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

Abstract

The application relates to a wind power prediction method and device, computer equipment and a readable storage medium. The method comprises the following steps: acquiring numerical weather prediction and measured power data, wherein the numerical weather prediction comprises multiple dimensions of meteorological characteristics; performing dimension reduction processing on the numerical weather prediction according to the measured power data to obtain dimension-reduced numerical weather prediction; and performing wind power prediction on the dimension-reduced numerical weather prediction through a wind power prediction model to obtain predicted wind power. The application can reduce the dimension of the numerical weather prediction, remove redundant information in the numerical weather prediction, and improve the wind power prediction accuracy.

Description

Technical Field

[0001] This application relates to the field of power system operation, and in particular to a wind power prediction method, device, electronic equipment and storage medium. Background Technology

[0002] With the continuous development of wind power, the proportion of wind power generation in the power system is gradually increasing. Due to the characteristics of large fluctuations and high uncertainty in wind power, it is particularly important to build accurate mathematical models to predict wind power to maintain the real-time energy supply and demand balance of the power system.

[0003] Currently, the industry typically uses numerical weather prediction combined with forecasting models to predict wind power. However, since numerical weather prediction itself is a form of forecasting information and has issues such as high data dimensionality and a lot of redundant information, the errors in numerical weather prediction are further amplified in the wind power prediction model, resulting in poor accuracy in wind power prediction. Summary of the Invention

[0004] Therefore, it is necessary to provide a wind power prediction method, device, electronic equipment, and storage medium to address the problem of low accuracy in existing wind power prediction models.

[0005] In a first aspect, this application provides a wind power prediction method, the method comprising:

[0006] Acquire numerical weather forecasts and measured power data, wherein the numerical weather forecasts include meteorological characteristics in multiple dimensions;

[0007] The numerical weather forecast is reduced in dimensionality based on the measured power data to obtain a dimensionality-reduced numerical weather forecast.

[0008] The wind power prediction model is used to predict the wind power from the dimensionality-reduced numerical weather forecast.

[0009] In one embodiment, the step of performing dimensionality reduction processing on the numerical weather forecast based on the measured power data to obtain a dimensionality-reduced numerical weather forecast includes:

[0010] Based on the measured power data and the meteorological characteristics, Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence are constructed.

[0011] Based on the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence, meteorological features in an abnormal state in the numerical weather forecast are determined, and the meteorological features in an abnormal state are deleted from the numerical weather forecast to obtain the first dimensionality-reduced numerical weather forecast.

[0012] In one embodiment, the step of performing dimensionality reduction processing on the numerical weather forecast based on the measured power data to obtain a dimensionality-reduced numerical weather forecast further includes:

[0013] Based on the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence of each meteorological feature in the first dimensionality reduction numerical weather forecast, a fusion correlation index sequence is constructed. The fusion correlation index sequence includes the fusion correlation index of each meteorological feature. The fusion correlation index is used to characterize the correlation between the meteorological feature and the measured power data.

[0014] Based on the fusion correlation index of each meteorological feature, the meteorological features are sorted from high to low to obtain a meteorological feature sequence;

[0015] Target meteorological features are determined sequentially from the meteorological feature sequence until the cumulative result of the fusion correlation index of the target meteorological features is greater than a threshold.

[0016] Based on the target meteorological characteristics, a second-dimensional reduced numerical weather forecast is obtained.

[0017] In one embodiment, the step of performing dimensionality reduction processing on the numerical weather forecast based on the measured power data to obtain a dimensionality-reduced numerical weather forecast further includes:

[0018] Based on the meteorological feature types, the measured data of wind farms and the meteorological features in the second dimensionality-reduced numerical weather forecast are classified to obtain the set of measured data of wind farms and the set of meteorological feature data corresponding to each meteorological feature type.

[0019] For any of the meteorological feature types, standard measured data are determined from the wind farm measured data set corresponding to the meteorological feature type, and the error between each meteorological feature in the meteorological feature data set corresponding to the meteorological feature type and the standard measured data is determined.

[0020] Based on the error corresponding to each meteorological feature in each meteorological feature data set, determine the meteorological feature fusion weight of each meteorological feature in each meteorological feature data set;

[0021] Based on the fusion weights of the meteorological features in each of the meteorological feature datasets, the meteorological features in each of the meteorological feature datasets are fused to obtain a third-dimensional reduced numerical weather forecast.

[0022] In one embodiment, the wind power prediction model includes at least two prediction models. The step of using the wind power prediction model to predict wind power from the dimensionality-reduced numerical weather forecast to obtain the predicted wind power includes:

[0023] The wind power forecast of the dimensionality-reduced numerical weather forecast is predicted by each of the prediction models to obtain multiple initial predicted wind power.

[0024] The multiple initial predicted wind power are fused according to the prediction model fusion weights of each prediction model to obtain the predicted wind power.

[0025] In one embodiment, the method further includes:

[0026] Acquire numerical weather forecasts and measured power data for a preset time period;

[0027] Based on the measured power data within the preset time period, the numerical weather forecast within the preset time period is subjected to dimensionality reduction processing to obtain the dimensionality-reduced numerical weather forecast within the preset time period.

[0028] A training set is constructed based on the dimensionality-reduced numerical weather forecasts within the preset time period;

[0029] Each prediction model is trained using the training set to obtain the loss value of each prediction model.

[0030] Based on the loss values ​​of each of the prediction models, the prediction model fusion weights of each of the prediction models are obtained.

[0031] Secondly, this application also provides a wind power prediction device, the device comprising:

[0032] The first acquisition module is used to acquire numerical weather forecasts and measured power data, wherein the numerical weather forecasts include meteorological characteristics in multiple dimensions.

[0033] The first dimensionality reduction module is used to perform dimensionality reduction processing on the numerical weather forecast based on the measured power data to obtain the dimensionality-reduced numerical weather forecast.

[0034] The prediction module is used to predict wind power from the dimensionality-reduced numerical weather forecast using a wind power prediction model, thereby obtaining the predicted wind power.

[0035] In one embodiment, the first dimensionality reduction module is further configured to:

[0036] Based on the measured power data and the meteorological characteristics, Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence are constructed.

[0037] Based on the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence, meteorological features in an abnormal state in the numerical weather forecast are determined, and the meteorological features in an abnormal state are deleted from the numerical weather forecast to obtain the first dimensionality-reduced numerical weather forecast.

[0038] In one embodiment, the first dimensionality reduction module is further configured to:

[0039] Based on the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence of each meteorological feature in the first dimensionality reduction numerical weather forecast, a fusion correlation index sequence is constructed. The fusion correlation index sequence includes the fusion correlation index of each meteorological feature. The fusion correlation index is used to characterize the correlation between the meteorological feature and the measured power data.

[0040] Based on the fusion correlation index of each meteorological feature, the meteorological features are sorted from high to low to obtain a meteorological feature sequence;

[0041] Target meteorological features are determined sequentially from the meteorological feature sequence until the cumulative result of the fusion correlation index of the target meteorological features is greater than a threshold.

[0042] Based on the target meteorological characteristics, a second-dimensional reduced numerical weather forecast is obtained.

