Method, device and computer equipment for predicting wind speed of wind turbine generator

Abstract

The present application relates to a method, apparatus, computer equipment, storage medium and computer program product for predicting the wind speed of a wind turbine. The method comprises: performing collective empirical mode decomposition on a wind speed sequence to be processed to obtain mode component data corresponding to each mode component. Obtaining the active power of the power grid and the ambient temperature corresponding to a historical time period, and combining the active power of the power grid, the ambient temperature, and the mode component data to obtain aggregate features. Performing time translation processing on the mode component data to obtain time features, and splicing and combining the aggregate features and the time features to obtain a feature subset. Performing feature cross-processing based on each combined feature in the feature subset to obtain a cross feature. Predicting the wind speed of the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within the target time period. In this way, the accuracy of the wind speed prediction of the wind turbine is greatly increased.

Description

Technical Field

[0001] The present application relates to the technical field of wind speed prediction, and in particular to a method, apparatus, computer equipment, storage medium, and computer program product for predicting wind speed of a wind turbine. Background Art

[0002] With the development of wind power generation technology, wind turbines are often used as the primary equipment for wind power generation. However, wind speed significantly affects wind turbine efficiency, necessitating wind speed prediction. This predicted wind speed can be used to determine grid dispatch plans and adjust dispatch frequency to ensure stable power supply quality.

[0003] Traditionally, recursive multi-step prediction methods have been used to predict wind speed. This method uses the previous step's prediction value as the true value and uses this true value to predict the next step, thereby achieving multi-step prediction. However, if the previous step's prediction value is inaccurate, the next step's prediction value will also be inaccurate, and as the number of prediction steps increases, the errors accumulate. This makes it difficult to accurately predict the wind speed of a wind turbine, resulting in low wind speed prediction accuracy. Summary of the Invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for predicting wind speed of a wind turbine in order to address the above technical problems.

[0005] In a first aspect, the present application provides a method for predicting wind speed of a wind turbine. The method comprises:

[0006] Performing ensemble empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period;

[0007] Acquiring grid active power and ambient temperature corresponding to the historical time period, and combining the grid active power, the ambient temperature, and the mode component data to obtain an aggregated feature;

[0008] Performing time translation processing on the pattern component data to obtain time features, and combining the aggregate features and the time features to obtain a feature subset;

[0009] Perform feature cross processing based on each combined feature in the feature subset to obtain a cross feature;

[0010] The wind speed is predicted for the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within a target time period.

[0011] In a second aspect, the present application also provides a device for predicting wind speed of a wind turbine. The device comprises:

[0012] A decomposition module is used to perform collective empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine within a historical time period;

[0013] a combining module, configured to obtain the grid active power and the ambient temperature corresponding to the historical time period, and combine the grid active power, the ambient temperature, and the mode component data to obtain an aggregated feature;

[0014] A translation module, configured to perform time translation processing on the pattern component data to obtain a time feature, and to combine the aggregate feature and the time feature to obtain a feature subset;

[0015] A cross module, configured to perform feature cross processing based on each combined feature in the feature subset to obtain a cross feature;

[0016] The prediction module is used to perform wind speed prediction on the cross-features and the feature subsets through a prediction model to obtain a wind speed prediction value of the wind turbine within a target time period.

[0017] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are performed:

[0018] Performing ensemble empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period;

[0019] Acquiring grid active power and ambient temperature corresponding to the historical time period, and combining the grid active power, the ambient temperature, and the mode component data to obtain an aggregated feature;

[0020] Performing time translation processing on the pattern component data to obtain time features, and combining the aggregate features and the time features to obtain a feature subset;

[0021] Perform feature cross processing based on each combined feature in the feature subset to obtain a cross feature;

[0022] The wind speed is predicted for the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within a target time period.

[0023] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0024] Performing ensemble empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period;

[0025] Acquiring grid active power and ambient temperature corresponding to the historical time period, and combining the grid active power, the ambient temperature, and the mode component data to obtain an aggregated feature;

[0026] Performing time translation processing on the pattern component data to obtain time features, and combining the aggregate features and the time features to obtain a feature subset;

[0027] Perform feature cross processing based on each combined feature in the feature subset to obtain a cross feature;

[0028] The wind speed is predicted for the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within a target time period.

