Construction method, system and equipment of wind power prediction model and medium

Abstract

The invention discloses a construction method, system and device of a wind power prediction model and a medium, and relates to the technical field of clean energy, and the method comprises the steps: obtaining historical meteorological data and historical power data of a target wind power station; constructing a time-frequency feature set based on the historical meteorological data and the historical power data; and inputting the time-frequency feature set into a pre-constructed preset prediction model for model training, and in the model training process, learning a mapping relationship between the first time-frequency feature and historical power data and between the second time-frequency feature and historical power data until the model converges, thereby obtaining a power prediction model. When the model is trained, periodic change characteristics, such as seasonal change and daily change, of the meteorological data and the power data can be considered, meanwhile, short-term fluctuation and long-term trend of the meteorological data and the power data are considered, and the finally obtained power prediction model improves the capacity of capturing changes of the meteorological characteristics and the power characteristics; and the precision of the prediction result and the adaptability of the power prediction model are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of clean energy, and particularly to a method, a system, a device and a medium for constructing a wind power prediction model. Background Art

[0002] In the face of the challenges of the global energy crisis, renewable energy sources such as photovoltaics, wind energy, and hydropower have received increasing attention; among them, wind power generation has become one of the most widely used renewable energy sources due to its important role in the power system; in order to improve the efficiency and reliability of wind power generation, accurate wind power prediction technology has become crucial.

[0003] In the related art, the wind power prediction method usually obtains the predicted power by fitting the corresponding relationship between historical meteorological data and historical power data, and then combining the predicted meteorological data and the corresponding relationship; however, in the existing wind power prediction method, the time domain characteristics and frequency domain characteristics of meteorological data and power data are not considered when fitting meteorological data and power data, resulting in inaccurate power prediction. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is: how to solve the problem that in the existing wind power prediction method, the time domain characteristics and frequency domain characteristics of meteorological data and power data are not considered when fitting meteorological data and power data, resulting in inaccurate power prediction.

[0006] To solve the above technical problems, the present invention provides the following technical solution: a method for constructing a wind power prediction model, including obtaining historical meteorological data and historical power data of a target wind farm; constructing a time-frequency feature set based on the historical meteorological data and the historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature, the first time-frequency feature is used to represent the periodic information contained in the historical meteorological data and the historical power data, and the second time-frequency feature is used to represent the local frequency information of the historical meteorological data and the historical power data at multiple time scales in the time-frequency domain; inputting the time-frequency feature set into a pre-constructed preset prediction model for model training, and during the model training process, learning the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model.

[0007] As a preferred embodiment of the method for constructing a wind power prediction model according to the present invention, wherein: constructing a time-frequency feature set based on historical meteorological data and historical power data includes performing Fourier transform on the historical meteorological data and historical power data to obtain historical meteorological frequency components and historical wind power frequency components; performing wavelet transform on the historical meteorological data and historical power data to obtain historical meteorological wavelet coefficients and historical wind power wavelet coefficients at each preset time scale; determining a first time-frequency feature based on the historical meteorological frequency components and historical meteorological wavelet coefficients; determining a second time-frequency feature based on the historical wind power frequency components and historical wind power wavelet coefficients; and fusing the first time-frequency feature and the second time-frequency feature to obtain a time-frequency feature set.

[0008] As a preferred embodiment of the method for constructing a wind power prediction model according to the present invention, wherein: performing Fourier transform on the historical meteorological data and historical power data includes performing Fourier transform on the historical meteorological data and historical power data to obtain a first frequency feature of the historical meteorological data and a second frequency feature of the historical power data; respectively extracting main frequency components from the first frequency feature and the second frequency feature to obtain historical meteorological frequency components and historical wind power frequency components; and the frequency components include amplitude features and phase features.

[0009] As a preferred embodiment of the method for constructing a wind power prediction model according to the present invention, wherein: performing wavelet transform on the historical meteorological data and historical power data includes respectively constructing a first wavelet function basis and a second wavelet function basis based on the historical meteorological data and historical power data; performing first multi-scale decomposition on the historical meteorological data based on the first wavelet function basis to obtain historical meteorological wavelet coefficients; and performing second multi-scale decomposition on the historical power data based on the second wavelet function basis to obtain historical wind power wavelet coefficients.

