Wind power prediction method and device, electronic equipment, product and storage medium

Abstract

The invention provides a wind power prediction method and device, electronic equipment, a product and a storage medium, and relates to the technical field of power prediction. The method comprises the following steps: inputting wind power time sequence data into a signal processing layer in a wind power prediction model to obtain a frequency component set output by the signal processing layer; inputting the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer; wherein the signal processing layer is used for extracting characteristics of the wind power time sequence data in different frequency ranges so as to obtain a frequency component set. According to the method, the defects that in the prior art, frequency components in the wind power time sequence data cannot be effectively extracted, potential information in the data cannot be fully mined, and the accuracy and reliability of a prediction model are affected can be overcome.

Description

Technical Field

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

[0002] With the rapid development of renewable energy, wind power, as one of the important clean energy sources, has been widely used around the world. The prediction of wind power is a key issue in the operation of wind power systems and grid dispatching. Accurate wind power prediction helps to improve the utilization efficiency of wind power, reduce energy waste, and promote the stability of the power system. However, due to the volatility, randomness and complexity of wind power, wind power prediction faces great challenges. Wind power time series data is usually affected by many factors, such as weather changes, wind speed changes, seasonal fluctuations, etc. These factors make the prediction results of wind power have large uncertainty and errors.

[0003] To address this problem, existing wind power prediction methods mostly use models based on time series analysis and machine learning algorithms, such as support vector machines (SVM), neural networks, long short-term memory networks (LSTM), etc. These methods are usually trained through historical wind power data to establish wind power prediction models. However, the frequency components in wind power time series data are extremely complex, and components of different frequencies have different effects on the prediction results. Traditional prediction models often cannot effectively extract the characteristics of these different frequencies, and thus cannot fully mine the potential information in the data, thus affecting the accuracy and reliability of the prediction model. Summary of the invention

[0004] The present invention provides a wind power prediction method, device, electronic device, product and storage medium to solve the defect in the prior art that the frequency components in wind power time series data cannot be effectively extracted, and the potential information in the data cannot be fully mined, thereby affecting the accuracy and reliability of the prediction model.

[0005] The present invention provides a wind power prediction method, comprising: Inputting wind power time series data into a signal processing layer in a wind power prediction model to obtain a frequency component set output by the signal processing layer; Inputting the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer; Among them, the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and the wind power prediction result labels corresponding to the sample wind power time series data.

[0006] According to a wind power prediction method provided by the present invention, the wind power time series data is input into a signal processing layer in a wind power prediction model to obtain a frequency component set output by the signal processing layer, including: Inputting wind power time series data into the signal decomposition layer in the signal processing layer to obtain an IMF sequence set output by the signal decomposition layer; the signal decomposition layer is used to decompose the wind power time series data into a plurality of characteristic sequences with different frequencies to obtain an IMF sequence set; The IMF sequence set is input to the signal reconstruction layer in the signal processing layer to obtain a frequency component set output by the signal reconstruction layer.

[0007] According to a wind power prediction method provided by the present invention, the wind power time series data is input into the signal decomposition layer in the signal processing layer to obtain the IMF sequence set output by the signal decomposition layer, including: Adding noise of different frequencies to the wind power time series data to obtain a noisy wind power data set; Decomposing each noise wind power data in the noise wind power data set to obtain a plurality of decomposition results; wherein each noise wind power data corresponds to one decomposition result; The multiple decomposition results are averaged to obtain an IMF sequence set.

[0008] According to a wind power prediction method provided by the present invention, the step of inputting the IMF sequence set to a signal reconstruction layer in the signal processing layer to obtain a frequency component set output by the signal reconstruction layer includes: Performing logarithmic average calculation on the volatility of each IMF sequence in the IMF sequence set to determine the volatility characteristics of each IMF sequence in the IMF sequence set; Based on the sorting results of the fluctuation characteristics of each of the IMF sequences, determining a high-frequency reconstruction component, a medium-frequency reconstruction component and a low-frequency reconstruction component; A frequency component set is determined based on the high frequency reconstruction component, the medium frequency reconstruction component and the low frequency reconstruction component.