[0029] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the following steps:

[0030] Performing ensemble empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period;

[0031] Acquiring grid active power and ambient temperature corresponding to the historical time period, and combining the grid active power, the ambient temperature, and the mode component data to obtain an aggregated feature;

[0032] Performing time translation processing on the pattern component data to obtain time features, and combining the aggregate features and the time features to obtain a feature subset;

[0033] Perform feature cross processing based on each combined feature in the feature subset to obtain a cross feature;

[0034] The wind speed is predicted for the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within a target time period.

[0035] The aforementioned wind turbine wind speed prediction method, apparatus, computer device, storage medium, and computer program product perform ensemble empirical mode decomposition (EEMD) on a processed wind speed sequence to obtain mode component data corresponding to each mode component. The processed wind speed sequence is obtained by collecting wind speed data from wind turbines over a historical time period. Thus, processing the processed wind speed sequence based on EEMD effectively avoids modal chaos and ensures the validity of the mode component data. The grid active power and ambient temperature corresponding to the historical time period are obtained and combined with the mode component data to obtain aggregated features. The aggregated features obtained by combining information on the three variables of power, temperature, and wind speed reflect multi-dimensional information and significantly enhance the validity of the aggregated features. The mode component data is time-shifted to obtain temporal features, which are then concatenated and combined with the temporal features to obtain a feature subset. Feature cross-processing is performed on each combined feature in the feature subset to obtain cross-features, thereby effectively expanding the number of features used for wind speed prediction. The prediction model uses this cross-feature and this feature subset to predict wind speed, obtaining a predicted wind speed value for the wind turbine within the target time period. In this way, the sufficient and highly effective cross-features and feature subsets significantly improve the quality of the features used for wind speed prediction. Therefore, based on these high-quality cross-features and feature subsets, the prediction model can produce highly accurate wind speed predictions, significantly increasing the accuracy of wind turbine wind speed predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 FIG. 1 is an application environment diagram of a method for predicting wind speed of a wind turbine generator according to an embodiment;

[0037] Figure 2 1 is a flow chart of a method for predicting wind speed of a wind turbine according to an embodiment;

[0038] Figure 3 A schematic flow chart of a step of determining aggregation features in one embodiment;

[0039] Figure 4 A schematic diagram illustrating a translation operation in one embodiment;

[0040] Figure 5 FIG. 1 is a flow chart of a step of determining a feature subset in one embodiment;

[0041] Figure 6 A schematic flow chart of a step of determining cross-features in one embodiment;

[0042] Figure 7 Schematic diagram of a flow chart of the steps of determining a prediction model in one embodiment;

[0043] Figure 8 A schematic diagram of determining a sample prediction sequence label in one embodiment;

[0044] Figure 9 A schematic diagram of a process for constructing a sample prediction sequence in one embodiment;

[0045] Figure 10 A schematic diagram of a process for training a prediction model in one embodiment;

[0046] Figure 11 A comparative schematic diagram of a test prediction model in one embodiment;

[0047] Figure 12 is a flow chart of a method for predicting wind speed of a wind turbine according to another embodiment;

[0048] Figure 13 is a flow chart of a method for predicting wind speed of a wind turbine according to another embodiment;

[0049] Figure 14 is a structural block diagram of a device for predicting wind speed of a wind turbine generator in one embodiment;

[0050] Figure 15 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0052] In one embodiment, Figure 1 As shown, a method for predicting the wind speed of a wind turbine is provided. This embodiment uses the method applied to a computer device as an example for illustration. The computer device can specifically be a terminal or a server. It can be understood that the method can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. Among them, the terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server can be implemented as an independent server or a server cluster composed of multiple servers. In this embodiment, the method includes the following steps:

[0053] Step S102 , performing ensemble empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period.

[0054] A wind turbine is an electrical device that converts wind energy into mechanical work, which drives the rotor to rotate and ultimately outputs alternating current. The wind speed sequence to be processed contains wind speeds at multiple moments. Ensemble empirical mode decomposition (EEMD) adds one or more sets of white noise signals to EMD, treating the combination of the signal and noise as the signal to be decomposed. By utilizing the uniform distribution of the white noise spectrum, when the signal is loaded onto a white noise background, signals of different time scales are automatically distributed to an appropriate reference scale. EMD (empirical mode decomposition) is then performed on each signal to obtain the true mode.

[0055] Specifically, the computer device obtains an original wind speed sequence, wherein the original wind speed sequence includes the wind speed of the wind turbine at each historical moment in a historical time period. The computer device combines the original wind speed sequence with white noise to obtain a processed wind speed sequence. The computer device performs ensemble empirical mode decomposition on the processed wind speed sequence to obtain mode component data corresponding to each modal component.