[0010] As a preferred solution of the method for constructing a wind power prediction model according to the present invention, wherein: the preset prediction model includes a first feature extraction module, a second feature extraction module, and a power output module; the input end of the first feature extraction module is used to input the time-frequency feature set, and the output end is connected to the input end of the second feature extraction module; the output end of the second feature extraction module is connected to the input end of the power output module; the model training includes inputting the time-frequency feature set into the preset prediction model for at least one round of model training until the model converges to obtain a power prediction model; during each round of model training, the module parameters of the first feature extraction module are adjusted so that the first feature extraction module extracts the local spatial features of the first time-frequency feature data and the second time-frequency feature data at different time scales; the module parameters of the second feature extraction module are adjusted so that the second feature extraction module can extract the temporal dependence features of the local spatial features at a preset time scale; the model parameters of the power output module are adjusted so that the power prediction model module outputs a power prediction result based on the local spatial features and the temporal dependence features.

[0011] As a preferred solution of the method for constructing a wind power prediction model according to the present invention, wherein: the first feature extraction module includes a TCN model, and the second feature extraction module includes at least one of an LSTM network, a Transformer, and an Informer.

[0012] As a preferred solution of the method for constructing a wind power prediction model according to the present invention, wherein: the model training further includes updating the model parameters of the preset prediction model based on the long-nosed coati optimization algorithm during the model training process.

[0013] Another object of the present invention is to provide a device for the method of constructing a wind power prediction model, which can solve the problem of constructing a wind power prediction model by constructing a device for the method of constructing a wind power prediction model.

[0014] To solve the above technical problems, the present invention provides the following technical solution: A system for constructing a wind power prediction model, including an acquisition module, a time-frequency feature set construction module, and a model training module; the acquisition module is used to acquire historical meteorological data and historical power data of a target wind farm; the time-frequency feature set construction module is used to construct a time-frequency feature set based on the historical meteorological data and historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature, the first time-frequency feature is used to represent the periodic information contained in the historical meteorological data and the historical power data, and the second time-frequency feature is used to represent the local frequency information of the historical meteorological data and the historical power data at multiple time scales in the time-frequency domain; the model training module is used to input the time-frequency feature set into a pre-constructed preset prediction model for model training, and during the model training process, learn the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model.

[0015] A computer device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for constructing a wind power prediction model are implemented.

[0016] A computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the steps of the above-mentioned method for constructing a wind power prediction model are implemented.

[0017] The beneficial effect of the present invention is as follows: When the model of the method for constructing a wind power prediction model provided by the present invention is trained, it can consider the periodic change characteristics of meteorological data and power data, such as seasonal changes and daily changes, and at the same time consider the short-term fluctuations and long-term trends of meteorological data and power data. The finally obtained power prediction model improves the ability to capture changes in meteorological characteristics and power characteristics, thereby improving the accuracy of the prediction result and the adaptability of the power prediction model. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0019] Figure 1 It is a flowchart of a method for constructing a wind power prediction model provided by the first embodiment of the present invention.

[0020] Figure 2 It is a structural diagram of a device for constructing a wind power prediction model provided by the second embodiment of the present invention.

[0021] Figure 3 Schematic diagram of the hardware structure of the computer device provided for the third embodiment of the present invention.

[0022] In the figure, 501 is an acquisition module; 502 is a time-frequency feature set construction module; 503 is a model training module. Specific implementation manners

[0023] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific implementation manners of the present invention will be described in detail below with reference to the accompanying drawings of the specification.

[0024] In the following description, many specific details are set forth to facilitate a thorough understanding of the present invention. However, the present invention may be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0025] Embodiment 1, referring to Figure 1 , which is the first embodiment of the present invention. This embodiment provides a method for constructing a wind power prediction model, including: acquiring historical meteorological data and historical power data of a target wind farm; constructing a time-frequency feature set based on the historical meteorological data and historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature. The first time-frequency feature is used to represent the periodic information contained in the historical meteorological data and historical power data, and the second time-frequency feature is used to represent the local frequency information of the historical meteorological data and historical power data at multiple time scales in the time-frequency domain; inputting the time-frequency feature set into a pre-constructed preset prediction model for model training. During the model training process, learn the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model.

[0026] Affected by the randomness, volatility, and intermittency of wind power generation, the meteorological-related features of wind power data, especially the wind speed-related features and wind power-related features, also have randomness, volatility, and intermittency; affected by the above, the meteorological-related features and wind power features are manifested as time series with time domain characteristics and frequency domain characteristics, and there is a time series dependence relationship between the meteorological-related features and wind power features.

