Wind power prediction method and device and storage medium

Abstract

The invention provides a wind power prediction method and device and a storage medium, and belongs to the technical field of power prediction.The method comprises the steps that multiple pieces of original wind power plant operation data are imported, the original wind power plant operation data are preprocessed, and preprocessed wind power plant operation data are obtained; performing normalization processing on the preprocessed wind power plant operation data to obtain an original wind power time sequence; and performing modal component analysis on each original wind power time sequence to obtain a residual component, a target high-frequency intrinsic mode function, a target intermediate-frequency intrinsic mode function and a target low-frequency intrinsic mode function. According to the method, the calculation efficiency and the solution quality are effectively improved, the problem of high volatility of wind power data is solved, the frequency characteristics are improved, the influence of secondary decomposition is reduced, the noise influence is eliminated, the prediction performance of the model is improved, and more representative characteristic components can be extracted.

Description

Technical Field

[0001] This invention relates to the field of power prediction technology, specifically to a wind power prediction method, device, and storage medium. Background Technology

[0002] Currently, existing wind power forecasting methods, when addressing the non-stationarity and multi-scale characteristics of wind power series, decompose the original sequence into several physically meaningful intrinsic mode functions (IMFs) using signal decomposition techniques. This effectively suppresses noise and mode aliasing, enhancing the interpretability of the input data and modeling performance. However, the nonlinear disturbances and strong noise in wind power time series are mainly concentrated in high-frequency components. The high-frequency IMFs after preliminary decomposition still contain a large number of non-stationary components, while the contribution of low-frequency IMFs to the prediction is limited.

[0003] To further suppress noise and effectively extract latent features, conventional methods typically involve secondary decomposition of the high-frequency IMF. For example, after initially reducing the nonlinearity of the sequence using wavelet transform (WT), variational mode decomposition (VMD) can be used to extract high-frequency mode functions and residual signals, effectively improving the predictability of the high-frequency component. Applying Local Mean Decomposition (LMD) to the high-frequency IMF using CEEMDAN and VMD effectively enhances the model's ability to identify non-stationary signals. However, VMD still faces two key problems in practical applications: first, its noise reduction effect is affected by parameters such as the mode number λ and the penalty coefficient φ, and it relies on manual settings, lacking adaptive capabilities and easily leading to improper mode number settings or frequency domain aliasing, affecting the decomposition quality; second, when facing wind power data with strong nonlinearity and high dimensionality, conventional optimization algorithms are prone to getting trapped in local optima during VMD parameter search, making it difficult to obtain the global optimum. Therefore, the existing solution is to try to introduce intelligent optimization algorithms to improve the parameter optimization capability of VMD, such as using whale optimization algorithm and butterfly optimization algorithm to optimize the modality number 𝐾 and penalty coefficient 𝛼, but there are problems such as slow convergence speed and getting stuck in local optima. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a wind power prediction method, device and storage medium to address the shortcomings of the prior art.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: A wind power prediction method, comprising the following steps: Import multiple original wind farm operation data, and preprocess each of the original wind farm operation data to obtain preprocessed wind farm operation data corresponding to each of the original wind farm operation data. Normalize the preprocessed wind farm operation data of each wind farm to obtain the original wind power time series corresponding to the original wind farm operation data of each wind farm. Modal component analysis is performed on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. A prediction model is constructed, and the prediction model is used to predict and analyze each of the residual components, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The analysis results are used as the wind power prediction results.

[0006] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: A wind power prediction device, comprising: The import module is used to import multiple raw wind farm operation data. The preprocessing module is used to preprocess the original wind farm operation data of each wind farm to obtain preprocessed wind farm operation data corresponding to the original wind farm operation data of each wind farm. The normalization processing module is used to normalize the preprocessed wind farm operation data of each wind farm to obtain the original wind power time series corresponding to the original wind farm operation data. The modal component analysis module is used to perform modal component analysis on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The prediction result acquisition module is used to construct a prediction model. The prediction model is used to perform prediction analysis on each of the residual components, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The analysis results are used as the wind power prediction results.

