Two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving

Abstract

The invention relates to a two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving. The method comprises the following steps: preprocessing measured data of a wind power plant to obtain a data set; constructing a power prediction sample according to the data set, and performing variational mode decomposition on a standardized power sequence in the sample to obtain an intrinsic mode function; predicting each intrinsic mode function through gating circulation to obtain a power prediction result of the first stage, and constructing an error sequence and an error prediction sample according to the result and real data; the error prediction sample is decomposed through CEEMDAN, a multi-level wavelet channel cross attention mechanism is introduced for modeling, and a total error prediction sequence is obtained; and carrying out dynamic weighting correction on the total prediction error through a fuzzy correction mechanism in combination with meteorological factors to obtain an annual prediction sequence. According to the method, relatively consistent prediction performance is kept under different time scales and different season conditions, and the prediction reliability and interpretability are greatly improved.

Description

TECHNICAL FIELD

[0001] The application relates to a wind power prediction method, in particular to a two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving, and belongs to the technical field of wind power prediction. BACKGROUND

[0002] With the transformation of energy structure, wind power as a clean and renewable energy source has a growing proportion in the power system. However, wind power output is affected by many factors such as meteorological environment, topography, etc., and has strong nonlinear, non-stationary and random fluctuation characteristics. This uncertainty not only increases the difficulty of power grid dispatching and operation, but also puts higher requirements on the high proportion of wind power grid connection and consumption capacity. Therefore, how to improve the accuracy and stability of wind power prediction has become an important research direction and engineering problem in the field of new energy.

[0003] At present, wind power prediction methods mainly include physical methods, statistical methods, machine learning methods and deep learning methods, etc. Physical methods are usually based on atmospheric physical laws, using meteorological data and topographic parameters to simulate wind field distribution and wind turbine response, but they are highly dependent on input data, and the model is complex and the calculation cost is high. Statistical methods establish a prediction model by digging the linear correlation of historical data, but this kind of method is generally based on linear assumption, and it is difficult to accurately describe the complex nonlinear relationship between wind speed and power.

[0004] With the development of machine learning technology, support vector machine (SVM), decision tree, random forest and artificial neural network (ANN) algorithms are widely used in the field of wind power prediction. Related research institutions have proposed a multi-input short-term load prediction model based on random forest algorithm, and further proposed an improved short-term load prediction model (RF-DL-RST) combined with rough set theory. These methods have alleviated the difficulty of nonlinear modeling to some extent, but still have problems such as high dependence on sample quality and insufficient generalization ability.

[0005] In recent years, deep learning models have become a research hotspot due to their strong feature extraction and time series modeling capabilities. Recurrent neural network (RNN) and its variants, such as long short-term memory network (LSTM) and gated recurrent unit (GRU), have shown significant advantages in processing time series data. For example, genetic algorithm is used to optimize the structure of LSTM network to achieve high-precision short-term prediction; or the grey wolf optimization algorithm (GWO) and data compression method are used to improve the wind speed prediction accuracy; in addition, K-means++, snake optimization algorithm and GRU can be combined for photovoltaic power prediction under different weather conditions. These studies show that deep learning can effectively capture the complex dynamic characteristics of wind speed and power, but it still faces problems such as high parameter sensitivity and lack of interpretability.

[0006] Meanwhile, the combination of signal processing and decomposition-reconstruction technology (such as VMD, EMD and CEEMDAN) and deep learning further improves the modeling ability of the model for non-stationary signals. Among them, the short-term wind power prediction model based on the combination of variable mode decomposition (VMD) and ConvLSTM has a significant effect in capturing spatio-temporal features; the decomposition effect of the combination of empirical mode decomposition (EMD) and echo state network (ESN) is better than that of a single decomposition method, and further, the use of CEEMDAN and attention mechanism for feature extraction and noise suppression can also improve the robustness of the prediction results. However, the decomposition-prediction-reconstruction type model usually has a complex structure and large amount of calculation, and the prediction accuracy is still insufficient when facing extreme weather and other sudden conditions.

[0007] In addition, the introduction of physical information neural network (PINN) and its extended structure in the original prediction model can enhance the physical interpretability of the prediction model. Among them, the introduction of multi-kernel pseudo-inverse neural network (MKPINN) can improve the prediction efficiency and accuracy; the combination of physical constraints and deep learning framework can realize high-precision modeling of wind farms and photovoltaic temperatures. This method performs well in mechanism fusion and generalization, but still has problems such as complex model design and difficult parameter optimization.