[0027] In the related art, the wind power prediction method realizes the prediction power by fitting the corresponding relationship between historical meteorological data and historical power data, and then combining the predicted meteorological data and the corresponding relationship. Among them, the corresponding relationship between historical meteorological data and historical power data usually includes numerical corresponding relationship, time series corresponding relationship, and time series dependence corresponding relationship. This method can realize the prediction of wind power by considering the corresponding relationship contained in the data. However, when predicting power, due to the influence of the time domain characteristics and frequency domain characteristics of meteorological related features and wind power related features not being considered, the sensitivity and response speed to wind speed changes and wind power fluctuations are not ideal when predicting wind power.

[0028] Figure 1 It is a flowchart of a method for constructing a wind power prediction model according to an embodiment of the present invention. As Figure 1 shown, the process includes the following steps:

[0029] Step S101, obtain the historical meteorological data and historical power data of the target wind farm.

[0030] In this embodiment, the historical meteorological data of the target wind farm can be obtained by obtaining the data recorded in the meteorological acquisition data, and the historical power data can be obtained by collecting the power data corresponding to the time series of the historical meteorological data.

[0031] Exemplarily, the historical meteorological data may include wind-related data such as historical wind speed data and historical wind direction data, and may also include environment-related data such as historical temperature, historical humidity, and historical air pressure.

[0032] In one embodiment, after obtaining the historical meteorological data and historical power data, preprocess the historical meteorological data and historical power data; preliminarily clean and reconstruct the abnormal values such as zero power values and unreasonable values in the original data.

[0033] Furthermore, perform standardization processing on the historical meteorological data and historical power data after cleaning and reconstruction, so that the historical meteorological data and historical power data have zero mean and unit variance.

[0034] Step S102, construct a time-frequency feature set based on the historical meteorological data and historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature. The first time-frequency feature is used to represent the periodic information contained in the historical meteorological data and historical power data, and the second time-frequency feature is used to represent the localization information of the historical meteorological data and historical power data in the time-frequency domain.

[0035] As described above, in the wind power prediction method in the related art, when predicting power, the influence of the time domain characteristics and frequency domain characteristics included in the meteorological related characteristics and wind power related characteristics is not considered, resulting in unsatisfactory sensitivity and response speed to wind speed changes and wind power fluctuations during wind power prediction; to solve this problem, in this embodiment, a time-frequency feature set is constructed based on historical meteorological data and historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature, the first time-frequency feature is used to represent the periodic information included in the historical meteorological data and historical power data, and the second time-frequency feature is used to represent the local frequency information of the historical meteorological data and historical power data at multiple time scales in the time-frequency domain.

[0036] Exemplarily, the first time-frequency feature can be obtained by respectively performing frequency domain feature extraction on the historical meteorological data and historical power data, and the second time-frequency feature can be obtained by respectively performing signal characteristic decomposition on the historical meteorological data and historical power data at different time scales.

[0037] In one embodiment, the first time-frequency feature can be obtained by using the method of autocorrelation function and partial autocorrelation function for the historical meteorological data and historical power data including historical wind speed data; the method of autocorrelation function and partial autocorrelation function identifies the periodic patterns of meteorological data and power data by calculating the correlation between the time series and its own lagged version, so as to obtain periodic change characteristics such as seasonal changes and daily changes for characterizing meteorological data and power data.

[0038] In one embodiment, the first time-frequency feature can be obtained by using the method of Fourier transform for the historical meteorological data and historical power data including historical wind speed data; the Fourier transform is used to convert the time domain signal into a frequency domain signal. For a periodic function, the Fourier series can be expressed as a linear combination of different frequency components, while for a non-periodic function, the Fourier transform represents the distribution of the signal in the frequency domain through an integral form; therefore, through the Fourier transform, not only can the frequency characteristics of meteorological data and power data be revealed, but also various important properties of meteorological data and power data can be identified.

[0039] In one embodiment, the second time-frequency feature can be obtained by using the method of empirical mode decomposition for the historical meteorological data and historical power data including historical wind speed data; the empirical mode decomposition can adaptively decompose a complex time series signal into several intrinsic mode functions, and each intrinsic mode function represents the local characteristics of the signal at different frequencies and time scales; through this decomposition, the characteristics of meteorological data and power data at different time scales can be extracted, and the short-term fluctuations and long-term trends of meteorological data and wind power data can be identified.

