Short-term wind power prediction method, system, medium and equipment under typhoon scenario
By determining the meteorological factors and maximum wind power generation capacity in typhoon scenarios and adjusting the power prediction model in real time, the problem of wind power power prediction deviation in typhoon weather is solved, and the prediction accuracy and stability are improved.
Patent Information
- Application Number
- CN202410280082.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-12
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-03-12
AI Technical Summary
The existing technology has large deviations in the prediction of wind power in typhoon weather, and the impact of extreme events such as fan cutting machines is not effectively considered, resulting in low prediction accuracy.
By determining the set of meteorological factors, screening key meteorological factors, and modifying the upper limit activation function of the power prediction model in real time based on the prediction results of wind power maximum power generation capacity, a short-term wind power power prediction model in typhoon scenarios is constructed to improve prediction accuracy.
It significantly improves the prediction accuracy of short-term wind power in typhoon scenarios, reduces prediction time, avoids the pollution of the model by data of high wind speed and low output, and reduces the difficulty of modeling.
Smart Images

Figure CN118232318B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of short-term wind power prediction, and in particular relates to a short-term wind power prediction method, system, medium and equipment in a typhoon scenario. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] At present, the power prediction technology for typhoon scenarios is divided into three types: beneficial, defensive and destructive, according to the impact of tropical cyclones on wind farm power generation. In areas affected by weaker tropical cyclones and their outer circulations, the impact of typhoons on wind power may be beneficial, and wind turbines can operate near the rated wind speed for a long time, more than doubling the power generation of wind turbines. Based on the above risks and difficulties, it is of great significance to improve the accuracy of wind power prediction in typhoon extreme weather scenarios.
[0004] At present, the accuracy of wind power prediction is mainly improved from three aspects: prediction model input, output and prediction model itself. Regarding the prediction model itself, although the traditional typical architecture is good at extracting global features, it lacks control over local features, especially in the scenario where the uncertainty of wind power prediction increases significantly in typhoon weather, there is an urgent need to enhance the ability to extract local features to cope with the scenario of new energy prediction in typhoon weather. Existing studies have failed to consider the impact of extreme events such as wind turbine cutting caused by typhoons, and mainly conducted research from the perspective of event prediction, or analyzed and established prediction models from the perspective of mathematical statistics. The prediction results mainly serve as risk warnings. Since the current prediction model fails to fully consider the high wind cutting situation, the abnormal situation of high wind speed and low output caused by cutting the machine during the training process leads to pollution in the training process, which causes the large deviation of wind power prediction in typhoon weather. Summary of the invention
[0005] In order to solve the technical problem of large deviation in wind power prediction under typhoon weather existing in the above-mentioned background technology, the present invention provides a short-term wind power prediction method, system, medium and equipment under typhoon scenario, which can comprehensively consider the characteristic dependence between the predicted value of wind power maximum generating capacity and the wind power prediction error, and effectively improve the accuracy of short-term wind power prediction under typhoon scenario.
[0006] In order to achieve the above object, the present invention adopts the following technical solution:
[0007] A first aspect of the present invention provides a short-term wind power prediction method in a typhoon scenario.
[0008] A short-term wind power prediction method in a typhoon scenario includes:
[0009] Based on the relationship between wind power prediction error under typhoon conditions and typhoon weather process, a set of meteorological factors that affect wind power under typhoon conditions is determined.
[0010] According to the correlation coefficients between different meteorological characteristics in the meteorological factor set and the typhoon data set, key meteorological factors are selected from the above meteorological factor set to obtain the model input characteristics under the typhoon scenario;
[0011] Based on the predefined wind power maximum power generation capacity, the theoretical value of wind power maximum power generation capacity is calibrated, and the meteorological data corresponding to the model input characteristics and the theoretical value of wind power maximum power generation capacity are used to form a prediction data set, and then combined with the preset power generation capacity prediction model to obtain the wind power maximum power generation capacity prediction result;
[0012] Based on the prediction result of the maximum power generation capacity of wind power, the upper limit activation function of the preset power prediction model is modified in real time to determine the final power prediction model; wherein the power prediction model is jointly constructed by the feature extraction model and the power generation capacity prediction model;
[0013] The short-term wind power forecast results under the typhoon scenario are obtained by using the final power forecast model and the meteorological data corresponding to the model input characteristics under the typhoon scenario.
