Wind power prediction method and related device

Abstract

The invention discloses a wind power prediction method and a related device, and relates to the technical field of data prediction, and the wind power prediction method comprises the steps: obtaining historical wind power data; performing empirical mode decomposition on the historical wind power data to obtain a plurality of target components, the plurality of target components including a plurality of intrinsic mode function (IMF) components and a residual component; obtaining target component features corresponding to the plurality of target components; using a power prediction model based on a relevance vector machine obtained by pre-training, and taking the target component characteristics corresponding to the plurality of target components as a basis to obtain prediction values corresponding to the plurality of target components at the prediction time; and adding the predicted values respectively corresponding to the plurality of target components at the same moment to obtain a wind power data predicted value of the predicted time. The wind power prediction method disclosed by the invention has relatively high prediction precision and calculation efficiency, and also has relatively high stability and generalization ability.

Description

Technical Field

[0001] This application relates to the technical field of data prediction, and particularly to a wind power prediction method and related devices. Background Art

[0002] With the continuous growth of the global demand for renewable energy, wind power generation, as a clean energy source, has been widely used. However, the generated power of a wind farm, i.e., wind power, has significant volatility and uncertainty, which poses a huge challenge to the dispatching and stable operation of the power system. Therefore, accurately predicting the wind power at future times has become an important research direction for ensuring the safe operation of the power grid and optimizing resource allocation, and how to accurately predict the wind power at future times is an urgent problem to be solved at present. Summary of the Invention

[0003] In view of this, this application provides a wind power prediction method and related devices for accurately predicting the wind power at future times, and its technical solutions are as follows:

[0004] The first aspect of this application provides a wind power prediction method, including:

[0005] Obtain historical wind power data;

[0006] Perform empirical mode decomposition on the historical wind power data to obtain a number of target components, where the number of target components includes a number of intrinsic mode function IMF components and a residual component;

[0007] Obtain the target component features corresponding to the number of target components respectively;

[0008] Using a pre-trained power prediction model, based on the target component features corresponding to the number of target components respectively, obtain the predicted values corresponding to the number of target components at the prediction time, where the power prediction model is trained by using the component features corresponding to a number of components obtained by performing empirical mode decomposition on historical wind power data samples as training samples and the real wind power data at the prediction time as sample labels to train a relevant vector machine;

[0009] Add the predicted values corresponding to the number of target components at the same moment to obtain the predicted value of the wind power data at the prediction time.

[0010] In a possible implementation manner, the step of using a pre-trained power prediction model, based on the target component features corresponding to the number of target components respectively, to obtain the predicted values corresponding to the number of target components at the prediction time includes:

[0011] Combine the target component features corresponding to the number of target components respectively to obtain a combined feature;

[0012] Input the combined features into a pre-trained power prediction model to obtain the probability distribution of the predicted values corresponding to each target component output by the power prediction model at the prediction time;

[0013] Calculate the expected value of the probability distribution of the predicted values corresponding to each target component at the prediction time to obtain the predicted value corresponding to each target component at the prediction time.

[0014] In a possible implementation manner, the empirical mode decomposition of the historical wind power data to obtain a number of target components includes:

[0015] Preprocess the historical wind power data, where the preprocessing includes one or more of the following processes: denoising, missing value filling, outlier deletion or correction;

[0016] Perform empirical mode decomposition on the preprocessed historical wind power data to obtain a number of target components.

[0017] In a possible implementation manner, the empirical mode decomposition of the preprocessed historical wind power data to obtain a number of target components includes:

[0018] Determine the extension length N according to the preprocessed historical wind power data, where N is an integer greater than 0;

[0019] Mirror-copy the first N data points of the preprocessed historical wind power data to the left end of the preprocessed historical wind power data, and mirror-copy the last N data points of the preprocessed historical wind power data to the right end of the preprocessed historical wind power data to obtain the extended historical wind power data;

[0020] Perform empirical mode decomposition on the extended historical wind power data to obtain a number of components;

[0021] Extract the components corresponding to the preprocessed historical wind power data from the number of components to obtain a number of target components.

[0022] In a possible implementation manner, the empirical mode decomposition of a historical wind power data includes:

[0023] Add white noise with different amplitudes to the historical wind power data multiple times to obtain multiple noise-assisted data;

[0024] Perform empirical mode decomposition on the multiple noise-assisted data respectively to obtain the component sets corresponding to the multiple noise-assisted data respectively, where the component set corresponding to any noise-assisted data includes a number of IMF components and a residual component;

[0025] Average the corresponding components in the component sets respectively corresponding to the multiple noise-assisted data to obtain a final decomposition result.

[0026] In a possible implementation, obtaining the component features respectively corresponding to several IMF components includes:

[0027] For each IMF component:

[0028] Extract one or more of the following features from this IMF component: statistical features, time-domain features, frequency-domain features, time-frequency features, to obtain the initial component features corresponding to this IMF component;

[0029] Select the features useful for wind power prediction from the initial component features corresponding to this IMF component to obtain the target component features corresponding to this IMF component.

[0030] In a possible implementation, the statistical features include some or all of the following features: mean, variance, skewness, kurtosis;

[0031] The time-domain features include some or all of the following features: autocorrelation function, zero-crossing rate, wave peak and wave valley features;

[0032] The frequency-domain features include some or all of the following features: power spectral density, dominant frequency, spectral centroid;

[0033] The time-frequency features include some or all of the following features: features characterizing the frequency characteristics of the IMF component in different time periods, features characterizing the time-frequency characteristics of the IMF at different scales and positions.

[0034] In a possible implementation, the selecting the features useful for wind power prediction from the initial component features corresponding to this IMF component includes:

[0035] Adopt one or more of the following feature selection methods to select the features useful for wind power prediction from the initial component features corresponding to this IMF component: feature selection method based on principal component analysis, feature selection method based on least absolute shrinkage and selection operator, feature selection method based on stepwise regression, feature selection method based on random forest.

[0036] The second aspect of this application provides a wind power prediction device, including: a data acquisition module, a data decomposition module, a feature acquisition module, a power prediction module, and a power prediction value determination module;

[0037] The data acquisition module is used to acquire historical wind power data;

[0038] The data decomposition module is used to perform empirical mode decomposition on the historical wind power data to obtain a number of target components, where the number of target components includes a number of intrinsic mode function (IMF) components and a residual component;

[0039] The feature acquisition module is used to acquire the target component features corresponding to the number of target components respectively;

[0040] The power prediction module is used to utilize a pre-trained power prediction model, and based on the target component features corresponding to the number of target components respectively, obtain the predicted values corresponding to the number of target components at the prediction time. Among them, the power prediction model is trained by using the component features corresponding to a number of components obtained by performing empirical mode decomposition on historical wind power data samples as training samples, and the true wind power data at the prediction time as sample labels to train a relevant vector machine;

[0041] The power prediction value determination module is used to add the predicted values corresponding to the number of target components at the same moment respectively to obtain the predicted value of the wind power data at the prediction time.

