Wind power generation output power prediction method, device, equipment and medium
By combining spatiotemporal graph neural network and potential seasonal-trend characterization model, the existing wind power output power prediction methods are solved in terms of accuracy and complexity, and higher prediction performance, stability and controllability are achieved.
Patent Information
- Application Number
- CN202411635354.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-15
- Publication Date
- 2025-02-14
AI Technical Summary
The existing wind power output power prediction methods have shortcomings in accuracy, complexity, performance, stability and controllability, especially inaccurate predictions in occasional and sudden wind conditions.
Using a combination of spatiotemporal graph neural network and potential seasonal-trend representation model, the training set is input into the spatiotemporal graph neural network by collecting and preprocessing wind power related data, and then the output result is encoded graph and the potential seasonal-trend representation model is input. , generate a seasonal-trend characterization array and input it into the wind power output power predictor for prediction.
It improves the accuracy of wind power output power prediction, reduces the complexity of prediction, improves prediction performance, stability and controllability, and can better deal with occasional and emergencies of wind conditions and reduces uncertain risks.
Smart Images
Figure CN119448275A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of wind power technology, and in particular to a method, device, equipment and medium for predicting wind power generation output power. Background Art
[0002] As one of the important renewable energy sources that is safe, reliable, pollution-free, fuel-free, and grid-connected, wind power generation has developed rapidly in China in recent years. However, wind power generation is greatly affected by factors such as meteorology and has strong volatility and intermittency. The high proportion of wind power generation access will bring challenges to the safety and stability of the power system, and also increase the difficulty of dispatching and operation. Wind power prediction is a key technology to reduce the impact of uncertainty. It is generally based on meteorological forecasts, station data and meteorological observations, combined with wind farm location and equipment parameters to model and predict the output change trend of wind farms in the future. At present, the research on wind power prediction using machine learning and deep learning requires a lot of computing time and computing resources, which is not suitable for short-term forecasting tasks; traditional statistical methods have low prediction accuracy and are difficult to deal with complex variables; traditional machine learning is difficult to effectively capture temporary information and has difficulty in processing complex and nonlinear wind power data. Existing wind power prediction methods are not accurate for occasional and sudden wind conditions.
[0003] It can be seen from the above that how to improve the accuracy of wind power generation output power prediction, reduce the complexity of wind power generation output power prediction, and improve prediction performance, stability and controllability are problems to be solved in this field. Summary of the invention
[0004] In view of this, the purpose of the present invention is to provide a method, device, equipment and medium for predicting wind power output power, which can improve the accuracy of wind power output power prediction, reduce the complexity of wind power output power prediction, and improve prediction performance, stability and controllability. The specific scheme is as follows:
[0005] In a first aspect, the present application discloses a method for predicting wind power output power, comprising:
[0006] Collecting wind power related data, preprocessing the wind power related data to obtain the preprocessed wind power related data, and dividing the preprocessed wind power related data into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing;
[0007] Inputting the training set into a preset spatiotemporal graph neural network to output a result encoding graph, inputting the result encoding graph into a preset potential seasonality-trend characterization model to output a seasonality-trend characterization array, and inputting the seasonality-trend characterization array into a wind power generation output power predictor to output a prediction result;
[0008] The prediction result is evaluated using the test set, and if the evaluation passes, the prediction result is used as the wind power generation output power prediction result.
[0009] Optionally, the collecting wind power related data and preprocessing the wind power related data to obtain the preprocessed wind power related data includes:
[0010] Collecting wind power related data; the wind power related data includes wind power data, weather forecast data, actual wind power output data, wind tower data, wind turbine location and surrounding location geographic information;
[0011] The wind power related data are respectively subjected to missing value processing, duplicate data elimination processing and numerical variable normalization processing to obtain the pre-processed wind power related data.
[0012] Optionally, the performing missing value processing, duplicate data elimination processing and numerical variable normalization processing on the wind power related data respectively includes:
[0013] Using the KNN interpolation algorithm to process missing values of the wind power related data, and performing duplicate data elimination processing on the wind power related data after the missing value processing;
[0014] The minimum-maximum normalization method is used to perform variable normalization processing on the wind power related data after the elimination processing.