[0014] A second aspect of the present invention provides a short-term wind power prediction system in a typhoon scenario.
[0015] A short-term wind power prediction system in a typhoon scenario, comprising:
[0016] A meteorological factor set determination module, which is used to determine the meteorological factor set that affects the wind power under typhoon conditions based on the relationship between the wind power prediction error and the typhoon weather process;
[0017] A model input feature determination module is used to select key meteorological factors from the meteorological factor set according to the correlation coefficient between different meteorological features in the meteorological factor set and the typhoon data set, so as to obtain the model input features under the typhoon scenario;
[0018] The maximum power generation capacity prediction module is used to calibrate the theoretical value of the maximum wind power generation capacity based on the predefined maximum wind power generation capacity, form a prediction data set with the meteorological data corresponding to the model input characteristics and the theoretical value of the maximum wind power generation capacity, and then combine it with the preset power generation capacity prediction model to obtain the prediction result of the maximum wind power generation capacity;
[0019] A power prediction model determination module is used to modify the upper limit activation function of the preset power prediction model in real time based on the wind power maximum power generation capacity prediction result, and determine the final power prediction model; wherein the power prediction model is jointly constructed by the feature extraction model and the power generation capacity prediction model;
[0020] The short-term wind power prediction module is used to obtain the short-term wind power prediction result under the typhoon scenario by using the final power prediction model and the meteorological data corresponding to the model input characteristics under the typhoon scenario.
[0021] A third aspect of the present invention provides a computer-readable storage medium.
[0022] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the short-term wind power prediction method under a typhoon scenario as described above.
[0023] A fourth aspect of the present invention provides an electronic device.
[0024] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for short-term wind power prediction in a typhoon scenario as described above are implemented.
[0025] Compared with the prior art, the present invention has the following beneficial effects:
[0026] (1) The present invention selects key meteorological factors from the meteorological factor set according to the correlation coefficient between different meteorological characteristics in the meteorological factor set and the typhoon data set, and obtains the model input characteristics under the typhoon scenario, which can reduce the prediction time while ensuring the accuracy of short-term wind power prediction under the typhoon scenario.
[0027] (2) The present invention proposes a definition of the maximum power generation capacity of wind power. By modifying the upper limit activation function of the prediction model in real time through the prediction results, the prediction power boundary is constrained, which can effectively avoid abnormal data such as high wind speed and low output from polluting the neural network model and reduce the difficulty of modeling.
[0028] (3) The uncertainty of power prediction comes from the random errors of meteorological prediction and the inherent errors of the model itself, which leads to the phenomenon that the uncertainty of power prediction increases with the extension of prediction time. The present invention introduces a power prediction method considering the maximum power generation capacity of wind power, and associates the influence of the prediction results of the maximum power generation capacity of wind power on the power prediction model to reduce the uncertainty of meteorological prediction such as wind speed, significantly improve the accuracy of power prediction, and thus determine the final power prediction model.
[0029] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0031] Figure 1 is a schematic diagram of an electronic device according to an embodiment of the present invention;
[0032] Figure 2 This is a flow chart of a method for predicting short-term wind power in a typhoon scenario according to an embodiment of the present invention;
[0033] Figure 3 Schematic diagram of a CNN convolutional network in an embodiment of the present invention;
[0034] Figure 4 This is a flow chart of CNN-Transformer power prediction in an embodiment of the present invention;
[0035] Figure 5 This is a diagram showing the comparison results of wind power maximum power generation capacity prediction in an embodiment of the present invention;
[0036] Figure 6 This is a diagram of the power prediction results of a station without a large wind shedding event in an embodiment of the present invention;
[0037] Figure 7 This is a diagram showing the power prediction results of a station when a large wind shedding event occurs in an embodiment of the present invention;
[0038] Figure 8 Schematic diagram of the structure of a short-term wind power prediction system in a typhoon scenario in an embodiment of the present invention. DETAILED DESCRIPTION
[0039] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.
[0040] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.
[0042] Reference Figure 1 , a schematic diagram of an electronic device is given. It should be noted that, Figure 1 The electronic device 100 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.
[0043] The electronic device 100 may be a server or other terminal device with communication and program execution functions, which will not be described in detail here.