[0042] A third aspect of the present application provides an electronic device, including at least one processor and a memory connected to the processor, where:

[0043] The memory is used to store a computer program;

[0044] The processor is used to execute the computer program so that the electronic device can implement the steps of any one of the above-mentioned wind power prediction methods.

[0045] A fourth aspect of the present application provides a computer storage medium, where the storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can be enabled to implement the steps of any one of the above-mentioned wind power prediction methods.

[0046] A fifth aspect of the present application provides a computer program product, including computer-readable instructions, and when the computer-readable instructions run on an electronic device, the electronic device is enabled to implement the steps of any one of the above-mentioned wind power prediction methods.

[0047] With the above technical solution, for the wind power prediction method provided by this application, after obtaining historical wind power data, the historical wind power data is first subjected to empirical mode decomposition to obtain several target components (i.e., several IMF components and one residual component). Then, the target component features corresponding to the several target components are obtained. Next, using the pre-trained power prediction model based on the relevance vector machine, based on the target component features corresponding to the several target components, the predicted values corresponding to the several target components at the prediction time are obtained. Finally, the predicted values corresponding to the several target components at the same moment are added together to obtain the predicted value of the wind power data at the prediction time. Considering that the historical wind power data has significant non-linear and non-stationary characteristics, the wind power prediction method provided by this application utilizes the adaptive decomposition ability of empirical mode decomposition to decompose the complex historical wind power data into several stationary IMF components and residual components. These components correspond to fluctuations on different time scales, significantly reducing the non-stationarity and complexity of the original historical wind power data, which helps with subsequent power prediction. At the same time, the wind power prediction method provided by this application uses the relevance vector machine for prediction modeling. The relevance vector machine is a sparse probability model based on Bayesian theory, which can automatically determine the relevant vectors to achieve sparsity. Thus, while ensuring high prediction accuracy, it can reduce the complexity of the model and improve the calculation efficiency. In summary, the wind power prediction method provided by this application has high prediction accuracy and calculation efficiency, and also has high stability and generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0049] Figure 1 It is a schematic diagram of a system architecture related to this application;

[0050] Figure 2 It is a schematic diagram of a hardware structure of a terminal provided by an embodiment of this application;

[0051] Figure 3 It is a schematic diagram of a hardware structure of a server provided by an embodiment of this application;

[0052] Figure 4 It is a schematic flowchart of the wind power prediction method provided by an embodiment of this application;

[0053] Figure 5A schematic flowchart of a process for obtaining predicted values corresponding to a plurality of target components at a prediction time based on component features corresponding to the plurality of target components by using a pre-trained power prediction model provided by an embodiment of the present application;

[0054] Figure 6 A schematic structural diagram of a wind power prediction device provided by an embodiment of the present application. Detailed implementation manners

[0055] The embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application. The terms used in the implementation manners part of the present application are only used to explain the specific embodiments of the present application, rather than intended to limit the present application.

[0056] The embodiments of the present application will be described below with reference to the accompanying drawings. Those skilled in the art will know that with the development of technology and the emergence of new scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

[0057] The terms "first", "second", etc. in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such terms can be interchanged under appropriate circumstances, which is only a way of distinguishing when describing objects with the same attributes in the embodiments of the present application. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, so that a process, method, system, product or device comprising a series of units does not have to be limited to those units, but may include other units not clearly listed or inherent to these processes, methods, products or devices.

[0058] In one possible implementation manner, as Figure 1 shown, the system architecture involved in the present application may include a terminal 101 and a server 102. The terminal 101 can interact with the server 102 through a network (wired network or wireless network). Among them, the server 102 may include one or more servers ( Figure 1 illustrated by taking one server as an example). The terminal can obtain historical wind power data and transmit the historical wind power data to the server through the network. The server predicts the wind power at a future time according to the historical wind power data.

[0059] In another possible implementation manner, the system architecture involved in the present application may include a terminal. The terminal has strong data processing capabilities. The terminal can obtain historical wind power data and then predict the wind power at a future time according to the historical wind power data.

[0060] Next, the product form of the above terminal will be described.

[0061] The above-mentioned terminal may be a mobile phone, a tablet computer, a wearable device, a vehicle-mounted device, a robot, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiments of the present application do not make any restrictions thereon.

[0062] Figure 2 FIG. shows an optional hardware structure diagram of the terminal.

[0063] Referring to Figure 2 as shown, the terminal may include a radio frequency unit 210, a memory 220, an input unit 230, a display unit 240, a camera 250 (optional), an audio circuit 260 (optional), a speaker 261 (optional), a microphone 262 (optional), a headphone jack 263 (optional), a processor 270, an external interface 280, a power supply 290, and other components. Those skilled in the art can understand that Figure 2 this is only an example of the terminal and does not constitute a limitation on the terminal. It may include more or fewer components than shown in the figure, or combine some components, or different components.

[0064] The input unit 230 can be used to receive input digital or character information, and generate key signal inputs related to the user settings and function controls of the terminal. Specifically, the input unit 230 may include a touch screen 231 (optional) and / or other input devices 232. The touch screen 231 can collect touch operations of the user thereon or nearby (such as operations of the user using any suitable object such as a finger, a joint, a stylus, etc. on or near the touch screen), and drive the corresponding connection device according to a pre-set program. The touch screen can detect the touch action of the user on the touch screen, convert the touch action into a touch signal and send it to the processor 270, and can receive and execute the command sent by the processor 270; the touch signal at least includes contact coordinate information. The touch screen 231 can provide an input interface and an output interface between the terminal and the user. In addition, multiple types such as resistive, capacitive, infrared, and surface acoustic wave can be used to implement the touch screen. In addition to the touch screen 231, the input unit 230 may further include other input devices. Specifically, the other input devices 232 may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, etc.

[0065] The display unit 240 can be used to display information input by the user or information provided to the user, various menus of the terminal, an interactive interface, file display, and / or the playback of any multimedia file.

[0066] The memory 220 can be used to store instructions and data. The memory 220 mainly includes a storage instruction area and a storage data area. The storage data area can store various data, such as multimedia files, texts, etc.; the storage instruction area can store software units such as an operating system, applications, and instructions required for at least one function, or subsets or extended sets thereof. It can also include a non-volatile random access memory; it provides management of hardware, software, and data resources in the computing processing device, supports control software and applications. It is also used for the storage of multimedia files, as well as the storage of running programs and applications.

