Wind power prediction method and device, electronic equipment and storage medium
By adopting a model of a recurrent neural network integrated with collaborative time characterization learning in wind power prediction, the problem of difficulty in capturing the complex time dependence of wind power data is solved in the prior art, and a more accurate and reliable wind power prediction results are achieved.
Patent Information
- Application Number
- CN202411461868.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-18
- Publication Date
- 2025-02-14
AI Technical Summary
Existing wind power power prediction methods are difficult to accurately capture the complex time dependence in wind power data, affecting the accuracy of prediction results.
Using a wind power power prediction model with a recurrent neural network integrated with collaborative time characterization learning, multi-directional time dependencies are captured by preprocessing and training of historical power generation power timing data, including complex dynamics across different time scales and variables.
It improves the accuracy and reliability of wind power prediction results, and can understand wind power generation patterns more comprehensively, thereby better improving effective integration capabilities. It is suitable for wind farm power prediction, power system scheduling and energy management.
Smart Images
Figure CN119448222A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of wind power technology, and in particular to a wind power prediction method, device, electronic equipment and storage medium. Background Art
[0002] The increasing reliance on wind energy as a renewable energy source requires accurate forecasting techniques to ensure grid stability and optimize energy management. However, the inherent variability and intermittency of wind energy, influenced by the complex and nonlinear relationships between weather variables such as wind speed and temperature, poses significant challenges for accurate forecasting.
[0003] Accurate prediction of wind power is crucial to the efficient operation of the power grid, including grid dispatch, power market transactions and backup power management, which ultimately contributes to the stability and security of the power system. However, existing methods find it difficult to capture the complex time dependencies in wind power data, thus affecting the accuracy of the prediction results. Summary of the invention
[0004] The purpose of the present application is to provide a wind power prediction method, device, electronic device and storage medium, which can improve the accuracy and reliability of wind power prediction results.
[0005] In a first aspect, an embodiment of the present application provides a wind power prediction method, comprising:
[0006] Obtain the time series data of the power generation of wind turbines in the historical time period;
[0007] After preprocessing the power generation time series data, a pre-trained wind power prediction model is input to obtain the predicted power generation of the wind turbine generator in a future set time period output by the wind power prediction model;
[0008] The wind power prediction model is obtained by training a recurrent neural network integrated with collaborative time representation learning through historical power generation time series data samples of wind turbines.
[0009] In some embodiments of the present application, the wind power prediction model is pre-trained in the following manner:
[0010] Obtain the historical power generation of the wind turbine generator, and count it as a time series within the same time interval to obtain a time series data sample of the historical power generation of the wind turbine generator;
[0011] After the preprocessing is performed on the historical power generation time series data samples, the historical power generation time series data samples are converted into power generation supervised data;
[0012] An objective function is set, and the supervised data of power generation is input into the recurrent neural network integrated with collaborative time representation learning for training, and a trained wind power prediction model is obtained after a preset training cutoff condition is reached.
[0013] In some embodiments of the present application, converting the historical power generation time series data samples into power generation supervised data includes:
[0014] The first several variables in the historical power generation time series data sample are used as the input variables of the model, and the last variable is aggregated as the output of the model to obtain the supervised data of power generation.
[0015] In some embodiments of the present application, the preprocessing includes maximum value processing, minimum value processing, data averaging and normalization processing.
[0016] In some embodiments of the present application, the normalization process uses a reversible instance normalization process.
[0017] In some embodiments of the present application, the recurrent neural network integrated with collaborative temporal representation learning includes:
[0018] Gated recurrent unit, which is used to capture long-term temporal dependencies in time series data and obtain temporal representation of time series data;
[0019] A collaboration representation unit, used for extracting the time dependency relationship in the temporal representation from different directions to obtain collaboration representations in different directions;
[0020] A representation fusion unit, used for serially splicing the temporal representation and the collaborative representations of different directions to obtain a fused representation;
[0021] A mapping unit is used to map the fusion representation to the predicted generated power.
[0022] In some implementations of the present application, the mapping unit includes:
[0023] An autoregressive layer is used to perform autoregressive processing on the fused representation to obtain an autoregressive fused representation;
[0024] A multi-layer perceptron is used to convert the autoregressive fusion representation into predicted generated power.