[0009] According to a wind power prediction method provided by the present invention, the step of inputting the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer includes: Predicting the frequency component set based on the prediction layer to obtain a high-frequency component prediction result, a mid-frequency component prediction result, and a low-frequency component prediction result; The high frequency component prediction result, the medium frequency component prediction result and the low frequency component prediction result are added together to obtain a wind power prediction result.

[0010] According to a wind power prediction method provided by the present invention, the prediction layer in the wind power prediction model is obtained based on SSA algorithm optimization; the prediction layer also includes an attention mechanism.

[0011] The present invention also provides a wind power prediction device, comprising: A signal processing module, which is used to input the wind power time series data into the signal processing layer in the wind power prediction model to obtain a frequency component set output by the signal processing layer; A prediction module, which is used to input the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer; Among them, the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and the wind power prediction result labels corresponding to the sample wind power time series data.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, any of the wind power prediction methods described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the wind power prediction method as described in any one of the above is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the wind power prediction methods described above.

[0015] The wind power prediction method, device, electronic device, product and storage medium provided by the present invention input wind power time series data into the signal processing layer in the wind power prediction model to obtain the frequency component set output by the signal processing layer; input the frequency component set into the prediction layer in the wind power prediction model to obtain the wind power prediction result output by the prediction layer; wherein the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain the frequency component set; the wind power prediction model is obtained by training based on sample wind power time series data and the wind power prediction result labels corresponding to the sample wind power time series data; the present invention is achieved by Features in different frequency ranges are extracted to obtain frequency component sets; the problem that traditional methods are difficult to effectively extract frequency features is solved, so that the potential information in the data can be more fully mined, so that the wind power prediction model can effectively cope with the complexity, volatility and randomness of wind power data, and improve the model's understanding and prediction capabilities of wind power time series; further, the frequency component set is input into the prediction layer of the wind power prediction model, which enables the model to separately predict the predicted contribution of components of different frequencies to wind power, avoids ignoring these detailed components, overcomes the problem of insufficient frequency component extraction in traditional wind power prediction methods when processing wind power time series data, and significantly improves the prediction accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 This is one of the flow charts of the wind power prediction method provided by the present invention.

[0018] Figure 2 This is the second flow chart of the wind power prediction method provided by the present invention.

[0019] Figure 3 It is a structural schematic diagram of the wind power prediction device provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] In the field of wind power forecasting, the volatility, randomness and complexity of wind power have always been the core challenges faced by forecasting models. When processing wind power time series data, traditional wind power forecasting models are unable to effectively extract and distinguish different frequency components in the data, resulting in limited recognition of key features. Patent applicants have found in actual applications that wind power time series data often contain complex frequency characteristics. Fluctuations of different frequencies are affected by factors such as weather and wind speed, and there are significant differences in short-term and long-term trends. Since existing models cannot effectively decompose and extract these frequency characteristics, the forecast results have large errors, especially when wind power changes dramatically or long-term fluctuations are strong, the reliability of model predictions is further reduced.

[0023] In order to address this issue, the applicant conducted in-depth research on time series analysis and machine learning models commonly used in the prior art (such as support vector machines, neural networks, and LSTM, etc.), and found that although these methods can capture the trend information of time series to a certain extent, they still have significant deficiencies in processing the frequency components of wind power data. Most of these methods rely on overall data characteristics and lack the ability to effectively extract and separate high-frequency, medium-frequency, and low-frequency components in wind power data. Therefore, in complex wind power change scenarios, it is impossible to fully explore the potential frequency feature information in the data. This deficiency directly affects the prediction accuracy of the model, making it difficult to cope with short-term drastic fluctuations and long-term trend changes in wind power.

[0024] In view of the above problems, the present invention proposes the following embodiments.

[0025] Figure 1 is a flow chart of the wind power prediction method provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 110, inputting the wind power time series data into a signal processing layer in a wind power prediction model to obtain a frequency component set output by the signal processing layer; Here, wind power time series data refers to continuous time series data generated based on the actual power generation of wind farms, which is usually sampled at fixed time intervals. This data reflects the changing trend of wind power generation over time and is affected by external factors such as wind speed and climate.