[0040] In one embodiment, the second time-frequency feature can be obtained by applying wavelet transform to historical meteorological data including historical wind speed data and historical power data; wavelet transform can provide a localized representation of a signal in both time and frequency; through multi-scale decomposition of wavelet transform, features of meteorological data and power data at different time scales can be extracted, and short-term fluctuations and long-term trends of meteorological data and wind power data can be identified.

[0041] Exemplarily, after obtaining the first time-frequency feature and the second time-frequency feature, a time-frequency feature set is constructed; exemplarily, the time-frequency feature set includes the first time-frequency feature and the second time-frequency feature.

[0042] In the above embodiment, the time-frequency feature set includes the first time-frequency feature and the second time-frequency feature. The first time-frequency feature can characterize periodic change features of meteorological data and power data such as seasonal change and daily change, and the second time-frequency feature can characterize short-term fluctuations and long-term trends of meteorological data and power data. Therefore, the time-frequency feature data set can comprehensively characterize the periodic and non-periodic changes of wind speed, as well as other meteorological features and power features.

[0043] Step S103: Input the time-frequency feature set into a pre-constructed preset prediction model for model training. During the model training process, learn the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model.

[0044] In this embodiment, the preset prediction model is used to learn the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data; since the first time-frequency feature can characterize periodic change features of meteorological data and power data such as seasonal change and daily change, and the second time-frequency feature can characterize short-term fluctuations and long-term trends of meteorological data and power data, therefore, by using the method of inputting the time-frequency feature set into a pre-constructed preset prediction model for model training, the model training can consider periodic change features of meteorological data and power data such as seasonal change and daily change, and at the same time consider short-term fluctuations and long-term trends of meteorological data and power data. The finally obtained power prediction model improves the ability to capture changes in meteorological features and power features, thereby improving the accuracy of the prediction result and the adaptability of the power prediction model.

[0045] Method for constructing wind power prediction model of the present invention, which obtains historical meteorological data and historical power data of a target wind farm; constructs a time-frequency feature set based on the historical meteorological data and the historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature, the first time-frequency feature is used to represent the periodic information contained in the historical meteorological data and the historical power data, and the second time-frequency feature is used to represent the localization information of the historical meteorological data and the historical power data in the time-frequency domain; inputs the time-frequency feature set into a pre-constructed preset prediction model for model training, and during the model training process, learns the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model; during model training, it is possible to consider the periodic change characteristics of meteorological data and power data, such as seasonal changes and daily changes, and at the same time consider the short-term fluctuations and long-term trends of meteorological data and power data. The finally obtained power prediction model improves the ability to capture changes in meteorological features and power features, thereby improving the accuracy of the prediction results and the adaptability of the power prediction model.

[0046] As an exemplary embodiment, constructing a time-frequency feature set based on historical meteorological data and historical power data includes: performing Fourier transform on the historical meteorological data and the historical power data to obtain historical meteorological frequency components and historical wind power frequency components; performing wavelet transform on the historical meteorological data and the historical power data to obtain historical meteorological wavelet coefficients and historical wind power wavelet coefficients at each preset time scale; determining the first time-frequency feature based on the historical meteorological frequency components and the historical meteorological wavelet coefficients; determining the second time-frequency feature based on the historical wind power frequency components and the historical wind power wavelet coefficients; and fusing the first time-frequency feature and the second time-frequency feature to obtain the time-frequency feature set.

[0047] In the present invention, Fourier transform is used to process the historical meteorological data and the historical power data to obtain historical meteorological frequency components and historical wind power frequency components; specifically, Fourier transform decomposes the time series composed of the historical meteorological data and the historical power data into a combination of sine waves and cosine waves of different frequencies; for a periodic function, the Fourier series is expressed as a linear combination of different frequency components, and for a non-periodic function, Fourier transform shows the distribution of the signal in the frequency domain through an integral form.

[0048] As an exemplary embodiment, performing Fourier transform on the historical meteorological data and the historical power data to obtain historical meteorological frequency components and historical wind power frequency components includes: performing Fourier transform on the historical meteorological data and the historical power data to obtain a first frequency feature of the historical meteorological data and a second frequency feature of the historical power data; respectively extracting the main frequency components from the first frequency feature and the second frequency feature to obtain the historical meteorological frequency components and the historical wind power frequency components; wherein, the frequency components include amplitude features and phase features.