[0025] In a second aspect, an embodiment of the present application provides a wind power prediction device, comprising:
[0026] An acquisition module is used to acquire the time series data of the power generation of the wind turbine in a historical time period;
[0027] A prediction module is used to pre-process the power generation time series data, input a pre-trained wind power prediction model, and obtain the predicted power generation of the wind turbine generator in a future set time period output by the wind power prediction model;
[0028] The wind power prediction model is obtained by training a recurrent neural network integrated with collaborative time representation learning through historical power generation time series data samples of wind turbines.
[0029] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect.
[0030] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer-readable instructions stored thereon, and the computer-readable instructions can be executed by a processor to implement the method described in the first aspect.
[0031] Compared with the prior art, the wind power prediction method provided by the present application obtains the power generation time series data of the wind turbine in the historical time period; after preprocessing the power generation time series data, the pre-trained wind power prediction model is input to obtain the predicted power generation of the wind turbine in the future set time period output by the wind power prediction model; wherein the wind power prediction model is obtained by training the recurrent neural network integrated with collaborative time representation learning through the historical power generation time series data samples of the wind turbine. The wind power prediction model of the present application is obtained based on the recurrent neural network integrated with collaborative time representation learning, and can capture multi-directional time dependencies through collaborative representation learning, including complex dynamics across different time scales and variables, so that it can more comprehensively understand the wind power generation mode, thereby better improving the effective integration capability. Compared with the traditional wind power prediction method, the method proposed in the present application can generate accurate wind power prediction results, providing a reliable basis for efficient management of the power grid and resource optimization. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:
[0033] Figure 1 A flow chart of a wind power prediction method provided in an embodiment of the present application is shown;
[0034] Figure 2 A schematic diagram of a prediction process of a wind power prediction model provided in an embodiment of the present application is shown;
[0035] Figure 3 A schematic diagram of a prediction process of another wind power prediction model provided in an embodiment of the present application is shown;
[0036] Figure 4 A schematic diagram of a wind power prediction device provided in an embodiment of the present application is shown;
[0037] Figure 5 A schematic diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION
[0038] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0039] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by technicians in the field to which this application belongs.
[0040] In addition, the terms "first" and "second" etc. are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or devices.
[0041] Traditional wind power forecasting methods, such as numerical weather prediction (NWP) models, often struggle to capture the nuances of these factors, especially local variations that can have a significant impact on short-term power generation, and may have difficulty predicting sudden gusts or wind direction changes that affect specific wind farms. In addition, existing forecasting methods, including machine learning and deep learning techniques, also face limitations. While machine learning methods are able to learn complex patterns from historical data, they often require extensive feature engineering and may not fully capture the complex temporal dependencies in wind data. Deep learning methods, including convolutional neural networks (CNNs), recurrent neural networks (RNNs), graph neural networks (GNNs), and attention mechanisms, offer great potential due to their ability to automatically learn hierarchical features and temporal dependencies. However, when these models are applied to wind power forecasting, they may encounter problems such as effectively capturing spatial correlations between wind farms, handling long-term dependencies in wind power generation patterns, and adapting to features of non-stationary wind. In addition, existing methods have not fully explored the advantages of collaborative representation learning in capturing multi-directional temporal dynamics in wind power data.
[0042] To address the shortcomings of existing methods, this application introduces a new deep learning model - collaborative temporal representation learning, a new method designed to improve the accuracy of wind power forecasting. This model leverages the power of collaborative representation learning, an approach that focuses on capturing multi-directional temporal dependencies, that is, dependencies that exist between different time scales and variables, to provide a more comprehensive understanding of wind power generation patterns. Specifically, the collaborative temporal representation learning model integrates reversible instance normalization to normalize input data and denormalize predictions, RNN-based hidden state learning to capture long-term temporal dependencies, and a new collaborative representation learning part to extract and highlight multi-directional temporal dynamics. These collaborative temporal representations are then combined with the original data, similar to residual connections, and processed through multiple layers of mapping to improve predictions. Finally, the output is denormalized using reversible instance normalization to obtain the final wind power forecast result. This comprehensive approach aims to address the limitations of existing methods by effectively leveraging the collaborative and temporal nature of wind data, ultimately achieving more accurate and reliable wind power forecasts.