[0008] As can be seen from the above, in the prior art, the hybrid prediction method is obviously an important research direction for improving the performance of wind power prediction, but there are still problems such as insufficient accuracy, complex model and difficult optimization, and static compensation for only system error often cannot cope with non-stationary conditions and volatility under extreme weather, which not only leads to error accumulation and prediction deviation, but also cannot achieve the technical purposes of simplifying the model and continuous optimization. SUMMARY

[0009] In view of the problems existing in the prior art, the present application provides a two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving. The method realizes multi-scale wind power prediction through the two-stage framework of "main prediction-error compensation", strengthens multi-scale feature extraction through multi-level wavelet channel cross attention (WCXA), and realizes dynamic error compensation through meteorological fuzzy factor adaptive correction (FC). Further, the method mainly focuses on the WCXA attention mechanism and the FC fuzzy correction module, which cooperatively solves the feature extraction and error correction defects of the prior art.

[0010] In order to achieve the above technical purposes, the present application provides a two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving; Step S1, collect the wind farm measured data, and obtain the data set after preprocessing; Step S2, constructing a power prediction sample according to the data set, and performing variational mode decomposition on the normalized power sequence in the sample to obtain an intrinsic mode function with a limited bandwidth; Step S3, predicting each intrinsic mode function through a gated recurrent unit to obtain a power prediction result in the first stage, and constructing an error sequence and an error prediction sample according to the prediction result and the real data; Step S4, decomposing the error prediction sample through CEEMDAN, performing a multi-level wavelet channel cross-attention encoder-decoder GRU time series modeling, and obtaining a total error prediction sequence; Step S5, dynamically weighting and correcting the total prediction error through a fuzzy correction mechanism combined with meteorological factors to obtain a full-year prediction sequence.

[0011] The method provided by the application mainly divides the prediction process into two stages, wherein the first stage is the process of establishing a main prediction model, that is, adopting VMD decomposition + GRU structure to extract the main trend of the wind power time sequence; the second stage is to establish an error compensation model, that is, on the basis of the main prediction model, wavelet multi-scale decomposition + channel cross-attention is adopted to deeply model the error sequence in the main prediction model, so that the high-frequency disturbance and low-frequency structure in the error can be fully extracted and utilized, and the prediction error is dynamically weighted and corrected through the introduction of a meteorological fuzzy factor adaptive correction mechanism, so that the method maintains relatively consistent prediction performance under different time scales and different seasonal conditions, greatly improving the prediction reliability and interpretability.

[0012] As a preferred scheme, the pre-processing process is: Step S1-1, adopting adjacent time average to interpolate the missing power in the wind farm measured data about The calculation process is: Formula 1: ; Step S1-2, dividing the full year into spring, summer, autumn and winter four parts according to the season, and dividing each part of data into training set 1, training set 2 and test set according to 6:1:3; In formula 1, is the wind power value at time is the wind power value at time is the wind power value at time is the wind power value at time is the wind power value at time is the wind power value at time

[0013] It is important to note that wind power exhibits significant seasonal variations. Meteorological conditions, such as temperature, air pressure, wind speed, and humidity, differ considerably across seasons, leading to significant differences in the operating status and output characteristics of wind turbines. If the data from the entire year is directly mixed for modeling, the model will struggle to accurately capture the patterns within each season, resulting in decreased prediction accuracy. Therefore, this invention divides and models the data according to the seasons, enabling the model to adaptively learn for the meteorological characteristics of different seasons, thereby improving overall prediction performance. Spring is from March to May, summer is from June to August, autumn is from September to November, and winter is from December to February.

[0014] As a preferred embodiment, the power prediction sample construction process is as follows: Supervised learning samples are generated using a sliding window. The input features only contain historical standardized power values ​​within the sliding window, and the output is the standardized power values ​​for multiple consecutive steps after the window. Assuming the window length is S and the prediction step size is H, the sample format is: Formula 2: ; Formula 3: ; The process of performing variational mode decomposition on the standardized power sequence in the sample is as follows: separating the multi-scale features of the power to obtain K intrinsic modes (IMFs), and the decomposition objective function and constraints are: Formula 4: ; In equations 2-4: The window for inputting the model has a length of [length missing]. ; The multi-step prediction results output by the model have a length of S is the input window length; H is the prediction step size. Standardized power sequence, For the first One IMF, For its center frequency, This represents the convolution operation; K is the number of modes in the VMD decomposition. Let be the Dirac distribution function.

[0015] Wind power signals exhibit significant nonlinear and nonstationary characteristics, influenced by factors such as sudden wind speed changes, meteorological conditions, and terrain disturbances. Power sequences often simultaneously contain low-frequency trend terms, high-frequency random fluctuations, and mid-frequency transition features. Directly modeling the original power sequence makes it difficult for the model to effectively distinguish the dynamic patterns of different frequency components, leading to decreased prediction accuracy and stability. Therefore, this invention introduces Variational Mode Decomposition (VMD) to adaptively decompose the complex wind power signal into several intrinsic mode functions (IMFs) with finite bandwidth. Each mode corresponds to the characteristic changes of the signal at different time scales. After VMD decomposition, the model can model and predict different mode features separately, and then reconstruct the overall prediction result, thereby significantly improving the modeling accuracy and generalization ability of nonstationary sequences.