[0044] like Figure 1 As shown, the electronic device 100 includes a central processing unit (CPU) 101, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 102 or a program loaded from a storage part 108 into a random access memory (RAM) 103. In the RAM 103, various programs and data required for system operation are also stored. The central processing unit 101, the ROM 102, and the RAM 103 are connected to each other via a bus 104. An input / output (I / O) interface 105 is also connected to the bus 104.
[0045] The following components are connected to the I / O interface 105: an input section 106 including a keyboard, a mouse, etc.; an output section 107 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 108 including a hard disk, etc.; and a communication section 109 including a network interface card such as a local area network (LAN) card, a modem, etc. The communication section 109 performs communication processing via a network such as the Internet. A drive 110 is also connected to the I / O interface 105 as needed. A removable medium 111, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 110 as needed, so that a computer program read therefrom is installed into the storage section 108 as needed.
[0046] When the central processing unit 101 in the electronic device 100 of this embodiment executes the program, the following is achieved: Figure 2 The steps in the short-term wind power prediction method under the typhoon scenario are shown.
[0047] Figure 2 A flow chart of a short-term wind power forecasting method under typhoon scenario is given. Figure 2 , a short-term wind power prediction method under typhoon scenario, which specifically includes the following steps:
[0048] S201: Based on the relationship between the wind power prediction error under typhoon conditions and the typhoon weather process, a set of meteorological factors affecting the typhoon on wind power is determined.
[0049] In some specific implementation processes, the relationship between wind power prediction error under typhoon conditions and typhoon weather process is studied from the single machine-station multi-level.
[0050] S202: According to the correlation coefficients between different meteorological characteristics in the meteorological factor set and the typhoon data set, key meteorological factors are selected from the meteorological factor set to obtain model input features under the typhoon scenario.
[0051] In some specific implementation processes, Gaussian retrieval analysis methods are used to screen out key meteorological factors from a set of meteorological factors that affect typhoons on wind power.
[0052] Among them, the process of selecting key meteorological factors from the set of meteorological factors affecting typhoon on wind power by using Gaussian retrieval analysis method is as follows:
[0053] Based on the meteorological data corresponding to the set of meteorological factors affecting wind power, a Gaussian similarity coefficient matrix is constructed;
[0054] The absolute difference between the Gaussian similarity coefficient matrix and the selected meteorological sequence elements including wind speed, wind direction, temperature and humidity is calculated, and the grey correlation degree is further calculated to construct the grey correlation degree matrix;
[0055] The weighted grey correlation matrix is summed and normalized to obtain the weighted meteorological characteristic similarity coefficients and sort them to determine the order sequence of similar characteristics of typhoons and finally determine the key meteorological factors.
[0056] Among them, the Gaussian retrieval analysis method is a method that imitates human thinking and reasoning process, using previous experience to reason and make decisions. Therefore, in essence, the similarity search method is a process of guiding future reasoning by calculating similarity.
[0057] Gaussian search uses Gaussian index to calculate the similarity of the required features. By calculating the distance between two feature vectors, each feature distance between the required features is converted into a Gaussian index. The specific conversion function is:
[0058]
[0059] Where: s is the bending point. By setting the bending point reasonably, different similarities can be explored. n is the Gaussian similarity coefficient, 0 <sg n <1; k is the dimension of meteorology; t is the number of each type of meteorology; n is the number of Gaussian indicators after transformation; is the meteorological feature to be compared; The meteorological characteristics are selected.
[0060] The grey correlation matrix is solved by grey correlation analysis, and the shape similarity between each sequence is calculated, so as to more accurately describe the importance of different meteorological characteristics. First, the absolute difference between the matrix SG and the selected sequence elements is calculated, and then the grey correlation degree is further calculated. And the grey correlation matrix SG':
[0061]
[0062]
[0063]
[0064] Where: Represent the minimum and maximum values of the jth meteorological feature in the ith data set and represents the Gaussian similarity coefficient of the jth meteorological feature in the initial 0th data set, represents the Gaussian similarity coefficient of the jth meteorological feature in the ith data set, ρ is the resolution coefficient, which is used to adjust the difference between correlation coefficients. In this embodiment, ρ=0.5.