[0007] Based on the above-mentioned wind power prediction method, the present invention also provides a wind power prediction system.

[0008] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a wind power prediction system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the wind power prediction method as described above.

[0009] Based on the above-mentioned wind power prediction method, the present invention also provides a computer-readable storage medium.

[0010] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the wind power prediction method as described above.

[0011] The beneficial effects of this invention are as follows: Preprocessing the original wind farm operation data yields preprocessed wind farm operation data; normalizing the preprocessed wind farm operation data yields the original wind power time series; modal component analysis of the original wind power time series yields residual components, target high-frequency intrinsic mode functions (IMFs), target mid-frequency IMFs, and target low-frequency IMFs; prediction analysis of the residual components, target high-frequency IMFs, target mid-frequency IMFs, and target low-frequency IMFs is performed using a prediction model, and the analysis results are used as wind power prediction results. This effectively improves computational efficiency and solution quality, solves the problem of high volatility in wind power data, enhances frequency characteristics, reduces the impact of secondary decomposition, eliminates noise effects, improves the predictive performance of the model, and enables the extraction of more representative feature components. Attached Figure Description

[0012] Figure 1 A flowchart illustrating the wind power prediction method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the decomposition process of the high-frequency intrinsic mode function to be processed in the wind power prediction method provided in the embodiments of the present invention; Figure 3 A schematic diagram of the BiLSTM model of the wind power prediction method provided in the embodiments of the present invention; Figure 4 This is a block diagram of a wind power prediction device provided in an embodiment of the present invention. Detailed Implementation

[0013] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0014] Figure 1 This is a flowchart illustrating a wind power prediction method provided in an embodiment of the present invention.

[0015] like Figure 1 As shown, a wind power prediction method includes the following steps: S1: Import multiple raw wind farm operation data, preprocess each of the raw wind farm operation data to obtain preprocessed wind farm operation data corresponding to each of the raw wind farm operation data; S2: Normalize the preprocessed wind farm operation data of each wind farm to obtain the original wind power time series corresponding to the original wind farm operation data of each wind farm. S3: Perform modal component analysis on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. S4: Construct a prediction model, and use the prediction model to predict and analyze each of the residual components, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and use the analysis results as the wind power prediction results.

[0016] In the above embodiments, preprocessing of the original wind farm operation data yields preprocessed wind farm operation data. Normalization of the preprocessed wind farm operation data yields the original wind power time series. Modal component analysis of the original wind power time series yields residual components, target high-frequency intrinsic mode functions (IMFs), target mid-frequency IMFs, and target low-frequency IMFs. Prediction analysis of the residual components, target high-frequency IMFs, target mid-frequency IMFs, and target low-frequency IMFs is performed using a prediction model. The analysis results are then used as wind power prediction results. This effectively improves computational efficiency and solution quality, addresses the high volatility of wind power data, enhances frequency characteristics, reduces the impact of secondary decomposition, eliminates noise, improves the model's prediction performance, and enables the extraction of more representative feature components.

[0017] Optionally, as an embodiment of the present invention, the process of normalizing the preprocessed wind farm operation data to obtain the original wind power time series corresponding to the original wind farm operation data includes: The preprocessed wind farm operation data are normalized using the maximum-minimum normalization algorithm to obtain the original wind power time series corresponding to the original wind farm operation data.

[0018] It should be understood that the maximum-minimum normalization method (i.e., the maximum-minimum normalization algorithm) is used for processing.

[0019] In the above embodiments, the preprocessed wind farm operation data are normalized to obtain the original wind power time series, thus eliminating the influence of data dimensions.

[0020] Optionally, as an embodiment of the present invention, the process of performing modal component analysis on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operating data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operating data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operating data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operating data includes: Each of the original wind power time series is decomposed to obtain the residual components corresponding to the operation data of each of the original wind farms and multiple original intrinsic mode functions corresponding to the operation data of each of the original wind farms. Based on each of the original wind power time series, the multiple original intrinsic mode functions corresponding to the original wind farm operation data are classified to obtain the high-frequency intrinsic mode function to be processed, the target mid-frequency intrinsic mode function, and the target low-frequency intrinsic mode function corresponding to each of the original wind farm operation data. Each of the high-frequency intrinsic mode functions to be processed is decomposed to obtain the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data.