[0043] In one embodiment, the first dimensionality reduction module is further configured to:

[0044] Based on the meteorological feature types, the measured data of wind farms and the meteorological features in the second dimensionality-reduced numerical weather forecast are classified to obtain the set of measured data of wind farms and the set of meteorological feature data corresponding to each meteorological feature type.

[0045] For any of the meteorological feature types, standard measured data are determined from the wind farm measured data set corresponding to the meteorological feature type, and the error between each meteorological feature in the meteorological feature data set corresponding to the meteorological feature type and the standard measured data is determined.

[0046] Based on the error corresponding to each meteorological feature in each meteorological feature data set, determine the meteorological feature fusion weight of each meteorological feature in each meteorological feature data set;

[0047] Based on the fusion weights of the meteorological features in each of the meteorological feature datasets, the meteorological features in each of the meteorological feature datasets are fused to obtain a third-dimensional reduced numerical weather forecast.

[0048] In one embodiment, the prediction module is further configured to:

[0049] The wind power forecast of the dimensionality-reduced numerical weather forecast is predicted by each of the prediction models to obtain multiple initial predicted wind power.

[0050] The multiple initial predicted wind power are fused according to the prediction model fusion weights of each prediction model to obtain the predicted wind power.

[0051] In one embodiment, the device further includes:

[0052] The second acquisition module is used to acquire numerical weather forecasts and measured power data within a preset time period;

[0053] The second dimensionality reduction module is used to perform dimensionality reduction processing on the numerical weather forecast within the preset time period based on the measured power data within the preset time period, so as to obtain the dimensionality-reduced numerical weather forecast within the preset time period.

[0054] The construction module is used to construct a training set based on the dimensionality-reduced numerical weather forecasts within the preset time period;

[0055] The training module is used to train each of the prediction models using the training set to obtain the loss value of each prediction model.

[0056] The fusion weight module is used to obtain the prediction model fusion weight of each prediction model based on the loss value of each prediction model.

[0057] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any of the above-described wind power prediction methods.

[0058] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements any of the above wind power prediction methods.

[0059] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements any of the above-described wind power prediction methods.

[0060] The aforementioned wind power prediction method, apparatus, electronic device, and storage medium can, after acquiring numerical weather forecasts and measured power data, perform dimensionality reduction processing on the numerical weather forecasts based on the measured power data. Then, a wind power prediction model can be used to predict wind power from the dimensionality-reduced numerical weather forecasts. Because the embodiments of this application perform dimensionality reduction processing on the numerical weather forecasts, the dimensionality of the numerical weather forecasts can be reduced, the data dimension processed by the wind power prediction model can be reduced, and redundant information in the numerical weather forecasts can be reduced, thereby effectively improving the prediction accuracy and efficiency of wind power. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a wind power prediction method in one embodiment;

[0062] Figure 2 This is a flowchart illustrating step 104 in one embodiment;

[0063] Figure 3 This is a flowchart illustrating step 104 in one embodiment;

[0064] Figure 4 This is a flowchart illustrating step 104 in one embodiment;

[0065] Figure 5 This is a flowchart illustrating step 106 in one embodiment;

[0066] Figure 6 This is a flowchart illustrating a wind power prediction method in one embodiment;

[0067] Figure 7 This is a flowchart illustrating a wind power prediction method in one embodiment;

[0068] Figure 8 This is a schematic diagram of a wind power prediction method in one embodiment;

[0069] Figure 9 This is a schematic diagram of a wind power prediction method in one embodiment;

[0070] Figure 10 This is a structural block diagram of a wind power prediction device in one embodiment;

[0071] Figure 11 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0072] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0073] In one embodiment, such as Figure 1 As shown, a wind power prediction method is provided. This embodiment illustrates the method's application to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0074] Step 102: Obtain numerical weather forecast and measured power data. Numerical weather forecast includes meteorological characteristics in multiple dimensions.

[0075] Numerical weather forecasting is a predictive information system that includes various predicted meteorological features for the forecast date. The forecast date can be the day the weather forecast is to be conducted; for example, if the numerical weather forecast predicts the weather for March 20th on March 19th, then the forecast date is March 20th. For instance, numerical weather forecasts can include various meteorological features such as temperature, wind direction, and cloud cover. For example, when the time resolution of the numerical weather forecast is 15 minutes and it includes 22 meteorological features, the numerical weather forecast is a 96×22 array, and the meteorological features are 96×1 arrays. The measured power data is the wind power data measured within the forecast date corresponding to the numerical weather forecast.

[0076] Step 104: Perform dimensionality reduction processing on the numerical weather forecast based on the measured power data to obtain the dimensionality-reduced numerical weather forecast.

[0077] In this embodiment, the dimensionality reduction of numerical weather forecasts can be achieved by analyzing the correlation between measured power and numerical weather prediction. For example, the correlation index between each meteorological feature and measured power data in the numerical weather prediction can be determined. Dimensionality reduction can be achieved by removing meteorological features with abnormal correlation indices, filtering meteorological features with correlation indices that meet preset conditions, and fusing meteorological features belonging to the same type of meteorological feature.

[0078] For example, after removing meteorological features with abnormal correlation indices and screening meteorological features with correlation indices that meet preset conditions, meteorological features belonging to the same type of meteorological feature can be fused to achieve dimensionality reduction of numerical weather forecasts, thereby obtaining dimensionality-reduced numerical weather forecasts.

[0079] This application does not specifically limit the method of dimensionality reduction of numerical weather forecasts based on the correlation index of meteorological characteristics and measured power data. Any method that can reduce the dimensionality of numerical weather forecasts based on the correlation index of meteorological characteristics and measured power data is applicable to this application.

[0080] Step 106: Use the wind power prediction model to predict the wind power of the dimensionality-reduced numerical weather forecast, and obtain the predicted wind power.

[0081] In this embodiment of the application, the wind power prediction model is a pre-trained model for predicting wind power. This embodiment of the application does not specifically limit the model structure and training process of the wind power prediction model.

[0082] After obtaining the dimensionality-reduced numerical weather forecast, it can be used as input information for the wind power prediction model to predict wind power and thus obtain the predicted wind power.

[0083] The wind power prediction method provided in this application can, after acquiring numerical weather forecasts and measured power data, perform dimensionality reduction processing on the numerical weather forecasts based on the measured power data. Then, a wind power prediction model can be used to predict wind power from the dimensionality-reduced numerical weather forecasts. Because this application embodiment performs dimensionality reduction processing on the numerical weather forecasts, it can reduce the dimensionality of the numerical weather forecasts, reduce the data dimensionality processed by the wind power prediction model, and reduce redundant information in the numerical weather forecasts, thereby effectively improving the prediction accuracy and efficiency of wind power.

[0084] In one embodiment, such as Figure 2 As shown, in step 104, the numerical weather forecast is dimensionality-reduced based on the measured power data to obtain the dimensionality-reduced numerical weather forecast, including:

[0085] Step 202: Based on the measured power data and meteorological characteristics, construct the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence.

[0086] For example, in this application embodiment, a correlation index sequence can be constructed based on measured power data and meteorological characteristics. The correlation index sequence may include Pearson correlation index sequence, mutual information entropy sequence, Spearman correlation index sequence, etc.