[0067] The processor 270 is the control center of the terminal. It connects various parts of the entire terminal using various interfaces and lines. By running or executing instructions stored in the memory 220 and invoking data stored in the memory 220, it executes various functions of the terminal and processes data, thereby controlling the terminal as a whole. Optionally, the processor 270 can include one or more processing units; preferably, the processor 270 can integrate an application processor and a modem processor. Among them, the application processor mainly processes the operating system, user interface, and application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor may not be integrated into the processor 270. In some embodiments, the processor and the memory can be implemented on a single chip. In some embodiments, they can also be separately implemented on independent chips. The processor 270 can also be used to generate corresponding operation control signals, send them to corresponding components of the computing processing device, read and process data in the software, especially read and process data and programs in the memory 220, so that each functional module therein executes corresponding functions, thereby controlling the corresponding components to act according to the requirements of the instructions.

[0068] Among them, the memory 220 can be used to store software codes related to the wind power prediction method. The processor 270 can execute the software codes in the memory 220 or can also schedule other units (such as the above-mentioned input unit 230 and display unit 240) to implement corresponding functions.

[0069] The radio frequency unit 210 (optional) can be used for receiving and transmitting information or signals during a call. For example, after receiving the downlink information from the base station, it is sent to the processor 270 for processing; in addition, the uplink data designed is sent to the base station. Generally, the radio frequency unit 210 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LNA), a duplexer, etc. In addition, the radio frequency unit 210 can also communicate with network devices and other devices through wireless communication. This wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

[0070] Among them, in the embodiment of this application, the radio frequency unit 210 can send data to other devices and can also receive data sent by other devices. It should be understood that the radio frequency unit 210 is optional and can be replaced by other communication interfaces, such as a network interface.

[0071] The terminal also includes a power supply 290 (such as a battery) for powering each component. Preferably, the power supply can be logically connected to the processor 270 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system.

[0072] The terminal also includes an external interface 280. This external interface can be a standard Micro USB interface or a multi-pin connector, and can be used to connect the terminal to other devices for communication and can also be used to connect a charger to charge the terminal.

[0073] Although not shown, the terminal may also include a flashlight, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which will not be elaborated here.

[0074] Next, the product form of the above server will be described.

[0075] Figure 3 A schematic structural diagram of the above server is provided, as Figure 3As shown, the server may include a bus 301, a processor 302, a communication interface 303, and a memory 304. The processor 302, the memory 304, and the communication interface 303 communicate with each other via the bus 301.

[0076] The bus 301 may be a peripheral component interconnect (PCI) bus, an extended industry standard architecture (EISA) bus, or the like. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease 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.

[0077] The processor 302 can be any one or more of processors such as a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0078] The memory 304 may include volatile memory, such as random access memory (RAM). The memory 304 may also include non-volatile memory, such as read-only memory (ROM), flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0079] The memory 304 can be used to store software code related to the wind power prediction method. The processor 302 can call the software code stored in the memory 304 or schedule other units to implement corresponding functions.

[0080] The processors in the above-mentioned terminal and server (such as processor 270 and processor 302) can be hardware circuits (such as application specific integrated circuit (ASIC), field-programmable gate array (FPGA), general-purpose processor, digital signal processing (DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits. For example, the processor can be a hardware system with the function of executing instructions, such as CPU, DSP, etc., or a hardware system without the function of executing instructions, such as ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without the function of executing instructions and the hardware system with the function of executing instructions.

[0081] In the process of implementing this case, the inventors of this case found that there are currently some wind power prediction schemes, such as prediction schemes based on ARIMA (Autoregressive Integrated Moving Average Model), prediction schemes based on SVM, prediction schemes based on a hybrid model of ARIMA and neural network, etc. Among them, in the prediction scheme based on ARIMA, ARIMA captures the trends and seasonality of time series data through autoregression, differencing, and moving average. Although ARIMA performs well in dealing with linear and stationary data, its effect is limited when facing non-linear and non-stationary wind power data. In the prediction scheme based on SVM, SVM classifies and regresses data by constructing an optimal hyperplane. Although SVM has advantages in dealing with small sample and high-dimensional data, its computational complexity is relatively high. The prediction scheme based on a hybrid model of ARIMA and neural network combines the linear processing ability of ARIMA and the non-linear processing ability of neural network, and improves the prediction accuracy by processing data in stages. However, the design and optimization of the hybrid model are relatively complex.

[0082] In view of the many problems existing in the existing wind power prediction schemes, the inventors of this case conducted research. Through continuous research, a wind power prediction method with better effect was finally proposed. Next, the wind power prediction method provided in this application will be introduced through the following embodiments.

[0083] Please refer to Figure 4 , which shows a schematic flowchart of the wind power prediction method provided in the embodiment of this application. The wind power prediction method can include:

[0084] Step S401: Obtain historical wind power data.

[0085] Among them, the historical wind power data is a power time series, which includes power values at several historical moments. For example, the power value at "2024-01-01 01:00", the power value at "2024-01-01 02:00", the power value at "2024-01-01 03:00", and so on.

[0086] Step S402: Perform empirical mode decomposition on the historical wind power data to obtain several target components.

[0087] The several target components obtained by performing empirical mode decomposition on the historical wind power data include several intrinsic mode function IMF components and a residual component.

[0088] The historical wind power data has significant non-linear and non-stationary characteristics. Considering that empirical mode decomposition has an adaptive decomposition ability, in this embodiment, the adaptive decomposition ability of empirical mode decomposition is utilized to decompose the complex historical wind power data into several stationary IMF components and a residual component. These components respectively correspond to fluctuations at different time scales, significantly reducing the non-stationarity and complexity of the original historical wind power data, which is helpful for subsequent power prediction.

[0089] It should be noted that empirical mode decomposition does not require a priori determination or forced specification of basis functions, but decomposes adaptively according to the characteristics of the data to be decomposed itself. That is, empirical mode decomposition can automatically adapt to the non-linear and non-stationary characteristics of historical wind power data. By decomposing the historical wind power data into IMF components in this embodiment, the fluctuation characteristics of historical wind power data at different time scales can be captured, thereby improving the accuracy and stability of wind power prediction.

[0090] Step S403: Obtain the target component features corresponding to the several target components respectively.

[0091] In this embodiment, after obtaining several target components (several IMF components and a residual component), instead of directly performing wind power prediction based on the several target components, the target component features corresponding to the several target components are first obtained, and then wind power prediction is performed based on the target component features corresponding to the several target components respectively. In this way, the prediction effect of wind power can be improved.

[0092] Step S404: Use the pre-trained power prediction model, based on the target component features corresponding to the several target components respectively, to obtain the predicted values corresponding to the several target components at the prediction time.

[0093] Among them, the prediction time can be a future moment (i.e., a future time point), or a future time period.

[0094] In this embodiment, the power prediction model is obtained by training a relevance vector machine with the component features corresponding to several components obtained by performing empirical mode decomposition on historical wind power data samples as training samples and the true wind power at the prediction time as the sample label. When training the relevance vector machine, the training objective is to make the predicted wind power data (the component features corresponding to several components obtained by performing empirical mode decomposition on historical wind power data samples) consistent with the sample label (the true wind power data at the prediction time).