[0021] In the above embodiments, modal component analysis is performed on each original wind power time series to obtain residual components, target high-frequency intrinsic mode functions, target mid-frequency intrinsic mode functions, and target low-frequency intrinsic mode functions. This effectively improves computational efficiency and solution quality, solves the problem of high volatility in wind power data, enhances frequency characteristics, reduces the impact of secondary decomposition, eliminates noise effects, improves the predictive performance of the model, and enables the extraction of more representative feature components.

[0022] Optionally, as an embodiment of the present invention, the process of decomposing each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data and the multiple original intrinsic mode functions corresponding to each of the original wind farm operation data includes: Preset Gaussian white noise is input into each of the original wind power time series to obtain the noisy wind power time series corresponding to the operation data of each of the original wind farms; Each of the noise-added wind power time series is decomposed to obtain the residual components corresponding to each of the original wind farm operation data and multiple original intrinsic mode functions corresponding to each of the original wind farm operation data.

[0023] It should be understood that the original data (i.e., the original wind power time series) is initially decomposed using the improved ensemble empirical mode decomposition (ICEEMDAN) to obtain multiple intrinsic mode functions (IMFs) (i.e., original intrinsic mode functions) and residual components Res.

[0024] Specifically, Gaussian white noise (i.e., preset Gaussian white noise) is added to the wind power time series to be decomposed (i.e., the original wind power time series) to form a new series (i.e., the wind power time series after noise addition); based on the local mean of the generated signal, the residual components and modal components (i.e., the original intrinsic mode functions) are calculated; using the same method, all modal components (i.e., the original intrinsic mode functions) and residuals (i.e., residual components) are repeatedly calculated iteratively.

[0025] In the above embodiments, each original wind power time series is decomposed to obtain residual components and original intrinsic mode functions, which solves the problem of high volatility in wind power data and improves frequency characteristics.

[0026] Optionally, as an embodiment of the present invention, the process of classifying multiple original intrinsic mode functions corresponding to the operating data of each original wind farm according to each original wind power time series, and obtaining the high-frequency intrinsic mode function to be processed corresponding to each original wind farm operating data, the target mid-frequency intrinsic mode function corresponding to each original wind farm operating data, and the target low-frequency intrinsic mode function corresponding to each original wind farm operating data includes: Sample entropy is calculated for each of the original wind power time series and multiple original intrinsic mode functions corresponding to the operation data of each of the original wind farms, to obtain multiple first sample entropies corresponding to the operation data of each of the original wind farms. Based on the K-means clustering algorithm and the entropy of multiple first samples corresponding to the original wind farm operation data, the original intrinsic mode functions corresponding to the original wind farm operation data are divided into multiple original high-frequency intrinsic mode functions, multiple original mid-frequency intrinsic mode functions, and multiple original low-frequency intrinsic mode functions corresponding to the original wind farm operation data. The original high-frequency intrinsic mode functions corresponding to the original wind farm operation data are merged and processed to obtain the high-frequency intrinsic mode functions to be processed corresponding to the original wind farm operation data. The original mid-frequency intrinsic mode functions corresponding to the original wind farm operation data are merged to obtain the target mid-frequency intrinsic mode function corresponding to the original wind farm operation data. The original low-frequency intrinsic mode functions corresponding to the original wind farm operation data are merged to obtain the target low-frequency intrinsic mode function corresponding to the original wind farm operation data.

[0027] It should be understood that sample entropy (SE) is a metric used to assess the complexity of a time series. Introducing SE as an evaluation metric assesses the complexity of a wind power time series and thus determines the magnitude of noise. The SE values ​​of each subsequence obtained from the decomposition (i.e., the original intrinsic mode functions) and the original sequence (i.e., the original wind power time series) are measured to determine the complexity of the modal components.