[0043] The following is a detailed introduction to this application.
[0044] Please refer to Figure 1 , Figure 1 This is a flow chart of a wind power prediction method provided in an embodiment of the present application, comprising the following steps S101-S102:
[0045] S101. Obtaining time series data of power generation of wind turbines in a historical period.
[0046] S102. After preprocessing the power generation time series data, a pre-trained wind power prediction model is input to obtain the predicted power generation of the wind turbine generator in a future set time period output by the wind power prediction model; wherein the wind power prediction model is obtained by training a recurrent neural network integrated with collaborative time representation learning through samples of the historical power generation time series data of the wind turbine generator.
[0047] The following describes how to pre-train a wind power prediction model. Specifically, the wind power prediction model can be pre-trained in the following manner, including the following steps:
[0048] S1. Obtain the historical power generation of the wind turbine generator and make statistics into a time series within the same time interval to obtain a time series data sample of the historical power generation of the wind turbine generator.
[0049] Historical power generation refers to the historical data of wind turbine power generation. Determine the time interval: Determine the time interval based on the time span of the data and the time granularity that needs to be counted. It can be set to hours, days, weeks, or months.
[0050] S2. After preprocessing the historical power generation time series data samples, convert the historical power generation time series data samples into supervised power generation data.
[0051] Specifically, the historical power generation time series data samples can be preprocessed through preliminary data preprocessing operations such as maximum value processing, minimum value processing, and data averaging, and then the time series data after preliminary preprocessing can be normalized.
[0052] Specifically, the time series data after preliminary preprocessing can be normalized using the reversible instance normalization (RevIN) method to solve the problem of different scales in the time series data and improve the stability of model training.
[0053] Furthermore, the first several variables in the preprocessed historical power generation time series data samples are used as the input variables of the model, and the last variable is aggregated as the output of the model to obtain the supervised data of power generation. That is, the time series data is preprocessed using the one-step forward segmentation technique to transform it into supervised data. This technique pairs each input sequence with its subsequent value as the output target. The input and output data are X∈R B×T×D and Y∈R B×H×D , where B is the batch size, T is the time step, D is the variable dimension, and H is the length of time to predict the future.
[0054] Specifically, in order to learn time series characteristics from time series, this application converts time series data into supervised data. Given a time series data, in the conversion process, the first several variables are used as input variables of the model, and the last variable is regarded as the output variable of the model. The wind power prediction model trained with supervised data can provide more accurate predictions.
[0055] There is no order in which the above normalization steps and the steps of converting to supervised data occur.
[0056] When the input data is normalized using the reversible instance normalization (RevIN) method, the mean of the current batch of data is first subtracted and divided by the standard deviation. The data is then further adjusted by learnable parameters to learn the affine transformation of each channel. This method is more effective than the simple time series mode normalization method and solves the problem of different scales in time series data. For example, the supervised data X is normalized:
[0057]
[0058] Among them, X'∈R is obtained by normalization B×T×D , μ and σ are the mean and standard deviation of the input data respectively.
[0059] S3. Setting an objective function, inputting the supervised data of power generation into the recurrent neural network integrated with collaborative time representation learning for training, and obtaining a trained wind power prediction model after reaching a preset training cutoff condition.
[0060] The preset training cutoff condition may be that the number of training cycles reaches a set number, or the prediction error is less than a preset threshold. The objective function may be set to a mean square error loss function (MSE).
[0061] Specifically, the above-mentioned recurrent neural network integrated with collaborative temporal representation learning includes: a gated recurrent unit, a collaborative representation unit, a representation fusion unit and a mapping unit.
[0062] Gated recurrent unit, which is used to capture long-term temporal dependencies in time series data and obtain temporal representation of time series data;
[0063] A collaboration representation unit, used for extracting the time dependency relationship in the temporal representation from different directions to obtain collaboration representations in different directions;
[0064] A representation fusion unit, used for serially splicing the temporal representation and the collaborative representations of different directions to obtain a fused representation;
[0065] A mapping unit is used to map the fusion representation to the predicted generated power.
