire wind farm.

[0125] The pre-training steps of the pre-trained wind power prediction model based on the WTC-Informer algorithm include steps S1 to S6.

[0126] S1. Obtain historical operating data of wind turbines.

[0127] The historical operation data of wind turbines are collected through sensors distributed on wind turbines. The historical operation data includes: including timestamps , wind speed , the angle between the wind direction and the position of the turbine nacelle , ambient temperature , the temperature inside the turbine nacelle , Nacelle direction (yaw angle) , blade 1 pitch angle , blade 2 pitch angle 、Blade 3 pitch angle , reactive power , active power Among them, active power It serves as both input feature and predicted target value.

[0128] Specifically, the fan The running state at any moment is represented by the feature vector: The goal is to use the past time step data , to predict the future Active power in time steps The learning goal is to find a mapping relationship , so that the predicted value As close to the true value as possible .in The wind power prediction model based on the WTC-Informer algorithm proposed in this invention (such as Figure 3 shown), is a trainable parameter, .

[0129] S2. Preprocess the historical operation data. The preprocessing includes detecting the historical operation data according to preset abnormal value judgment conditions, deleting all abnormal data, and filling missing data with the value of the previous moment to ensure data integrity.

[0130] Specifically, the abnormal value determination condition is:

[0131] When the active power is less than 0, it is considered abnormal.

[0132] When the wind speed is less than 1 m / s and the active power is greater than 10 kilowatts, it is considered abnormal.

[0133] When the wind speed is less than 2 m / s and the active power is greater than 100 kilowatts, it is considered abnormal.

[0134] When the wind speed is less than 3 m / s and the active power is greater than 200 kilowatts, it is considered abnormal.

[0135] When the wind speed is greater than 2.5 m / s and the active power is equal to 0, it is considered abnormal.

[0136] When the wind speed is equal to 0, the angle between the wind direction and the turbine nacelle position is equal to 0, and the ambient temperature is equal to 0, it is considered abnormal.

[0137] When the ambient temperature is less than -21 degrees Celsius or greater than 60 degrees Celsius, it is considered abnormal.

[0138] When the temperature inside the turbine compartment is less than -21 degrees Celsius or greater than 70 degrees Celsius, it is considered abnormal.

[0139] When the angle between the wind direction and the turbine nacelle position is greater than 180 degrees or less than -180 degrees, it is considered abnormal.

[0140] When the cabin direction is greater than 720 degrees or less than -720 degrees, it is considered abnormal.

[0141] When any one of blade pitch angle 1, blade pitch angle 2 or blade pitch angle 3 is greater than 89 degrees, it is considered abnormal.

[0142] After removing outliers, check for missing values ​​in the dataset. If a missing value appears in the first data point, delete the record at that moment. Otherwise, fill in the missing value with the value from the previous moment.

[0143] S3. Divide the preprocessed historical operation data into a training set, a validation set, and a test set, and perform standardization.

[0144] In this example, historical operational data is divided into training, validation, and test sets according to a certain ratio. The training set is used to train the model, the validation set is used to adjust hyperparameters and monitor model performance during training, and the test set is used to ultimately evaluate the model's generalization ability. The historical data is then normalized based on the statistics of the training set to eliminate the dimensionality effects of different features.

[0145] Specifically, the wind turbine operation data is divided into training set, validation set and test set according to the time sequence and a certain ratio. The mean and standard deviation of all features and active power except timestamp in training set, validation set and test set The features outside the dataset are Z-score standardized and mapped to a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0146] Furthermore, the fan is Features , its standardized formula can be expressed as:

[0147]

[0148] in, For the standardized data, is the original value, For the Features The mean value on the training set, Characterized by The standard deviation on the training set.

[0149] and The calculation method is:

[0150]

[0151]

[0152] in, Represents the total number of time steps in the training set, that is, the number of original data items contained in the training set.

[0153] Fan at all times Active power The standardized formula can be expressed as:

[0154]

[0155] in, is the normalized active power, is the original value of active power, is the mean of active power in the training set, is the standard deviation of active power in the training set.

[0156] After standardization, the fan Operation data at the moment Expressed as: .

[0157] S4. Perform correlation analysis on various features of the standardized historical operating data, and select features whose correlation with wind power is higher than a preset value.

[0158] Specifically, we conduct correlation analysis on various features in historical data to assess their impact on wind power. We screen out features that are highly correlated with wind power and construct a new dataset as input for model training. This reduces redundant features, mitigates the risk of overfitting, and improves model training efficiency and computational speed.