[0065] By normalizing the weighted grey correlation matrix, we can get the weighted meteorological feature similarity coefficient. By sorting them, we can get the comprehensive correlation coefficients between different meteorological features and different datasets of this transitional weather. By performing the above processing on the typhoon datasets, we can get the sequence of similar features of typhoon transitional weather.
[0066] It should be noted here that, in other embodiments, other existing methods may also be used to screen out key meteorological factors from the set of meteorological factors that affect typhoons on wind power. Those skilled in the art may make a selection based on actual conditions, which will not be described in detail here.
[0067] Table 1 shows the correlation coefficients between meteorological factors and wind power output in this embodiment.
[0068] Table 1 Correlation coefficients between meteorological factors and wind power output
[0069] Meteorological variables Typhoon similarity coefficient Wind speed 0.743 Air pressure -0.359 temperature 0.110 wind direction 0.377 humidity -0.235
[0070] According to Table 1, wind speed, wind direction and air pressure can be screened out as key meteorological factors.
[0071] S203: Calibrate the theoretical value of the maximum wind power generation capacity based on the predefined maximum wind power generation capacity, form a prediction data set with the meteorological data corresponding to the model input characteristics and the theoretical value of the maximum wind power generation capacity, and then combine it with the preset power generation capacity prediction model to obtain the prediction result of the maximum wind power generation capacity.
[0072] In one or more embodiments, the preset power generation capacity prediction model is a Transformer neural network model.
[0073] Transformer neural network model:
[0074] The Transformer neural network model is essentially a model based on a sequence-to-sequence encoder-decoder structure, which consists of three parts: input embedding, encoder-decoder module, and model output.
[0075] Compared with the RNN model, the Transformer model has the advantage of parallel operation, but it cannot represent the sequence order of the original input, so position encoding is introduced. The encoder of the Transformer model is composed of multiple independent encoding layers stacked together, each encoding layer contains a multi-head attention layer and a fully connected layer; similarly, the decoder is also composed of multiple independent decoding layers stacked together. The position encoding formula is as follows:
[0076]
[0077]
[0078] Where P s is the position index of the input sequence where the data is located, d m is the dimension of the word embedding of the input sequence, i is the vector dimension, P E (P s ,2i) and P E (P s ,2i+1) are the sequence position index P s Positional encoding formula on even or odd number of positions.
[0079] The attention mechanism is the core of the Transformer network and the main component of the encoder and decoder. When the query vector, key vector, and value vector input in the operation are all the same value, it is called the self-attention mechanism. In the decoder, in addition to the need to perform self-attention mechanism calculations on the decoder input, it is also necessary to merge the encoder output for multi-head attention mechanism calculations. The attention mechanism obtains the correlation between a data in the sequence and other data by calculating the attention score. The attention mechanism formula is as follows:
[0080]
[0081]
[0082] Where: q is the query vector; k is the key vector; v is the value vector; W q , W k , W v The distribution is a matrix of three vectors; the output matrix is H; d k represents the dimension of the input and output keys. The attention function includes the scaled dot product of the query and the key, and is divided by Attention is the attention mechanism, and softmax is the softmax function.
[0083] In order to further improve the generalization ability of the Transformer network, multiple attention units are often stacked side by side, so that the network can focus on different features in the subspace. The attention structure is spliced through dimensions to produce a d-dimensional output, and then projected through a fully connected layer to complete the mapping of features. The final output formula is as follows:
[0084] M(Q,K,V)=Concat(H1,...,H n )W Out (9)
[0085] Where n is the number of stacked attention mechanisms, Concat is the dimension concatenation, and W Out Output is a three-dimensional vector matrix.
[0086] It can be understood here that in other embodiments, other existing neural network models can also be used to predict the maximum wind power generation capacity. Those skilled in the art can make a choice based on actual conditions, which will not be described in detail here.
[0087] S204: Based on the prediction result of the maximum power generation capacity of wind power, modify the upper limit activation function of the preset power prediction model in real time to determine the final power prediction model; wherein the power prediction model is jointly constructed by the feature extraction model and the power generation capacity prediction model.
[0088] The power prediction model of this embodiment is a CNN-Transformer neural network prediction model, which is formed by fusing the CNN neural network model and the Transformer neural network model through the fully connected layer of the implicit feature fusion network.