[0087] Taking variable X as a meteorological characteristic and variable Y as measured power data as an example, the Pearson correlation index can be determined using formula (I), the mutual information entropy using formula (II), and the Spearman correlation index using formula (III). It should be noted that the obtained Pearson correlation index, mutual information entropy, and Spearman correlation index are all real numbers in the interval -1 to 1.

[0088]

[0089] Where, ρ XY It is the Pearson correlation index between meteorological characteristics and measured power data, cov(X,Y) is the covariance between meteorological characteristics and measured power data, and σ is the Pearson correlation index between meteorological characteristics and measured power data. X It is the variance of meteorological characteristics, σ Y It is the variance of the measured power data.

[0090]

[0091] Where I(X;Y) is the mutual information entropy between meteorological features and measured power data, p(X,Y) is the joint distribution of meteorological features and measured power data, p(X) is the distribution of meteorological features, and p(Y) is the distribution of measured power data.

[0092]

[0093] Where, r s It is the Spearman correlation index between meteorological characteristics and measured power data, cov(R(X),R(Y)) is the covariance between the ordinal variable meteorological characteristics and the ordinal variable measured power data, and σ is the Spearman correlation index between the meteorological characteristics and measured power data. R(X) It is the variance of the ordinal variable meteorological characteristics, σ R(Y) It is the variance of the measured power data of the ordinal variable.

[0094] In this embodiment, during the construction of the correlation index sequence, each meteorological feature and measured power data in the numerical weather prediction can be normalized. Then, based on the normalized meteorological features and measured power data, the Pearson correlation index for each meteorological feature and measured power data is calculated. Finally, based on the Pearson correlation indices of each meteorological feature and measured power data, a Pearson correlation index sequence is constructed. Based on the normalized meteorological features and the normalized measured power data, the mutual information entropy of each meteorological feature and the measured power data is calculated. Then, based on the mutual information entropy of each meteorological feature and the measured power data, a mutual information entropy sequence is constructed. Based on the normalized meteorological characteristics and normalized measured power data, the Spearman correlation index for each meteorological characteristic and measured power data is calculated. Then, based on the Spearman correlation indices for each meteorological characteristic and measured power data, a Spearman correlation index sequence is constructed. For example, taking the above example again, when the numerical weather forecast used has 22 dimensions, the resulting Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence are all 22×1 sequences.

[0095] Step 204: Based on the Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence, determine the meteorological features in the numerical weather forecast that are in an abnormal state, delete the meteorological features in the abnormal state from the numerical weather forecast, and obtain the first dimensionality-reduced numerical weather forecast.

[0096] In this embodiment of the application, anomaly judgments can be made on various meteorological features in numerical weather prediction using Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence, so as to identify meteorological features in an abnormal state from numerical weather prediction.

[0097] For example, when the three evaluation systems corresponding to a meteorological feature (Pearson correlation index, mutual information entropy, and Spearman correlation index) contradict each other in determining whether the meteorological feature and the measured power data are positively or negatively correlated—that is, when not all of the Pearson correlation index, mutual information entropy, and Spearman correlation index corresponding to the meteorological feature are positive or not all of them are negative—then the meteorological feature can be determined to be in an abnormal state, i.e., it is considered an anomalous meteorological feature. For example, when the Pearson correlation index and mutual information entropy are positive, and the Spearman correlation index is negative, the meteorological feature can be determined to be in an abnormal state.

[0098] After completing the anomaly assessment of all meteorological features, this embodiment of the application can remove the abnormal meteorological features from the numerical weather prediction to obtain a first-dimensionality reduced numerical weather prediction. Correspondingly, the Pearson correlation index of the abnormal meteorological features can also be removed from the Pearson correlation index sequence, the mutual information entropy of the abnormal meteorological features can be removed from the mutual information entropy sequence, and the Spearman correlation index of the abnormal meteorological features can be removed from the Spearman correlation index sequence.

[0099] For example, if the numerical weather forecast used has 22 dimensions and two abnormal meteorological features are identified, the first-dimensional numerical weather forecast after deleting the abnormal meteorological features has 20 dimensions. The Pearson correlation index, mutual information entropy, and Spearman correlation index sequences after deleting the abnormal meteorological features are all 20×1 sequences.

[0100] The wind power prediction method provided in this application can construct Pearson correlation index sequences, mutual information entropy sequences, and Spearman correlation index sequences based on measured power data and meteorological characteristics. Then, based on these sequences, meteorological characteristics in an abnormal state in the numerical weather prediction can be identified and removed, resulting in a first dimensionality-reduced numerical weather prediction. Because this application performs dimensionality reduction on the numerical weather prediction, it reduces the dimensionality of the numerical weather prediction, decreases the data dimension processed by the wind power prediction model, and reduces redundant information in the numerical weather prediction, thereby effectively improving the prediction accuracy and efficiency of wind power.

[0101] In one embodiment, such as Figure 3 As shown, in step 204, the numerical weather forecast is dimensionality-reduced based on the measured power data to obtain the dimensionality-reduced numerical weather forecast, including:

[0102] Step 302: Based on the Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence of each meteorological feature in the first dimensionality reduction numerical weather forecast, construct a fusion correlation index sequence. The fusion correlation index sequence includes the fusion correlation index of each meteorological feature. The fusion correlation index is used to characterize the correlation between meteorological features and measured power data.

[0103] The fusion correlation index of meteorological characteristics can be obtained by fusing the Pearson correlation index, mutual information entropy, and Spearman correlation index of the meteorological characteristic using the Dempster-Shafer theory (see formula (IV)).

[0104]

[0105] DS evidence theory is used to fuse the probabilities of two trust functions m1 and m2 assigned to set A, where m 1,2 (A) represents the fused probabilities. Set A is any subset of the recognition frame Θ. The sum of probabilities assigned by the trust function m1 to all subsets A of the recognition frame Θ is 1, and the sum of probabilities assigned by the trust function m2 to all subsets A of the recognition frame Θ is also 1.

[0106] For example, before adopting the DS evidence theory, the Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence can be preprocessed, including: using formula (v) to calculate the proportion of each Pearson correlation index in the Pearson correlation index sequence, the proportion of each mutual information entropy in the mutual information entropy sequence, and the proportion of each Spearman correlation index in the Spearman correlation index sequence.

[0107]

[0108] in, This refers to the weighting of the correlation index. `j` indicates the formula used to calculate the correlation index, where `j=1` represents the Pearson correlation index, `j=2` represents the mutual information entropy, and `j=3` represents the Spearman correlation index. For example, when `j=1`, The Pearson correlation index refers to the correlation between the i-th meteorological feature and the measured power data. This refers to the proportion of the Pearson correlation index in the sum of the Pearson correlation index series; when j=2, The mutual information entropy between the i-th meteorological feature and the measured power data. This refers to the proportion of the mutual information entropy in the sum of the mutual information entropy sequences; when j=3, The Spearman correlation index refers to the correlation between the i-th meteorological feature and the measured power data. This refers to the proportion of the Spearman correlation index in the sum of the Spearman correlation index series.