[0159] In this embodiment, the Pearson correlation analysis is performed using the normalized training set data to obtain the Pearson correlation matrix. Since the data has been normalized by Z-score,

[0160] Pearson correlation coefficient The calculation formula can be expressed as:

[0161]

[0162] in, represents the total number of time steps in the training set, For the standardized data, is the normalized active power.

[0163] The Pearson correlation coefficient can be used to measure the correlation between two variables, and its value range is . Indicates a perfect positive correlation. Indicates a perfect negative correlation. Indicates that there is no linear correlation between the two. By calculating the Pearson correlation matrix, we screen out the N features with the highest correlation with active power. We use this to construct a new dataset as input for the model to improve model performance and computational efficiency.

[0164] The data after feature selection is represented as:

[0165] training set ,in Indicates that the nth feature is from t=1 to t= sequence on.

[0166] Validation set ,in Indicates that the nth feature is at t= +1 to t= sequence on;

[0167] Validation set ,in Indicates that the nth feature is at t= +1 to t= sequence on.

[0168] For the new training set, validation set, and test set, the sliding window technique is used to generate the input matrix of the final model, where the sliding window size is .

[0169] S5. Build a wind power prediction model based on the WTC-Informer algorithm and train it based on the final training set. The architecture of the wind power prediction model is as follows: Figure 3 As shown in Figure 1, the model consists of a multi-scale feature extraction unit, an encoder unit, and a decoder unit. The multi-scale feature extraction unit includes a wavelet transform sub-unit and an ECA channel attention sub-unit.

[0170] The multi-scale feature extraction unit uses wavelet transforms, depthwise separable convolutions, and channel attention to extract multi-scale features from the input sequence. Furthermore, a positional encoding module in the embedding layer introduces positional information into the time series data and maps the features to a higher dimension to compensate for the self-attention mechanism's inability to directly perceive sequence order. The encoder unit, composed of multiple stacked encoding modules, extracts global dependencies and local features from the historical sequence and compresses them into an efficient contextual representation. Based on the context vector output by the encoder, the decoder unit generates a prediction sequence for the wind turbine's active power over a period of time.

[0171] The training set is used to train a wind power prediction model based on the WTC-Informer algorithm, establishing a mapping relationship between input features and wind power. Backpropagation is performed using a pre-defined loss function to gradually adjust the model parameters, enabling the model to more accurately predict wind power.

[0172] The validation set is used to evaluate the performance of the model after each training cycle. The early stopping strategy is used to terminate the training early when the performance on the validation set does not improve over several consecutive training cycles to ensure the generalization ability of the model.

[0173] In this embodiment, a pre-defined loss function is used to continuously optimize model parameters using data from the training set. The present invention uses mean squared error (MSE) as the loss function. During model training, an early stopping strategy based on validation set monitoring is employed to balance model performance and training efficiency. After each training cycle, the model's generalization ability is evaluated using the pre-defined MSE metric based on the validation set, and the current optimal model parameters are dynamically saved.

[0174] If the validation set performance does not improve over several consecutive training cycles or reaches the maximum iteration limit, training is terminated. After training is complete, the model parameters are rolled back to the best-performing validation set, ensuring the final model has optimal generalization capabilities and can be used for test set evaluation or actual wind farm power forecasting, achieving a mapping from historical features to future power.

[0175] S6. The trained model is evaluated by inputting the final test set. When the model meets the preset standards, the model parameters are saved and the pre-trained wind power prediction model based on the WTC-Informer algorithm is obtained.

[0176] Specifically, the test data set is fed into the trained prediction model to generate prediction results. By comparing the results with the real data, model evaluation metrics are calculated to further assess the model's predictive performance. By analyzing the deviation between the prediction results and the actual situation, the model design is adjusted and retrained to improve its accuracy and robustness.

[0177] The test set is input into the trained wind power prediction model, and forward reasoning is performed to obtain the future The wind turbine active power prediction sequence is calculated based on the time step. By comparing it with the real data, evaluation indicators such as mean square error (MSE) and mean absolute error (MAE) are calculated.

[0178] The final generalization capability of the model is evaluated based on the evaluation indicators. If the main indicators meet the preset standards, the model is deemed qualified and the model parameters are saved simultaneously, forming a deployable pre-trained prediction model. If it does not meet expectations, the model is retrained.

[0179] Analyze the evaluation results, check the performance of the model in different scenarios, and understand its strengths and limitations. Based on the evaluation results, the model may need to be adjusted and optimized to improve its performance and obtain the final wind power prediction model based on the WTC-Informer algorithm.