[0089] Among them, the CNN neural network model is used to obtain local time series feature information within the receptive field window, and learn shallow implicit features such as power time series variation law, periodicity and external meteorological characteristics in typhoon scenarios.
[0090] CNN(Temporal Convolutional Network):
[0091] CNN is parallel, unlike RNN where the prediction of subsequent time steps must wait for its predecessor to complete. As one of the most effective deep learning models, convolutional neural network (CNN) has unique advantages in data dimension reduction and feature extraction. Compared with traditional neural networks, the weight sharing technology of convolutional neural network can reduce network parameters and shorten training time. The advantages of convolutional neural network determine that it is suitable for extracting key features from a large amount of input data, which is just suitable for the characteristics of massive data in wind power cluster power generation prediction. As an effective feature extraction method, convolutional neural network has been widely used in various fields, such as pattern classification, image processing and model recognition. Generally, convolutional neural network consists of input layer, convolution layer, pooling layer and fully connected layer. Among them, convolution layer and pooling layer are the key network layers for dimension reduction and feature extraction. Convolutional neural network such as Figure 3 shown.
[0092] Given a one-dimensional input sequence x, the output of the convolution operation can be expressed as:
[0093]
[0094] Where: is the i-th element of the l-th feature graph in the convolution layer obtained by convolution calculation, x i is the i-th element of the input feature graph, such as wind speed, wind direction, cluster wind power, etc. The lth element of the convolution kernel, b j is the bias, f1 is the activation function, usually the Relu function, and * represents the convolution operation.
[0095] Each feature graph in the convolutional layer can be downsampled to form a new low-dimensional feature graph in the pooling layer, which can be expressed as:
[0096]
[0097] Where: is the jth element of the lth feature graph in the pooling layer, is the i-th element of the j-th submodule of the l-th feature graph in the convolutional layer, M is the dimension of the submodule, and b h is the bias, f2 is the activation function, generally the proportion function.
[0098] In wind power forecasting, especially in extreme weather scenarios with strong winds, the impact of wind speed and other features on the accuracy of power forecasting is further increased. At the same time, since the CNN network can only obtain local temporal feature information within the receptive field window, it is unable to connect the context information to highlight the impact of key features. In this regard, the self-attention layer in the transformer is used to automatically learn key features through the attention mechanism, while weakening the interference of other input feature variables on the model prediction results, so that the prediction model can be applied to wind power forecasting tasks in strong wind scenarios.
[0099] S205: Obtain a short-term wind power prediction result under the typhoon scenario by using the final power prediction model and the meteorological data corresponding to the model input characteristics under the typhoon scenario.
[0100] In this embodiment, the power prediction model is taken as an example to illustrate the process of obtaining the short-term wind power prediction result under the typhoon scenario, wherein the wind power prediction model framework based on CNN-Transformer is as follows: Figure 4 As shown:
[0101] 1) Firstly, Gaussian retrieval-grey correlation weighting is used as an indicator for meteorological feature selection to obtain the input data for power prediction under typhoon extreme weather;
[0102] 2) CNN (temporal convolutional network) is used to obtain local time series feature information within the receptive field window, and to learn shallow implicit features such as the power time series variation law, periodicity and external meteorological characteristics under extreme weather conditions.
[0103] 3) Utilizing the powerful nonlinear and implicit feature learning capabilities of deep learning, CNN and Transformer are fused through the fully connected layer of the implicit feature fusion network to construct a wind power prediction model based on CNN-Transformer.
[0104] The data set used below includes the daily forecast values of 6 types of meteorological factors and measured values of wind power in the typhoon weather of a certain region in the past five years, with a time resolution of 15 minutes. Among them, the data of 9 wind farms in a certain region under the influence of Typhoon "Tali" in 2023 during the typhoon transit period are used for example verification, with July 17, 2023 as the typhoon entry day, and July 18-20 as the days to be predicted. .
[0105] Based on the correlation analysis method, the wind power maximum power generation capacity calibration data set is used as the prediction input data set, and the input is wind speed, wind direction and air pressure. Transformer wind power maximum power generation capacity is used for prediction. In terms of available capacity prediction, the prediction method of this embodiment is set as scheme 1, and scheme 2 is set as LSTM neural network method and scheme 3 is set as TCN-Transformer neural network method as comparison schemes. Comparison of wind power maximum power generation capacity prediction results Figure 5 The evaluation indicators are shown in Table 2.