[0095] The relevance vector machine (RVM) is a sparse probability model based on Bayesian theory. Structurally, it is similar to the support vector machine (SVM). However, in the prediction modeling process, by introducing a Bayesian framework, the relevance vector machine can automatically determine the relevance vectors, thereby achieving sparsity. Using the relevance vector machine for prediction modeling can reduce the complexity of the model and improve the computational efficiency while ensuring high prediction accuracy. In addition, the Bayesian characteristics of the relevance vector machine enable it to maintain good generalization ability in small-sample scenarios, especially suitable for the situation where the historical wind power data samples are insufficient.

[0096] Step S405: Add the predicted values corresponding to several target components at the same moment to obtain the predicted value of the wind power data at the prediction time.

[0097] Exemplarily, performing empirical mode decomposition on historical wind power data yields 4 IMF components (IMF component 1, IMF component 2, IMF component 3, IMF component 4) and 1 residual component. The predicted values corresponding to the 4 IMF components at a certain moment are 100, 50, 30, and 20 respectively, and the predicted value of the residual component at this moment is 10. Then the predicted value of the wind power data at this moment is 100 + 50 + 30 + 20 + 10 = 210. If the prediction time is a future time period, the predicted values of the wind power at each moment of this time period are obtained in the above manner, and the predicted values of the wind power at each moment of this time period form a wind power prediction sequence.

[0098] The wind power prediction method provided by the embodiment of the present application, after obtaining historical wind power data, first performs empirical mode decomposition on the historical wind power data to obtain a number of target components (i.e., a number of IMF components and a residual component), then obtains the target component features corresponding to the number of target components respectively, and then uses the pre-trained power prediction model based on the relevance vector machine. Based on the target component features corresponding to the number of target components respectively, the predicted values corresponding to the number of target components at the prediction time are obtained. Finally, the predicted values corresponding to the number of target components at the same moment are added together to obtain the predicted value of the wind power data at the prediction time. The wind power prediction method provided by the present application uses the adaptive decomposition ability of empirical mode decomposition to decompose complex historical wind power data into a number of stationary IMF components and residual components, significantly reducing the non-stationarity and complexity of the original historical wind power data, which is helpful for subsequent prediction. At the same time, the wind power prediction method provided by the embodiment of the present application uses the relevance vector machine for prediction modeling. The relevance vector machine can automatically determine the relevant vectors to achieve sparsity. Thus, while ensuring high prediction accuracy, it can reduce the complexity of the model and improve the calculation efficiency. The wind power prediction method provided by the embodiment of the present application has high prediction accuracy and calculation efficiency, and also has high stability and generalization ability. This wind power prediction method provides reliable technical support for the operation and scheduling of wind farms and the stability of power systems, and helps to further promote the application and development of wind power in the energy structure.

[0099] In another embodiment of the present application, the implementation process of "Step S401: Perform empirical mode decomposition on the historical wind power data to obtain a number of target components" in the above embodiment is introduced.

[0100] In a possible implementation manner, empirical mode decomposition can be directly performed on the historical wind power data to obtain a number of IMF components and a residual component.

[0101] In order to improve the effect of empirical mode decomposition and further improve the subsequent wind power prediction effect, this embodiment provides another implementation manner of performing empirical mode decomposition on the historical wind power data to obtain a number of target components: Step a1, preprocess the historical wind power data to obtain the preprocessed historical wind power data.

[0102] In this embodiment, the preprocessing performed on the historical wind power data may include one or more of the following processes: denoising, missing value filling, outlier deletion or correction.

[0103] Historical wind power data may contain noise, which mainly comes from sensor errors and environmental interference. To obtain better prediction results subsequently, the historical wind power data can be denoised. Optionally, a denoising method based on wavelet transform can be used to denoise the historical wind power data. By selecting appropriate wavelet basis functions and decomposition levels, wavelet transform can effectively remove high-frequency noise and retain the main features of the data. This embodiment does not limit the use of the denoising method based on wavelet transform to denoise the historical wind power data, and other denoising methods can also be used, such as denoising methods based on moving average, denoising methods based on Kalman filter, and so on.

[0104] There may be missing values in the historical wind power data (such as the missing power value at a certain moment), and the missing values may be caused by equipment failures or data transmission problems. To ensure the stability of empirical mode decomposition, the missing values in the historical wind power data can be filled. The filling method can be but is not limited to the mean filling method, interpolation method, multiple imputation method, etc. Among them, the interpolation method infers the missing value based on adjacent data points. The interpolation method can specifically be the linear interpolation method, spline interpolation method, etc. The core idea of the multiple imputation method is to generate multiple complete data sets containing different imputed values through estimation, then analyze them separately, and finally summarize the results.

[0105] There may be outliers in the historical wind power data. Outliers refer to extreme values that deviate from the normal range, and they may be caused by sensor failures or external interference. For historical wind power data, statistical methods (such as Z-score, box plot method), machine learning methods (such as isolation forest, support vector machine), etc. can be used for outlier detection. After detecting the outliers, the outliers can be deleted, or a preset correction method can be used to correct the outliers.

[0106] Step a2: Perform empirical mode decomposition on the preprocessed historical wind power data to obtain several target components.

[0107] After preprocessing the historical wind power data, empirical mode decomposition can be performed on the preprocessed historical wind power data to obtain several target components (several IMF components and one residual component).

[0108] In a possible implementation, empirical mode decomposition can be directly performed on the preprocessed historical wind power data. Considering that the preprocessed historical wind power data lacks sufficient information at the endpoints, and the lack of sufficient information at the endpoints will lead to inaccurate decomposition results near the boundaries. In view of this, this embodiment provides another implementation:

[0109] Step a21: Determine the extension length N according to the preprocessed historical wind power data.

[0110] In a possible implementation, the extension length N can be determined according to the length of the preprocessed historical wind power data. For example, the extension length N is determined to be 10%-20% of the length of the preprocessed historical wind power data. To avoid over-extension or under-extension, in another possible implementation, the extension length can be adaptively selected according to the local characteristics of the preprocessed historical wind power data. Step a22: Mirror-copy the first N data points of the preprocessed historical wind power data to the left end of the preprocessed historical wind power data, and mirror-copy the last N data points of the preprocessed historical wind power data to the right end of the preprocessed historical wind power data to obtain the extended historical wind power data.

[0111] Exemplarily, the preprocessed historical wind power data is x = [x1, x2, x3, x4, x5 x6, x7,x8], and the extension length is 2. Then, mirror-copy the mirror images [x2, x1] of the first two data points of x to the left end of x, and mirror-copy the mirror images [x8, x7] of the last two data points of x to the right end of x to obtain the extended historical wind power data x′ = [x2, x1, x1, x2, x3,x4, x5 x6, x7, x8, x8, x7].

[0112] Step a23: Perform empirical mode decomposition on the extended historical wind power data to obtain several components.

[0113] Step a24: Extract the components corresponding to the preprocessed historical wind power data from the several components to obtain several target components.