[0109] After calculating the Pearson correlation index weight for each Pearson correlation index, the mutual information entropy weight for each mutual information entropy, and the Spearman correlation index weight for each Spearman correlation index, the Pearson correlation index weight sequence can be obtained. Mutual information entropy weight sequence Spearman correlation index weight series

[0110] In this embodiment of the application, after obtaining the Pearson correlation index weight sequence, the mutual information entropy weight sequence, and the Spearman correlation index weight sequence, for the i-th meteorological feature, this embodiment of the application can first use the DS evidence theory to fuse the Pearson correlation index weight and the mutual information entropy weight of the meteorological feature to obtain a preliminary fusion index (see formula (vi)). Then, the preliminary fusion index and the Spearman correlation index weight of the meteorological feature are fused (see formula (vii)) to obtain the fused correlation index:

[0111]

[0112]

[0113] in, It is a preliminary integration index. It is a fusion correlation index, where n is the number of meteorological features in the first-dimensional numerical weather forecast.

[0114] Step 304: Based on the fusion correlation index of each meteorological feature, sort the meteorological features from high to low to obtain the meteorological feature sequence.

[0115] According to the embodiments of this application, the meteorological features in the first dimensionality reduced numerical weather forecast can be sorted from high to low according to the magnitude of the fusion correlation index of the meteorological features. The meteorological feature with the highest fusion correlation index is arranged in the first position, and the meteorological feature with the lowest fusion correlation index is arranged in the last position, so as to obtain the meteorological feature sequence.

[0116] Step 306: Sequentially determine the target meteorological features from the meteorological feature sequence until the cumulative result of the fusion correlation index of the target meteorological features is greater than the threshold.

[0117] The threshold is a predefined percentage value. In this embodiment, the specific value of the threshold is not limited. In this example, the threshold value is 90% to illustrate the embodiment of this application.

[0118] In this embodiment, the proportion of each fusion correlation index in all fusion correlation indices is calculated, and the proportion of each meteorological feature is used as the contribution rate. The contribution rate is calculated according to formula (VIII):

[0119]

[0120] in, It is the contribution rate of the i-th meteorological feature. It is the fusion correlation index of the i-th meteorological feature, and n is the number of all meteorological features in the first-dimensional numerical weather forecast.

[0121] This embodiment of the application selects the first meteorological feature as the target meteorological feature, starting from the first meteorological feature sequence, and takes the contribution rate corresponding to the fusion correlation index of the first meteorological feature. If the contribution rate is not greater than a threshold, the second meteorological feature is selected as the target meteorological feature, and the contribution rate of the second meteorological feature is added to the contribution rate of the first meteorological feature. If the cumulative result is still not greater than the threshold, the third meteorological feature is selected as the target meteorological feature, and the contribution rate of the third meteorological feature is added to the aforementioned cumulative result... until the sum of the contribution rates of all meteorological features selected as target meteorological features is greater than the threshold. For example, when the threshold is set to 90%, if the contribution rate of the first meteorological feature is 91%, this embodiment of the application selects only the first meteorological feature as the target meteorological feature; if the contribution rate of the last meteorological feature is 11%, that is, the sum of the contribution rates of all target meteorological features without selecting the last meteorological feature is only 89%, which is less than the threshold, this embodiment of the application needs to select all meteorological features as target meteorological features.

[0122] It should be noted that, in this embodiment, the meteorological features in the first dimensionality reduction numerical weather forecast can also be sorted from low to high according to the magnitude of the fusion correlation index of the meteorological features. When selecting target meteorological features, the target meteorological features are selected sequentially starting from the last position of the meteorological feature sequence until the sum of the contribution rates of all target meteorological features is greater than the threshold. In this embodiment, no specific limitation is made on the sorting method of meteorological features or the selection method of target meteorological features.

[0123] Step 308: Based on the target meteorological characteristics, obtain the second-dimensional reduced numerical weather forecast.

[0124] In this embodiment of the application, after obtaining the target meteorological features, a second dimensionality-reduced numerical weather forecast can be constructed based on the target meteorological features. For example, if the first dimensionality-reduced numerical weather forecast has 20 dimensions, and 17 target meteorological features are selected from the first dimensionality-reduced numerical weather forecast, the second dimensionality-reduced numerical weather forecast will have 17 dimensions.

[0125] It should be noted that the embodiments of this application can further reduce the dimensionality of the first dimensionality-reduced numerical weather forecast to obtain the second dimensionality-reduced numerical weather forecast. In fact, the numerical weather forecast can also be directly reduced in dimensionality to obtain the second dimensionality-reduced numerical weather forecast. The specific process can be referred to the relevant description of the foregoing embodiments, and will not be repeated here.

[0126] The wind power prediction method provided in this application can obtain the fusion correlation index of each meteorological feature based on the Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence of each meteorological feature in the first dimensionality-reduced numerical weather forecast. Then, meteorological features whose fusion correlation index in the first dimensionality-reduced numerical weather forecast meets preset conditions are selected as target meteorological features. A second dimensionality-reduced numerical weather forecast is obtained based on the target meteorological features. Since this application embodiment performs dimensionality reduction processing on the first dimensionality-reduced numerical weather forecast again, that is, performs dimensionality reduction processing twice on the numerical weather forecast, it can further reduce the dimensionality of the numerical weather forecast, reduce the data dimension processed by the wind power prediction model, and further reduce redundant information in the numerical weather forecast, thereby effectively improving the prediction accuracy and efficiency of wind power.

[0127] In one embodiment, such as Figure 4 As shown, step 204, which involves dimensionality reduction of the numerical weather forecast based on measured power data to obtain the dimensionality-reduced numerical weather forecast, also includes:

[0128] Step 402: Classify the measured data of wind farms and the meteorological features in the second-dimensional reduced numerical weather forecast according to the meteorological feature type, and obtain the set of measured data of wind farms and the set of meteorological feature data corresponding to each meteorological feature type.

[0129] Among them, the measured data index of the wind farm is the measured weather data for the forecast date. For example, if the numerical weather forecast predicts the weather for March 20th on March 19th, then the forecast date is March 20th.

[0130] This application embodiment categorizes meteorological features into multiple meteorological feature types based on their types. For example, if meteorological features include 10-meter wind speed, 30-meter wind speed, 70-meter wind speed, 10-meter wind direction, 70-meter wind direction, 100-meter wind direction, sea level pressure, and cloud cover, then the meteorological feature type can include wind speed, wind direction, sea level pressure, and cloud cover. This application embodiment can classify wind farm measured data according to meteorological feature types to obtain a wind farm measured data set corresponding to each meteorological feature type. It also classifies meteorological features in the second-dimensionality reduced numerical weather forecast according to meteorological feature types to obtain a meteorological feature data set corresponding to each meteorological feature type. For any given meteorological feature type, its corresponding wind farm measured data set includes at least one wind farm measured data corresponding to that meteorological feature type, and its corresponding meteorological feature data set includes at least one meteorological feature in the second-dimensionality reduced numerical weather forecast corresponding to that meteorological feature type.

[0131] Step 404: For any meteorological feature type, determine the standard measured data from the wind farm measured data set corresponding to the meteorological feature type, and determine the error between each meteorological feature in the meteorological feature data set corresponding to the meteorological feature type and the standard measured data.

[0132] For example, for any meteorological feature type, the measured data and measured power data of each wind farm in the wind farm measured data set corresponding to that meteorological feature type can be normalized. Based on the normalized wind farm measured data and normalized measured power data, the Pearson correlation index between the measured data and measured power data of each wind farm in the wind farm measured data set corresponding to that meteorological feature type can be calculated. The measured data of the wind farm with the largest Pearson correlation index is determined as the standard measured data corresponding to that meteorological feature type.