[0015] Optionally, the spatiotemporal graph neural network includes a graph neural network and a sequence model; the sequence model includes a recurrent neural network, a gated recurrent unit and a Transformers large model; the potential seasonality-trend representation model includes corresponding encoders and decoders.
[0016] Optionally, the seasonality-trend characterization array is input into a wind power output power predictor to output a prediction result, including:
[0017] The seasonality-trend characterization array is input into a wind power generation output power predictor, so that the wind power generation output power predictor performs seasonal prediction and trend prediction on the seasonality-trend characterization array respectively to output prediction results.
[0018] Optionally, the step of inputting the seasonality-trend characterization array into a wind power generation output power predictor so that the wind power generation output power predictor performs seasonality prediction and trend prediction on the seasonality-trend characterization array respectively comprises:
[0019] The seasonality-trend characterization array is input into a wind power generation output power predictor so that the wind power generation output power predictor uses a discrete Fourier transform algorithm to perform seasonal prediction on the seasonality-trend characterization array and uses a feedforward neural network to perform trend prediction on the seasonality-trend characterization array.
[0020] Optionally, the using the test set to evaluate the prediction result, and if the evaluation passes, using the prediction result as the wind power generation output power prediction result, includes:
[0021] The prediction results are evaluated using the test set to obtain evaluation indicators, wherein the evaluation indicators include mean absolute error, mean absolute percentage error, and root mean square error.
[0022] If the evaluation index is passed, the prediction result is used as the wind power generation output power prediction result.
[0023] In a second aspect, the present application discloses a wind power generation output power prediction device, comprising:
[0024] A data acquisition and processing module, used for acquiring wind power related data, preprocessing the wind power related data to obtain the preprocessed wind power related data, and dividing the preprocessed wind power related data into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing;
[0025] A prediction result output module, used for inputting the training set into a preset spatiotemporal graph neural network to output a result coding map, inputting the result coding map into a preset potential seasonality-trend characterization model to output a seasonality-trend characterization array, and inputting the seasonality-trend characterization array into a wind power generation output power predictor to output a prediction result;
[0026] The prediction result evaluation module is used to evaluate the prediction result using the test set, and if the evaluation passes, the prediction result is used as the wind power generation output power prediction result.
[0027] In a third aspect, the present application discloses an electronic device, comprising:
[0028] Memory, used to store computer programs;
[0029] The processor is used to execute the computer program to implement the above-mentioned wind power generation output power prediction method.
[0030] In a fourth aspect, the present application discloses a computer storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the wind power generation output power prediction method disclosed above are implemented.
[0031] It can be seen that the present application provides a method for predicting wind power output power, including collecting wind power related data, preprocessing the wind power related data to obtain the preprocessed wind power related data, and dividing the preprocessed wind power related data into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing; the training set is input into a preset spatiotemporal graph neural network to output a result encoding map, the result encoding map is input into a preset potential seasonal-trend characterization model to output a seasonal-trend characterization array, and the seasonal-trend characterization array is input into a wind power output power predictor to output a prediction result; the prediction result is evaluated using the test set, and if the evaluation passes, the prediction result is used as the wind power output power prediction result. The present application preprocesses wind power related data, and divides the preprocessed wind power related data into a training set and a test set, inputs the training set into a space-time graph neural network, improves the fitting ability and prediction ability of the space-time graph neural network, obtains a result coding map, uses a potential seasonal-trend characterization model to obtain a seasonal-trend characterization array at the current moment, combines the space-time neural network with the potential seasonal-trend characterization model, can make full use of time and space information, distinguishes it from common time series model processing methods, improves prediction performance and stability, uses a predictor to predict future moments, obtains prediction results, can reduce uncertainty risks, and achieve better combined scheduling of wind power generation systems, providing guarantees for the safe and stable operation of wind farms and power systems. In addition, compared with other models, the potential seasonal-trend characterization model has stronger robust performance, can improve the accuracy of wind power output power prediction, reduce the complexity of wind power output power prediction, and improve prediction performance, stability and controllability. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0033] Figure 1A flow chart of a method for predicting wind power output power disclosed in this application;