[0114] The above method copies data in a mirror-symmetric manner to extend the preprocessed historical wind power data, which can reduce the boundary effect and improve the empirical mode decomposition effect.

[0115] Next, the process of performing empirical mode decomposition on a historical wind power data (such as the above preprocessed historical wind power data or the extended historical wind power data) is introduced.

[0116] In a possible implementation, the process of performing empirical mode decomposition on a historical wind power data x(t) may include:

[0117] Step b1: Initialize the residual component as the historical wind power data x(t).

[0118] Represent the residual component as r(t), and initialize r(t) = x(t).

[0119] Step b2: Identify the local maxima and local minima of the residual component.

[0120] Identify all local maxima and all local minima of the residual component r(t).

[0121] Step b3: Construct the upper envelope and the lower envelope based on the local maxima and local minima.

[0122] Using the interpolation method (such as cubic spline interpolation), construct the upper envelope based on all local maxima, and using the interpolation method (such as cubic spline interpolation), construct the lower envelope based on all local minima.

[0123] Step b4: Calculate the mean of the upper envelope and the lower envelope to obtain the envelope mean.

[0124] Step b5: Subtract the envelope mean from the residual component, and use the result as the candidate IMF component.

[0125] If the envelope mean is denoted as m(t), then subtract m(t) from the residual component r(t), and denote the result as h(t), i.e., h(t) = r(t) - m(t), then use h(t) as the candidate IMF component.

[0126] Step b6: Determine whether the candidate IMF component meets the IMF conditions. If the candidate IMF component meets the IMF conditions, then execute step b7-a; if the candidate IMF component does not meet the IMF conditions, then execute step b7-b.

[0127] Among them, the IMF conditions are: the number of extreme points is equal to or at most differs by 1 from the number of zero-crossing points, and the mean of the upper and lower envelopes is zero.

[0128] That is, if the number of extreme points of the candidate IMF component h(t) is equal to or at most differs by 1 from the number of zero-crossing points, and the mean of the upper and lower envelopes is zero, then it is determined that the candidate IMF component meets the IMF conditions, and the candidate IMF component is the IMF component.

[0129] When screening IMF components, a screening stop criterion is usually set. In one possible implementation, the screening stop criterion can be to stop screening after reaching the set maximum number of screening times. Considering that the above screening stop criterion cannot adapt to the complexity of different data, in another possible implementation, the screening stop criterion can be an adaptive screening stop criterion based on entropy or energy ratio. The introduction of the adaptive screening stop criterion based on entropy (the entropy of the current screening result) or energy ratio (the energy ratio of the current screening result to the original data) enables the decomposition process to automatically adjust according to the internal fluctuations of the data to ensure the physical interpretability of each IMF component.

[0130] Step b7-a: Subtract the IMF component from the residual component, and use the result as the new residual component.

[0131] If the candidate IMF component h(t) satisfies the IMF condition, update the residual component to r(t) - h(t).

[0132] Step b7-b: Take the candidate IMF component as the new residual component.

[0133] If the candidate IMF component h(t) does not satisfy the IMF condition, update the residual component to h(t).

[0134] Step b8: Determine whether the current decomposition termination condition is satisfied. If the current condition is not satisfied, execute Step b2 and subsequent steps until the decomposition termination condition is met. If the current condition is satisfied, end the decomposition process.

[0135] Among them, the decomposition termination condition is that the residual component is a monotonic function, or the number of extreme points of the residual component is less than the preset quantity threshold.

[0136] Through empirical mode decomposition, the historical wind power data x(t) is decomposed into several IMF components with different time-scale characteristics and a residual component. Each IMF component obtained by decomposition represents the fluctuation characteristics of x(t) at a specific time scale, and the residual component is the final remaining part of the empirical mode decomposition, which represents the long-term trend information of x(t).

[0137] To improve the decomposition effect, this embodiment provides another implementation method for empirical mode decomposition of a historical wind power data x(t): Add white noise with different amplitudes to the historical wind power data x(t) multiple times to obtain multiple noise-assisted data; perform empirical mode decomposition on each noise-assisted data (using the decomposition process of Step b1 to Step b8 above to perform empirical mode decomposition on each noise-assisted data), obtain the component sets corresponding to the multiple noise-assisted data respectively (a component combination corresponding to a noise-assisted data includes several IMF components and a residual component), and average the corresponding components in the component sets corresponding to the multiple noise-assisted data respectively to obtain the final several components (IMF components and residual components).

[0138] The above method can reduce the mode mixing phenomenon and thus improve the decomposition stability by introducing white noise-assisted decomposition. Since the scale characteristics of white noise are evenly distributed in the entire time-frequency domain, when performing empirical mode decomposition, each scale characteristic will contain the component of white noise. After adding white noise to the historical wind power data x(t) and performing decomposition, a certain scale characteristic contained in x(t) will be decomposed to the scale characteristic corresponding to the white noise. Although the uncertainty of the noise may affect the decomposition result in a single decomposition, according to the statistical characteristics of white noise, if different white noises are added and the means of a large number of decomposition results are calculated, the noise components in each scale characteristic will cancel each other out, and finally only the signal to be decomposed remains.

[0139] In another embodiment of the present application, the specific implementation process of "step S403: Obtain the target component features corresponding to several target components" in the above embodiment is introduced.

[0140] As mentioned in the above embodiment, several target components include several IMF components and a residual component. This embodiment focuses on introducing the specific implementation process of obtaining the target component features corresponding to several IMF components.

[0141] In a possible implementation manner, the process of obtaining the target component features corresponding to several IMF components may include: for each IMF component, extract some or all (preferably all) of the statistical features, time-domain features, frequency-domain features, and time-frequency features from the IMF component, and use the extracted features as the component features corresponding to the IMF component.

[0142] Considering that there are many features obtained in the above manner and the features obtained in the above manner are not all useful for power prediction, this embodiment provides another implementation manner of obtaining the target component features corresponding to several IMF components:

[0143] Step c1: For each IMF component, extract one or more (preferably multiple) of the following features from the IMF component: statistical features, time-domain features, frequency-domain features, and time-frequency features, to obtain the initial component features corresponding to the IMF component.

[0144] The statistical features may include some or all (preferably all) of the following features: mean, variance, skewness, and kurtosis. Among them, the mean can reflect the central tendency of the IMF component, the variance can reflect the volatility of the IMF component, the skewness can reflect the symmetry of the data distribution of the IMF component, and the kurtosis can reflect the peakedness of the data distribution of the IMF.

[0145] The time-domain features may include some or all (preferably all) of the following features: autocorrelation function, zero-crossing rate, and peak and valley features. Among them, the autocorrelation function can reflect the correlation of the IMF component at different time lags, the zero-crossing rate can reflect the frequency at which the IMF component crosses zero, reflecting the frequency components of the IMF component, and the peak and valley features include the number, position, and amplitude of peaks and valleys.