[0028] Specifically, using the K-means clustering algorithm, the subsequences (i.e., the original intrinsic mode functions) are divided into three categories: high frequency (i.e., the original high-frequency intrinsic mode functions), mid-frequency (i.e., the original mid-frequency intrinsic mode functions), and low frequency (i.e., the original low-frequency intrinsic mode functions). The sequences of each category are then directly superimposed to generate three new modal components, denoted as high frequency (i.e., the high-frequency intrinsic mode functions to be processed), mid-frequency (i.e., the target mid-frequency intrinsic mode functions), and low-frequency IMFs (i.e., the target low-frequency intrinsic mode functions).

[0029] Specifically, the modal components (i.e., the original intrinsic mode functions) are classified and defined as high-frequency (i.e., the high-frequency intrinsic mode functions to be processed), mid-frequency (i.e., the target mid-frequency intrinsic mode functions), and low-frequency IMFs (i.e., the target low-frequency intrinsic mode functions). Although the sample entropy of high-frequency IMFs is lower than that of the original sequence, their noise is large and their fluctuations are large, indicating that their internal structure is the most complex and that there are still high-frequency noise components that have not been completely eliminated. Mid-frequency and low-frequency IMFs meet the requirements of the prediction model and can be directly used for prediction.

[0030] In the above embodiments, the original intrinsic mode functions are classified according to the original wind power time series to obtain the high-frequency intrinsic mode function to be processed, the target mid-frequency intrinsic mode function, and the target low-frequency intrinsic mode function. This solves the problem of high volatility in wind power data, improves frequency characteristics, reduces the impact of secondary decomposition, eliminates noise, improves the predictive performance of the model, and can extract more representative feature components.

[0031] Optionally, as an embodiment of the present invention, such as Figure 2 As shown, the process of decomposing each of the high-frequency intrinsic mode functions to be processed to obtain the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data includes: S331: Import the original PLO parameters and randomly initialize the original PLO parameters to obtain the initialized PLO parameters; S332: Initialize the particle swarm, and decompose the high-frequency intrinsic mode function to be processed according to the initialized PLO parameters and the particle parameters in the initialized particle swarm to obtain multiple decomposed high-frequency intrinsic mode function components corresponding to each of the original wind farm operation data. S333: Calculate the sample entropy of each of the decomposed high-frequency intrinsic mode function components to obtain multiple second sample entropies corresponding to each of the original wind farm operation data. S334: Perform kurtosis calculation on each of the decomposed high-frequency intrinsic mode function components to obtain multiple kurtosis corresponding to each of the original wind farm operation data; S335: The weighted sum of multiple second sample entropies corresponding to each of the original wind farm operation data and multiple kurtosis corresponding to each of the original wind farm operation data is calculated to obtain the fitness value corresponding to each of the original wind farm operation data. S336: Determine whether all the fitness values ​​are greater than or equal to the fitness threshold. If yes, proceed to S337; otherwise, proceed to S339. S337: Perform a global search on the particle parameters in the initialized particle swarm to obtain the global target position; S338: Update the particle parameters in the initialized particle swarm according to the global target position to obtain the updated particle swarm, and then execute S3311; S339: Perform a local search on the particle parameters in the initialized particle swarm to obtain the local target position; S3310: Update the particle parameters in the initialized particle swarm according to the local target position to obtain the updated particle swarm, and then execute S3311; S3311: Until the preset maximum number of iterations is reached, the high-frequency intrinsic mode function to be processed is decomposed according to the initialized PLO parameters and the updated particle swarm to obtain the target high-frequency intrinsic mode function corresponding to each of the original wind farm operation data.

[0032] It should be understood that the PLO parameters refer to the parameters of the Aurora Optimization Algorithm.

[0033] It should be understood that a secondary decomposition is performed on the high-frequency IMFs sequence (i.e., the high-frequency intrinsic mode functions to be processed) to further suppress noise.

[0034] It should be understood that initializing the particle swarm refers to the VMD parameters K and α.