[0026] Here, the signal processing layer refers to a functional module in the wind power prediction model, which is specially used to process time series data. Its function is to decompose the complex original time series into multiple frequency components. The frequency component set refers to the set of multiple frequency components extracted from the wind power data. Each frequency component represents the fluctuation characteristics of different time scales in the data. For example, the high-frequency component corresponds to short-term fluctuations and reflects the rapidly changing part; the low-frequency component corresponds to the long-term trend and reflects the overall trend of change.

[0027] It should be noted that the signal processing layer deconstructs the complex wind power time series data into a set of frequency components that are easy to analyze and process, laying the foundation for the subsequent prediction steps. This signal processing method makes wind power prediction more accurate at different time scales, effectively solving the problem of insufficient processing of data non-stationarity in traditional methods.

[0028] For example, wind power time series data is highly nonlinear and non-stationary, showing periodic changes, random fluctuations, and dramatic short-term fluctuations. Traditional models are difficult to directly extract effective features from these complex data, so the signal processing layer here decomposes the multiple frequency components in the original time series data so that each frequency component contains only a relatively single frequency feature.

[0029] For the specific steps of using the signal processing layer, please refer to the following embodiments, which will not be elaborated here.

[0030] Step 120, inputting the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer; Among them, the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and the wind power prediction result labels corresponding to the sample wind power time series data.

[0031] Here, the prediction layer is a functional module used to generate prediction results in the wind power prediction model. It builds a prediction model for the wind power time series based on the extracted frequency features and historical data through a training process. The sample wind power time series data and the wind power prediction result label are the data required for training the prediction model. The sample wind power time series data contains the wind power time series over a period of time in the past, and the wind power prediction result label is the actual wind power output value. Through model training based on these data, the prediction layer learns to recognize patterns in the time series and can predict future data.

[0032] The wind power prediction method provided by the embodiment of the present invention obtains a frequency component set by extracting features of different frequency ranges in the signal processing layer; it solves the problem that traditional methods are difficult to effectively extract frequency features, thereby being able to more fully mine the potential information in the data, so that the wind power prediction model can effectively cope with the complexity, volatility and randomness of wind power data, and improves the model's understanding and prediction capabilities of wind power time series; further, the frequency component set is input into the prediction layer in the wind power prediction model, so that the model can separately predict the predicted contributions of components of different frequencies to the wind power, avoiding ignoring these detailed components, overcoming the problem of insufficient frequency component extraction in traditional wind power prediction methods when processing wind power time series data, and significantly improving the prediction accuracy and reliability.

[0033] Figure 2 is a flow chart of the wind power prediction method provided by the present invention, such as Figure 2 As shown, Figure 2 The processing flow of wind power time series data by the signal processing layer is demonstrated.

[0034] Based on any of the above embodiments, in the method, inputting the wind power time series data into the signal processing layer in the wind power prediction model to obtain the frequency component set output by the signal processing layer includes: Inputting wind power time series data into the signal decomposition layer in the signal processing layer to obtain an IMF sequence set output by the signal decomposition layer; the signal decomposition layer is used to decompose the wind power time series data into a plurality of characteristic sequences with different frequencies to obtain an IMF sequence set; The IMF sequence set is input to the signal reconstruction layer in the signal processing layer to obtain a frequency component set output by the signal reconstruction layer.

[0035] Here, the IMF sequence set wind power time series data is processed by the signal decomposition layer to obtain multiple intrinsic mode function sequences, namely multiple IMF sequences. Each IMF sequence represents the fluctuation characteristics of the data at different frequencies. Each IMF captures a certain frequency band of the data, often including detailed descriptions of different components such as high frequency and low frequency.

[0036] It should be noted that the signal decomposition layer is responsible for decomposing the original wind power time series and extracting IMF sequences of different frequencies, while the signal reconstruction layer reconstructs and filters the decomposed IMF sequences to obtain a more useful set of frequency components. These two layers play the role of decomposition and feature extraction respectively.