[0146] Frequency domain features may include some or all (preferably all) of the following features: power spectral density (PSD), dominant frequency, and spectral centroid. The power spectral density can reflect the energy distribution of IMF components in the frequency domain and reveal the frequency characteristics of IMF components. The dominant frequency is the main frequency component of the IMF component, and the spectral centroid is the centroid position of the spectrum of the IMF component, which can reflect the concentration degree of the frequency distribution of the IMF component.

[0147] Time-frequency features may include some or all (preferably all) of the following features: features characterizing the frequency characteristics of IMF components in different time periods, and features characterizing the time-frequency characteristics of IMF components at different scales and positions. The features characterizing the frequency characteristics of IMF components in different time periods can be obtained by performing short-time Fourier transform (STFT) on the IMF components, and the features characterizing the time-frequency characteristics of IMF components at different scales and positions can be obtained by performing wavelet transform (WT) on the IMF components.

[0148] Step c2: Select the features useful for wind power prediction from the initial component features corresponding to the IMF component to obtain the target component features corresponding to the IMF component.

[0149] Specifically, the process of selecting the features useful for wind power prediction from the initial component features corresponding to the IMF component may include: using one or more of the following feature selection methods to select the features useful for power prediction from the extracted features: feature selection method based on principal component analysis, feature selection method based on least absolute shrinkage and selection operator, feature selection method based on stepwise regression, and feature selection method based on random forest.

[0150] Preferably, the feature selection method based on principal component analysis, the feature selection method based on least absolute shrinkage and selection operator, the feature selection method based on stepwise regression, and the feature selection method based on random forest can be used simultaneously to select the features useful for power prediction from the extracted features respectively. Furthermore, the features selected by the above several feature selection methods are combined and de-duplicated.

[0151] Principal component analysis (PCA) is a dimensionality reduction technique that projects high-dimensional data into a low-dimensional space to extract the main feature components. PCA can maximize the variance of the data, retain the main information of the original data, and reduce redundant features. Least absolute shrinkage and selection operator (LASSO) tends to produce a sparse solution when selecting features by introducing L1 regularization, that is, making some regression coefficients zero, thereby achieving feature selection. Stepwise regression selects the features that have the greatest impact on power prediction by gradually adding or removing features. Random forest constructs multiple decision trees and calculates the importance scores of features based on the splitting of features in the trees, thereby performing feature selection.

[0152] In the process of selecting features using each feature selection method, a contribution value (or importance score) is assigned to each feature to quantify its impact on prediction. Then, based on these contribution values (or importance scores), feature selection is completed. Through feature selection, features with high information content, high correlation, strong generalization ability, and low redundancy can be obtained. It can be seen that through feature selection, high-quality input data can be provided for the power prediction model, thereby improving the accuracy, stability, and generalization ability of power prediction.

[0153] In another embodiment of the present application, the specific implementation process of "step S404: Using the pre-trained power prediction model, based on the target component features corresponding to several target components, obtain the predicted values corresponding to several target components at the prediction time" in the above embodiment is introduced.

[0154] As Figure 5 shown, the process of using the pre-trained power prediction model and based on the target component features corresponding to several target components to obtain the predicted values corresponding to several target components at the prediction time may include:

[0155] Step S501: Combine the target component features corresponding to several target components to obtain the combined features.

[0156] Exemplarily, if there are a total of 20 component features corresponding to several target components, then these 20 component features are combined together.

[0157] Step S502: Input the combined features into the pre-trained power prediction model to obtain the probability distribution of the predicted value corresponding to each target component output by the power prediction model at the prediction time.

[0158] Input the combined features into the pre-trained power prediction model based on the relevance vector machine. The power prediction model makes predictions according to the input features and outputs the probability distribution corresponding to each target component. The probability distribution corresponding to any target component is the probability distribution of the predicted value corresponding to this target component at the prediction time.

[0159] The power prediction model based on the relevance vector machine provides the probability distribution of the prediction result, which can quantify the uncertainty of the prediction and provide richer information for decision-making.

[0160] Step S503: Calculate the expected value of the probability distribution of the predicted value corresponding to each target component at the prediction time to obtain the predicted value corresponding to each target component at the prediction time.

[0161] The prediction of wind power is realized based on the pre-trained power prediction model. Next, the process of training the power prediction model is introduced.

[0162] The process of training the power prediction model may include:

[0163] Step d1: Obtain a plurality of historical wind power data samples and the true wind power corresponding to each of the plurality of historical wind power data samples at the prediction time.

[0164] Step d2: For each historical wind power data sample, perform empirical mode decomposition on it to obtain a number of components (a number of IMF components and a residual component), and obtain the component features corresponding to each of the components. Use the component features corresponding to the obtained components as feature samples, and use the true wind power data corresponding to the historical wind power data sample as the sample label corresponding to the feature sample.

[0165] Through step d2, multiple pieces of sample data including feature samples and corresponding sample labels can be obtained.

[0166] Step d3: Divide the multiple pieces of sample data into two parts, one part forms the training set, and the other part forms the test set.

[0167] The training set is used to train the model, and the test set is used to test the trained model.

[0168] Step d4: Use the sample data in the training set to train the relevance vector machine.

[0169] Specifically, first initialize the hyperparameters, including the noise variance and the prior distribution parameters, and then use the sample data in the training set to train the relevance vector machine. During training, the goal is to make the wind power data predicted based on the sample features in the training set tend to be consistent with the corresponding sample labels (i.e., the true wind power data).

[0170] During the training process, the relevance vector machine learns the relationship between the feature samples in the training set and the corresponding sample labels. In addition, during the training process, the relevance vector machine will automatically select the relevant feature vectors (i.e., the feature vectors that are most important for prediction) and eliminate the unimportant feature vectors, thereby achieving sparsity.

[0171] After the training is completed, the trained relevance vector machine can be tested using the test set. Specifically, use the trained relevance vector machine to predict the wind power data based on the feature samples in the test set, calculate the mean square error or the mean absolute error between the predicted wind power data and the corresponding true wind power data, and evaluate whether the trained relevance vector machine meets the requirements according to the calculated mean square error or mean absolute error. If it meets the requirements, use the trained relevance vector machine as the power prediction model. If it does not meet the requirements, continue training until the relevance vector machine meets the requirements.

[0172] The wind power prediction method provided by the embodiments of the present application is introduced above. The device corresponding to the above wind power prediction method will be introduced below.

[0173] Please refer to Figure 6 , Figure 6 which is a schematic structural diagram of a wind power prediction device provided by an embodiment of the present application. The wind power prediction device may include: a data acquisition module 601, a data decomposition module 602, a feature acquisition module 603, a power prediction module 604, and a power prediction value determination module 605.

[0174] The data acquisition module 601 is configured to acquire historical wind power data;

[0175] The data decomposition module 602 is configured to perform empirical mode decomposition on the historical wind power data to obtain a number of target components, and the number of target components includes a number of intrinsic mode function IMF components and a residual component.