[0035] Specifically, to reduce the complexity of high-frequency IMFs, the VMD algorithm is used to decompose the sequence, and the Aurora Optimization (PLO) algorithm is used to jointly and adaptively optimize its key control parameters to avoid mode aliasing and getting trapped in local optima.

[0036] Specifically, in VMD decomposition, the number of modes K is crucial to ensuring sufficient coverage of frequency components while avoiding overfitting; the penalty coefficient α controls the similarity between adjacent modes. Too large an α leads to excessive smoothing, while too small an α may cause overfitting. Therefore, appropriately selecting K and α is essential for obtaining accurate and smooth decomposition results. Thus, the PLO algorithm is used to optimize VMD parameters, and its principle is as follows: Data initialization involves randomizing the PLO parameters (i.e., the original PLO parameters); generating an initial particle swarm where particles are randomly positioned in space, defined as X[K,α]; decomposing historical wind power data (i.e., the high-frequency intrinsic mode functions to be processed) using the current parameters (i.e., the initialized PLO parameters and the particle parameters in the initialized particle swarm); constructing a new fitness function by fusing sample entropy and kurtosis to calculate the fitness of each particle; selecting the search space (global search or local search) based on the fitness value to find the optimal solution (i.e., the global target position or the local target position). This is achieved by simulating the diffusion and propagation of aurora, extensively exploring the entire search space to avoid getting trapped in local optima; determining the currently found optimal solution, updating the particle swarm position and iterative formula; repeating the following steps until the maximum number of iterations is reached, outputting the optimal solution, i.e., the VMD parameters K and α.

[0037] In the above embodiments, each high-frequency intrinsic mode function to be processed is decomposed to obtain the target high-frequency intrinsic mode function, which reduces the complexity of high-frequency IMFs, suppresses noise, and avoids the problems of mode mixing and getting trapped in local optima.

[0038] Optionally, as an embodiment of the present invention, such as Figure 3 As shown, the prediction model includes a BiLSTM model. The process of using the prediction model to perform prediction analysis on each of the residual components, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operating data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operating data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operating data, and using the analysis results as the wind power prediction results, includes: The BiLSTM model is used to predict the high-frequency intrinsic mode functions of each target, and the prediction results of the high-frequency intrinsic mode functions corresponding to the original wind farm operation data are obtained. The BiLSTM model is used to predict the mid-frequency intrinsic mode functions of each target, and the prediction results of the mid-frequency intrinsic mode functions corresponding to the original wind farm operation data are obtained. The BiLSTM model is used to predict the low-frequency intrinsic mode functions of each target, and the prediction results of the low-frequency intrinsic mode functions corresponding to the original wind farm operation data are obtained. The prediction results of each high-frequency intrinsic mode function, the prediction results of the mid-frequency intrinsic mode functions corresponding to each original wind farm operation data, and the prediction results of the low-frequency intrinsic mode functions corresponding to each original wind farm operation data are accumulated to obtain the target intrinsic mode function corresponding to each original wind farm operation data. The residual components corresponding to each of the target intrinsic mode functions and each of the original wind farm operation data are reconstructed, and all reconstruction results are used as wind power prediction results.

[0039] It should be understood that the BiLSTM model is used to predict each modal component (i.e., the target high-frequency intrinsic mode function, the target mid-frequency intrinsic mode function, and the target low-frequency intrinsic mode function).

[0040] Specifically, the prediction results of each modal component (i.e., the prediction results of high-frequency intrinsic mode functions, mid-frequency intrinsic mode functions, and low-frequency intrinsic mode functions) are accumulated and reconstructed to obtain the wind power prediction result for the day to be measured (i.e., the wind power prediction result).

[0041] Specifically, the BiLSTM model is used as the prediction model to predict each modal component (i.e., the target high-frequency intrinsic mode function, the target mid-frequency intrinsic mode function, and the target low-frequency intrinsic mode function) to obtain the wind power prediction result for the day to be measured (i.e., the wind power prediction result). The principle of the BiLSTM model is as follows: BiLSTM (BiLSTM model) applies two independent LSTM networks to the input sequence: one along the positive time direction (forward LSTM) and the other along the negative time direction (backward LSTM). These two networks learn the forward and backward dependencies of the sequence respectively, and then their outputs are combined to obtain a more comprehensive sequence representation. Input layer: Takes the wind power time series data after quadratic decomposition as input; Forward LSTM layer: Processes the input sequence along the positive time direction; Backward LSTM layer: Processes the input sequence along the negative time direction; Output layer: Combines the outputs of the forward and backward LSTM layers to form the final sequence representation.