[0106] Table 2 Evaluation indicators for available capacity prediction
[0107] Evaluation indicators / schemes Method 1 Method 2 Method 3 nRMSE 0.17 0.31 0.33 nMAE 0.16 0.28 0.27
[0108] nRMSE: Normalized Root Mean Squared Error, normalized root mean square error.
[0109] nMAE: Normalized Mean Absolute Error, normalized mean absolute error.
[0110] Depend on Figure 5 It can be seen that in the period when no machine cutting occurs, the calibration measurement results of the maximum power generation capacity of wind power tend to be close to the rated power, and the available capacity prediction results during the typhoon period show the same trend as the actual power, which further illustrates that the use of the maximum power generation capacity of wind power can effectively constrain the power prediction boundary and improve the prediction accuracy of the model. The method used in this embodiment predicts results that are closer to the theoretical calibration results than the two comparative methods, and at the same time reduces the impact of the uncertainty of meteorological factors such as wind speed that are not considered in the calibration results.
[0111] Depend on Figure 7 It can be seen that the comparison proves that the Transformer method is better than LSTM and slightly better than TCN-Transformer. This is because the maximum power generation capacity of wind power is based on the station as the research object, and TCN does not need to focus on the local effect of extraction, but the Transformer neural network has a better prediction effect. The evaluation indicators of Solution 1 proposed in this embodiment are better than the benchmark method, which provides a basis for the next step of improving power prediction accuracy.
[0112] In terms of power prediction, the method adopted in this embodiment is set as Scheme 1, and two comparative schemes are set up to test the model effect of the scheme proposed in this embodiment. Scheme 2 is a Transformer neural network power prediction model used alone; Scheme 3 is an LSTM neural network prediction model that considers the maximum power generation capacity of wind power. Figure 6 and Figure 7The prediction results of the stations with and without large wind shear events are shown respectively. Table 3 shows the evaluation indicators of power prediction.
[0113] Table 3 Evaluation indicators of power prediction
[0114] Evaluation indicators / schemes Method 1 Method 2 Method 3 nRMSE 0.14 0.24 0.26 nMAE 0.13 0.23 0.27
[0115] Depend on Figure 6 and Figure 7 It can be seen that Figure 6 and Figure 7 It indicates the change of ReLU upper limit activation function under the conditions of maximum power generation capacity of 9.56 and 4.58 respectively. In the normal period before the arrival of typhoon, the wind power prediction effect is better.
[0116] According to Table 3, Scheme 1 is slightly more accurate than Scheme 3, which shows that the Transformer network has a good prediction effect, and the overall nRMSE and nMAE are improved by 3.257 and 3.59 respectively. The accuracy of Scheme 1 is significantly higher than that of Scheme 2, which shows that the maximum power generation capacity prediction proposed in this embodiment has a strong guiding role in wind power prediction and can effectively improve the accuracy of wind power prediction. The overall nRMSE and nMAE are improved by 4.324 and 5.367 respectively. Especially in the case of machine cutting, the method used in this embodiment improves the accuracy more than other comparison methods, proving that considering the maximum power generation capacity of wind power can effectively improve the accuracy of power prediction.
[0117] Figure 8 is a schematic diagram of the structure of a short-term wind power prediction system in a typhoon scenario in an embodiment of the present invention. Figure 2 The short-term wind power prediction method under typhoon scenario corresponds to Figure 8 As shown, the short-term wind power prediction system under the typhoon scenario in this embodiment may include: a meteorological factor set determination module 301, a model input feature determination module 302, a maximum power generation capacity prediction module 303, a power prediction model determination module 304 and a short-term wind power prediction module 305.
[0118] In the specific implementation process, the meteorological factor set determination module 301 is used to determine the meteorological factor set that affects the wind power under typhoon conditions based on the relationship between the wind power prediction error and the typhoon weather process.
[0119] In some specific implementation processes, the relationship between wind power prediction error under typhoon conditions and typhoon weather process is studied from the single machine-station multi-level.