[0049] In this embodiment, after performing Fourier transform on historical meteorological data and historical power data to obtain the first frequency feature and the second frequency feature, the main frequency components of the first frequency feature and the second frequency feature are extracted respectively to obtain the main frequency and its corresponding amplitude and phase.

[0050] Exemplarily, the method of fast Fourier transform can be used to process historical meteorological data and historical power data; specifically, the time series composed of meteorological data and power data is subjected to fast Fourier transform to obtain the spectra corresponding to the meteorological data and the power data; further, the main frequency is extracted based on the spectra, and the amplitude and phase of the main frequency are extracted to obtain the first frequency feature of the historical meteorological data and the second frequency feature of the historical power data.

[0051] In the present invention, wavelet transform is performed on historical meteorological data and historical power data simultaneously to obtain historical meteorological wavelet coefficients and historical wind power wavelet coefficients at each preset time scale, and the frequency components of the meteorological data and the power data at different preset time scales are characterized in the form of wavelet coefficients, which can represent the instantaneous changes and detailed features of the meteorological data and the power data.

[0052] As an exemplary embodiment, performing wavelet transform on historical meteorological data and historical power data to obtain historical meteorological wavelet coefficients and historical wind power wavelet coefficients at each preset time scale includes: constructing a first wavelet function basis and a second wavelet function basis based on the historical meteorological data and the historical power data respectively; performing first multi-scale decomposition on the historical meteorological data based on the first wavelet function basis to obtain historical meteorological wavelet coefficients; performing second multi-scale decomposition on the historical power data based on the second wavelet function basis to obtain historical wind power wavelet coefficients.

[0053] Specifically, wavelet transform is used to decompose a signal into wavelets at different time scales to reveal the characteristics of the signal at different time scales and frequency scales; specifically, when performing wavelet transform on historical meteorological data and historical power data, a first wavelet function basis and a second wavelet function basis are constructed based on the historical meteorological data and the historical power data respectively. Further, multi-scale decomposition is performed on the historical meteorological data and the historical power data based on the first wavelet function basis and the second wavelet function basis respectively to obtain wavelet coefficients at the time scale and position corresponding to each scale; during decomposition, by continuously changing the scale and translation amount of the wavelet function and convolving with the signal, the frequency components of the signal are analyzed at different scales, and the instantaneous changes and detailed features of the historical meteorological data and the historical power data are captured.

[0054] As an exemplary embodiment, the preset prediction model includes a first feature extraction module, a second feature extraction module, and a power output module. The input end of the first feature extraction module is used to input the time-frequency feature set, and the output end is connected to the input end of the second feature extraction module. The output end of the second feature extraction module is connected to the input end of the power output module. Training the model by inputting the time-frequency feature set into the pre-constructed preset prediction model includes: inputting the time-frequency feature set into the preset prediction model for at least one round of model training until the model converges to obtain a power prediction model. During each round of model training, adjust the module parameters of the first feature extraction module so that the first feature extraction module can extract the local spatial features of the first time-frequency feature data and the second time-frequency feature data at different time scales; adjust the module parameters of the second feature extraction module so that the second feature extraction module can extract the temporal dependence features of the local spatial features at the preset time scale; adjust the model parameters of the power output module so that the power prediction model module outputs a power prediction result based on the local spatial features and the temporal dependence features.

[0055] In the present invention, the preset prediction model is used to obtain power data according to the local spatial features, global spatial features of meteorological data at different time scales, and the temporal dependence relationship of meteorological data.

[0056] Specifically, the preset prediction model includes a first feature extraction module, a second feature extraction module, and a power output module. The input end of the first feature extraction module is used to input the time-frequency feature set, and the output end is connected to the input end of the second feature extraction module. The output end of the second feature extraction module is connected to the input end of the power output module. When performing model training, input the time-frequency feature set into the preset prediction model for at least one round of model training until the model converges to obtain a power prediction model. During each round of model training, adjust the module parameters of the first feature extraction module so that the first feature extraction module can extract the local spatial features of the first time-frequency feature data and the second time-frequency feature data at different time scales; adjust the module parameters of the second feature extraction module so that the second feature extraction module can extract the temporal dependence features of the local spatial features at the preset time scale; adjust the model parameters of the power output module so that the power prediction model module outputs a power prediction result based on the local spatial features and the temporal dependence features.