[0133] Furthermore, the difference between each meteorological feature in the meteorological feature data set corresponding to the meteorological feature type and the standard measured data corresponding to the meteorological feature type can be calculated to obtain the error sequence between each meteorological feature and the standard measured data. For any meteorological feature, the error sequence can be summed to obtain the error between the meteorological feature and the standard measured data. The error sequences corresponding to other meteorological features can be processed in the same way as the error sequence of the meteorological feature to obtain the error between each meteorological feature and the standard measured data.

[0134] Step 406: Determine the meteorological feature fusion weight of each meteorological feature in each meteorological feature data set based on the error corresponding to each meteorological feature in each meteorological feature data set.

[0135] In this embodiment of the application, for each meteorological feature data set, the meteorological feature fusion weight of each meteorological feature is calculated based on the error corresponding to each meteorological feature in the meteorological feature data set, so that the fusion weight of the meteorological feature with the largest error from the standard measured data is minimized, and the fusion weight of the meteorological feature with the smallest error from the standard measured data is maximized.

[0136] The formula for calculating the fusion weight is shown in Formula (IX):

[0137]

[0138] Where, λ i The meteorological feature fusion weight of the i-th meteorological feature in the meteorological feature dataset, e i The error between the i-th meteorological feature and the standard measured data is denoted as n, and the number of meteorological features in the meteorological feature data set is denoted as n.

[0139] Step 408: Based on the fusion weights of the meteorological features in each meteorological feature dataset, the meteorological features in each meteorological feature dataset are fused to obtain the third-dimensional reduced numerical weather forecast.

[0140] For example, for any meteorological feature, after obtaining the fusion weight of each meteorological feature in the meteorological feature data set, the meteorological features in the meteorological feature data set can be fused according to the meteorological feature fusion weight of each meteorological feature in the meteorological feature data set. After processing other meteorological feature data sets by referring to the fusion process, a third-dimensionality reduced numerical weather forecast can be obtained.

[0141] This application does not specifically limit the fusion method. For example, a linear weighting method can be used to fuse meteorological features from various meteorological feature datasets. The formula for linear weighting is shown in Formula (X):

[0142]

[0143] Among them, nwp j The meteorological feature refers to the meteorological feature obtained by fusing meteorological features from the j-th meteorological feature dataset. The meteorological feature fusion weight refers to the meteorological feature fusion weight of the i-th meteorological feature in the j-th meteorological feature dataset. It refers to the i-th meteorological feature in the j-th meteorological feature data set.

[0144] This application embodiment fuses meteorological features from various meteorological feature datasets to obtain fused meteorological features, the number of which equals the number of meteorological feature datasets. After obtaining the fused meteorological features, a third-dimensionality reduced numerical weather forecast can be constructed based on them. For example, if a second-dimensionality reduced numerical weather forecast has 17 dimensions, and this application embodiment divides these 17 meteorological features into 10 meteorological feature types and fuses the meteorological features from these 10 types, then the third-dimensionality reduced numerical weather forecast has 10 dimensions.

[0145] It should be noted that the embodiments of this application can further reduce the dimensionality of the second dimensionality-reduced numerical weather forecast to obtain the third dimensionality-reduced numerical weather forecast. In fact, the numerical weather forecast can also be directly reduced in dimensionality to obtain the third dimensionality-reduced numerical weather forecast. Alternatively, the first dimensionality-reduced numerical weather forecast can be directly reduced in dimensionality to obtain the third dimensionality-reduced numerical weather forecast. The specific process can be referred to the relevant descriptions of the foregoing embodiments, and the embodiments of this application will not be repeated here.

[0146] The wind power prediction method provided in this application can classify the meteorological features in the wind farm measured data and the second dimensionality-reduced numerical weather prediction according to the meteorological feature type. After determining the standard measured data from the wind farm measured data set corresponding to each meteorological feature type, the meteorological features in each meteorological feature data set can be fused according to the standard measured data corresponding to each meteorological feature type to obtain the third dimensionality-reduced numerical weather prediction. Since this application embodiment performs dimensionality reduction processing on the second dimensionality-reduced numerical weather prediction again, that is, performs dimensionality reduction processing on the numerical weather prediction three times, it can further reduce the dimensionality of the numerical weather prediction, reduce the data dimension processed by the wind power prediction model, and further reduce the redundant information in the numerical weather prediction, thereby effectively improving the prediction accuracy and efficiency of wind power.

[0147] In one embodiment, such as Figure 5 As shown, the wind power prediction model includes at least two prediction models. In step 106 above, the wind power prediction model is used to predict the wind power from the dimensionality-reduced numerical weather forecast to obtain the predicted wind power, including:

[0148] Step 502: Use each prediction model to predict wind power from the dimensionality-reduced numerical weather forecast, and obtain multiple initial predicted wind power values.

[0149] For example, the wind power prediction model in this application includes at least two prediction models. The number and type of prediction models are not specifically limited in this application. The following uses the example of a wind power prediction model including at least three prediction models to illustrate the embodiments of this application.

[0150] For example, the wind power prediction model in this application embodiment includes three prediction models: an Extreme Learning Machine (ELM) optimized by the Cloud Multi-objective Slime Mould Algorithm (CMOSMA), namely the MOSMA-ELM model; a Back Propagation (BP) model optimized by the CMOSMA algorithm, namely the MOSMA-BP model; and an ELMAN neural network optimized by the CMOSMA algorithm, namely the MOSMA-ELMAN model.

[0151] In this embodiment of the application, the dimensionality-reduced numerical weather forecast is input into each prediction model for prediction processing to obtain multiple initial predicted wind power. For example, taking the aforementioned example, the initial predicted wind power p1 of the MOSMA-ELM model, the initial predicted wind power p2 of the MOSMA-BP model, and the initial predicted wind power p3 of the MOSMA-ELMAN model can be obtained.

[0152] Step 504: Based on the prediction model fusion weights of each prediction model, the multiple initial predicted wind power are fused to obtain the predicted wind power.

[0153] For example, after obtaining the initial predicted wind power of each prediction model, the initial predicted wind power can be fused according to the prediction model fusion weight of each prediction model.

[0154] This application does not specifically limit the fusion method. For example, a linear weighting method can be used to fuse the initial predicted wind power of each prediction model. Taking the prediction models used as MOSMA-ELM, MOSMA-BP, and MOSMA-ELMAN models as examples, the formula for linear weighting is shown in formula (XI):

[0155]

[0156] Where p is the predicted wind power, m refers to the prediction model used, m=1 refers to the MOSMA-ELM model, m=2 refers to the MOSMA-BP model, m=3 refers to the MOSMA-ELMAN model, and λ m It is the fusion weight of each prediction model, p m This represents the initial predicted wind power output of each prediction model.

[0157] The wind power prediction method provided in this application uses multiple prediction models to perform prediction processing on the dimensionality-reduced numerical weather forecast, calculates the prediction model fusion weight corresponding to each prediction model, and fuses the initial predicted wind power of each prediction model based on the prediction model fusion weight. This effectively makes up for the shortcomings of each prediction model in wind power prediction, improves the problem of low prediction accuracy of a single prediction model, and further improves the accuracy of wind power prediction.