[0176] The feature acquisition module 603 is configured to acquire target component features corresponding to the number of target components respectively.

[0177] The power prediction module 604 is configured to use a pre-trained power prediction model, and based on the target component features corresponding to the number of target components respectively, obtain the predicted values corresponding to the number of target components at the prediction time respectively.

[0178] Among them, the power prediction model is trained by using the component features corresponding to the number of components obtained by performing empirical mode decomposition on the historical wind power data samples as training samples, and the real wind power data at the prediction time as sample labels, and training a relevant vector machine.

[0179] The power prediction value determination module 605 is configured to add the predicted values corresponding to the number of target components at the same moment respectively to obtain the predicted value of the wind power data at the prediction time.

[0180] In a possible implementation manner, when the power prediction module 604 uses a pre-trained power prediction model and based on the target component features corresponding to the number of target components respectively, to obtain the predicted values corresponding to the number of target components at the prediction time respectively, it is specifically configured to:

[0181] Combine the target component features corresponding to the number of target components respectively to obtain a combined feature;

[0182] Input the combined feature into the pre-trained power prediction model to obtain the probability distribution of the predicted value corresponding to each target component output by the power prediction model at the prediction time;

[0183] Calculate the expected value of the probability distribution of the predicted value corresponding to each target component at the prediction time, and obtain the predicted value corresponding to each target component at the prediction time.

[0184] In a possible implementation manner, when the data decomposition module 602 performs empirical mode decomposition on historical wind power data to obtain a number of target components, it specifically is used for:

[0185] Preprocess the historical wind power data, where the preprocessing includes one or more of the following processes: denoising, missing value filling, outlier deletion or correction;

[0186] Perform empirical mode decomposition on the preprocessed historical wind power data to obtain a number of target components.

[0187] In a possible implementation manner, when the data decomposition module 602 performs empirical mode decomposition on the preprocessed historical wind power data to obtain a number of target components, it specifically is used for:

[0188] Determine the extension length N according to the preprocessed historical wind power data, where N is an integer greater than 0;

[0189] Mirror-copy the first N data points of the preprocessed historical wind power data to the left end of the preprocessed historical wind power data, and mirror-copy the last N data points of the preprocessed historical wind power data to the right end of the preprocessed historical wind power data to obtain the extended historical wind power data;

[0190] Perform empirical mode decomposition on the extended historical wind power data to obtain a number of components;

[0191] Extract the components corresponding to the preprocessed historical wind power data from the number of components to obtain a number of target components.

[0192] In a possible implementation manner, when the data decomposition module 602 performs empirical mode decomposition on a historical wind power data, it specifically is used for:

[0193] Initialize the residual component as the historical wind power data;

[0194] Identify the local maxima and local minima of the residual component;

[0195] Construct an upper envelope and a lower envelope according to the local maxima and local minima, and calculate the mean of the upper envelope and the lower envelope to obtain the envelope mean;

[0196] Subtract the envelope mean from the residual component, and use the obtained result as the candidate IMF component;

[0197] Determine whether the candidate IMF component satisfies the IMF condition, where the IMF condition is that the number of extreme points is equal to or at most differs by 1 from the number of zero-crossing points, and the mean of the upper and lower envelope lines is zero;

[0198] If the candidate IMF component satisfies the IMF condition, subtract the IMF component from the residual component to obtain a new residual component. If the candidate IMF component is not an IMF component, use the candidate IMF component as the new residual component;

[0199] Determine whether the current decomposition termination condition is satisfied, where the decomposition termination condition is that the residual component is a monotonic function, or the number of extreme points of the residual component is less than a preset number threshold;

[0200] If the current does not satisfy the decomposition termination condition, perform the steps of identifying the local maxima and local minima of the residual component and subsequent steps until the decomposition termination condition is satisfied.

[0201] In a possible implementation manner, when the data decomposition module 602 performs empirical mode decomposition on a historical wind power data, it is specifically used for:

[0202] Add white noise with different amplitudes to the historical wind power data multiple times to obtain multiple noise-assisted data;

[0203] Perform empirical mode decomposition on multiple noise-assisted data respectively to obtain component sets corresponding to the multiple noise-assisted data respectively, where the component set corresponding to any noise-assisted data includes several IMF components and a residual component;

[0204] Average the corresponding components in the component sets corresponding to the multiple noise-assisted data respectively to obtain the final decomposition result.

[0205] In a possible implementation manner, the feature acquisition module 603 includes: a feature extraction module and a feature selection module.

[0206] The feature extraction module is used to extract one or more of the following features for each IMF component: statistical features, time-domain features, frequency-domain features, time-frequency features, to obtain the initial component features corresponding to the IMF component.

[0207] The feature selection module is used to select the features useful for wind power prediction from the initial component features corresponding to the IMF component to obtain the target component features corresponding to the IMF component.

[0208] In a possible implementation manner, the statistical features include some or all of the following features: mean, variance, skewness, kurtosis;

[0209] The time-domain features include some or all of the following features: autocorrelation function, zero-crossing rate, crest and trough features;

[0210] The frequency-domain features include some or all of the following features: power spectral density, main frequency, spectral centroid;

[0211] The time-frequency features include some or all of the following features: features characterizing the frequency characteristics of IMF components in different time periods, features characterizing the time-frequency characteristics of IMF at different scales and positions.

[0212] In a possible implementation manner, when the feature selection module selects features useful for wind power prediction from the initial component features corresponding to the IMF component, it is specifically used for:

[0213] Select one or more of the following several feature selection methods to select features useful for wind power prediction from the initial component features corresponding to the IMF component: feature selection method based on principal component analysis, feature selection method based on least absolute shrinkage and selection operator, feature selection method based on stepwise regression, feature selection method based on random forest.

[0214] The wind power prediction device provided by the embodiments of the present application utilizes the adaptive decomposition ability of empirical mode decomposition to decompose complex historical wind power data into several stationary IMF components and residual components, significantly reducing the non-stationarity and complexity of the original historical wind power data, which is helpful for subsequent prediction. At the same time, the wind power prediction device provided by the embodiments of the present application uses a relevance vector machine for prediction modeling. The relevance vector machine can automatically determine the relevance vectors to achieve sparsity. Thus, while ensuring high prediction accuracy, it can reduce the complexity of the model and improve the calculation efficiency. The wind power prediction device provided by the embodiments of the present application has high prediction accuracy and calculation efficiency, and also has high stability and generalization ability.

[0215] The embodiments of the present application also provide an electronic device, which may include: at least one processor, at least one communication interface, at least one memory, and at least one communication bus.

[0216] In the embodiments of the present application, the number of the processor, the communication interface, the memory, and the communication bus is at least one, and the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0217] The processor may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application, etc.;

[0218] The memory may include high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory;

[0219] Among them, the memory stores a program, and the processor can call the program stored in the memory, and the program is used to implement the steps of the wind power prediction method provided in the above embodiments.