[0066] Specifically, the mapping unit includes: an autoregressive layer and a multilayer perceptron; the autoregressive layer is used to perform autoregressive processing on the fusion representation to obtain an autoregressive fusion representation; the multilayer perceptron is used to convert the autoregressive fusion representation into predicted power generation.
[0067] Please refer to Figure 2 ,The wind power prediction model is introduced in detail as follows.
[0068] Temporal representation is performed on the normalized data, and the temporal representation uses a gated recurrent unit (GRU) to capture long-term temporal dependencies in the normalized input data. Due to its gating mechanism for controlling the flow of information, GRU is particularly effective in processing long sequences, allowing it to retain relevant past information while alleviating the gradient vanishing problem:
[0069] T g = GRU(X′);
[0070] Among them, T g ∈R B×T×G is the hidden state H of all time steps t ∈R B×G The GRU output is obtained by concatenating them. G represents the hidden layer size of GRU. Then the hidden state T is converted to g Mapping obtains temporal representation:
[0071] T l =W l T g +b l ;
[0072] Among them, the tense representation T l ∈R B×T×D Captures the features of all time steps, W l and b l are the weight matrix and bias vector.
[0073] Collaborative Representation Unit (CRU) is a key innovation of this application, which transforms and highlights the temporal patterns in the data to enhance the model's ability to learn multi-directional temporal dependencies. By extracting information not only from the temporal dimension but also from different batches and variables, the collaborative representation unit can more fully understand the underlying data patterns. The collaborative representation unit first uses a learnable weight matrix W c Modify tense representation T l :
[0074] C=T l ·W c ;
[0075] Where C∈RB×T×D is the converted data, W c ∈R D×T Learn the weights for each time step and variable. The differences in the data C along different dimensions (d=0, 1, 2 for batch, sequence, and variable) are:
[0076]
[0077] This process focuses the model's attention on the most relevant aspects of the data.
[0078] The representation fusion unit splices the obtained collaborative representations and temporal representations in the three directions to converge the collaborative time representation in the wind power generation data, namely the fused representation:
[0079] P c =[C 0 ; C 1 ; C 2 ; T l ];
[0080] Among them, P c ∈R B×T×4D .
[0081] This concatenated tensor is processed using a mapping unit consisting of autoregressive layers and multi-layer perceptrons (MLPs) to transform the data and generate predictions.
[0082] First, an autoregressive step is performed:
[0083]
[0084] Among them, P a ∈R B×1×4D , W a ∈R T×4D and b a are weights and biases respectively. Next, use MLP to achieve the final scale transformation of the data:
[0085] P m =W3(W2(W1P a +b1)+b2)+b3,
[0086] So we get the prediction result P m ∈R B×1×D .
[0087] Finally, we use reversible instance normalization to normalize P m Perform inverse normalization to obtain the final predicted power generation Y. Through inverse normalization, the data is restored to the original scale and distribution. The inverse normalization step is defined as:
[0088] X = X′·σ+μ;
[0089] Where μ and σ are the mean and standard deviation of the input data.
[0090] During model training, the model parameters are adjusted according to the comparison results of the loss function, and the optimal wind power prediction model is obtained after reaching the preset training cutoff condition.
[0091] After obtaining the wind power prediction model, the real-time wind power data of the wind power to be detected can be input into the wind power prediction model to predict the wind power of the wind power in a future set time period, and finally the prediction result is sent to the power grid operator.
[0092] like Figure 3 As shown, wind turbine data (i.e., wind turbine power generation data) is acquired in real time, the wind turbine data is preprocessed, and then model training and real-time prediction are performed, and the prediction results are sent to the grid operator. The grid operator can adjust the power supply of the wind power network, which provides a reliable basis for efficient management and resource optimization of the grid, and effectively improves the stability and management efficiency of the grid.
[0093] The wind power prediction model provided in this application aims to improve the accuracy and robustness of wind power prediction. This application standardizes wind power data by introducing reversible instance normalization, uses recurrent neural networks (RNNs) to capture long-term temporal dependencies, and extracts and enhances multi-dimensional temporal dynamic features through collaborative representation units (CRUs). Experimental results show that the wind power prediction method provided in this application has significantly better prediction accuracy than many existing prediction methods on multiple actual wind power data sets, has better adaptability and generalization ability, can effectively cope with the spatiotemporal complexity in wind power prediction, and improve the stability of the power grid and the integration efficiency of renewable energy. This application is applicable to fields such as wind farm power prediction, power system scheduling, and energy management.