[0158] In one embodiment, such as Figure 6 As shown, the above method also includes:

[0159] Step 602: Obtain numerical weather forecasts and measured power data for the preset time period.

[0160] In this embodiment of the application, during the process of predicting wind power on the prediction date, numerical weather forecasts and measured power data within a preset period before the prediction date can be obtained. Then, based on the numerical weather forecasts and measured power data within the preset period, the prediction model fusion weights corresponding to each prediction model are determined during the process of predicting wind power on the prediction date.

[0161] The preset time period refers to a pre-defined time range. This embodiment does not specifically limit the value of the preset time period. For example, the preset time period could be three weeks before the prediction date; that is, if the prediction date is the 22nd, the preset time period could be the 1st to the 21st of the same month.

[0162] Step 604: Based on the measured power data within the preset time period, perform dimensionality reduction processing on the numerical weather forecast within the preset time period to obtain the dimensionality-reduced numerical weather forecast.

[0163] In this embodiment, the process of dimensionality reduction of numerical weather forecasts within a preset time period can be referred to the relevant description in the foregoing embodiments, and will not be repeated here.

[0164] Step 606: Construct a training set based on the dimensionality-reduced numerical weather forecasts.

[0165] After dimensionality reduction of the numerical weather forecasts within a preset time period, a training set can be constructed based on the dimensionality-reduced numerical weather forecasts. This training set can then be used to calculate the loss of each prediction model in order to obtain the prediction model fusion weights corresponding to each prediction model.

[0166] To facilitate obtaining the fusion weights of the prediction models for each model, embodiments of this application can divide the constructed training set into multiple sub-training sets. For example, when the preset time period is three weeks before the date to be predicted, the sub-training sets can be divided by week, i.e., the first sub-training set is from the 1st to the 7th, the second sub-training set is from the 8th to the 14th, and the third sub-training set is from the 15th to the 21st.

[0167] Step 608: Train each prediction model using the training set to obtain the loss value of each prediction model.

[0168] Each prediction model will obtain a loss value on each sub-training set, based on which the training error sequence for each sub-training set can be constructed. For example, when the prediction model is a MOSMA-ELM model, a MOSMA-BP model, or a MOSMA-ELMAN model, the training error sequence for the i-th sub-training set can be... in The training error of the MOSMA-ELM model on the i-th sub-training set. The training error of the MOSMA-BP model on the i-th sub-training set. This refers to the training error of the MOSMA-ELMAN model on the i-th sub-training set.

[0169] Step 610: Based on the loss values ​​of each prediction model, obtain the prediction model fusion weights of each prediction model.

[0170] The prediction model fusion weights for each prediction model can be obtained by fusing the training errors of the prediction model on each sub-training set using the DS evidence theory, resulting in a correlation index.

[0171] Taking the aforementioned example, this embodiment of the application preprocesses each training error sequence using formula (xii) to obtain the training error weight sequence of each sub-training set.

[0172]

[0173] Where m represents the prediction model used, m=1 represents the MOSMA-ELM model, m=2 represents the MOSMA-BP model, and m=3 represents the MOSMA-ELMAN model. For example, The proportion of training error of the MOSMA-ELM model on the i-th sub-training set. The proportion of training error of the MOSMA-BP model on the i-th sub-training set. This refers to the proportion of training error of the MOSMA-ELMAN model on the i-th sub-training set.

[0174] In this embodiment, formulas (xiii) and (xiv) are used to fuse the training error weights of each training model across all sub-training sets to obtain the prediction model fusion weights λ1 for the MOSMA-ELM model, λ2 for the MOSMA-BP model, and λ3 for the MOSMA-ELMAN model.

[0175]

[0176] in, This is the first fused training error weight, which is the result of fusing the training error weights of the first and second sub-training sets. In this embodiment, the first fused training error weight can be further fused with the training error weight of the third sub-training set to obtain the second fused training error weight, and so on, until the (n-2)th fused training error weight is fused with the training error weight of the nth sub-training set to obtain the prediction model fusion weight. The final fusion formula for the prediction model fusion weight is as follows:

[0177]

[0178] The embodiments of this application can then use formula (xi) to obtain the final predicted value based on λ1, λ2, and λ3.

[0179] The wind power prediction method provided in this application can calculate the training error of each prediction model on each sub-training set, and obtain the prediction model fusion weight of each prediction model by fusing the training errors of each prediction model on each sub-training set. Then, the initial predicted wind power of each prediction model can be fused based on the prediction model fusion weight, which effectively makes up for the shortcomings of each prediction model in wind power prediction, improves the problem of low prediction accuracy of a single prediction model, and further improves the accuracy of wind power prediction.

[0180] To enable those skilled in the art to better understand the embodiments of this application, the embodiments of this application are described below through specific examples.

[0181] For example, such as Figure 7 , Figure 8 The diagram shows a flowchart of a wind power prediction method. When the prediction date is the 22nd, numerical weather forecasts from the three weeks prior to the prediction date (1st to 7th, 8th to 14th, and 15th to 21st) can be selected as the training dataset. During actual training, the dimensionality-reduced numerical weather forecasts can be used as input, and the corresponding measured power data can be used as output to train the wind power prediction model.

[0182] Firstly, dimensionality reduction can be performed on numerical weather prediction. For each meteorological feature included in the numerical weather prediction, the correlation index between each meteorological feature and the corresponding measured power data can be calculated using the Pearson correlation index, mutual information entropy, and Spearman correlation index. Based on the results, a Pearson correlation index sequence, a mutual information entropy sequence, and a Spearman correlation index sequence can be constructed.

[0183] Among the three correlation indices used (Pearson correlation index, mutual information entropy, and Spearman correlation index), the Pearson correlation index reflects the waveform similarity between two datasets and can be used to measure the conflict between meteorological characteristics and measured power data; mutual information entropy measures the uncertainty between variables; and the Spearman correlation index is a non-parametric ordinal statistical indicator that can be used to measure the consistency of changes in two variables. All three correlation indices have the ability to measure the correlation between meteorological characteristics and power.

[0184] Furthermore, meteorological features in an anomalous state can be identified based on the Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence, and these anomalous meteorological features can be removed from the numerical weather prediction to obtain the first dimensionality-reduced numerical weather prediction. The specific steps for identifying meteorological features in an anomalous state can be referred to the relevant descriptions in the foregoing embodiments, and will not be repeated here.

[0185] This application embodiment further identifies meteorological features with correlation indices greater than a threshold as target meteorological features based on the correlation indices of each meteorological feature. Since a single correlation index calculation method cannot comprehensively reflect the mapping relationship between meteorological features and measured power data, this application embodiment uses DS evidence theory to fuse three correlation indices (Pearson correlation index, mutual information entropy, and Spearman correlation index) to obtain a fused correlation index. Based on the fused correlation index of each meteorological feature, meteorological features with fused correlation indices greater than a threshold are identified as target meteorological features. A second-dimensionality reduced numerical weather forecast is then constructed based on the target meteorological features. The specific steps for determining the target meteorological features can be referred to the relevant descriptions in the foregoing embodiments, and will not be repeated here.