[0056] For example, let the original wind speed sequence be x(t) and the white noise be n(t). The computer device then adds and combines the original wind speed sequence and the white noise sequence to obtain the wind speed sequence to be processed X(t):

[0057] X(t)=x(t)+n(t)

[0058] The computer equipment performs ensemble empirical mode decomposition on the wind speed sequence to be processed and obtains the mode component data IMF corresponding to each mode component. k .

[0059] Step S104 : acquiring the grid active power and ambient temperature corresponding to the historical time period, and combining the grid active power, the ambient temperature, and the mode component data to obtain an aggregated feature.

[0060] Specifically, the computer device obtains the grid active power and ambient temperature corresponding to the historical time period and, based on the time information of the historical time period, divides the grid active power, ambient temperature, and mode component data to obtain sub-feature sets. Each sub-feature set is composed of multiple components with the same time information, and the time information of each sub-feature set is different. The computer device performs statistical calculations on each sub-feature set to obtain statistical results for the corresponding sub-feature set. The computer device concatenates and combines the statistical results corresponding to each sub-feature set to obtain an aggregated feature.

[0061] For example, the computer device obtains the grid active power and ambient temperature corresponding to the historical time period, and based on the time information of the historical time period, divides the grid active power, ambient temperature, and mode component data to obtain various sub-feature sets. Each sub-feature set includes multiple grid active power, ambient temperature, and mode component data corresponding to the corresponding sub-feature set. For each sub-feature set, the computer device performs statistical processing on the multiple grid active power, ambient temperature, and mode component data corresponding to the corresponding sub-feature set to obtain power statistics, temperature statistics, and mode component statistics corresponding to each sub-feature set. The computer device splices and combines the power statistics, temperature statistics, and mode component statistics corresponding to each sub-feature set to obtain an aggregated feature.

[0062] Step S106: performing time translation processing on the pattern component data to obtain a time feature, and concatenating and combining the aggregate feature and the time feature to obtain a feature subset.

[0063] Specifically, the computer device performs multiple time shifts on each pattern component data, and combines the shift results of the multiple time shifts to obtain a time feature. The computer device obtains an aggregate feature, and combines the aggregate feature with the time feature to obtain an aggregate feature.

[0064] Step S108: performing feature cross processing based on each combined feature in the feature subset to obtain cross features.

[0065] Specifically, the computer device obtains the feature subset and performs cross-processing on each combined feature in the feature subset with each other combined feature to obtain multiple cross-processing results. The computer device filters each cross-processing result to obtain a filtered result. The computer device determines the cross-feature based on the multiple filtered results.

[0066] For example, the feature subset contains five combined features: L1, L2, L3, L4, and L5. L1 is cross-processed with the other four combined features to obtain a cross-processed result corresponding to L1. The computer device characterizes the cross-processed result as a missing value or deletes an infinite number of cross-processed results to obtain a filtered result. The computer device uses the collection of multiple filtered results as a cross-feature.

[0067] Step S110 , performing wind speed prediction on the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within a target time period.

[0068] Specifically, the computer device determines a target time period for wind speed prediction, obtains a prediction model corresponding to the target time period, and uses the cross-feature and the feature subset as prediction features. The computer device inputs the prediction features into the prediction model to obtain a predicted wind speed value for the wind turbine during the target time period.

[0069] The prediction model is built based on the Catboost algorithm, an open-source machine learning library. The Catboost algorithm is a GBDT (Gradient Boosting Decision Tree) framework based on a symmetric decision tree as a base learner, with low parameters, support for categorical variables, and high accuracy.

[0070] For example, if the computer device determines that the target time period is 20 minutes, a trained Catboost prediction model for 20 minutes of prediction is obtained. The computer device inputs the cross-features and feature subsets corresponding to the target time period into the Catboost prediction model for prediction, and obtains the wind speed prediction values ​​corresponding to each target moment in the target time period.