[0180] Save the best model and load it directly for future use, avoiding the time and computing resource overhead of retraining.

[0181] To validate the effectiveness of the proposed model and solution, we selected operational data from a wind turbine at a specific wind farm as our research object. This dataset covers 245 consecutive days of operation, providing sufficient data to ensure statistical significance and reliability of the experimental results.

[0182] In experiments, the proposed model was used to predict wind turbine active power for different time periods in the future, including forecasts for the next 2, 4, 8, and 12 hours. For each forecast period, two key evaluation metrics, the mean squared error (MSE) and mean absolute error (MAE), were calculated to quantify the accuracy of the predictions.

[0183] In order to comprehensively evaluate the performance of the model proposed in this invention, this application selects four types of representative time series prediction architectures as benchmark models: the basic Informer model, the classic Transformer model, and the LightTS model.

[0184] The Transformer model uses the original self-attention mechanism as its core component. Its multi-head attention layer models global sequence dependencies through parallel computation. The Informer model optimizes computational complexity through the ProbSparse self-attention mechanism and designs a distillation operation for hierarchical feature extraction. The LightTS model builds a lightweight temporal network based on depthwise separable convolutions. Its temporal convolution module captures multi-scale temporal patterns through dilated causal convolutions.

[0185] Experimental results demonstrate that the WTC-Informer model offers specific advantages in multi-period wind power forecasting tasks: within the 2-12 hour forecast range, its MSE and MAE indicators significantly outperform competing models such as the Transformer and Informer models. In particular, in the 12-hour long-range forecast, the MSE is 4.8% lower than the suboptimal model, and the error growth rate is the slowest. Through the cross-scale fusion mechanism of wavelet processors and deep networks, the model accurately captures the transient fluctuations and slow-changing trends of wind power in the joint time-frequency domain, effectively addressing the error accumulation problem of traditional methods in long-term forecasting. This demonstrates the high adaptability of the multi-scale feature learning architecture to complex time series patterns, providing more reliable forecasting support for wind power grid-connected scheduling.

[0186] Table 1 Comparative experimental results

[0187]

[0188] In summary, the embodiment of the present invention introduces the idea of ​​wavelet transform into the processing of wind power and its related variable data, and constructs a multi-scale feature extraction module. Through the layer-by-layer low-frequency decomposition of wavelet transform, differential feature enhancement and convolution are performed on the low-frequency and high-frequency results at different levels, respectively, to efficiently capture local and global features in the time series, enhance the modeling ability of multi-scale information, and avoid the problem of a sharp increase in the number of parameters caused by directly using large convolution kernels to extract features from the original sequence. Through the sparse attention mechanism designed specifically for long time series modeling, only a small number of point pairs that contribute the most to the output are calculated, achieving efficient attention calculation.

[0189] By combining wavelet convolution with the Informer algorithm, this paper provides a more powerful wind power forecasting model. This model, with its powerful multi-scale feature extraction capabilities and attention mechanism, can fully leverage both long-term trends and short-term fluctuations in variables, establishing a precise nonlinear mapping between historical features and future generated power, thus providing a new and effective approach for wind power forecasting.

[0190] Embodiment 2: A wind power prediction device based on the WTC-Informer algorithm, which is used to implement the wind power prediction method based on the WTC-Informer algorithm as described in any paragraph of Embodiment 1. The wind power prediction device includes: an operation data acquisition module, a prediction module, a wavelet transform subunit, an ECA channel attention subunit, an encoder unit, and a decoder unit.

[0191] The operation data acquisition module is used to obtain the operation data of the wind turbine.

[0192] The prediction module is used to input the operating data into a pre-trained wind power prediction model based on the WTC-Informer algorithm, perform data processing through the following units, and obtain a prediction sequence of the active power of the wind turbine.

[0193] The wavelet transform subunit is used to extract high-frequency components and low-frequency components through wavelet transform according to the operating data and to splice them to obtain multi-scale features.

[0194] The ECA channel attention subunit is used to dynamically assign weights to each feature channel through the ECA channel attention mechanism to obtain weighted multi-scale features.

[0195] The encoder unit is used to stack multiple sparse self-attention encoding layers into an Informer encoder, which reduces the computational complexity of attention from the standard Transformer down to , obtain the connection feature graph; where, Indicates the time complexity, is the length of the input sequence.

[0196] The decoder unit is used to extract the temporal pattern using sparse attention through a generative decoder, and combine it with the connection feature map output by the cross-attention fusion encoder to output the future wind power prediction sequence.