[0037] Exemplarily, the signal processing layer may be obtained based on CEEMDAN (Complete Ensemble Empirical Mode Decomposition of Adaptive Noise).

[0038] For example, the wind power time series data is input into the signal decomposition layer, and the decomposition layer performs time-frequency decomposition on the data. This process can use Empirical Mode Decomposition (EMD) to decompose the data into multiple components with different frequencies, called Intrinsic Mode Functions (IMF), that is, multiple IMF sequences. Each IMF sequence corresponds to a specific frequency component and describes the change pattern of the data within a certain frequency band.

[0039] In the wind power prediction method provided by the embodiment of the present invention, the wind power time series data first enters the signal decomposition layer, and the decomposition layer decomposes the data into multiple intrinsic mode function (IMF) sequences of different frequencies to form an IMF sequence set. This decomposition process can separate the different frequency components of the wind power data, so that different features such as high frequency and low frequency can be extracted and analyzed independently. Subsequently, the IMF sequence set enters the signal reconstruction layer, and the reconstruction layer screens and combines these frequency components to generate a frequency component set; this decomposition and reconstruction process enables the frequency component set to more accurately reflect the characteristic information in the wind power data, effectively improving the adaptability of the prediction model to complex fluctuations, thereby improving the accuracy and reliability of wind power prediction.

[0040] Based on any of the above embodiments, in the method, inputting the wind power time series data into the signal decomposition layer in the signal processing layer to obtain the IMF sequence set output by the signal decomposition layer includes: Adding noise of different frequencies to the wind power time series data to obtain a noisy wind power data set; Decomposing each noise wind power data in the noise wind power data set to obtain a plurality of decomposition results; wherein each noise wind power data corresponds to one decomposition result; The multiple decomposition results are averaged to obtain an IMF sequence set.

[0041] Exemplarily, noises of different frequencies are added to the wind power time series data, and each wind power time series data is added with a frequency to form a noisy wind power data. By adding noises of different frequencies, multiple noisy wind power data are obtained to obtain a noisy wind power data set; further, each noisy wind power data in the noisy wind power data set is decomposed to obtain multiple decomposition results, one noisy wind power data corresponds to one decomposition result; one decomposition result is an IMF sequence set; further, the decomposition results corresponding to all the noisy wind power data are averaged to obtain the final IMF sequence set. This averaging operation can smooth the data fluctuations introduced by noise, eliminate the random components generated by noise during the decomposition process, and make the final IMF sequence more representative and truly reflect the frequency characteristics of the original wind power data.

[0042] Specifically, for each decomposition result, that is, each IMF sequence set, it is necessary to ensure that the frequency level of its IMF sequence is consistent. Usually, the alignment is based on the frequency or volatility characteristics of the IMF to ensure that the i-th IMF in each IMF sequence set has the same or similar frequency components. The i-th IMF obtained from all decompositions is averaged, that is, the IMFs at the same position in each group of IMF sequences are averaged point by point to obtain the final i-th IMF. By performing point-by-point averaging on each IMF component, the random components introduced by noise can be smoothed out, so that the frequency components retained in the IMF sequence set are more stable and accurate, and truly reflect the characteristics of the original time series.

[0043] The wind power prediction method provided by the embodiment of the present invention generates multiple noisy wind power data by adding noise of different frequencies to the wind power time series data, and then decomposes each noise wind power data to obtain a corresponding decomposition result. This noise-assisted decomposition process helps to reveal the essential characteristics of the frequency components in the original data and suppress the random noise interference in the traditional decomposition process. Subsequently, the IMF sequences of the multiple decomposition results are averaged to obtain a stable IMF sequence set, thereby eliminating the random disturbance introduced by the noise, so that the IMF sequence set more accurately reflects the frequency characteristics of the original wind power data. This process significantly improves the accuracy of wind power time series feature extraction and provides a more stable and reliable feature input for subsequent wind power prediction.