[0057] In the above embodiments, during the model training process, the first feature extraction module can learn to extract local spatial features based on the input meteorological features and determined by the first time-frequency feature data of historical meteorological features and the second time-frequency feature data of historical power data. The second feature extraction module can learn to extract the temporal dependence relationship at a preset time scale corresponding to the local spatial features determined by the first time-frequency feature data of historical meteorological features and the second time-frequency feature data of historical power data based on the input meteorological features. Therefore, when using the trained model for power prediction, the first feature extraction module extracts the time-frequency features of the input predicted meteorological data and outputs local spatial features according to the time-frequency features. Further, based on the local spatial features, the temporal dependence features of historical meteorological data and historical power data are determined, and power prediction is performed according to the temporal dependence features. The above steps can determine the local spatial features of historical meteorological data and historical power data corresponding to the predicted meteorological data, and can perform power prediction with the temporal dependence relationship corresponding to the local features. The trained power prediction model, through the combination of the first feature extraction module and the second feature extraction module, can identify and learn complex time series patterns, improve the ability to capture changes in meteorological features and power features, and thus improve the accuracy of the prediction results and the adaptability of the power prediction model.

[0058] As an exemplary embodiment, the first feature extraction module includes a TCN model, and the second feature extraction module includes at least one of an LSTM network, a Transformer, and an Informer.

[0059] Exemplarily, taking the first feature extraction module as a TCN model, the time convolution extraction module as an Informer model, and the power output module as a fully connected layer as an example, the technical solution of the embodiment is described:

[0060] In this embodiment, the preset prediction model is constructed based on the TCN-Informer module. Among them, the TCN module is used to extract the spatial features in the time-frequency feature set, and the Informer part is used to capture the long-term dependence relationship in the time series data corresponding to the spatial features.

[0061] As an exemplary embodiment, when inputting the time-frequency feature set into the pre-constructed preset prediction model for model training, it further includes: during the model training process, updating the model parameters of the preset prediction model based on the kinkajou optimization algorithm.

[0062] In this embodiment, the Coati Optimization Algorithm (COA) is selected as a tool for optimizing the hyperparameters of the preset prediction model to improve the performance of the model. Specifically, when implementing hyperparameter optimization based on COA, the first thing to do is to determine the optimization objective, that is, to identify the hyperparameters to be adjusted and their value ranges. In this embodiment, the optimization objectives may include the convolutional kernel size, number of layers, dilation rate, etc. of TCN, the number of attention heads, number of encoder and decoder layers, etc. of Informer, as well as the learning rate, batch size, and regularization parameter, etc. Further, a group of initial hyperparameter combinations are randomly generated as the starting point for optimization. The core of COA lies in its heuristic search strategy, which efficiently explores the hyperparameter space by simulating natural phenomena or animal behaviors, while balancing local search and global search to avoid getting trapped in local optima and improve search efficiency. During the optimization process, the TCN-Informer model is trained using each set of hyperparameters, and its performance is evaluated on the validation set, such as measured by metrics like Mean Absolute Percentage Error (MAPE) or Root Mean Square Error (RMSE). According to the performance feedback, COA dynamically adjusts the hyperparameter combinations, retains the combinations with better performance, and performs small-scale random mutations on them to explore new hyperparameter spaces. This process is continuously iterated until the termination condition is met, such as reaching the preset maximum number of iterations or the performance no longer improves significantly. Finally, the set of hyperparameters with the best performance is selected from all the evaluated hyperparameter combinations, and the model is retrained using this set of hyperparameters, and its final performance is evaluated on the test set. In this way, COA can efficiently optimize hyperparameters, improve the prediction performance of the model, while maintaining the flexibility and adaptability of the algorithm.

[0063] The method for constructing the power prediction model in this embodiment adopts the method of inputting the time-frequency feature set into a pre-constructed preset prediction model for model training. When training the model, the periodic variation characteristics such as seasonal variation and daily variation of meteorological data and power data can be considered, and at the same time, the short-term fluctuations and long-term trends of meteorological data and power data are considered. At the same time, the COA algorithm is used to finely adjust the hyperparameters of the power prediction model, so that the model can not only identify and learn complex time series patterns, but also adapt to different geographical locations and changing meteorological conditions. Finally, the obtained power prediction model improves the ability to capture changes in meteorological features and power features, thereby improving the accuracy of the prediction results and the adaptability of the power prediction model.

[0064] Example 2, refer to Figure 2, which is the second embodiment of the present invention. Different from the previous embodiment, it provides a device for constructing a wind power prediction model, including: an acquisition module 501, a time-frequency feature set construction module 502, and a model training module 503.