[0071] In the aforementioned wind turbine wind speed prediction method, ensemble empirical mode decomposition (EEMD) is performed on a processing wind speed sequence to obtain mode component data corresponding to each mode component. The processing wind speed sequence is obtained by collecting wind turbine wind speeds over a historical time period. Thus, processing the processing wind speed sequence based on EEMD effectively avoids modal chaos and ensures the validity of the mode component data. The grid active power and ambient temperature corresponding to the historical time period are obtained, and the grid active power, ambient temperature, and mode component data are combined to obtain aggregated features. The aggregated features obtained by combining information on the three variables of power, temperature, and wind speed reflect multi-dimensional information, significantly enhancing the validity of the aggregated features. The mode component data is time-shifted to obtain temporal features, and the aggregated features are concatenated with the temporal features to obtain a feature subset. Feature cross-processing is performed on each combined feature in the feature subset to obtain cross-features, thereby effectively expanding the number of features used for wind speed prediction. The prediction model uses this cross-feature and this feature subset to predict wind speed, obtaining a predicted wind speed value for the wind turbine within the target time period. In this way, the sufficient and highly effective cross-features and feature subsets significantly improve the quality of the features used for wind speed prediction. Therefore, based on these high-quality cross-features and feature subsets, the prediction model can produce highly accurate wind speed predictions, significantly increasing the accuracy of wind turbine wind speed predictions.

[0072] In one embodiment, performing collective empirical mode decomposition on the processed wind speed sequence to obtain mode component data corresponding to each mode component includes: performing multiple empirical mode decompositions on the processed wind speed sequence to obtain multiple groups of mode component vectors; each group of mode component vectors includes multiple mode component values. For each mode component, the value of the corresponding mode component is obtained from the multiple groups of mode component vectors, and the mean of the multiple mode component values ​​corresponding to the corresponding mode component is calculated to obtain the mean corresponding to the corresponding mode component. The multiple mode component data are obtained by using the mean corresponding to the mode component as the mode component data corresponding to the corresponding mode component.

[0073] Specifically, a computer device performs multiple empirical mode decompositions on a to-be-processed wind speed sequence containing white noise, obtaining multiple groups of mode component vectors corresponding to the multiple empirical mode decompositions. Each group of mode component vectors includes multiple mode component values. For each mode component, the value of the corresponding mode component is obtained from the multiple groups of mode component vectors, and the mean of the multiple mode component values ​​corresponding to the corresponding mode component is calculated to obtain the mean corresponding to the corresponding mode component. For each mode component, the computer device uses the mean corresponding to the corresponding mode component as the mode component data corresponding to the corresponding mode component, thereby obtaining multiple mode component data.

[0074] For example, Figure 2 As shown, the computer device obtains the wind speed sequence to be processed based on the original wind speed sequence x(t) and the white noise n(t). The computer device initializes the number of cycles, that is, the number of cycles is set to j=1. The computer device performs a collective empirical mode decomposition on the wind speed sequence to be processed. Specifically, under the current number of cycles, the rapid wind speed sequence to be processed is subjected to empirical mode decomposition to obtain a set of mode component vectors. The computer device determines whether the current number of cycles j is equal to M. If not, the current number of cycles is updated, and the empirical mode decomposition is performed again until the current number of cycles is equal to M. The computer device obtains the M groups of mode component vectors obtained by decomposition, wherein the jth IMF component obtained by the mth empirical mode decomposition is set to C j,m The computer is set to average the IMF components corresponding to each stage obtained by M-times empirical mode decomposition. The final IMF components are obtained, namely IMF component 1, IMF component 2, ..., IMF component p. Among them, the mode component data IMF of the kth IMF component is k The calculation is as follows:

[0075]

[0076] In this embodiment, by performing ensemble empirical mode decomposition on the wind speed sequence to be processed containing white noise, the modal aliasing phenomenon occurring in the empirical mode decomposition process can be avoided, thereby ensuring the validity of the mode component data.

[0077] In one embodiment, Figure 3 As shown, the grid active power, ambient temperature, and mode component data all correspond to historical time periods. The grid active power, the ambient temperature, and the mode component data are combined to obtain aggregate features, including:

[0078] Step S302: divide the historical time period into multiple sub-historical time periods, and divide the grid active power, the ambient temperature, and the mode component data corresponding to the historical time period according to each sub-historical time period to obtain each sub-feature set; wherein each sub-feature set corresponds to a sub-historical time period, and each sub-feature set contains the grid active power, ambient temperature, and mode component data corresponding to the corresponding sub-historical time period.

[0079] Specifically, the computer device divides the historical time period into a plurality of sub-historical time periods based on time information in the historical time period. For each sub-historical time period, the computer device determines a sub-feature set corresponding to the corresponding sub-historical time period from the plurality of pieces of grid active power, ambient temperature, and mode component data corresponding to the historical time period. Each sub-feature set includes the grid active power, ambient temperature, and mode component data corresponding to the corresponding sub-historical time period.