[0044] Based on any of the above embodiments, in the method, inputting the IMF sequence set to a signal reconstruction layer in the signal processing layer to obtain a frequency component set output by the signal reconstruction layer includes: Performing logarithmic average calculation on the volatility of each IMF sequence in the IMF sequence set to determine the volatility characteristics of each IMF sequence in the IMF sequence set; Based on the sorting results of the fluctuation characteristics of each of the IMF sequences, determining a high-frequency reconstruction component, a medium-frequency reconstruction component and a low-frequency reconstruction component; A frequency component set is determined based on the high frequency reconstruction component, the medium frequency reconstruction component and the low frequency reconstruction component.

[0045] Here, volatility refers to the amplitude of vibration or the intensity of change in each IMF series, which is used to describe the volatility characteristics of the series.

[0046] In one embodiment, by calculating the logarithmic average of the volatility of each IMF series, the volatility characteristics of each IMF can be reflected more stably. This logarithmic average can reduce the influence of extreme values, obtain smoother results, and more effectively capture the characteristics of the IMF series at different frequencies; further, each IMF is sorted according to its volatility characteristics, and is divided into high-frequency reconstruction components, medium-frequency reconstruction components, and low-frequency reconstruction components. The sorting is based on the volatility of the IMF to distinguish the main frequency components. The high-frequency component usually represents rapidly changing fluctuations, while the low-frequency component reflects the trend or cyclical fluctuations in the data.

[0047] Obviously, this processing can effectively extract the main fluctuation information of each IMF series, so that when processing different frequency components, it can avoid the interference of some random noise, which helps to grasp the changing characteristics of each frequency component more accurately; by layering the processing of different frequency components, the model can better focus on the information of each frequency level, so as to more clearly distinguish short-term fluctuations and long-term trends in the prediction, and further improve the prediction performance of the model.

[0048] The wind power prediction method provided by the embodiment of the present invention extracts the volatility characteristics of each IMF by calculating the logarithmic average of the volatility of each IMF sequence in the IMF sequence set; this processing method can not only effectively reduce the impact of extreme values, but also more accurately retain the frequency characteristics of each IMF sequence. After obtaining the volatility characteristics of each IMF sequence, these characteristics are sorted and the sequences are divided into high-frequency reconstruction components, medium-frequency reconstruction components and low-frequency reconstruction components, so that the high-frequency components can reflect the rapidly changing short-term fluctuations, while the low-frequency components reveal the trend or periodic changes; finally, the complete frequency component set is determined through the three levels of frequency reconstruction components, and the refined and layered wind power characteristic information is obtained.

[0049] Based on any of the above embodiments, in the method, inputting the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer includes: Predicting the frequency component set based on the prediction layer to obtain a high-frequency component prediction result, a mid-frequency component prediction result, and a low-frequency component prediction result; The high frequency component prediction result, the medium frequency component prediction result and the low frequency component prediction result are added together to obtain a wind power prediction result.

[0050] Here, the high-frequency reconstruction component represents the rapidly changing part of the wind power time series, which is usually affected by short-term factors, such as instantaneous fluctuations in wind speed. The medium-frequency reconstruction component represents the medium-term changes in wind power, which are affected by factors such as weather changes. The low-frequency reconstruction component refers to the long-term trend component, which is often related to seasonal or geographical changes and provides long-term trend information in the series.

[0051] In one embodiment, the prediction layer predicts the high-frequency reconstruction component, the medium-frequency reconstruction component and the low-frequency reconstruction component in the frequency component concentration respectively. The high-frequency reconstruction component reflects short-term rapid fluctuations, so its prediction model focuses on processing the fluctuation characteristics of short-term data; the medium-frequency reconstruction component prediction focuses on the medium-term change characteristics, while the low-frequency reconstruction component pays more attention to long-term trends and cyclical changes; further, the high-frequency component prediction results, the medium-frequency component prediction results and the low-frequency component prediction results are added together to obtain the final wind power prediction result. This frequency division prediction and addition strategy can more comprehensively integrate different frequency characteristics, thereby obtaining an accurate and stable wind power prediction.