[0034] Figure 2 A spatiotemporal graph neural network structure diagram disclosed in this application;
[0035] Figure 3 A potential seasonality-trend characterization model structure diagram disclosed in this application;
[0036] Figure 4 A structural diagram of a wind power generation output power predictor disclosed in this application;
[0037] Figure 5 A specific wind power generation output power prediction flow chart disclosed in this application;
[0038] Figure 6 This is a schematic diagram of the structure of a wind power generation output power prediction device disclosed in this application;
[0039] Figure 7 A structural diagram of an electronic device provided for this application. DETAILED DESCRIPTION
[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0041] As one of the important renewable energy sources that is safe, reliable, pollution-free, fuel-free, and grid-connected, wind power generation has developed rapidly in China in recent years. However, wind power generation is greatly affected by factors such as meteorology and has strong volatility and intermittency. The high proportion of wind power generation access will bring challenges to the safety and stability of the power system, and also increase the difficulty of dispatching and operation. Wind power prediction is a key technology to reduce the impact of uncertainty. It is generally based on meteorological forecasts, station data and meteorological observations, combined with wind farm location and equipment parameters to model and predict the output change trend of wind farms in the future. At present, the research on wind power prediction using machine learning and deep learning requires a lot of computing time and computing resources, which is not suitable for short-term forecasting tasks; traditional statistical methods have low prediction accuracy and are difficult to deal with complex variables; traditional machine learning is difficult to effectively capture temporary information and has difficulty in processing complex and nonlinear wind power data. Existing wind power prediction methods are not accurate for occasional and sudden wind conditions. It can be seen from the above that how to improve the accuracy of wind power generation output power prediction, reduce the complexity of wind power generation output power prediction, and improve prediction performance, stability and controllability are problems to be solved in this field.
[0042] See also Figure 1 As shown, the embodiment of the present invention discloses a method for predicting wind power output power, which may specifically include:
[0043] Step S11: Collect wind power related data, preprocess the wind power related data to obtain the preprocessed wind power related data, and divide the preprocessed wind power related data into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing.
[0044] In this embodiment, wind power related data are collected, and the wind power related data are respectively subjected to missing value processing, duplicate data elimination processing and numerical variable normalization processing to obtain the pre-processed wind power related data; the wind power related data include wind power data, weather forecast data, actual wind power output data, wind tower data, wind turbine location and surrounding location geographic information; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing.
[0045] Among them, the specific process of data preprocessing is: using the KNN interpolation algorithm to process the missing values of the wind power-related data, and removing duplicate data from the wind power-related data after the missing value processing; using the minimum-maximum normalization method to normalize the variables of the wind power-related data after the elimination processing, so as to obtain the preprocessed wind power-related data.
[0046] In this embodiment, the missing value processing uses KNN (K-Nearest Neighbors) interpolation technology to replace the missing values of a given variable with the average of adjacent observations. The missing value is determined by calculating the distance between the missing observation and other observations using the Euclidean distance calculation formula:
[0047] ;
[0048] in, For variable i in the target observation The value in is the value of variable i in another observation The value in is the distance between the target observation value and the observation value;
[0049] Normalization of numerical variables uses the min-max normalization method to transform the data into the range (0, 1):
[0050] ;
[0051] in, is the standardized X value, is the current value of variable X, and are the maximum and minimum values of the variable in the input data set.
[0052] Step S12: Input the training set into a preset spatiotemporal graph neural network to output a result coding map, input the result coding map into a preset potential seasonality-trend characterization model to output a seasonality-trend characterization array, and input the seasonality-trend characterization array into a wind power generation output power predictor to output a prediction result.
[0053] Among them, ST-GNN (Space-time Graph Neural Networks) includes a series of GNN (Graph Neural Network) and sequence models; the sequence model includes RNN (Recurrent Neural Network), GRU (Gated Recurrent Unit) and Transformers large model; the potential seasonality-trend representation model includes corresponding encoders and decoders, namely seasonal encoder-decoder and trend encoder-decoder, and follows the variational inference theory to learn some decoupled representation arrays that describe the seasonality and trend in time series data. The inference method uses variational inference to maximize the confidence lower limit. , to obtain the potential representation and Taking GRU as an example, the structure of the spatiotemporal graph neural network is as follows: Figure 2 As shown, the structure of the potential seasonality-trend characterization model is as follows Figure 3 shown.