[0042] In the above embodiments, the prediction model is used to predict and analyze the residual components, the target high-frequency intrinsic mode functions, the target mid-frequency intrinsic mode functions, and the target low-frequency intrinsic mode functions, respectively. The analysis results are used as the wind power prediction results, which effectively improves the computational efficiency and the quality of the solution, solves the problem of high volatility in wind power data, improves frequency characteristics, reduces the impact of secondary decomposition, eliminates noise, improves the prediction performance of the model, and can extract more representative feature components.

[0043] Optionally, as another embodiment of the present invention, the present invention relates to a wind power prediction method. Wind power generation is characterized by high randomness, volatility, and uncertainty. Coupled with the inherent noise factor of wind turbines, accurate wind power prediction is extremely challenging. Furthermore, when wind turbines operate under special environmental conditions, such as blade icing, the dynamic changes in wind power are further aggravated, making it difficult for traditional prediction models to meet the power system's requirements for wind power prediction accuracy.

[0044] Optionally, as another embodiment of the present invention, the present invention proposes a multi-stage modal processing and complexity degradation method for processing raw data, the steps of which are as follows: (1) In the data processing stage, construct the phase modal processing and complexity degradation, and combine ICEEMDAN and VMD decomposition models to reduce noise interference and uncertainty of the original data and improve the ability to separate time series features. (2) In the decomposition model, a K-means clustering fusion mechanism driven by sample entropy is constructed to reduce the computational complexity and computational cost of the model while maintaining the integrity of feature information. (3) Based on the VMD algorithm, PLO is used to adaptively optimize the VMD parameters. By escaping local optima, the global search capability is enhanced, the VMD decomposition effect is optimized, and the input quality of the prediction model is improved.

[0045] Alternatively, as another embodiment of the present invention, the protection points of the present invention are as follows: 1. An improved set empirical mode decomposition (ICEEMDAN) initial decomposition method is adopted.

[0046] 2. A method for adaptively optimizing VMD parameters based on PLO.

[0047] Alternatively, as another embodiment of the present invention, the advantages of the present invention are as follows: First, an improved ensemble empirical mode decomposition (ICEEMDAN) is used to initially decompose the wind power time series, which solves the problem of high volatility in wind power data.

[0048] Second, the K-means algorithm is used to divide and aggregate each modal component (i.e., the original intrinsic mode function) to improve the frequency characteristics and reduce the impact of the secondary decomposition.

[0049] Third, based on the PLO adaptive optimization method for VMD parameters, the high-frequency intrinsic mode components in wind power data are decomposed in two stages. By simulating the movement of particles in the Earth's magnetic field, the model avoids getting trapped in local optima, eliminates the influence of noise, and improves the model's predictive performance.

[0050] Figure 4 This is a block diagram of a wind power prediction device provided in an embodiment of the present invention.

[0051] Alternatively, as another embodiment of the present invention, such as Figure 4 As shown, a wind power prediction device includes: The import module is used to import multiple raw wind farm operation data. The preprocessing module is used to preprocess the original wind farm operation data of each wind farm to obtain preprocessed wind farm operation data corresponding to the original wind farm operation data of each wind farm. The normalization processing module is used to normalize the preprocessed wind farm operation data of each wind farm to obtain the original wind power time series corresponding to the original wind farm operation data. The modal component analysis module is used to perform modal component analysis on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The prediction result acquisition module is used to construct a prediction model. The prediction model is used to perform prediction analysis on each of the residual components, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The analysis results are used as the wind power prediction results.

[0052] Optionally, another embodiment of the present invention provides a wind power prediction system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the wind power prediction method as described above. This system can be a computer or similar system.