[0220] The embodiment of the present application also provides a computer storage medium, and the storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can be enabled to implement the steps of the wind power prediction method provided in the above embodiments.

[0221] The embodiment of the present application also provides a computer program product, including computer-readable instructions. When the computer-readable instructions run on an electronic device, the electronic device is enabled to implement the steps of the wind power prediction method provided in the above embodiments.

[0222] In addition, it should be noted that the device embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided in the present application, the connection relationship between the modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines.

[0223] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by dedicated hardware, including dedicated integrated circuits, dedicated CPUs, dedicated memories, dedicated components, etc. Generally, functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structures for implementing the same function can also be various, such as analog circuits, digital circuits or dedicated circuits. However, for the present application, in more cases, software program implementation is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a readable storage medium, such as a floppy disk, a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk or an optical disc of a computer, and includes several instructions for enabling a computer device (which may be a personal computer, a training device, or a network device, etc.) to execute the methods described in various embodiments of the present application.

[0224] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product.

[0225] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device or data center that includes one or more integrated available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid state disk (SSD)), etc.

Claims

1. A wind power prediction method, characterized in that: include: Obtain historical wind power data; Performing empirical mode decomposition on the historical wind power data to obtain a number of target components, wherein the number of target components includes a number of intrinsic mode function IMF components and a residual component; Obtaining target component features corresponding to the plurality of target components respectively; Using the pre-trained power prediction model, based on the target component features corresponding to the target components, the predicted values ​​corresponding to the target components at the prediction time are obtained, wherein the power prediction model uses the component features corresponding to the target components obtained by performing empirical mode decomposition on historical wind power data samples as training samples, and uses the real wind power data at the prediction time as sample labels, and trains the relevant vector machine to obtain the results; The predicted values ​​corresponding to the several target components at the same time are added together to obtain the predicted value of wind power data at the predicted time.

2. The wind power prediction method according to claim 1, characterized in that: The power prediction model obtained by pre-training is used to obtain the prediction values ​​corresponding to the plurality of target components at the prediction time based on the target component features corresponding to the plurality of target components, including: Combining the target component features corresponding to the plurality of target components to obtain a combined feature; Inputting the combined features into a pre-trained power prediction model to obtain a probability distribution of a predicted value corresponding to each target component output by the power prediction model at a prediction time; The expected value of the probability distribution of the predicted value corresponding to each target component at the prediction time is calculated to obtain the predicted value corresponding to each target component at the prediction time.

3. The wind power prediction method according to claim 1, characterized in that: The historical wind power data is subjected to empirical mode decomposition to obtain several target components, including: Preprocessing the historical wind power data, wherein the preprocessing includes one or more of the following processes: denoising, missing value filling, outlier deletion or correction; The preprocessed historical wind power data is subjected to empirical mode decomposition to obtain several target components.

4. The wind power prediction method according to claim 3, characterized in that: The empirical mode decomposition is performed on the pre-processed historical wind power data to obtain several target components, including: According to the pre-processed historical wind power data, an extension length N is determined, where N is an integer greater than 0; Mirroring the first N data points of the preprocessed historical wind power data to the left end of the preprocessed historical wind power data, and mirroring the last N data points of the preprocessed historical wind power data to the right end of the preprocessed historical wind power data, to obtain extended historical wind power data; Performing empirical mode decomposition on the extended historical wind power data to obtain several components; Components corresponding to the pre-processed historical wind power data are extracted from the plurality of components to obtain a plurality of target components.

5. The wind power prediction method according to any one of claims 1 to 4, characterized in that: Perform empirical mode decomposition on historical wind power data, including: Adding white noise of different amplitudes to the historical wind power data multiple times to obtain multiple noise auxiliary data; Performing empirical mode decomposition on the multiple noise auxiliary data respectively to obtain component sets corresponding to the multiple noise auxiliary data respectively, wherein the component set corresponding to any noise auxiliary data includes a plurality of IMF components and one residual component; Corresponding components in the component sets corresponding to the multiple noise auxiliary data are averaged to obtain a final decomposition result.

6. The wind power prediction method according to claim 1, characterized in that: Get the component features corresponding to several IMF components, including: For each IMF component: Extract one or more of the following features from the IMF component: statistical features, time domain features, frequency domain features, and time-frequency features, to obtain initial component features corresponding to the IMF component; Features useful for wind power prediction are selected from the initial component features corresponding to the IMF component to obtain the target component features corresponding to the IMF component.

7. The wind power prediction method according to claim 6, characterized in that: The statistical features include some or all of the following features: mean, variance, skewness, kurtosis; The time domain features include some or all of the following features: autocorrelation function, zero crossing rate, peak and trough features; The frequency domain features include some or all of the following features: power spectrum density, main frequency, spectrum centroid; The time-frequency characteristics include part or all of the following characteristics: characteristics characterizing the frequency characteristics of IMF components in different time periods, and characteristics characterizing the time-frequency characteristics of IMF at different scales and positions.

8. The wind power prediction method according to claim 6, characterized in that: The selecting of features useful for wind power prediction from the initial component features corresponding to the IMF component includes: One or more of the following feature selection methods are used to select features useful for wind power prediction from the initial component features corresponding to the IMF component: a feature selection method based on principal component analysis, a feature selection method based on minimum absolute shrinkage and selection operator, a feature selection method based on stepwise regression, and a feature selection method based on random forest.

9. A wind power prediction device, characterized in that: include: Data acquisition module, data decomposition module, feature acquisition module, power prediction module and power prediction value determination module; The data acquisition module is used to acquire historical wind power data; The data decomposition module is used to perform empirical mode decomposition on the historical wind power data to obtain a number of target components, wherein the number of target components includes a number of intrinsic mode function IMF components and a residual component; The feature acquisition module is used to acquire target component features corresponding to the plurality of target components respectively; The power prediction module is used to use the pre-trained power prediction model and obtain the prediction values ​​corresponding to the target components at the prediction time based on the target component features corresponding to the target components, wherein the power prediction model uses the component features corresponding to the target components obtained by performing empirical mode decomposition on historical wind power data samples as training samples, and uses the real wind power data at the prediction time as sample labels to train the relevant vector machine; The power prediction value determination module is used to add the prediction values ​​corresponding to the several target components at the same time to obtain the wind power data prediction value at the prediction time.

10. An electronic device, characterized in that: The method comprises at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program so that the electronic device can implement the steps of the wind power prediction method as described in any one of claims 1 to 8.

11. A computer storage medium, characterized in that: The storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement the steps of the wind power prediction method as described in any one of claims 1 to 8.

12. A computer program product, characterized in that The method comprises computer-readable instructions, and when the computer-readable instructions are executed on an electronic device, the electronic device implements the steps of the wind power prediction method as claimed in any one of claims 1 to 8.

Images