[0016] As a preferred approach, the process for obtaining the power prediction results in the first stage is as follows: For each IMF, a GRU predictor is constructed to learn and predict the results of the next H steps, and trained using power prediction samples. The steps are as follows: Step S3-1: The GRU predictor learns the historical dependencies of the IMF through the update gate and the reset gate, and extracts time-series features. The process is as follows: Formula 5: ; Formula 6: ; Formula 7: ; Formula 8: ; Step S3-2: Map the GRU hidden states to H-step predicted values ​​using a fully connected layer, simplifying the structure while maintaining the accuracy of the main trend. The process is as follows: Formula 9: ; Step S3-3: Linearly superimpose all IMF prediction sequences, and then perform inverse normalization to obtain the power prediction result. The process is as follows: Formula 10: ; In equations 5-10, To update the door, To reset the door, In the candidate hidden state, This is the final hidden state. It is a trainable weight matrix that is input into the gating matrix. For bias terms, For the GRU hidden layer dimension, for Activation function This is element-wise multiplication; For the first A predicted sequence of IMFs, Indicates a fully connected layer; This is the power prediction result.

[0017] As a preferred embodiment, the process for constructing the error sequence and error prediction samples is as follows: Through real power With power prediction results The error sequence is obtained. The process of constructing error prediction samples is as follows: Formula 11: ; Formula 12: ; Formula 13: ; In equations 11-13, This represents the actual power. For power prediction results, It is an error sequence; The window for inputting the model has a length of [length missing]. ; The multi-step prediction results output by the model have a length of S is the input window length; H is the prediction step size.

[0018] As a preferred embodiment, the process of time-series modeling of the encoder-decoder GRU with multi-level wavelet channel cross-attention is as follows: Step S4-1, error prediction samples Error IMFs components were obtained through Ceemdan decomposition. ; Step S4-2: Perform discrete wavelet transform on the error sequence to obtain time-frequency features of different frequency bands, and map them to a unified latent space representation to provide multi-layer feature channels for the cross-attention mechanism. Step S4-3: Perform multi-level channel cross-attention modeling, assign weights across error IMF channels and wavelet scales, and further decode after weighted fusion to obtain the final prediction error. Then reconstruct the total error prediction sequence.

[0019] As a preferred embodiment, the process of assigning weights to the cross-error IMF channel and wavelet scale is as follows: Formula 14: ; Formula 15: ; Formula 16: ; Formula 17: ; Formula 18: ; Formula 19: ; Formula 20: ; The weighted fusion process is as follows: Equation 21: ; Equation 22: ; The reconstruction process of the total error prediction sequence is as follows: Equation 23: ; In equations 14-22, Q represents the coding feature. For time step, For feature dimension, For a moment Feature representation; The wavelet decomposition level is denoted as . For the first Each wavelet channel at time Feature representation; These are the projection matrices for the query, key, and value, respectively. The projected query vector, For the first Key and value representation of layer wavelet channels; The feature dimensions are divided into The dimension of attention head; For the first The input of attention head; The attention weight matrix characterizes the correlation between time steps; It is a scaling factor used to stabilize the gradient; For the first The height in Output on each channel; g i For channel output The gating factor For learnable parameters, For bias terms; for The final prediction error is obtained by further processing the data through a decoder. This is the total error prediction sequence. Further optimization, to reduce computational overhead, sets parameter sharing, i.e., all wavelet layers share the same parameters. .

[0020] After initial prediction, the error sequence between the model output and the true value usually still contains nonlinear and non-stationary features not captured by the main model. These residual errors often exhibit complex multi-scale characteristics, making it difficult to achieve ideal results by directly performing secondary modeling. Virtual Dynamics (VMD) performs excellently in modeling the original wind power sequence, achieving stable frequency domain separation and clear mode division, making it suitable for extracting main trends and energy features. However, the error sequence usually has small amplitude, violent fluctuations, and strong randomness, exhibiting obvious nonlinear and non-stationary characteristics. In this case, VMD is sensitive to noise and struggles to fully decompose complex high-frequency disturbances. In contrast, CEEMDAN, by introducing an adaptive white noise ensemble averaging strategy, can effectively suppress mode aliasing while preserving signal details, improving the stability and completeness of the decomposition. It is more suitable for analyzing error signals dominated by noise and with irregular spectral distribution. Therefore, this invention uses CEEMDAN to finely decompose the residuals in the error correction stage, achieving complementarity between trend modeling and stochastic correction, thereby improving the overall prediction accuracy and robustness.

[0021] As a preferred option, the process of dynamically weighting and correcting the total prediction error by combining meteorological factors is as follows: Step S5-1, by integrating the combined effects of wind shear and wind direction diffusion among meteorological factors through a two-dimensional rule matrix, the dynamic correction coefficient is mapped out. Step S5-2: Based on the dynamic correction coefficient, generate a gated correction factor by weighted averaging, so that the compensation range can be adaptively adjusted according to the intensity of the working condition.