[0094] The wind power prediction method provided in the embodiment of the present application has the following beneficial effects:
[0095] a) Improve the accuracy and robustness of predictions
[0096] This application integrates reversible instance normalization, hidden state learning based on recurrent neural network (RNN), and specialized collaborative representation units, which can better capture important features in the data and reduce prediction errors, especially under complex time series data. By integrating collaborative representation units (CRU), the model can better emphasize and extract key features in the time series and enhance the robustness of the model.
[0097] b) Strong adaptability and better ability to improve effective integration
[0098] The present invention does not rely on external meteorological data, but learns from historical wind power generation data, and captures multi-directional time dependencies through collaborative representation learning, including complex dynamics across different time scales and variables, so that it can more comprehensively understand wind power generation patterns and thus better improve effective integration capabilities.
[0099] In the above embodiment, a wind power prediction method is provided. Correspondingly, the present application also provides a wind power prediction device. The wind power prediction device provided in the embodiment of the present application can implement the above wind power prediction method. The wind power prediction device can be implemented by software, hardware, or a combination of software and hardware. For example, the wind power prediction device can include integrated or separate functional modules or units to perform the corresponding steps in the above methods. Please refer to Figure 4 As shown, the wind power prediction device 10 includes:
[0100] An acquisition module 101 is used to acquire the power generation time series data of the wind turbine in a historical time period;
[0101] The prediction module 102 is used to pre-process the power generation time series data, input a pre-trained wind power prediction model, and obtain the predicted power generation of the wind turbine generator in a future set time period output by the wind power prediction model;
[0102] The wind power prediction model is obtained by training a recurrent neural network integrated with collaborative time representation learning through historical power generation time series data samples of wind turbines.
[0103] In some embodiments of the present application, the device further includes: a training module, configured to pre-train the wind power prediction model in the following manner:
[0104] Obtain the historical power generation of the wind turbine generator, and count it as a time series within the same time interval to obtain a time series data sample of the historical power generation of the wind turbine generator;
[0105] After the preprocessing is performed on the historical power generation time series data samples, the historical power generation time series data samples are converted into power generation supervised data;
[0106] An objective function is set, and the supervised data of power generation is input into the recurrent neural network integrated with collaborative time representation learning for training, and a trained wind power prediction model is obtained after a preset training cutoff condition is reached.
[0107] In some embodiments of the present application, the training module is specifically used for:
[0108] The first several variables in the historical power generation time series data sample are used as the input variables of the model, and the last variable is aggregated as the output of the model to obtain the supervised data of power generation.
[0109] In some embodiments of the present application, the preprocessing includes maximum value processing, minimum value processing, data averaging and normalization processing.
[0110] In some embodiments of the present application, the normalization process uses a reversible instance normalization process.
[0111] In some embodiments of the present application, the recurrent neural network integrated with collaborative temporal representation learning includes:
[0112] Gated recurrent unit, which is used to capture long-term temporal dependencies in time series data and obtain temporal representation of time series data;
[0113] A collaboration representation unit, used for extracting the time dependency relationship in the temporal representation from different directions to obtain collaboration representations in different directions;
[0114] A representation fusion unit, used for serially splicing the temporal representation and the collaborative representations of different directions to obtain a fused representation;
[0115] A mapping unit is used to map the fusion representation to the predicted generated power.
[0116] In some implementations of the present application, the mapping unit includes:
[0117] An autoregressive layer is used to perform autoregressive processing on the fused representation to obtain an autoregressive fused representation;
[0118] A multi-layer perceptron is used to convert the autoregressive fusion representation into predicted generated power.
[0119] The wind power prediction device provided in the embodiment of the present application and the wind power prediction method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.
[0120] An embodiment of the present application also provides an electronic device corresponding to the method provided in the aforementioned embodiment, and the electronic device may be an electronic device for a client, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the above-mentioned wind power prediction method.