[0054] Each time step of ST-GNN is a graph, which is passed through the GCN (Graph Convolutional Networks) network to obtain the resulting encoding graph that embeds the spatial interdependencies of the data. The resulting encoding graph can be modeled like time series data as input to the potential seasonality-trend representation model.
[0055] In this embodiment, after obtaining the seasonal-trend characterization array, the seasonal-trend characterization array is input into a wind power generation output power predictor so that the wind power generation output power predictor performs seasonal prediction and trend prediction on the seasonal-trend characterization array respectively to output prediction results.
[0056] That is, after obtaining the seasonality-trend characterization array, the seasonality-trend characterization array is input into the wind power generation output power predictor, so that the wind power generation output power predictor uses a DFT (DiscreteFourier Transform) algorithm to perform seasonal prediction on the seasonality-trend characterization array, and uses two FFNs (Feed-Forward Neural Networks) to perform trend prediction on the seasonality-trend characterization array. The structure of the wind power generation output power predictor is as follows: Figure 4 shown.
[0057] Seasonality and trend are common characteristics of time series. Trend is monotonicity over a period of time, and the slope is generally fixed; seasonality is a change of fixed length, just like the temperature change in spring, summer, autumn and winter. Due to the normal changes of fixed cycles such as the change of seasons and the alternation of night and day, as well as the inherent laws of wind speed changes and wind turbine characteristics, there are a lot of potential seasonality and trend in wind power data.
[0058] Consider the output spatiotemporal information encoding diagram after ST-GNN processing: ,in The univariate or multivariate value representing the observation at the current moment. The goal is to find an expression Model to predict future sequences (i.e. wind power forecast at a future time). The model for inferring the likelihood between observation X with potential representation Z and future Y is shown below:
[0059] ;
[0060] Using variational inference, the likelihood P(X,Y) is obtained by the posterior distribution To calculate and maximize it by using the lower confidence limit:
[0061] ;
[0062] in, Depend on , composition, represents the learned parameters;
[0063] Using separate reconstructions indirectly promotes the decoupling of seasonal and trend representations, so it is necessary to impose direct constraints on the representations themselves. This is done by introducing an additional mutual information regularization. Specifically, by increasing , The mutual information between X and X can solve the discrete reduction problem, which is conducive to further decoupling and representation.
[0064] Due to the information entanglement problem in spatiotemporal series data, it is necessary to incorporate the decoupling strategy into the potential seasonality-trend characterization model. The spatiotemporal signals X and Y are formulated as the sum of the seasonality and trend components, that is: Therefore, the latent representation Z is decomposed into and ,in and Independent of each other, that is . The maximization objective is expressed as:
[0065] ;
[0066] Where I(,) represents the mutual information between two representations.
[0067] Furthermore, in order to optimize the training effect of the seasonality-trend representation model, and The lower limit and The upper limit is limited.
[0068] Specifically, for and We use the KL (Kullback-Leibler divergence) between the joint distribution and the boundary, and define a family of energy-based variational inference to achieve a flexible and scalable lower bound:
[0069] ;
[0070] Specifically, for By introducing an energy-based variational inference family , using the normalized critical To establish a retroactive upper bound on the season-trend mutual information:
[0071] ;
[0072] By inputting the spatiotemporal information encoding map into the latent seasonal-trend representation model (LLST, Learning Latent Seasonal-Trend representations), the seasonal-trend representation ( and ) as input to the forecaster. The forecaster consists of two separate components: seasonal forecast ( ) and trend forecast ( ). Given a seasonal representation ,Seasonal prediction first uses the Discrete Fourier Transform (DFT) algorithm to detect seasonal frequencies in the input, i.e. , where due to the Nyquist theorem, The frequency is then inverted back to the time domain to extend it to future predictions, i.e. Given a trend characterization , trend prediction via feed-forward network (FFN) To produce a predictable representation . And use two FFNs to get the final prediction, and Mapped to and .
[0073] Step S13: using the test set to evaluate the prediction result, if the evaluation passes, the prediction result is used as the wind power generation output power prediction result.