[0022] As a preferred embodiment, the process for obtaining the dynamic correction coefficient is as follows: Formula 24: ; Formula 25: ; Equation 26: ; Equation 27: ; Equation 28: ; Equation 29: ; Formula 30: ; The process of adaptively adjusting the compensation range according to the intensity of the working condition is as follows: Equation 31: ; Equation 32: ; Equation 33: ; In equations 24 to 33, For wind speed shear, To spread according to wind direction, and Indicates height , Wind speed at the location; among which and Indicates height , The wind direction at that location Represents modulo operation; and These represent the horizontal mappings of wind speed shear and wind direction diffusion between 0 and 1; Based on the two-dimensional rule matrix, , It is the left boundary. It is a vertex with a membership degree of 1. It is the right boundary; The combined weights under each fuzzy rule, As a weighting adjustment factor, The corrected power prediction results; This represents the wind speed shear at point i. Indicate the wind direction diffusion at point i; For fuzzy membership functions; Let l represent the fuzzy membership degree of the wind speed shear at point l; Let m be the fuzzy membership degree of the wind speed shear at point m; Let h be the fuzzy membership degree of the wind speed shear at point h; This is the total error prediction sequence.

[0023] In wind power forecasting, model errors are often not independent random disturbances, but rather exhibit significant operating condition dependence and seasonality. When wind shear increases or wind direction diffusion intensifies, flow field stability decreases and turbulence intensifies, leading to changes in the residual structure and amplitude. Traditional linear correction or hard thresholding methods are difficult to adapt to different operating conditions and may even introduce abrupt changes and boundary sensitivity issues.

[0024] To address this, this invention employs a fuzzy rule-based approach for adaptive error correction. It represents "low-medium-high" operating conditions with continuous weights through membership functions, ensuring a smooth transition in correction as operating conditions change. A two-dimensional rule matrix is ​​used to synthesize the combined effects of wind shear and wind direction diffusion, mapping out dynamic correction coefficients. Finally, a weighted average is used to generate a gated correction factor, allowing the compensation magnitude to adaptively adjust with the intensity of the operating condition. Compared to black-box models, fuzzy rules offer a clear structure, interpretable parameters, ease of incorporating domain knowledge, and effectively reduce the risk of overfitting.

[0025] As a preferred embodiment, the process of obtaining the full-year forecast sequence is as follows: the power forecast results obtained in spring, summer, autumn, and winter are concatenated to obtain the sequence, and the formula is: Equation 34: ; in, For the full year forecast series, For spring power forecast results, The power forecast results for summer. For the autumn power forecast results, This is the power forecast result for winter.

[0026] Compared with the prior art, the beneficial technical effects of the technical solution provided by the present invention are as follows: 1) The method provided by this invention achieves a comprehensive improvement in prediction stability, model efficiency and engineering deployability through a two-stage collaborative framework consisting of "lightweight master forecast + multi-scale error compensation + meteorological fuzz correction". This structure can maintain relatively consistent prediction performance under different time scales and different seasonal conditions, greatly improve prediction reliability and interpretability, and has significant engineering applicability and wide applicability.

[0027] 2) In the technical solution provided by the present invention, the “VMD decomposition + GRU structure” is adopted in the first stage and used as the main prediction model. It has the advantages of being lightweight, efficient and suitable for real-time deployment on site. Among them, since VMD can effectively separate the main trend components of the power sequence and GRU can stably model the time series relationship, the combination of the two can accurately capture the overall trend of wind power change with low model complexity, providing a stable and reliable basic prediction result for subsequent error compensation.

[0028] 3) In the technical solution provided by this invention, in the second stage, the error sequence in the main prediction model is deeply modeled by “wavelet multi-scale decomposition + channel cross attention”, so that high-frequency disturbances and low-frequency structures in the error can be fully extracted and utilized; among them, the cross attention mechanism can automatically identify the importance of different error channels, dynamically allocate feature weights, and realize the emphasis on key error information, thereby improving the fine-grained characterization of error compensation, so that the final prediction result can maintain better stability under the condition of drastic fluctuations or complex weather changes.