[0186] After obtaining the second-dimensionality reduced numerical weather forecast, it can be classified according to meteorological feature types. The meteorological features in the corresponding meteorological feature data sets for each meteorological feature type are then fused to obtain the fused meteorological features. A third-dimensionality reduced numerical weather forecast is then constructed based on these fused meteorological features. The specific steps for determining the fused meteorological features can be referred to the relevant descriptions in the foregoing embodiments, and will not be repeated here.

[0187] In this application embodiment, multiple prediction models are used to perform prediction processing on the dimensionality-reduced numerical weather forecast to obtain the initial predicted wind power. Then, the multiple initial predicted wind power are fused according to the prediction model fusion weight of each prediction model to obtain the predicted wind power.

[0188] For example, embodiments of this application can use the MOSMA-ELM model, MOSMA-BP model, and MOSMA-ELMAN model for prediction. ELM is a type of single-hidden-layer feedforward network that can train network weights with only a small amount of computation, greatly improving the network's generalization ability and learning speed, and possessing strong nonlinear fitting ability. BP is a forward propagation model with strong nonlinear mapping ability. ELMAN is a recursively optimized neural network with local feedback and local memory modules, possessing global stability and time-varying adaptability. All three prediction models (MOSMA-ELM, MOSMA-BP, and MOSMA-ELMAN) can predict wind power, but each model has limited inference and fitting capabilities in wind power prediction. Therefore, embodiments of this application require fusing the initial predicted power of each prediction model to reduce prediction error and improve prediction accuracy. The specific steps for fusing the initial predicted wind power of each prediction model can be referred to the relevant descriptions in the foregoing embodiments, and will not be repeated here.

[0189] For example, such as Figure 9 As shown, the predicted value curve and the measured power data curve of the wind power prediction method provided in the embodiments of this application are illustrated.

[0190] Based on the same inventive concept, this application also provides a wind power prediction device for implementing the wind power prediction method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more wind power prediction device embodiments provided below can be found in the limitations of the wind power prediction method described above, and will not be repeated here.

[0191] In one embodiment, such as Figure 10 As shown, a wind power prediction device is provided, comprising: a first acquisition module 1002, a first dimensionality reduction module 1004, and a prediction module 1006, wherein:

[0192] The first acquisition module 1002 is used to acquire numerical weather forecasts and measured power data, wherein the numerical weather forecasts include meteorological characteristics in multiple dimensions.

[0193] The first dimensionality reduction module 1004 is used to perform dimensionality reduction processing on the numerical weather forecast based on the measured power data to obtain the dimensionality-reduced numerical weather forecast.

[0194] The prediction module 1006 is used to predict wind power from the dimensionality-reduced numerical weather forecast using a wind power prediction model, thereby obtaining the predicted wind power.

[0195] Based on the wind power prediction device provided in this application embodiment, by acquiring numerical weather forecasts and measured power data, the numerical weather forecasts can be dimensionality-reduced according to the measured power data. This application embodiment can then use a wind power prediction model to predict wind power from the dimensionality-reduced numerical weather forecasts, obtaining the predicted wind power. This application embodiment reduces the dimensionality of numerical weather forecasts, removes redundant information in numerical weather forecasts, and effectively improves the prediction accuracy of wind power.

[0196] In one embodiment, the first dimensionality reduction module 1004 is further configured to:

[0197] Based on the measured power data and the meteorological characteristics, Pearson correlation index sequence, mutual information entropy sequence, and Spearman correlation index sequence are constructed.

[0198] Based on the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence, meteorological features in an abnormal state in the numerical weather forecast are determined, and the meteorological features in an abnormal state are deleted from the numerical weather forecast to obtain the first dimensionality-reduced numerical weather forecast.

[0199] In one embodiment, the first dimensionality reduction module 1004 is further configured to:

[0200] Based on the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence of each meteorological feature in the first dimensionality reduction numerical weather forecast, a fusion correlation index sequence is constructed. The fusion correlation index sequence includes the fusion correlation index of each meteorological feature. The fusion correlation index is used to characterize the correlation between the meteorological feature and the measured power data.

[0201] Based on the fusion correlation index of each meteorological feature, the meteorological features are sorted from high to low to obtain a meteorological feature sequence;

[0202] Target meteorological features are determined sequentially from the meteorological feature sequence until the cumulative result of the fusion correlation index of the target meteorological features is greater than a threshold.

[0203] Based on the target meteorological characteristics, a second-dimensional reduced numerical weather forecast is obtained.

[0204] In one embodiment, the first dimensionality reduction module 1004 is further configured to:

[0205] Based on the meteorological feature types, the measured data of wind farms and the meteorological features in the second dimensionality-reduced numerical weather forecast are classified to obtain the set of measured data of wind farms and the set of meteorological feature data corresponding to each meteorological feature type.

[0206] For any of the meteorological feature types, standard measured data are determined from the wind farm measured data set corresponding to the meteorological feature type, and the error between each meteorological feature in the meteorological feature data set corresponding to the meteorological feature type and the standard measured data is determined.

[0207] Based on the error corresponding to each meteorological feature in each meteorological feature data set, determine the meteorological feature fusion weight of each meteorological feature in each meteorological feature data set;

[0208] Based on the fusion weights of the meteorological features in each of the meteorological feature datasets, the meteorological features in each of the meteorological feature datasets are fused to obtain a third-dimensional reduced numerical weather forecast.

[0209] In one embodiment, the prediction module 1006 is further configured to:

[0210] The wind power forecast of the dimensionality-reduced numerical weather forecast is predicted by each of the prediction models to obtain multiple initial predicted wind power.

[0211] The multiple initial predicted wind power are fused according to the prediction model fusion weights of each prediction model to obtain the predicted wind power.

[0212] In one embodiment, the device further includes:

[0213] The second acquisition module is used to acquire numerical weather forecasts and measured power data within a preset time period;

[0214] The second dimensionality reduction module is used to perform dimensionality reduction processing on the numerical weather forecast within the preset time period based on the measured power data within the preset time period, so as to obtain the dimensionality-reduced numerical weather forecast within the preset time period.

[0215] The construction module is used to construct a training set based on the dimensionality-reduced numerical weather forecasts within the preset time period;

[0216] The training module is used to train each of the prediction models using the training set to obtain the loss value of each prediction model.

[0217] The fusion weight module is used to obtain the prediction model fusion weight of each prediction model based on the loss value of each prediction model.

[0218] The modules in the aforementioned wind power prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0219] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 11 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a wind power prediction method.