[0065] The acquisition module 501 is configured to acquire historical meteorological data and historical power data of a target wind farm.

[0066] The time-frequency feature set construction module 502 is configured to construct a time-frequency feature set based on the historical meteorological data and historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature. The first time-frequency feature is used to represent the periodic information included in the historical meteorological data and historical power data, and the second time-frequency feature is used to represent the local frequency information of the historical meteorological data and historical power data at multiple time scales in the time-frequency domain.

[0067] The model training module 503 is configured to input the time-frequency feature set into a pre-constructed preset prediction model for model training. During the model training process, it learns the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model.

[0068] It should be noted here that the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiments.

[0069] It should be noted that the above modules, as part of the device, can be implemented by software or hardware. Among them, the hardware environment includes a network environment.

[0070] Embodiment 3, refer to Figure 3 , which is the third embodiment of the present invention. Different from the previous two embodiments,

[0071] The embodiment of the present invention further provides a computer device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus. The memory is used to store a computer program; the processor is configured to execute the method in any of the above embodiments by running the computer program stored in the memory.

[0072] Figure 3 is a structural block diagram of an optional computer device according to an embodiment of the present application. As Figure 3 shown, it includes a processor 10, a communication interface 20, a memory 30, and a communication bus 40. Among them, the processor 10, the communication interface 20, and the memory 30 complete communication with each other through the communication bus 40. Among them,

[0073] The memory 30 is used to store a computer program;

[0074] The processor 10, when executing the computer program stored in the memory 30, implements the method of any of the above embodiments.

[0075] Optionally, in this embodiment, the above communication bus may be a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 3 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0076] The communication interface is used for communication between the above computer device and other devices.

[0077] The memory may include a RAM, and may also include a non-volatile memory, for example, at least one disk memory. Optionally, the memory may also be at least one storage device located far from the aforementioned processor.

[0078] The above processor may be a general-purpose processor, which may include but is not limited to: a CPU (Central Processing Unit), an NP (Network Processor), etc.; it may also be a DSP (Digital Signal Processing), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0079] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and will not be elaborated here.

[0080] Those of ordinary skill in the art can understand that Figure 3 the structure shown is only schematic, and the device for implementing the method of any one of the above embodiments may be a terminal device, and the terminal device may be a smart phone (such as an Android phone, an IOS phone, etc.), a tablet computer, a palm computer, and a mobile Internet device (MID), a PAD and other terminal devices. Figure 3It does not limit the structure of the above electronic device. For example, the terminal device may further include more or fewer components (such as a network interface, a display device, etc.) than those shown in Figure 3 , or have a different configuration from that shown in Figure 3 .

[0081] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware of the terminal device through a program, and the program can be stored in a computer-readable storage medium. The storage medium may include: a flash drive, a ROM, a RAM, a magnetic disk, or an optical disc, etc.

[0082] As an exemplary embodiment, the present application also provides a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the method steps of any one of the embodiments when running.

[0083] Optionally, in this embodiment, the above storage medium may be used to execute the program code of the method steps of the embodiments of the present application.

[0084] Optionally, in this embodiment, the above storage medium may be located on at least one of the multiple network devices in the network shown in the above embodiments.

[0085] Optionally, in this embodiment, the storage medium is configured to store the method for executing the above embodiments.

[0086] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and will not be elaborated herein.

[0087] Optionally, in this embodiment, the above storage medium may include, but is not limited to: various media such as a USB flash drive, a ROM, a RAM, a mobile hard disk, a magnetic disk, or an optical disc that can store program code.

[0088] The serial numbers of the embodiments of the present application above are only for description and do not represent the advantages or disadvantages of the embodiments.

[0089] If the integrated unit in the above embodiments is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in the above computer-readable storage medium. Based on such an understanding, the technical solution of the present application, 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. The computer software product is stored in the storage medium and includes several instructions for causing one or more computer devices (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method in the above embodiments.

[0090] In several embodiments provided by this application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0091] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution provided in this embodiment.

[0092] In addition, in each embodiment of this application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0093] In the above embodiments of this application, the descriptions of each embodiment have their own focuses. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0094] The above is only the preferred embodiment of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this application.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for constructing a wind power prediction model, characterized in that: Including, Obtaining historical meteorological data and historical power data of a target wind farm; Constructing a time-frequency feature set based on the historical meteorological data and the historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature, the first time-frequency feature is used to represent the periodic information contained in the historical meteorological data and the historical power data, and the second time-frequency feature is used to represent the local frequency information of the historical meteorological data and the historical power data at multiple time scales in the time-frequency domain; Inputting the time-frequency feature set into a pre-constructed preset prediction model for model training. During the model training process, learn the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model.