[0052] The wind power prediction method provided by the embodiment of the present invention realizes multi-level modeling of wind power through the method of frequency division prediction and component addition, which can focus on short-term fluctuations, medium-term changes and long-term trends respectively, and significantly improves the accuracy and reliability of the prediction. The frequency band addition structure also improves the robustness of the model, making the final prediction result more sensitive to the characteristics of different frequency bands, adapting to the strong volatility of wind power, and improving the comprehensive effect of wind power prediction.

[0053] Based on any of the above embodiments, in the method, the prediction layer in the wind power prediction model is optimized based on the SSA (Sparrow Search Algorithm); the prediction layer also includes an attention mechanism.

[0054] Here, the SSA algorithm is used to optimize the parameters in the prediction layer and find the best prediction model configuration. Through SSA optimization, the model can fit the wind power data more accurately and improve the prediction accuracy. SSA is an optimization algorithm based on swarm intelligence, which simulates the process of sparrow foraging behavior and has strong global search capabilities. SSA finds the optimal solution by simulating the self-regulating behavior of sparrow groups in the process of finding food. It is an optimization method that can effectively avoid local optimal solutions and is suitable for complex and high-dimensional prediction tasks.

[0055] Here, the attention mechanism is a mechanism that simulates the human attention concentration process and can dynamically adjust the degree of attention to different information. Simply put, the attention mechanism can dynamically assign weights according to the importance of the input data, giving more attention to important features and less attention to unimportant information. In the prediction layer, the attention mechanism is used to adjust the influence of each part of the data in the wind power prediction process according to the importance of different time steps and different frequency components. Through this mechanism, the model can focus more on key data features, thereby enhancing the accuracy of the prediction.

[0056] For example, the prediction layer can be LSTM (Long Short-Term Memory), which is a special recurrent neural network (RNN) structure. Traditional RNN models often face the problem of gradient vanishing or gradient exploding when processing long sequences, resulting in the model being unable to effectively learn information over a long time span. LSTM can avoid gradient vanishing and maintain a stable flow of information through its unique gating mechanism (such as forget gate, input gate, and output gate), ensuring that it can process longer time series and can stably update the parameters of the model.

[0057] Obviously, wind power data has significant time dependence, that is, the change of wind power is affected by factors such as wind speed and meteorological conditions in the past. The long-term memory ability of LSTM makes it particularly suitable for processing such time series data. Through the LSTM model, the long-term dependency in historical wind power data can be effectively captured, thereby improving the prediction accuracy.

[0058] The wind power prediction method provided by the embodiment of the present invention can effectively process the time dependency and complexity of wind power data by introducing LSTM into the wind power prediction model, overcome the defects of the traditional RNN model, and thus provide a more accurate and reliable wind power prediction. This not only improves the prediction accuracy of the model, but also enhances the system's adaptability to long-term dependencies and external influences, which helps to improve the efficiency of the wind power generation system and the stability of the power grid dispatch.

[0059] The wind power prediction device provided by the present invention is described below. The wind power prediction device described below and the wind power prediction method described above can be referred to each other.

[0060] Figure 3 The schematic diagram of the wind power prediction device provided by the present invention is as follows: Figure 3 As shown, the wind power prediction device comprises: A signal processing module 310, which is used to input the wind power time series data into the signal processing layer in the wind power prediction model to obtain a frequency component set output by the signal processing layer; A prediction module 320, which is used to input the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer; Among them, the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and the wind power prediction result labels corresponding to the sample wind power time series data.

[0061] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the wind power prediction method, which includes: inputting the wind power time series data into the signal processing layer in the wind power prediction model to obtain the frequency component set output by the signal processing layer; inputting the frequency component set into the prediction layer in the wind power prediction model to obtain the wind power prediction result output by the prediction layer; wherein the signal processing layer is used to extract the features of the wind power time series data in different frequency ranges to obtain the frequency component set; the wind power prediction model is trained based on the sample wind power time series data and the wind power prediction result label corresponding to the sample wind power time series data.