[0029] 4) The technical solution provided by this invention introduces a meteorological fuzzy correction mechanism based on wind shear and wind direction diffusion. Through continuous membership functions and rule reasoning, the error compensation weight is automatically adjusted, so that the correction process can change smoothly with meteorological conditions. This avoids the over- or under-correction caused by the traditional fixed coefficient method when the conditions change abruptly. In addition, since the fuzzy rules are based on the physical mechanism of the wind field, this correction method has good interpretability, which is conducive to engineers' understanding and optimization, and improves the adaptability and robustness of the model under complex meteorological conditions. Attached Figure Description

[0030] Figure 1This is a schematic diagram of the overall flow of the prediction method provided in Embodiments 1-3 of the present invention; Figure 2 This is a schematic diagram of the sliding window principle in the prediction methods provided in Embodiments 1-3 of the present invention; Figure 3 This is a schematic diagram of the GRU model structure in the prediction methods provided in Embodiments 1-3 of the present invention; Figure 4 The flowchart of CEEMDAN in the prediction method provided in Embodiments 1-3 of this invention is shown below; Figure 5 This is a schematic diagram of the multi-level channel cross-attention mechanism in the prediction methods provided in Embodiments 1-3 of the present invention; Where n is the input sequence Length, For the input sequence Length, The dimension of the input feature. For the dimension of the query vector, Let be the dimension of the key vector. The dimension of the value vector. To represent the weight matrix, The projection weight of the query. The projection weights of the Key (the first one) (size) Projection weights for Value (the first) (size) For querying the matrix, The key matrix (the first one) (size) Value matrix (the first) (size) Let Q be the dot product of Q and K, and Softmax be the normalization function. Detailed Implementation

[0031] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. Through these descriptions, the features and advantages of the present invention will become clearer and more apparent. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them.

[0032] The implementation environment of the embodiments provided by this invention is as follows: Hardware: Intel Core i5-13400KF processor, 32GB DDR4 memory, NVIDIA GTX 5060 graphics card (8GB VRAM); Software: Python 3.9, PyTorch 2.7.1, PyWavelets 1.3.0.

[0033] Embodiments 1-3 of this invention all employ a two-stage multi-step wind power prediction method driven by meteorological fuzzy factors, specifically including the following process: Step S1: Collect measured data from the wind farm, preprocess it, and obtain the dataset; Step S2: Construct power prediction samples based on the dataset, and perform variational mode decomposition on the standardized power sequences in the samples to obtain eigenmode functions with finite bandwidth. Step S3: Predict each intrinsic mode function through a gated cyclic unit to obtain the power prediction result of the first stage, and construct an error sequence and error prediction sample based on the prediction result and the real data. Step S4: Decompose the error prediction samples using CEEMDAN, perform encoder-decoder GRU time-series modeling with multi-level wavelet channel cross-attention, and obtain the total error prediction sequence. Step S5: Through the fuzzy correction mechanism, the total forecast error is dynamically weighted and corrected in combination with meteorological factors to obtain the full-year forecast sequence.

[0034] As a preferred option, the measured data from wind farms includes wind power series and meteorological factor series; among which, the wind power series... Sampling interval 15 min, N is the total number of samples; meteorological factor sequence: wind speed at 10m altitude 50m height wind speed 10m height wind direction 50m height wind direction .

[0035] As a preferred embodiment, the preprocessing procedure is as follows: Step S1-1: Use the average value of adjacent time points to analyze the measured data of the wind farm regarding... The missing power at a given moment is interpolated, and the calculation process is as follows: Formula 1: ; Step S1-2: Divide the measured data of the wind farm into four parts according to the season: spring, summer, autumn and winter. Then divide the data of each part into training set 1, training set 2 and test set according to 7:2:1. In Equation 1, It is a moment The wind power output value; It is a moment The wind power output value; It is a moment The wind power values ​​are shown. Training set 1 is used for the first stage of modeling, and training set 2 is used for the second stage of modeling.

[0036] As a preferred embodiment, the power prediction sample construction process is as follows: Supervised learning samples are generated using a sliding window. The input features only contain historical standardized power values ​​within the sliding window, and the output is the standardized power values ​​for multiple consecutive steps after the window. Assuming the window length is S and the prediction step size is H, the sample format is: Formula 2: ; Formula 3: ; The process of performing variational mode decomposition on the standardized power sequence in the sample is as follows: separating the multi-scale features of the power to obtain K intrinsic modes (IMFs), and the decomposition objective function and constraints are: Formula 4: ; In equations 2-4: The window for inputting the model has a length of [length missing]. ; The multi-step prediction results output by the model have a length of S is the input window length; H is the prediction step size. Standardized power sequence, For the first One IMF, For its center frequency, This represents the convolution operation; K is the number of modes in the VMD decomposition. Let be the Dirac distribution function.

[0037] As a preferred approach, the process for obtaining the power prediction results in the first stage is as follows: For each IMF, a GRU predictor is constructed to learn and predict the results of the next H steps, and trained using power prediction samples. The steps are as follows: Step S3-1: The GRU predictor learns the historical dependencies of the IMF through the update gate and the reset gate, and extracts time-series features. The process is as follows: Formula 5: ; Formula 6: ; Formula 7: ; Formula 8: ; Step S3-2: Map the GRU hidden states to H-step predicted values ​​using a fully connected layer, simplifying the structure while maintaining the accuracy of the main trend. The process is as follows: Formula 9: ; Step S3-3: Linearly superimpose all IMF prediction sequences, and then perform inverse normalization to obtain the power prediction result. The process is as follows: Formula 10: ; In equations 5-10, To update the door, To reset the door, In the candidate hidden state, This is the final hidden state. It is a trainable weight matrix that is input into the gating matrix. For bias terms, For the GRU hidden layer dimension, for Activation function This is element-wise multiplication; For the first A predicted sequence of IMFs, Indicates a fully connected layer; This is the power prediction result.