[0121] Please refer to Figure 5 , which shows a schematic diagram of an electronic device provided by some embodiments of the present application. Figure 5As shown, the electronic device 20 includes: a processor 200, a memory 201, a bus 202 and a communication interface 203, and the processor 200, the communication interface 203 and the memory 201 are connected via the bus 202; the memory 201 stores a computer program that can be run on the processor 200, and when the processor 200 runs the computer program, the wind power prediction method provided in any of the aforementioned embodiments of the present application is executed.
[0122] The memory 201 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 203 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.
[0123] The bus 202 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory 201 is used to store programs, and the processor 200 executes the program after receiving an execution instruction. The wind power prediction method disclosed in any implementation of the embodiment of the present application may be applied to the processor 200, or implemented by the processor 200.
[0124] The processor 200 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 200. The above processor 200 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a readily available programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor to be executed, or the hardware and software modules in the decoding processor can be executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 201, and the processor 200 reads the information in the memory 201 and completes the steps of the above method in combination with its hardware.
[0125] The electronic device provided in the embodiment of the present application and the wind power prediction method provided in the embodiment of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, operated or implemented therein.
[0126] An embodiment of the present application also provides a computer-readable storage medium corresponding to the wind power prediction method provided in the aforementioned embodiment, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it will execute the wind power prediction method provided in any of the aforementioned embodiments.
[0127] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical or magnetic storage media, which are not listed here one by one.
[0128] The computer-readable storage medium provided in the above-mentioned embodiments of the present application and the wind power prediction method provided in the embodiments of the present application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.
[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application 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 replace some or all of the technical features therein by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application.
Claims
1. A wind power prediction method, characterized in that: include: Obtain the time series data of the power generation of wind turbines in the historical time period; After preprocessing the power generation time series data, a pre-trained wind power prediction model is input to obtain the predicted power generation of the wind turbine generator in a future set time period output by the wind power prediction model; The wind power prediction model is obtained by training a recurrent neural network integrated with collaborative time representation learning through historical power generation time series data samples of wind turbines.
2. The method according to claim 1, characterized in that The wind power prediction model is pre-trained in the following manner: Obtain the historical power generation of the wind turbine generator, and count it as a time series within the same time interval to obtain a time series data sample of the historical power generation of the wind turbine generator; After the preprocessing is performed on the historical power generation time series data samples, the historical power generation time series data samples are converted into power generation supervised data; An objective function is set, and the supervised data of power generation is input into the recurrent neural network integrated with collaborative time representation learning for training, and a trained wind power prediction model is obtained after a preset training cutoff condition is reached.
3. The method according to claim 2, characterized in that The converting the historical power generation time series data samples into the power generation supervised data includes: The first several variables in the historical power generation time series data sample are used as the input variables of the model, and the last variable is aggregated as the output of the model to obtain the supervised data of power generation.
4. The method according to claim 1, characterized in that The preprocessing includes maximum value processing, minimum value processing, data averaging and normalization processing.
5. The method according to claim 4, characterized in that The normalization process adopts reversible instance normalization process.
6. The method according to claim 1, characterized in that The recurrent neural network integrated with collaborative temporal representation learning includes: Gated recurrent unit, which is used to capture long-term temporal dependencies in time series data and obtain temporal representation of time series data; A collaboration representation unit, used for extracting the time dependency relationship in the temporal representation from different directions to obtain collaboration representations in different directions; A representation fusion unit, used for serially splicing the temporal representation and the collaborative representations of different directions to obtain a fused representation; A mapping unit is used to map the fusion representation to the predicted generated power.
7. The method according to claim 6, characterized in that The mapping unit comprises: An autoregressive layer is used to perform autoregressive processing on the fused representation to obtain an autoregressive fused representation; A multi-layer perceptron is used to convert the autoregressive fusion representation into predicted generated power.
8. A wind power prediction device, characterized in that: include: An acquisition module is used to acquire the time series data of the power generation of the wind turbine in a historical time period; A prediction module is used to pre-process the power generation time series data, input a pre-trained wind power prediction model, and obtain the predicted power generation of the wind turbine generator in a future set time period output by the wind power prediction model; The wind power prediction model is obtained by training a recurrent neural network integrated with collaborative time representation learning through historical power generation time series data samples of wind turbines.
9. An electronic device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that: Computer-readable instructions are stored thereon, and the computer-readable instructions can be executed by a processor to implement the method according to any one of claims 1 to 7.