[0062] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0063] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the wind power prediction method provided by the above methods, which method includes: inputting wind power time series data into a signal processing layer in a wind power prediction model to obtain a frequency component set output by the signal processing layer; inputting the frequency component set into a prediction layer in the wind power prediction model to obtain a wind power prediction result output by the prediction layer; wherein the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and wind power prediction result labels corresponding to the sample wind power time series data.

[0064] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the wind power prediction method provided by the above-mentioned methods, the method comprising: inputting wind power time series data into a signal processing layer in a wind power prediction model to obtain a frequency component set output by the signal processing layer; inputting the frequency component set into a prediction layer in the wind power prediction model to obtain a wind power prediction result output by the prediction layer; wherein the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and wind power prediction result labels corresponding to the sample wind power time series data.

[0065] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0066] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A wind power prediction method, characterized in that: include: Inputting wind power time series data into a signal processing layer in a wind power prediction model to obtain a frequency component set output by the signal processing layer; Inputting the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer; Among them, the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and the wind power prediction result labels corresponding to the sample wind power time series data.

2. The wind power prediction method according to claim 1, characterized in that: The step of inputting the wind power time series data into the signal processing layer in the wind power prediction model to obtain the frequency component set output by the signal processing layer includes: Inputting the wind power time series data into the signal decomposition layer in the signal processing layer to obtain an IMF sequence set output by the signal decomposition layer; the signal decomposition layer is used to decompose the wind power time series data into a plurality of characteristic sequences with different frequencies to obtain an IMF sequence set; The IMF sequence set is input to the signal reconstruction layer in the signal processing layer to obtain a frequency component set output by the signal reconstruction layer.

3. The wind power prediction method according to claim 2, characterized in that: The step of inputting the wind power time series data into the signal decomposition layer in the signal processing layer to obtain the IMF sequence set output by the signal decomposition layer includes: Adding noise of different frequencies to the wind power time series data to obtain a noisy wind power data set; Decomposing each noise wind power data in the noise wind power data set to obtain a plurality of decomposition results; wherein each noise wind power data corresponds to one decomposition result; The multiple decomposition results are averaged to obtain an IMF sequence set.

4. The wind power prediction method according to claim 2, characterized in that: The step of inputting the IMF sequence set to a signal reconstruction layer in the signal processing layer to obtain a frequency component set output by the signal reconstruction layer includes: Performing logarithmic average calculation on the volatility of each IMF sequence in the IMF sequence set to determine the volatility characteristics of each IMF sequence in the IMF sequence set; Based on the sorting results of the fluctuation characteristics of each of the IMF sequences, determining a high-frequency reconstruction component, a medium-frequency reconstruction component, and a low-frequency reconstruction component; A frequency component set is determined based on the high frequency reconstruction component, the medium frequency reconstruction component and the low frequency reconstruction component.

5. The wind power prediction method according to claim 1, characterized in that: The step of inputting the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer includes: Predicting the frequency component set based on the prediction layer to obtain a high-frequency component prediction result, a mid-frequency component prediction result, and a low-frequency component prediction result; The high frequency component prediction result, the medium frequency component prediction result and the low frequency component prediction result are added together to obtain a wind power prediction result.

6. The wind power prediction method according to claim 5, characterized in that: The prediction layer in the wind power prediction model is optimized based on the SSA algorithm; the prediction layer also includes an attention mechanism.

7. A wind power prediction device, characterized in that: include: A signal processing module, which is used to input the wind power time series data into the signal processing layer in the wind power prediction model to obtain a frequency component set output by the signal processing layer; A prediction module, which is used to input the frequency component set into a prediction layer in a wind power prediction model to obtain a wind power prediction result output by the prediction layer; Among them, the signal processing layer is used to extract the characteristics of the wind power time series data in different frequency ranges to obtain a frequency component set; the wind power prediction model is trained based on sample wind power time series data and the wind power prediction result labels corresponding to the sample wind power time series data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the wind power prediction method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the wind power prediction method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the wind power prediction method according to any one of claims 1 to 6 is implemented.

Images