[0080] For example, the computer device divides the historical time period into hours based on the time information in the historical time period to obtain sub-historical time periods. Each sub-historical time period is an hour. The computer device uses the grid active power, ambient temperature, and mode component data belonging to the first sub-historical time period as the first sub-feature set.

[0081] Step S304 : For each sub-feature set, multiple statistical calculations are performed on each grid active power corresponding to the corresponding sub-feature set to obtain multiple power statistical results corresponding to each sub-feature set.

[0082] Specifically, for each sub-feature set, the computer device performs at least one of median calculation, mean calculation, standard deviation calculation, maximum value calculation, minimum value calculation, and range calculation on each grid active power corresponding to the corresponding sub-feature set to obtain multiple power statistical results corresponding to each sub-feature set.

[0083] For example, for the first sub-feature set, the first sub-feature set includes multiple first active powers, such as P 1m The computer device performs at least one of median calculation, mean calculation, standard deviation calculation, maximum calculation, minimum calculation, and range calculation based on the multiple first active powers, and obtains power statistical results C corresponding to each calculation. P1,n , where C P1,n Represents n power statistics results corresponding to the first sub-feature set active power P1.

[0084] Step S306 : For each sub-feature set, perform multiple statistical calculations on each ambient temperature corresponding to the corresponding sub-feature set to obtain multiple temperature statistical results corresponding to each sub-feature set.

[0085] Specifically, for each sub-feature set, the computer device performs at least one of median calculation, mean calculation, standard deviation calculation, maximum calculation, minimum calculation, and range calculation on each ambient temperature corresponding to the corresponding sub-feature set to obtain multiple temperature statistical results corresponding to each sub-feature set.

[0086] For example, for the first sub-feature set, the first sub-feature set includes multiple first ambient temperatures, such as T 1m The computer device performs at least one of median calculation, mean calculation, standard deviation calculation, maximum calculation, minimum calculation, and range calculation based on the multiple first temperatures to obtain temperature statistical results C corresponding to each calculation. T1,n , where C T1,n Represents n temperature statistics corresponding to the first sub-feature set temperature T1.

[0087] Step S308: For each sub-feature set, perform multiple statistical calculations on the pattern component data corresponding to the corresponding sub-feature set to obtain multiple pattern component statistical results corresponding to each sub-feature set; wherein the statistical calculation includes at least one of median calculation, mean calculation, standard deviation calculation, maximum value calculation, minimum value calculation, and range calculation.

[0088] Specifically, for each sub-feature set, the computer device performs at least one of median calculation, mean calculation, standard deviation calculation, maximum value calculation, minimum value calculation, and range calculation on each pattern component data corresponding to the corresponding sub-feature set to obtain multiple pattern component statistical results corresponding to each sub-feature set.

[0089] For example, for the first sub-feature set, the first sub-feature set includes a plurality of first mode component data, such as I 1m The computer device performs at least one of median calculation, mean calculation, standard deviation calculation, maximum calculation, minimum calculation, and range calculation based on the plurality of first mode component data, and obtains the mode component statistical results C corresponding to each calculation. I1,n , where C I1,n Represents the statistical results of n pattern components corresponding to the first sub-feature set pattern component I1.

[0090] Step S310 : For each sub-feature set, the power statistics, temperature statistics, and mode component statistics corresponding to the corresponding sub-feature set are used as sub-aggregate features corresponding to the corresponding sub-feature set.

[0091] Specifically, the computer device obtains the power statistics, temperature statistics, and mode component statistics corresponding to each sub-feature set. For each sub-feature set, the computer device uses the power statistics, temperature statistics, and mode component statistics corresponding to the corresponding sub-feature set as the sub-aggregate feature corresponding to the corresponding sub-feature set.

[0092] Step S312: combining the sub-aggregation features corresponding to the sub-feature sets to obtain an aggregate feature.

[0093] Specifically, the computer device obtains the sub-aggregate features corresponding to each sub-feature set, and concatenates the sub-aggregate features corresponding to each sub-feature set to obtain the aggregate feature.