[0220] Those skilled in the art will understand that Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0221] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0222] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0223] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0224] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0225] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0226] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0227] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A wind power prediction method, characterized by, The method comprises: obtaining numerical weather prediction and measured power data, the numerical weather prediction comprising a plurality of dimensional meteorological characteristics; calculating the Pearson correlation index, mutual information entropy and Spearman correlation index of each meteorological characteristic and measured power data, constructing a Pearson correlation index sequence, a mutual information entropy sequence and a Spearman correlation index sequence according to the Pearson correlation index, mutual information entropy and Spearman correlation index of each meteorological characteristic, and calculating the first proportion of each Pearson correlation index in the Pearson correlation index sequence, the second proportion of each mutual information entropy in the mutual information entropy sequence, and the third proportion of each Spearman correlation index in the Spearman correlation index sequence; determining whether each meteorological characteristic is in an abnormal state according to whether the Pearson correlation index, mutual information entropy and Spearman correlation index of each meteorological characteristic are consistent in sign, deleting the meteorological characteristics in the abnormal state from the numerical weather prediction to obtain a first reduced-dimension numerical weather prediction; constructing a fused correlation index sequence according to the Pearson correlation index sequence, mutual information entropy sequence and Spearman correlation index sequence of each meteorological characteristic in the first reduced-dimension numerical weather prediction, the fused correlation index sequence comprising a fused correlation index of each meteorological characteristic, the fused correlation index being used to represent the correlation between the meteorological characteristic and the measured power data, and the fused correlation index being obtained by the following formula: wherein is a fusion correlation index of the ith meteorological feature, , is the first weight of the ith meteorological feature, is the second weight of the ith meteorological feature, is the third weight of the ith meteorological feature, and n is the total number of the meteorological features. sorting each meteorological characteristic in descending order according to the fused correlation index of each meteorological characteristic to obtain a meteorological characteristic sequence; determining target meteorological characteristics from the meteorological characteristic sequence in sequence until the cumulative result of the fused correlation index of the target meteorological characteristic is greater than a threshold value, and constructing a second reduced-dimension numerical weather prediction according to each target meteorological characteristic; performing wind power prediction on the second reduced-dimension numerical weather prediction by a wind power prediction model to obtain predicted wind power.

2. The method of claim 1, wherein, The method of performing wind power prediction on the second reduced-dimension numerical weather prediction by a wind power prediction model to obtain predicted wind power comprises: classifying the meteorological characteristics in the wind farm measured data and the second reduced-dimension numerical weather prediction according to meteorological characteristic types to obtain a wind farm measured data set and a meteorological characteristic data set corresponding to each meteorological characteristic type; for any meteorological characteristic type, determining standard measured data from the wind farm measured data set corresponding to the meteorological characteristic type, and determining the error between each meteorological characteristic in the meteorological characteristic data set corresponding to the meteorological characteristic type and the standard measured data; determining a meteorological characteristic fusion weight of each meteorological characteristic in each meteorological characteristic data set according to the error corresponding to each meteorological characteristic in each meteorological characteristic data set. According to the meteorological feature fusion weight of each meteorological feature in each meteorological feature data set, the meteorological features in each meteorological feature data set are fused to obtain a third reduced dimension numerical weather forecast; The third reduced dimension numerical weather forecast is subjected to wind power prediction through a wind power prediction model to obtain predicted wind power.

3. The method of claim 2, wherein, The wind power prediction model includes at least two prediction models, The third reduced dimension numerical weather forecast is subjected to wind power prediction through a wind power prediction model to obtain predicted wind power, including: The third reduced dimension numerical weather forecast is subjected to wind power prediction through each prediction model to obtain multiple initial predicted wind power; The multiple initial predicted wind power is subjected to fusion processing according to the prediction model fusion weight of each prediction model to obtain predicted wind power.

4. The method of claim 3, wherein, The method further includes: Obtaining numerical weather forecast and measured power data in a preset period; According to the measured power data in the preset period, the numerical weather forecast in the preset period is subjected to dimension reduction processing to obtain reduced dimension numerical weather forecast in the preset period; According to the reduced dimension numerical weather forecast in the preset period, a training set is constructed; Each prediction model is subjected to training processing through the training set to obtain a loss value of each prediction model; According to the loss value of each prediction model, a prediction model fusion weight of each prediction model is obtained.

5. A wind power prediction device, characterized by, The device includes a first acquisition module, a first dimension reduction module, and a prediction module, wherein: The first acquisition module is configured to obtain numerical weather forecast and measured power data, and the numerical weather forecast includes multiple dimensions of meteorological features; The first dimension reduction module is configured to calculate the Pearson correlation index, mutual information entropy, and Spearman correlation index of each meteorological feature and measured power data, construct a Pearson correlation index sequence, a mutual information entropy sequence, and a Spearman correlation index sequence according to the Pearson correlation index, mutual information entropy, and Spearman correlation index of each meteorological feature, and calculate a first proportion of each Pearson correlation index in the Pearson correlation index sequence, a second proportion of each mutual information entropy in the mutual information entropy sequence, and a third proportion of each Spearman correlation index in the Spearman correlation index sequence; According to whether the positive and negative of the Pearson correlation index, mutual information entropy, and Spearman correlation index of each meteorological feature are consistent, it is determined whether the meteorological feature is in an abnormal state, the meteorological feature in the abnormal state is deleted from the numerical weather forecast to obtain a first reduced dimension numerical weather forecast; According to the Pearson correlation index sequence, the mutual information entropy sequence, and the Spearman correlation index sequence of each meteorological feature in the first reduced-dimension numerical weather forecast, a fused correlation index sequence is constructed, the fused correlation index sequence including a fused correlation index of each meteorological feature, the fused correlation index being used to represent the correlation between the meteorological feature and the measured power data; the fused correlation index is obtained by the following formula: wherein is a fusion correlation index of the i-th meteorological feature, , is the first weight of the i-th meteorological feature, is the second weight of the i-th meteorological feature, is the third weight of the i-th meteorological feature, and n is the total number of the meteorological features. According to the fused correlation index of each meteorological feature, each meteorological feature is sorted from high to low to obtain a meteorological feature sequence; Target meteorological features are sequentially determined from the meteorological feature sequence until the cumulative result of the fused correlation index of the target meteorological feature is greater than a threshold value, and a second reduced-dimension numerical weather forecast is constructed according to each target meteorological feature. The prediction module is configured to perform wind power prediction on the second reduced-dimension numerical weather forecast by using a wind power prediction model to obtain predicted wind power.

6. The apparatus of claim 5, wherein, The first reduced-dimension module is further configured to: According to the meteorological feature types, the meteorological features in the wind farm measured data and the second reduced-dimension numerical weather forecast are classified to obtain a wind farm measured data set and a meteorological feature data set corresponding to each meteorological feature type; For any meteorological feature type, standard measured data are determined from the wind farm measured data set corresponding to the meteorological feature type, and errors between each meteorological feature in the meteorological feature data set corresponding to the meteorological feature type and the standard measured data are determined; According to the errors corresponding to each meteorological feature in each meteorological feature data set, meteorological feature fusion weights of each meteorological feature in each meteorological feature data set are determined; According to the meteorological feature fusion weights of each meteorological feature in each meteorological feature data set, the meteorological features in each meteorological feature data set are fused to obtain a third reduced-dimension numerical weather forecast; The third reduced-dimension numerical weather forecast is subjected to wind power prediction by using a wind power prediction model to obtain predicted wind power.

7. The apparatus of claim 6, wherein, The wind power prediction model includes at least two prediction models, The first reduced-dimension module is further configured to: The third reduced-dimension numerical weather forecast is subjected to wind power prediction by each prediction model to obtain a plurality of initial predicted wind power; The plurality of initial predicted wind power is subjected to fusion processing according to the prediction model fusion weights of each prediction model to obtain predicted wind power.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method of any one of claims 1 to 4.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 4.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 4.

Images