[0053] Optionally, another embodiment of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the wind power prediction method as described above.

[0054] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0055] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0056] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0057] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0058] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0059] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0060] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting wind power output, characterized in that, Includes the following steps: Import multiple original wind farm operation data, and preprocess each of the original wind farm operation data to obtain preprocessed wind farm operation data corresponding to each of the original wind farm operation data. Normalize the preprocessed wind farm operation data of each wind farm to obtain the original wind power time series corresponding to the original wind farm operation data of each wind farm. Modal component analysis is performed on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. A prediction model is constructed, and the prediction model is used to predict and analyze each of the residual components, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The analysis results are used as the wind power prediction results.

2. The wind power prediction method according to claim 1, characterized in that, The process of normalizing the preprocessed wind farm operation data to obtain the original wind power time series corresponding to the original wind farm operation data includes: The preprocessed wind farm operation data are normalized using the maximum-minimum normalization algorithm to obtain the original wind power time series corresponding to the original wind farm operation data.

3. The wind power prediction method according to claim 1, characterized in that, The process of performing modal component analysis on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operating data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operating data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operating data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operating data includes: Each of the original wind power time series is decomposed to obtain the residual components corresponding to the operation data of each of the original wind farms and multiple original intrinsic mode functions corresponding to the operation data of each of the original wind farms. Based on each of the original wind power time series, the multiple original intrinsic mode functions corresponding to the original wind farm operation data are classified to obtain the high-frequency intrinsic mode function to be processed, the target mid-frequency intrinsic mode function, and the target low-frequency intrinsic mode function corresponding to each of the original wind farm operation data. Each of the high-frequency intrinsic mode functions to be processed is decomposed to obtain the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data.

4. The wind power prediction method according to claim 3, characterized in that, The process of decomposing each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data and the multiple original intrinsic mode functions corresponding to each of the original wind farm operation data includes: Preset Gaussian white noise is input into each of the original wind power time series to obtain the noisy wind power time series corresponding to the operation data of each of the original wind farms; Each of the noise-added wind power time series is decomposed to obtain the residual components corresponding to each of the original wind farm operation data and multiple original intrinsic mode functions corresponding to each of the original wind farm operation data.

5. The wind power prediction method according to claim 3, characterized in that, The process of classifying multiple original intrinsic mode functions corresponding to the operating data of each original wind farm according to each original wind power time series, and obtaining the high-frequency intrinsic mode function to be processed, the target mid-frequency intrinsic mode function, and the target low-frequency intrinsic mode function corresponding to each original wind farm operating data, includes: Sample entropy is calculated for each of the original wind power time series and multiple original intrinsic mode functions corresponding to the operation data of each of the original wind farms, to obtain multiple first sample entropies corresponding to the operation data of each of the original wind farms. Based on the K-means clustering algorithm and the entropy of multiple first samples corresponding to the original wind farm operation data, the original intrinsic mode functions corresponding to the original wind farm operation data are divided into multiple original high-frequency intrinsic mode functions, multiple original mid-frequency intrinsic mode functions, and multiple original low-frequency intrinsic mode functions corresponding to the original wind farm operation data. The original high-frequency intrinsic mode functions corresponding to the original wind farm operation data are merged and processed to obtain the high-frequency intrinsic mode functions to be processed corresponding to the original wind farm operation data. The original mid-frequency intrinsic mode functions corresponding to the original wind farm operation data are merged to obtain the target mid-frequency intrinsic mode function corresponding to the original wind farm operation data. The original low-frequency intrinsic mode functions corresponding to the original wind farm operation data are merged to obtain the target low-frequency intrinsic mode function corresponding to the original wind farm operation data.