[0197] Embodiment 3: A wind power prediction device based on the WTC-Informer algorithm, comprising a processor, a memory, and a computer program stored in the memory. The computer program can be executed by the processor to implement the wind power prediction method based on the WTC-Informer algorithm as described in any paragraph of Embodiment 1.

[0198] Embodiment 4: A computer-readable storage medium. The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device containing the computer-readable storage medium is controlled to execute the wind power prediction method based on the WTC-Informer algorithm as described in any paragraph of Embodiment 1.

[0199] In the several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the devices, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, as well as the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or actions, or can be implemented using a combination of dedicated hardware and computer instructions.

[0200] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0201] If the functions are implemented as software modules and sold or used as standalone products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for causing a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories, random access memories, magnetic disks, or optical disks. It should be noted that, as used herein, the terms "comprise," "include," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. Without further limitation, the phrase "comprises a..." does not preclude the presence of other identical elements in the process, method, article, or device comprising the elements.

[0202] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "an", "the" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise.

[0203] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0204] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0205] The references to "first" and "second" in the embodiments merely distinguish similar objects and do not represent a specific ordering of the objects. It is understood that the specific order or precedence of "first" and "second" can be interchanged where appropriate. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein.

[0206] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A wind power prediction method based on the WTC-Informer algorithm, characterized in that: Include: Obtaining wind turbine operating data; Input the operating data into a pre-trained wind power prediction model based on the WTC-Informer algorithm, and perform the following data processing steps to obtain a prediction sequence of the active power of the wind turbine; According to the operating data, high-frequency components and low-frequency components are extracted by wavelet transform and spliced ​​to obtain multi-scale features; Through the ECA channel attention mechanism, weights are dynamically assigned to each feature channel to obtain weighted multi-scale features; The Informer encoder is composed of multiple sparse self-attention encoding layers, which reduces the computational complexity of attention from the standard Transformer to down to , obtain the connection feature graph; where, Indicates the time complexity, is the length of the input sequence; Through the generative decoder, sparse attention is used to extract temporal patterns, combined with the connection feature map output by the cross-attention fusion encoder to output the future wind power prediction sequence; According to the operating data, high-frequency components and low-frequency components are extracted and spliced ​​by wavelet transform to obtain multi-scale feature tensors, specifically including: Initialize the filter bank containing the multi-layer wavelet decomposition Wavelet transform module; among them, represents the low-frequency wavelet filter, represents a high-frequency wavelet filter; Inputting the operating data into the wavelet transform module, extracting high-frequency components through the high-frequency wavelet filters of the wavelet decomposition filter bank, extracting low-frequency components through the low-frequency wavelet filters of the wavelet decomposition filter bank, and calculating the coefficients of the current layer through a convolution operation; wherein the low-frequency components extracted by the wavelet decomposition filter bank of the previous layer are used as inputs of the wavelet decomposition filter bank of the next layer and are further decomposed into high-frequency components and low-frequency components; Based on the high-frequency components of all levels obtained by the wavelet transform module, noise suppression is performed on each high-frequency component through a depth-wise separable convolution and point-by-point convolution to capture local changes, and a residual connection is made with the original high-frequency component to obtain the high-frequency features after reducing the high-frequency noise; the processing model of the high-frequency component is: ； in, Indicates the The variable The high frequency components extracted by wavelet transform, Represents point-by-point convolution operation, Indicates splicing, represents a depth-wise separable convolution operation; The low-frequency components obtained by the last wavelet transform are interpolated to make their size consistent with the original input sequence. Then, the first-order difference is used to capture the change pattern of the sequence. Finally, the low-frequency components are enhanced through residual connection to obtain the enhanced low-frequency features. The processing model of the low-frequency components obtained by the last wavelet transform is: ； ； in, for The first-order difference of time, Indicates the variables The low-frequency components extracted at each moment, Indicates the variables The low-frequency components extracted at each moment, Represents the low-frequency component obtained by the last layer of wavelet transform; Concatenating the high-frequency features and low-frequency features of all levels to obtain the multi-scale feature tensor; The encoder reduces the sequence length by downsampling layer by layer; The encoder model is: ; in, For the layer Samples of time, represents the maximum pooling, is the activation function, represents one-dimensional convolution, Indicates the layer Samples of time, Represents a multi-head sparse attention block.

2. The wind power prediction method based on the WTC-Informer algorithm according to claim 1, characterized in that: Obtain wind turbine operating data, including: Obtain the original operating data of the wind turbine; The raw operating data is preprocessed to obtain preprocessed operating data; wherein the preprocessing includes outlier screening, missing value filling, data standardization, and feature selection.