[0120] In the specific implementation process, the model input feature determination module 302 is used to screen out key meteorological factors from the above meteorological factor set according to the correlation coefficient between different meteorological features in the meteorological factor set and the typhoon data set, so as to obtain the model input features under the typhoon scenario.
[0121] In some specific implementation processes, Gaussian retrieval analysis methods are used to screen out key meteorological factors from a set of meteorological factors that affect typhoons on wind power.
[0122] Among them, the process of selecting key meteorological factors from the set of meteorological factors affecting typhoon on wind power by using Gaussian retrieval analysis method is as follows:
[0123] Based on the meteorological data corresponding to the set of meteorological factors affecting wind power, a Gaussian similarity coefficient matrix is constructed;
[0124] The absolute difference between the Gaussian similarity coefficient matrix and the selected meteorological sequence elements including wind speed, wind direction, temperature and humidity is calculated, and the grey correlation degree is further calculated to construct the grey correlation degree matrix;
[0125] The weighted grey correlation matrix is summed and normalized to obtain the weighted meteorological characteristic similarity coefficients and sort them to determine the order sequence of similar characteristics of typhoons and finally determine the key meteorological factors.
[0126] In the specific implementation process, the maximum power generation capacity prediction module 303 is used to calibrate the theoretical value of the maximum wind power generation capacity based on the predefined maximum wind power generation capacity, form a prediction data set with the meteorological data corresponding to the model input characteristics and the theoretical value of the maximum wind power generation capacity, and then combine it with the preset power generation capacity prediction model to obtain the wind power maximum power generation capacity prediction result.
[0127] In one or more embodiments, the preset power generation capacity prediction model is a Transformer neural network model.
[0128] It can be understood here that in other embodiments, other existing neural network models can also be used to predict the maximum wind power generation capacity. Those skilled in the art can make a choice based on actual conditions, which will not be described in detail here.
[0129] In the specific implementation process, the power prediction model determination module 304 is used to modify the upper limit activation function of the preset power prediction model in real time based on the wind power maximum power generation capacity prediction result, and determine the final power prediction model; wherein the power prediction model is jointly constructed by the feature extraction model and the power generation capacity prediction model.
[0130] The power prediction model of this embodiment is a CNN-Transformer neural network prediction model, which is formed by fusing the CNN neural network model and the Transformer neural network model through the fully connected layer of the implicit feature fusion network.
[0131] Among them, the CNN neural network model is used to obtain local time series feature information within the receptive field window, and learn shallow implicit features such as power time series variation law, periodicity and external meteorological characteristics in typhoon scenarios.
[0132] In the specific implementation process, the short-term wind power prediction module 305 is used to obtain the short-term wind power prediction result under the typhoon scenario by using the final power prediction model and the meteorological data corresponding to the model input characteristics under the typhoon scenario.
[0133] In one aspect of the present invention, a computer program product or a computer program is also provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the above-mentioned Figure 2 Different embodiments and combinations of embodiments are provided in various optional implementations of the short-term wind power prediction method in a typhoon scenario.
[0134] In some other embodiments, a computer-readable storage medium is further provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned Figure 2 The steps in the short-term wind power prediction method under the typhoon scenario are shown.
[0135] The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products of the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of the processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0136] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A short-term wind power prediction method under typhoon scenarios, It is characterized in that include: Based on the relationship between wind power prediction error under typhoon conditions and typhoon weather process, a set of meteorological factors that affect wind power under typhoon conditions is determined. According to the correlation coefficients between different meteorological characteristics in the meteorological factor set and the typhoon data set, key meteorological factors are selected from the above meteorological factor set to obtain the model input characteristics under the typhoon scenario; The key meteorological factors are selected from the above meteorological factor set: the key meteorological factors are selected from the meteorological factor set of the typhoon on wind power by using Gaussian search analysis method; the Gaussian search uses Gaussian index to calculate the similarity of the required features, calculates the distance between two feature vectors, and then converts each feature distance between the required features into a Gaussian index; The theoretical value of the maximum wind power generation capacity is calibrated based on the predefined maximum wind power generation capacity, and the meteorological data corresponding to the model input characteristics and the theoretical value of the maximum wind power generation capacity are used to form a prediction data set, and then combined with a preset power generation capacity prediction model to obtain a prediction result of the maximum wind power generation capacity; the preset power generation capacity prediction model is a Transformer neural network model; Based on the prediction result of the maximum power generation capacity of wind power, the upper limit activation function of the preset power prediction model is modified in real time to determine the final power prediction model; wherein, the power prediction model is jointly constructed by the feature extraction model and the power generation capacity prediction model; the power prediction model is a CNN-Transformer neural network prediction model, which is formed by fusing the CNN neural network model and the Transformer neural network model through the fully connected layer of the implicit feature fusion network; the Transformer neural network model adopts the self-attention layer of the Transformer, automatically learns key features through the attention mechanism, and at the same time weakens the interference of input feature variables other than the key features on the model prediction results; The short-term wind power forecast results under the typhoon scenario are obtained by using the final power forecast model and the meteorological data corresponding to the model input characteristics under the typhoon scenario.