[0038] As a preferred embodiment, the process for constructing the error sequence and error prediction samples is as follows: Through real power With power prediction results The error sequence is obtained. The process of constructing error prediction samples is as follows: Formula 11: ; Formula 12: ; Formula 13: ; In equations 11-13, This represents the actual power. For power prediction results, It is an error sequence; The window for inputting the model has a length of [length missing]. ; The multi-step prediction results output by the model have a length of S is the input window length; H is the prediction step size.

[0039] As a preferred embodiment, the process of time-series modeling of the encoder-decoder GRU with multi-level wavelet channel cross-attention is as follows: Step S4-1, error prediction samples Error IMFs components were obtained through Ceemdan decomposition. ; Step S4-2: Perform discrete wavelet transform on the error sequence to obtain time-frequency features of different frequency bands, and map them to a unified latent space representation to provide multi-layer feature channels for the cross-attention mechanism.

[0040] Step S4-3: Perform multi-level channel cross-attention modeling, assign weights across error IMF channels and wavelet scales, and further decode after weighted fusion to obtain the final prediction error. Then reconstruct the total error prediction sequence.

[0041] As a preferred embodiment, the process of assigning weights to the cross-error IMF channel and wavelet scale is as follows: Formula 14: ; Formula 15: ; Formula 16: ; Formula 17: ; Formula 18: ; Formula 19: ; Formula 20: ; The weighted fusion process is as follows: Equation 21: ; Equation 22: ; The reconstruction process of the total error prediction sequence is as follows: Equation 23: ; In equations 14-22, Q represents the coding feature. For time step, For feature dimension, For a moment Feature representation; The wavelet decomposition level is denoted as . For the first Each wavelet channel at time Feature representation; These are the projection matrices for the query, key, and value, respectively. The projected query vector, For the first Key and value representation of layer wavelet channels; The feature dimensions are divided into The dimension of attention head; For the first The input of attention head; The attention weight matrix characterizes the correlation between time steps; It is a scaling factor used to stabilize the gradient; For the first The height in Output on each channel; g i For channel output The gating factor For learnable parameters, For bias terms; for The final prediction error is obtained by further processing the data through a decoder. This is the total error prediction sequence.

[0042] As a preferred option, the process of dynamically weighting and correcting the total prediction error by combining meteorological factors is as follows: Step S5-1, by integrating the combined effects of wind shear and wind direction diffusion among meteorological factors through a two-dimensional rule matrix, the dynamic correction coefficient is mapped out. Step S5-2: Based on the dynamic correction coefficient, generate a gated correction factor by weighted averaging, so that the compensation range can be adaptively adjusted according to the intensity of the working condition.

[0043] As a preferred embodiment, the process for obtaining the dynamic correction coefficient is as follows: 1. Meteorological factors: wind speed shear and wind direction diffusion The calculation is as follows: Formula 24: ; Formula 25: ; 2. Using functions Map the shear and diffusion levels to a range of 0-1.

[0044] Equation 26: ; 3. Multiply the shear level and wind direction diffusion level by the defined modified intensity base matrix. This results in... Basic two-dimensional rule matrix : Equation 27: ; 4. Based on quantiles (Q10 / Q50 / Q90) and trigonometric membership functions (Equation 22), wind shear and wind direction diffusion are classified into low / medium / high levels: Equation 28: ; For wind shear, the classification is defined as follows: Equation 29: ; For wind direction diffusion, the division is defined as follows: Formula 30: ; The process of adaptively adjusting the compensation range according to the intensity of the working condition is as follows: Equation 31: ; Equation 32: ; Equation 33: ; In equations 24 to 33, For wind speed shear, To spread according to wind direction, and Indicates height , Wind speed at the location; among which and Indicates height , The wind direction at that location Represents modulo operation; and These represent the horizontal mappings of wind speed shear and wind direction diffusion between 0 and 1; Based on the two-dimensional rule matrix, , It is the left boundary. It is a vertex with a membership degree of 1. It is the right boundary; The combined weights under each fuzzy rule, As a weighting adjustment factor, The corrected power prediction results; This represents the wind speed shear at point i. Indicate the wind direction diffusion at point i; For fuzzy membership functions; Let l represent the fuzzy membership degree of the wind speed shear at point l; Let m be the fuzzy membership degree of the wind speed shear at point m; Let h be the fuzzy membership degree of the wind speed shear at point h; This is the total error prediction sequence.