2. The construction method of a wind power prediction model according to claim 1, characterized in that: The constructing of the time-frequency feature set based on the historical meteorological data and the historical power data includes performing Fourier transform on the historical meteorological data and the historical power data to obtain historical meteorological frequency components and historical wind power frequency components; Performing wavelet transform on the historical meteorological data and the historical power data to obtain historical meteorological wavelet coefficients and historical wind power wavelet coefficients at each preset time scale; Determining the first time-frequency feature based on the historical meteorological frequency components and the historical meteorological wavelet coefficients; Determining the second time-frequency feature based on the historical wind power frequency components and the historical wind power wavelet coefficients; Fusing the first time-frequency feature and the second time-frequency feature to obtain a time-frequency feature set.

3. The construction method of a wind power prediction model according to claim 2, characterized in that: The performing of Fourier transform on the historical meteorological data and the historical power data includes, Performing Fourier transform on the historical meteorological data and the historical power data to obtain a first frequency feature of the historical meteorological data and a second frequency feature of the historical power data; Performing main frequency component extraction on the first frequency feature and the second frequency feature respectively to obtain historical meteorological frequency components and historical wind power frequency components; The frequency components include amplitude features and phase features.

4. The construction method of a wind power prediction model according to claim 3, characterized in that: The performing of wavelet transform on the historical meteorological data and the historical power data includes, Constructing a first wavelet function basis and a second wavelet function basis based on the historical meteorological data and the historical power data respectively; Performing first multi-scale decomposition on the historical meteorological data based on the first wavelet function basis to obtain historical meteorological wavelet coefficients; Performing second multi-scale decomposition on the historical power data based on the second wavelet function basis to obtain historical wind power wavelet coefficients.

5. The construction method of a wind power prediction model according to claim 4, wherein: The preset prediction model includes a first feature extraction module, a second feature extraction module, and a power output module; The input end of the first feature extraction module is used to input the time-frequency feature set, and the output end is connected to the input end of the second feature extraction module; The output end of the second feature extraction module is connected to the input end of the power output module; The performing of model training includes inputting the time-frequency feature set into the preset prediction model for at least one round of model training until the model converges to obtain a power prediction model; During each round of model training, adjusting the module parameters of the first feature extraction module so that the first feature extraction module extracts the local spatial features of the first time-frequency feature data and the second time-frequency feature data at different time scales; Adjust the module parameters of the second feature extraction module so that the second feature extraction module can extract the temporal dependence features of the local spatial features at a preset time scale; Adjust the model parameters of the power output module so that the power prediction model module outputs a power prediction result based on the local spatial features and the temporal dependence features.

6. The construction method of a wind power prediction model according to claim 5, characterized in that: The first feature extraction module includes a TCN model, and the second feature extraction module includes at least one of an LSTM network, a Transformer, and an Informer.

7. The construction method of a wind power prediction model according to claim 6, characterized in that: The model training further includes updating the model parameters of the preset prediction model based on the kinkajou optimization algorithm during the model training process.

8. A construction device using a construction method of a wind power prediction model as described in any one of claims 1 to 7, characterized in that: Including, An acquisition module (501), a time-frequency feature set construction module (502), and a model training module (503); The acquisition module (501) is configured to acquire historical meteorological data and historical power data of a target wind farm; The time-frequency feature set construction module (502) is configured to construct a time-frequency feature set based on the historical meteorological data and the historical power data; wherein, the time-frequency feature set includes a first time-frequency feature and a second time-frequency feature, the first time-frequency feature is used to represent the periodic information included in the historical meteorological data and the historical power data, and the second time-frequency feature is used to represent the local frequency information of the historical meteorological data and the historical power data at multiple time scales in the time-frequency domain; The model training module (503) is configured to input the time-frequency feature set into a preset prediction model constructed in advance for model training, and during the model training process, learn the mapping relationship between the first time-frequency feature and the second time-frequency feature and the historical power data until the model converges to obtain a power prediction model.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the method for constructing a wind power prediction model according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the method for constructing a wind power prediction model according to any one of claims 1 to 7 are implemented.

Images