[0074] In this embodiment, the prediction result is evaluated using a test set to obtain evaluation indicators; the evaluation indicators include MAE (Mean Absolute Error), MAPE (Mean Absolute Percentage), and RMSE (Root Mean Squard Error); if the evaluation indicators pass, the prediction result is used as the wind power output power prediction result.
[0075] The future moment and The wind power forecast result Y at the future time is obtained by summing up, that is: .
[0076] The prediction results are expressed by prediction evaluation indicators. The accuracy of the model can be evaluated by comparing the prediction results with the real data using regression indicators. The prediction evaluation indicators include mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE):
[0077] ;
[0078] ;
[0079] .
[0080] The specific process of wind power generation output power prediction in this application is as follows Figure 5 As shown, firstly, wind power related data is collected, and preprocessed, analyzed and converted, and data is divided; then the data of the training set is sent to the spatiotemporal graph neural network to obtain the result encoding map embedded in the data space interdependence; then the result encoding map is sent to the potential seasonality-trend characterization model to obtain the seasonal trend-potential representation form; then the potential seasonal trend-potential representation is sent to the predictor to obtain the model output value, that is, the ultra-short-term prediction and the medium-short-term prediction results; finally, the obtained prediction results are evaluated, and if the evaluation passes, the prediction results are used as the wind power output power prediction results. This application can reduce uncertainty risks and achieve better combined scheduling of wind power generation systems by predicting the output power of wind farms in the future, thus providing a guarantee for the safe and stable operation of wind farms and power systems. Compared with other models, the wind power prediction method based on the potential seasonality-trend characterization model proposed in this application shows better prediction results, is insensitive to noise in the data, and has stronger robust performance.
[0081] In this embodiment, wind power related data are collected, and the wind power related data are preprocessed to obtain the preprocessed wind power related data, and the preprocessed wind power related data are divided into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing; the training set is input into a preset spatiotemporal graph neural network to output a result encoding map, the result encoding map is input into a preset potential seasonal-trend characterization model to output a seasonal-trend characterization array, and the seasonal-trend characterization array is input into a wind power generation output power predictor to output a prediction result; the prediction result is evaluated using the test set, and if the evaluation passes, the prediction result is used as the wind power generation output power prediction result. The present application preprocesses wind power related data, and divides the preprocessed wind power related data into a training set and a test set, inputs the training set into a space-time graph neural network, improves the fitting ability and prediction ability of the space-time graph neural network, obtains a result coding map, uses a potential seasonal-trend characterization model to obtain a seasonal-trend characterization array at the current moment, combines the space-time neural network with the potential seasonal-trend characterization model, can make full use of time and space information, distinguishes it from common time series model processing methods, improves prediction performance and stability, uses a predictor to predict future moments, obtains prediction results, can reduce uncertainty risks, and achieve better combined scheduling of wind power generation systems, providing guarantees for the safe and stable operation of wind farms and power systems. In addition, compared with other models, the potential seasonal-trend characterization model has stronger robust performance, can improve the accuracy of wind power output power prediction, reduce the complexity of wind power output power prediction, and improve prediction performance, stability and controllability.
[0082] See also Figure 6 As shown, the embodiment of the present invention discloses a wind power generation output power prediction device, which may specifically include:
[0083] The data acquisition and processing module 11 is used to acquire wind power related data, pre-process the wind power related data to obtain the pre-processed wind power related data, and divide the pre-processed wind power related data into a training set and a test set; the pre-processing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing;
[0084] A prediction result output module 12 is used to input the training set into a preset spatiotemporal graph neural network to output a result coding map, input the result coding map into a preset potential seasonality-trend characterization model to output a seasonality-trend characterization array, and input the seasonality-trend characterization array into a wind power generation output power predictor to output a prediction result;
[0085] The prediction result evaluation module 13 is used to evaluate the prediction result using the test set, and if the evaluation passes, the prediction result is used as the wind power generation output power prediction result.