[0045] As a preferred embodiment, the process of obtaining the full-year forecast sequence is as follows: the power forecast results obtained in spring, summer, autumn, and winter are concatenated to obtain the sequence, and the formula is: Equation 34: ; in, For the full year forecast series, For spring power forecast results, The power forecast results for summer. For the autumn power forecast results, This is the power forecast result for winter.

[0046] Example 1

[0047] The dataset consists of 70,176 samples, with a time scale of 15 minutes, using power data from a 99MW wind farm in Inner Mongolia from 2019 to 2020. Wind speed and direction data were also collected simultaneously. The parameters were set as follows: MD parameter K=5, learning rate 0.001, batch size 32, number of training epochs 500; WCXA, wavelet scale 3.

[0048] Training was performed using the above prediction method. During sample generation, S=24 and H=1. The prediction results after 15 minutes (1-step) showed that the prediction method had NRMSE=1.69 and NMAE=0.98, which were 6.31% and 4.20% lower than the existing GRU method, respectively.

[0049] Example 2

[0050] This embodiment is exactly the same as Embodiment 1, except that S=24 and H=4 in the sample generation process. The prediction results after 1 hour (4-step) show that NRMSE=2.66 and NMAE=1.65, which are reduced by 9.11% and 6.71% respectively compared with the GRU method in the prior art.

[0051] Example 3

[0052] This embodiment is exactly the same as Embodiment 1, except that S=48 and H=16 in the sample generation process. In the 4h (16-step) prediction results, NRMSE=5.10 and NMAE=3.27, which are 10.10% and 8.75% lower than the GRU method in the prior art, respectively.

[0053] The present invention compares the effects of Examples 1-3 with those of common prediction methods in the prior art, and the results are shown in Table 1.

[0054] ; Table 1 shows the two-stage predictions. The first stage uses GRU, and the table indicates the method used in the second stage. CA represents channel attention, XA represents cross-attention, WCXA represents multi-level wavelet channel cross-attention, and FC represents fuzzy adaptive correction. The prediction results in Table 1 are from ablation experiments. For different steps of the experiment, the values ​​on the left are the evaluation results, and the values ​​on the right are the percentage improvement compared to GRU. As shown in Table 1, the prediction method provided by this invention has significant advantages over existing prediction methods. Under the same computing power, the evaluation result improvement rate can reach up to 10.1%.

Claims

1. A two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving, characterized in that: Step S1, collect the measured data of the wind farm, and obtain a data set after preprocessing; Step S2, construct a power prediction sample according to the data set, and perform variational mode decomposition on the normalized power sequence in the sample to obtain intrinsic mode functions with limited bandwidth; Step S3, predict each intrinsic mode function through a gated recurrent unit to obtain a power prediction result in the first stage, and construct an error sequence and an error prediction sample according to the prediction result and the true data; Step S4, decompose the error prediction sample through CEEMDAN, perform multi-level wavelet channel cross-attention encoding-decoding GRU time series modeling, and obtain a total error prediction sequence; Step S5, dynamically weight and correct the total prediction error by combining meteorological factors through a fuzzy correction mechanism to obtain a full-year prediction sequence. 2.The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 1, characterized in that: The preprocessing process is as follows: Step S1-1: Use the average value of adjacent time points to analyze the measured data of the wind farm regarding... The missing power at a given moment is interpolated, and the calculation process is as follows: Formula 1: ; Step S1-2, divide the full year into spring, summer, autumn and winter according to seasons, and divide each part of data into training set 1, training set 2 and test set according to 6:1:3:1; in formula 1, is the wind power value at time ; is the wind power value at time ; is the wind power value at time . 3.The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 2, characterized in that: The construction process of the power prediction sample is as follows: a sliding window is used to generate a supervised learning sample, the input features only include historical normalized power values in the sliding window, and the output is the normalized power value in the next continuous steps, assuming that the window length is S, the prediction step is H, and the sample format is: Formula 2: ; Formula 3: ; The process of variational mode decomposition on the normalized power sequence in the sample is as follows: the multi-scale features of the power are separated to obtain K intrinsic modes IMFs, and the target function and the constraint condition are: Formula 4: ; In formulas 2~4: is the window of input model, length of ; is the multi-step prediction result of model output, length of ; S is the input window length; H is the prediction step; is the normalized power sequence, is the th IMF, is the center frequency of represents the convolution operation; K is the modal number of VMD decomposition; is the Dirac distribution function.