2. The short-term wind power prediction method under typhoon scenario as claimed in claim 1, It is characterized in that The relationship between wind power prediction error under typhoon conditions and typhoon weather process is studied from the perspective of single machine-station multi-level.
3. The short-term wind power prediction method under typhoon scenario as claimed in claim 1, It is characterized in that The process of selecting key meteorological factors from the set of meteorological factors that affect wind power by typhoons using the Gaussian retrieval analysis method is as follows: Based on the meteorological data corresponding to the set of meteorological factors affecting wind power, a Gaussian similarity coefficient matrix is constructed; Calculate the absolute difference between the Gaussian similarity coefficient matrix and the selected meteorological sequence elements, further deduce the grey correlation degree, and then construct the grey correlation matrix; The weighted grey correlation matrix is summed and normalized to obtain the weighted meteorological characteristic similarity coefficients and sort them to determine the order sequence of similar characteristics of typhoons and finally determine the key meteorological factors.
4. The short-term wind power prediction method under typhoon scenario as claimed in claim 1, It is characterized in that The CNN neural network model is used to obtain local time series feature information within the receptive field window and learn shallow implicit features such as power time series variation law, periodicity and external meteorological characteristics in typhoon scenarios.
5. A short-term wind power prediction system under typhoon scenarios, It is characterized in that include: A meteorological factor set determination module, which is used to determine the meteorological factor set that affects the wind power under typhoon conditions based on the relationship between the wind power prediction error and the typhoon weather process; A model input feature determination module is used to select key meteorological factors from the meteorological factor set according to the correlation coefficient between different meteorological features in the meteorological factor set and the typhoon data set, so as to obtain the model input features under the typhoon scenario; The key meteorological factors are selected from the above meteorological factor set: the key meteorological factors are selected from the meteorological factor set of the typhoon on wind power by using Gaussian search analysis method; the Gaussian search uses Gaussian index to calculate the similarity of the required features, calculates the distance between two feature vectors, and then converts each feature distance between the required features into a Gaussian index; The maximum power generation capacity prediction module is used to calibrate the theoretical value of the maximum wind power generation capacity based on the predefined maximum wind power generation capacity, form a prediction data set with the meteorological data corresponding to the model input characteristics and the theoretical value of the maximum wind power generation capacity, and then combine it with the preset power generation capacity prediction model to obtain the prediction result of the maximum wind power generation capacity; the preset power generation capacity prediction model is a Transformer neural network model; A power prediction model determination module is used to modify the upper limit activation function of the preset power prediction model in real time based on the wind power maximum power generation capacity prediction result, and determine the final power prediction model; wherein the power prediction model is jointly constructed by the feature extraction model and the power generation capacity prediction model; the power prediction model is a CNN-Transformer neural network prediction model, which is formed by fusing the CNN neural network model and the Transformer neural network model through the fully connected layer of the implicit feature fusion network; the Transformer neural network model adopts the self-attention layer of the Transformer, automatically learns key features through the attention mechanism, and at the same time weakens the interference of input feature variables other than the key features on the model prediction results; The short-term wind power prediction module is used to obtain the short-term wind power prediction result under the typhoon scenario by using the final power prediction model and the meteorological data corresponding to the model input characteristics under the typhoon scenario.
6. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the steps in the short-term wind power prediction method in a typhoon scenario as described in any one of claims 1 to 4 are implemented.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the steps of the short-term wind power prediction method in a typhoon scenario as described in any one of claims 1 to 4 are implemented.