[0094] In this embodiment, grid active power, ambient temperature, and mode component data are divided into sub-historical time periods to generate sub-feature sets with the same time information. For each sub-feature set, statistical calculations are performed on the active power, ambient temperature, and mode component corresponding to that sub-feature set, resulting in aggregated features containing information about time, power, temperature, and mode components. This allows for intuitive and accurate representation of multi-dimensional information, significantly enhancing the effectiveness of aggregated features.

[0095] In one embodiment, the pattern component data is subjected to time shift processing to obtain a time feature, including: increasing the shift amount based on the shift amount of the previous shift to obtain the shift amount corresponding to the current shift. The pattern component data is shifted according to the shift amount corresponding to the current shift to obtain each sub-time feature corresponding to the current shift. The shift amount is increased based on the shift amount of the current shift to continue the next shift operation until a preset number of shifts is reached, thereby obtaining each sub-time feature corresponding to each shift. The sub-time features obtained by multiple shift operations are combined to obtain a time feature.

[0096] Specifically, the computer device adds a translation interval to the translation amount of the previous translation to obtain the translation amount corresponding to the current translation. The computer device obtains the pattern component data and translates the pattern component data according to the translation amount corresponding to the current translation to obtain each sub-time feature corresponding to the current translation. A translation interval is added to the translation amount of the current translation to continue the next translation operation until a preset number of translations is reached, thereby obtaining each sub-time feature corresponding to each translation. The computer device obtains the sub-time features obtained from multiple translation operations and splices and combines the sub-time features obtained from the multiple translation operations to obtain the time feature.

[0097] For example, if the translation amount of the last translation is T units, the computer device adds a translation interval T to the translation amount of the last translation, and obtains a translation amount of 2T corresponding to the current translation. The computer device translates each pattern component data backward by 2T units to obtain each sub-time feature corresponding to the current translation. A translation interval T is added to the translation amount of the current translation to obtain a translation amount of 3T for the next translation, so as to continue the next translation operation until the preset number of translations is reached, 10, and the sub-time features corresponding to the 10 translations are obtained. The specific presentation is as follows. Figure 4 The computer device combines the sub-time features corresponding to the 10 translations to obtain the time feature.

[0098] In this embodiment, by performing multiple translation operations on each mode component data, sub-time features corresponding to the number of translations are obtained, and the sub-time features are combined to obtain the time feature. This can further enhance the characteristics of the time dimension, which is conducive to improving the accuracy of wind turbine wind speed prediction.

[0099] In one embodiment, Figure 5As shown, the historical time period is divided into multiple sub-historical time periods, and the grid active power, ambient temperature, and pattern component data corresponding to the historical time period (i.e., corresponding to IMF component 1, IMF component 2, ..., IMF component p in the figure) are divided according to each sub-historical time period to obtain various sub-feature sets (i.e., corresponding to combination 1, combination 2, ..., combination q in the figure); wherein each sub-feature set corresponds to a sub-historical time period, and each sub-feature set contains the grid active power, ambient temperature, and pattern component data corresponding to the corresponding sub-historical time period. For each sub-feature set, multiple statistical calculations are performed on the grid active power corresponding to the corresponding sub-feature set to obtain multiple power statistical results corresponding to each sub-feature set. For each sub-feature set, multiple statistical calculations are performed on the ambient temperature corresponding to the corresponding sub-feature set to obtain multiple temperature statistical results corresponding to each sub-feature set. For each sub-feature set, multiple statistical calculations are performed on the pattern component data corresponding to the corresponding sub-feature set to obtain multiple pattern component statistical results corresponding to each sub-feature set. For each sub-feature set, the power statistics, temperature statistics, and mode component statistics corresponding to the corresponding sub-feature set are used as sub-aggregate features corresponding to the corresponding sub-feature set. The sub-aggregate features corresponding to each sub-feature set are combined to obtain an aggregate feature. The translation amount is increased based on the translation amount of the previous translation to obtain the translation amount corresponding to the current translation. The mode component data is translated according to the translation amount corresponding to the current translation to obtain each sub-time feature corresponding to the current translation. The translation amount is increased based on the translation amount of the current translation to continue the next translation operation until the preset number of translations of 10 is reached and the sub-time features corresponding to each translation are obtained. The sub-time features obtained by multiple translation operations are combined to obtain a time feature. The aggregate feature and the time feature are spliced ​​and combined to obtain a feature subset (i.e., the original feature set in the corresponding figure).