6. The wind power prediction method according to claim 3, characterized in that, The process of decomposing each of the high-frequency intrinsic mode functions to be processed to obtain the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data includes: S331: Import the original PLO parameters and randomly initialize the original PLO parameters to obtain the initialized PLO parameters; S332: Initialize the particle swarm, and decompose the high-frequency intrinsic mode function to be processed according to the initialized PLO parameters and the particle parameters in the initialized particle swarm to obtain multiple decomposed high-frequency intrinsic mode function components corresponding to each of the original wind farm operation data. S333: Calculate the sample entropy of each of the decomposed high-frequency intrinsic mode function components to obtain multiple second sample entropies corresponding to each of the original wind farm operation data. S334: Perform kurtosis calculation on each of the decomposed high-frequency intrinsic mode function components to obtain multiple kurtosis corresponding to each of the original wind farm operation data; S335: The weighted sum of multiple second sample entropies corresponding to each of the original wind farm operation data and multiple kurtosis corresponding to each of the original wind farm operation data is calculated to obtain the fitness value corresponding to each of the original wind farm operation data. S336: Determine whether all the fitness values ​​are greater than or equal to the fitness threshold. If yes, proceed to S337; otherwise, proceed to S339. S337: Perform a global search on the particle parameters in the initialized particle swarm to obtain the global target position; S338: Update the particle parameters in the initialized particle swarm according to the global target position to obtain the updated particle swarm, and then execute S3311; S339: Perform a local search on the particle parameters in the initialized particle swarm to obtain the local target position; S3310: Update the particle parameters in the initialized particle swarm according to the local target position to obtain the updated particle swarm, and then execute S3311; S3311: Until the preset maximum number of iterations is reached, the high-frequency intrinsic mode function to be processed is decomposed according to the initialized PLO parameters and the updated particle swarm to obtain the target high-frequency intrinsic mode function corresponding to each of the original wind farm operation data.

7. The wind power prediction method according to claim 1, characterized in that, The prediction model includes a BiLSTM model. The process of using the prediction model to perform prediction analysis on each residual component, the target high-frequency intrinsic mode function corresponding to each original wind farm operating data, the target mid-frequency intrinsic mode function corresponding to each original wind farm operating data, and the target low-frequency intrinsic mode function corresponding to each original wind farm operating data, and using the analysis results as the wind power prediction result includes: The BiLSTM model is used to predict the high-frequency intrinsic mode functions of each target, and the prediction results of the high-frequency intrinsic mode functions corresponding to the original wind farm operation data are obtained. The BiLSTM model is used to predict the mid-frequency intrinsic mode functions of each target, and the prediction results of the mid-frequency intrinsic mode functions corresponding to the original wind farm operation data are obtained. The BiLSTM model is used to predict the low-frequency intrinsic mode functions of each target, and the prediction results of the low-frequency intrinsic mode functions corresponding to the original wind farm operation data are obtained. The prediction results of each high-frequency intrinsic mode function, the prediction results of the mid-frequency intrinsic mode functions corresponding to each original wind farm operation data, and the prediction results of the low-frequency intrinsic mode functions corresponding to each original wind farm operation data are accumulated to obtain the target intrinsic mode function corresponding to each original wind farm operation data. The residual components corresponding to each of the target intrinsic mode functions and each of the original wind farm operation data are reconstructed, and all reconstruction results are used as wind power prediction results.

8. A wind power prediction device, characterized in that, include: The import module is used to import multiple raw wind farm operation data. The preprocessing module is used to preprocess the original wind farm operation data of each wind farm to obtain preprocessed wind farm operation data corresponding to the original wind farm operation data of each wind farm. The normalization processing module is used to normalize the preprocessed wind farm operation data of each wind farm to obtain the original wind power time series corresponding to the original wind farm operation data. The modal component analysis module is used to perform modal component analysis on each of the original wind power time series to obtain the residual components corresponding to each of the original wind farm operation data, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The prediction result acquisition module is used to construct a prediction model. The prediction model is used to perform prediction analysis on each of the residual components, the target high-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, the target mid-frequency intrinsic mode functions corresponding to each of the original wind farm operation data, and the target low-frequency intrinsic mode functions corresponding to each of the original wind farm operation data. The analysis results are used as the wind power prediction results.

9. A wind power prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the wind power prediction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the wind power prediction method as described in any one of claims 1 to 7.

Images