4. The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 3, characterized in that: The process of obtaining the power prediction result in the first stage is as follows: for each IMF, a GRU predictor is constructed to learn the result of the future H steps, and the predictor is trained through the power prediction sample, and the steps are as follows: Step S3-1, the GRU predictor learns the historical dependence of the IMF through the update gate and the reset gate to extract the time series features, and the process is as follows: Formula 5: ; Formula 6: ; Formula 7: ; Formula 8: ; Step S3-2, the GRU hidden state is mapped to the H-step prediction value through a fully connected layer to simplify the structure while ensuring the accuracy of the main trend, and the process is as follows: Formula 9: ; Step S3-3, linearly superimpose all IMF prediction sequences, and then restore them through inverse normalization to obtain the power prediction result, and the process is as follows: Formula 10: ; in formulas 5-10, is an update gate, is a reset gate, is a candidate hidden state, is a final hidden state, is a trainable input-to-gate matrix weight matrix, is a bias term, is a GRU hidden layer dimension, is an activation function, is an element-wise multiplication; is a predicted sequence of the th IMF, denotes a fully connected layer; is a power prediction result.

5. The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 4, characterized in that: The construction process of the error sequence and the error prediction sample is as follows: By real power With power prediction results , get error sequence And build error prediction sample structure, the process is: Formula 11: ; Formula 12: ; Formula 13: ; In formulas 11-13, is the true power, is the power prediction result, is the error sequence; is the input window of the model, with a length of ; is the multi-step prediction result of the model output, with a length of ; S is the input window length; H is the prediction step.

6. The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 5, characterized in that: The process of the fusion multi-level wavelet channel cross-attention encoder-decoder GRU time series modeling is: step S4-1, error prediction sample The error IMF component obtained by CEEMDAN decomposition ; Step S4-2, perform discrete wavelet transform on the error sequence to obtain time-frequency features in different frequency bands, and map them to a unified hidden space representation to provide multiple feature channels for the cross-attention mechanism; Step S4-3, perform multi-level channel cross-attention modeling, assign weights to the error IMF channels and wavelet scales, further decode after weighted fusion, obtain the final prediction error, and reconstruct the total error prediction sequence.

7. The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 6, characterized in that: The process of assigning weights to the error IMF channels and wavelet scales is as follows: Formula 14: ; Formula 15: ; Formula 16: ; Formula 17: ; Formula 18: ; Formula 19: ; Formula 20: ; The process of weighted fusion is as follows: Formula 21: ; Formula 22: ; The reconstruction process of the total error prediction sequence is as follows: Formula 23: ; Q is the encoding feature in formula 14~22, is the time step, is the feature dimension, is the feature representation at time ; is the wavelet decomposition level, is the feature representation of the th wavelet channel at time ; are the projection matrices of query, key and value respectively; is the projected query vector, is the key and value representation of the th wavelet channel; is the feature dimension divided into attention heads; is the input of the th attention head; is the attention weight matrix, which characterizes the correlation between time steps; is the scaling factor to stabilize the gradient; is the output of the th head on the th channel;g i is the gating factor of the channel output ; is the learnable parameter, is the bias term; is the further decoded by the decoder to obtain the final prediction error, is the total error prediction sequence. 8.The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 5, characterized in that: The process of dynamically weighting and correcting the total prediction error by combining meteorological factors is as follows: Step S5-1, the combined effect of wind shear and wind direction diffusion in the meteorological factors is mapped into a dynamic correction coefficient through a two-dimensional regular matrix; Step S5-2, generating a gating correction factor by weighted average according to the dynamic correction coefficient, so that the compensation amplitude is adaptively adjusted according to the working condition intensity.

9. The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 8, characterized in that: The acquisition process of the dynamic correction coefficient is as follows: Formula 24: ; Formula 25: ; Formula 26: ; Formula 27: ; Formula 28: ; Formula 29: ; Formula 30: ; The process of adaptively adjusting the compensation amplitude according to the working condition intensity is as follows: Formula 31: ; Formula 32: ; Formula 33: ; in the formulas 24~33, is the wind speed shear, is the wind direction spread, and represents the height , of the wind speed at the point; wherein and represents the height , of the wind direction at the point, represents the modulo operation; and are respectively the horizontal mapping of the wind speed shear and the wind direction spread between 0~1; is the basic two-dimensional regular matrix, , is the left boundary, is the vertex with membership degree 1, is the right boundary; is the combined weight under each fuzzy rule, is the weight adjustment factor, is the corrected power prediction result; represents the wind speed shear at the point i, represents the wind direction spread at the point i; is the fuzzy membership function; is the fuzzy membership of the wind speed shear at l; is the fuzzy membership of the wind speed shear at m; is the fuzzy membership of the wind speed shear at h; is the total error prediction sequence.

10. The two-stage wind power multi-step prediction method based on meteorological fuzzy factor driving according to claim 9, characterized in that: The acquisition process of the full-year prediction sequence is as follows: the power prediction results of spring, summer, autumn and winter are spliced, and the formula is as follows: The acquisition process of the dynamic correction coefficient is as follows: Formula 34: ; wherein, is the full year forecast sequence, is the power forecast result for spring, is the power forecast result for summer, is the power forecast result for autumn, is the power forecast result for winter.

Images