3. The wind power prediction method based on the WTC-Informer algorithm according to claim 1, characterized in that: Through the ECA channel attention mechanism, weights are dynamically assigned to each feature channel to obtain weighted multi-scale features, including: Adaptive average pooling is performed on each channel of the multi-scale feature tensor to compress the spatial dimension; wherein, Where, For the Adaptive average pooling results for channels, is the sequence number of the moment, For multi-scale features, is the sequence number of the channel; The local channel relationship is captured by one-dimensional convolution to obtain the weight; Where, For weight, represents convolution, is the multi-scale feature after adaptive average pooling, represents transpose; Use the Sigmoid function to normalize the weights, then broadcast the weights to the original spatiotemporal dimensions, perform channel-level scaling, and obtain weighted multi-scale features; where, Where, is the weighted multi-scale feature tensor, represents the channel-by-channel product, Represents the Sigmoid function; Sin-cosine position encoding is used to convert the weighted multi-scale feature timestamps into position information.

4. The wind power prediction method based on the WTC-Informer algorithm according to claim 1, characterized in that: Through the generative decoder, sparse attention is used to extract the temporal pattern, combined with the connection feature map output by the cross-attention fusion encoder, to output the future wind power prediction sequence, specifically including: Input a data sequence into the sparse self-attention module and capture the internal correlation features of the data sequence through the multi-head sparse self-attention mechanism; wherein the data sequence includes the running data sequence and the sequence to be predicted; Inputting the connection feature map into the cross-attention module, using cross-attention to interact the associated features with the connection feature map, integrating the key features captured by the encoder into the prediction generation process of the decoder through the attention mechanism, and obtaining a decoding vector of the hidden representation; The decoded vector is mapped to a wind power prediction result through a fully connected layer to obtain a target sequence.

5. The wind power prediction method based on the WTC-Informer algorithm according to claim 1, characterized in that: The wind power prediction model based on the WTC-Informer algorithm is trained through the following steps: Acquire historical operating data of the wind turbine; wherein the historical operating data includes characteristic data of a plurality of wind turbines; Preprocessing the historical operation data; wherein the preprocessing includes detecting the historical operation data according to a preset abnormal value judgment condition, deleting all abnormal data, and filling missing data with the value of the previous moment; Divide the preprocessed historical running data into training set, validation set and test set, and perform standardization; Perform correlation analysis on the various features of the standardized historical operating data, screen out features with a correlation with wind power higher than a preset value, and obtain the final training set, validation set, and test set; A wind power prediction model based on the WTC-Informer algorithm was constructed and trained using the final training set and a pre-defined loss function to establish a mapping relationship between input features and wind power. The model's performance was evaluated on the validation set after each training cycle. An early stopping strategy was used to terminate training early if performance on the validation set did not improve over several consecutive training cycles to ensure the model's generalization ability. The trained model is evaluated by inputting the final test set. When the model meets the preset standards, the model parameters are saved and the pre-trained wind power prediction model based on the WTC-Informer algorithm is obtained.

6. A wind power prediction device based on the WTC-Informer algorithm, characterized in that: Used to implement a wind power prediction method based on the WTC-Informer algorithm as described in any one of claims 1 to 5; The wind power prediction device comprises: An operation data acquisition module is used to obtain the operation data of the wind turbine; A prediction module is used to input the operating data into a pre-trained wind power prediction model based on the WTC-Informer algorithm, perform data processing through the following units, and obtain a prediction sequence of the active power of the wind turbine; A wavelet transform subunit is used to extract high-frequency components and low-frequency components according to the operating data through wavelet transform and splice them to obtain multi-scale features; The ECA channel attention subunit is used to dynamically assign weights to each feature channel through the ECA channel attention mechanism to obtain weighted multi-scale features; The encoder unit is used to stack multiple sparse self-attention encoding layers into an Informer encoder, which reduces the computational complexity of attention from the standard Transformer down to , obtain the connection feature graph; where, Indicates the time complexity, is the length of the input sequence; The decoder unit is used to extract the temporal pattern using sparse attention through a generative decoder, and combine it with the connection feature map output by the cross-attention fusion encoder to output the future wind power prediction sequence.

7. A wind power prediction device based on the WTC-Informer algorithm, characterized in that: The invention comprises a processor, a memory, and a computer program stored in the memory; the computer program can be executed by the processor to implement a wind power prediction method based on the WTC-Informer algorithm as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the wind power prediction method based on the WTC-Informer algorithm according to any one of claims 1 to 5.

Images