[0086] In this embodiment, wind power related data are collected, and the wind power related data are preprocessed to obtain the preprocessed wind power related data, and the preprocessed wind power related data are divided into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing; the training set is input into a preset spatiotemporal graph neural network to output a result encoding map, the result encoding map is input into a preset potential seasonal-trend characterization model to output a seasonal-trend characterization array, and the seasonal-trend characterization array is input into a wind power generation output power predictor to output a prediction result; the prediction result is evaluated using the test set, and if the evaluation passes, the prediction result is used as the wind power generation output power prediction result. The present application preprocesses wind power related data, and divides the preprocessed wind power related data into a training set and a test set, inputs the training set into a space-time graph neural network, improves the fitting ability and prediction ability of the space-time graph neural network, obtains a result coding map, uses a potential seasonal-trend characterization model to obtain a seasonal-trend characterization array at the current moment, combines the space-time neural network with the potential seasonal-trend characterization model, can make full use of time and space information, distinguishes it from common time series model processing methods, improves prediction performance and stability, uses a predictor to predict future moments, obtains prediction results, can reduce uncertainty risks, and achieve better combined scheduling of wind power generation systems, providing guarantees for the safe and stable operation of wind farms and power systems. In addition, compared with other models, the potential seasonal-trend characterization model has stronger robust performance, can improve the accuracy of wind power output power prediction, reduce the complexity of wind power output power prediction, and improve prediction performance, stability and controllability.
[0087] In some specific embodiments, the data collection and processing module 11 may specifically include:
[0088] A data acquisition module is used to collect wind power related data; the wind power related data includes wind power data, weather forecast data, actual wind power output data, wind tower data, wind turbine location and surrounding location geographical information;
[0089] The data processing module is used to perform missing value processing, duplicate data elimination processing and numerical variable normalization processing on the wind power related data to obtain the pre-processed wind power related data.
[0090] In some specific embodiments, the data collection and processing module 11 may specifically include:
[0091] A missing value and duplicate data elimination processing module is used to perform missing value processing on the wind power related data using a KNN interpolation algorithm, and to perform duplicate data elimination processing on the wind power related data after the missing value processing;
[0092] The variable normalization processing module is used to perform variable normalization processing on the wind power related data after elimination processing by using the minimum-maximum normalization method.
[0093] In some specific embodiments, the spatiotemporal graph neural network includes a graph neural network and a sequence model; the sequence model includes a recurrent neural network, a gated recurrent unit, and a Transformers large model; the potential seasonality-trend representation model includes corresponding encoders and decoders.
[0094] In some specific embodiments, the prediction result output module 12 may specifically include:
[0095] The prediction result output module is used to input the seasonal-trend characterization array into the wind power generation output power predictor, so that the wind power generation output power predictor performs seasonal prediction and trend prediction on the seasonal-trend characterization array respectively to output the prediction result.
[0096] In some specific embodiments, the prediction result output module 12 may specifically include:
[0097] A seasonal prediction module is used to input the seasonal-trend characterization array into a wind power generation output power predictor so that the wind power generation output power predictor uses a discrete Fourier transform algorithm to perform seasonal prediction on the seasonal-trend characterization array and uses a feedforward neural network to perform trend prediction on the seasonal-trend characterization array.
[0098] In some specific embodiments, the prediction result evaluation module 13 may specifically include:
[0099] An evaluation module is used to evaluate the prediction results using the test set to obtain evaluation indicators; the evaluation indicators include mean absolute error, mean absolute percentage error, and root mean square error;
[0100] The wind power generation output power prediction result determination module is used to use the prediction result as the wind power generation output power prediction result if the evaluation index is passed.
[0101] Figure 7A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the wind power output power prediction method performed by the electronic device disclosed in any of the aforementioned embodiments.
[0102] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.
[0103] In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon include an operating system 221, a computer program 222 and data 223, etc. The storage method can be temporary storage or permanent storage.
[0104] Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20, so as to realize the operation and processing of the data 223 in the memory 22 by the processor 21, which can be Windows, Unix, Linux, etc. In addition to including a computer program that can be used to complete the wind power output power prediction method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks. In addition to including data transmitted from an external device received by the wind power output power prediction device, the data 223 can also include data collected by its own input and output interface 25, etc.
[0105] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
[0106] Furthermore, an embodiment of the present application also discloses a computer-readable storage medium, in which a computer program is stored. When the computer program is loaded and executed by a processor, the steps of the wind power generation output power prediction method disclosed in any of the aforementioned embodiments are implemented.
[0107] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0108] The above is a detailed introduction to a wind power generation output power prediction method, device, equipment and storage medium provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.
Claims
1. A method for predicting wind power output power, characterized in that: include: Collecting wind power related data, preprocessing the wind power related data to obtain the preprocessed wind power related data, and dividing the preprocessed wind power related data into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing; Inputting the training set into a preset spatiotemporal graph neural network to output a result encoding graph, inputting the result encoding graph into a preset potential seasonality-trend characterization model to output a seasonality-trend characterization array, and inputting the seasonality-trend characterization array into a wind power generation output power predictor to output a prediction result; The prediction result is evaluated using the test set, and if the evaluation passes, the prediction result is used as the wind power generation output power prediction result.
2. The method for predicting wind power output according to claim 1, characterized in that: The collecting of wind power related data and preprocessing the wind power related data to obtain the preprocessed wind power related data include: Collecting wind power related data; the wind power related data includes wind power data, weather forecast data, actual wind power output data, wind tower data, wind turbine location and surrounding location geographic information; The wind power related data are respectively subjected to missing value processing, duplicate data elimination processing and numerical variable normalization processing to obtain the pre-processed wind power related data.
3. The method for predicting wind power output according to claim 2, characterized in that: The step of performing missing value processing, duplicate data elimination processing, and numerical variable normalization processing on the wind power related data includes: Using the KNN interpolation algorithm to process missing values of the wind power related data, and performing duplicate data elimination processing on the wind power related data after the missing value processing; The minimum-maximum normalization method is used to perform variable normalization processing on the wind power related data after the elimination processing.
4. The method for predicting wind power output according to claim 1, characterized in that: The spatiotemporal graph neural network includes a graph neural network and a sequence model; the sequence model includes a recurrent neural network, a gated recurrent unit and a Transformers large model; the potential seasonality-trend representation model includes corresponding encoders and decoders.
5. The method for predicting wind power output according to claim 1, characterized in that: The step of inputting the seasonality-trend characterization array into a wind power generation output power predictor to output a prediction result includes: The seasonality-trend characterization array is input into a wind power generation output power predictor, so that the wind power generation output power predictor performs seasonal prediction and trend prediction on the seasonality-trend characterization array respectively to output prediction results.
6. The method for predicting wind power output according to claim 5, characterized in that: The step of inputting the seasonality-trend characterization array into a wind power generation output power predictor so that the wind power generation output power predictor performs seasonality prediction and trend prediction on the seasonality-trend characterization array respectively comprises: The seasonality-trend characterization array is input into a wind power generation output power predictor so that the wind power generation output power predictor uses a discrete Fourier transform algorithm to perform seasonal prediction on the seasonality-trend characterization array and uses a feedforward neural network to perform trend prediction on the seasonality-trend characterization array.
7. The method for predicting wind power output according to any one of claims 1 to 6, characterized in that: The step of evaluating the prediction result by using the test set and taking the prediction result as the wind power generation output power prediction result if the evaluation passes includes: The prediction results are evaluated using the test set to obtain evaluation indicators, wherein the evaluation indicators include mean absolute error, mean absolute percentage error, and root mean square error. If the evaluation index is passed, the prediction result is used as the wind power generation output power prediction result.
8. A wind power generation output power prediction device, characterized in that: include: A data acquisition and processing module, used for acquiring wind power related data, preprocessing the wind power related data to obtain the preprocessed wind power related data, and dividing the preprocessed wind power related data into a training set and a test set; the preprocessing includes missing value processing, duplicate data elimination processing and numerical variable normalization processing; A prediction result output module, used for inputting the training set into a preset spatiotemporal graph neural network to output a result coding map, inputting the result coding map into a preset potential seasonality-trend characterization model to output a seasonality-trend characterization array, and inputting the seasonality-trend characterization array into a wind power generation output power predictor to output a prediction result; The prediction result evaluation module is used to evaluate the prediction result using the test set, and if the evaluation passes, the prediction result is used as the wind power generation output power prediction result.
9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the method for predicting wind power output power according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: Used to store a computer program; wherein, when the computer program is executed by a processor, the wind power generation output power prediction method according to any one of claims 1 to 7 is implemented.