[0100] In this embodiment, by dividing the grid active power, ambient temperature, and mode component data according to various sub-historical time periods, sub-feature sets with the same time information are obtained. For each sub-feature set, statistical calculations are performed on the active power, ambient temperature, and mode component corresponding to the corresponding sub-feature set, thereby obtaining aggregated features containing time information, power information, temperature information, and mode component information. In this way, based on the aggregated features, multi-dimensional information can be intuitively and accurately reflected. By performing multiple translation operations on each mode component data, sub-time features corresponding to the number of translations are obtained, and the sub-time features are spliced ​​and combined to obtain time features, thereby further enhancing the features of the time dimension. The aggregated features and time features are spliced ​​and combined to obtain a feature subset. In this way, by using aggregated features with multi-dimensional information and time features that strengthen the time dimension, the effectiveness of the feature subset can be greatly improved, which is conducive to improving the accuracy of wind speed prediction for wind turbines.

[0101] In one embodiment, the feature cross processing is performed based on each combination feature in the feature subset to obtain a cross feature, including: for the current combination feature in the feature subset, the other combination features in the feature subset except the current combination feature are compared with the current combination feature to obtain a sub-cross feature corresponding to the current combination feature; based on the sub-cross features corresponding to each combination feature in the feature subset, the cross feature is determined.

[0102] Specifically, the computer device obtains a feature subset consisting of combined features, and for a current combined feature in the feature subset, divides the current combined feature by each of the other combined features in the feature subset except the current combined feature, to obtain multiple sub-cross processing results corresponding to the current combined feature. The computer device filters each sub-cross processing result to obtain a sub-cross feature. The computer device determines a cross feature based on the multiple sub-cross features corresponding to each combined feature.

[0103] In one embodiment, Figure 6 As shown, the computer device obtains a feature subset consisting of combined features, wherein the feature subset is determined based on the aggregated features and the time features. The computer device initializes the outer loop number i and the inner loop number j, and the number of features N in the feature subset fea(i.e. N in the figure) is 10, and each combination feature is numbered, i.e. 1 to 10. Under the current outer loop number, the current inner loop number is obtained by initializing the inner loop number, and it is judged whether the current outer loop number is equal to the current inner loop number. If the current outer loop number is not equal to the inner loop number, then based on the current outer loop number, the current inner loop number, and the sequence number of each combination feature, the first combination feature corresponding to the current outer loop number and the second combination feature corresponding to the current inner loop number are determined. The first feature is divided by the second combination feature to obtain the sub-cross processing result corresponding to the current inner loop number. If the current inner loop number is less than the feature quantity, the current inner loop number is updated to obtain the next inner loop number, and the next inner loop number is used as the current inner loop number corresponding to the next inner loop iteration, and the next inner loop iteration is entered. The step of returning to the judgment whether the current outer loop number is equal to the current inner loop number continues to execute until the current inner loop number is equal to the feature quantity. If the current inner loop number is equal to the feature quantity, the sub-cross processing result obtained through multiple inner loop iterations is obtained under the current outer loop number. Determine whether the current number of outer loops is less than the number of features. If the current number of outer loops is less than the number of features, update the current number of outer loops to obtain the next number of outer loops, use the next number of outer loops as the current number of outer loops corresponding to the next outer loop iteration, enter the next outer loop iteration, and return to the step of obtaining the current number of inner loops by initializing the number of inner loops under the current number of outer loops and continue to execute until the current number of outer loops is equal to the number of features. If the current number of outer loops is equal to the number of features, obtain the sub-cross processing results obtained after multiple outer loop iterations, and determine the cross feature through each sub-cross processing result. The computer device filters the sub-cross processing results that represent missing values ​​or infinite quantities in each sub-cross processing result to obtain a sub-cross feature. The computer device determines the cross feature based on the multiple sub-cross features corresponding to each combined feature.

[0104] In this embodiment, for the current combined feature in the feature subset, all other combined features in the feature subset, excluding the current combined feature, are compared with the current combined feature to obtain a sub-cross feature corresponding to the current combined feature. A cross feature is then determined based on the sub-cross features corresponding to each combined feature in the feature subset. This cross feature processing can generate highly reliable cross features, effectively expanding the number of features used for wind speed prediction.

[0105] In one embodiment, Figure 7 As shown, the steps for determining the prediction model include:

[0106] Step S702: Obtain a sample wind speed sequence corresponding to the wind turbine, and determine the number of sample shifts based on a target time period and a data acquisition frequency; wherein the sample wind speed sequence includes sample wind speeds at each sample moment, and the target time period is used to represent the prediction time period of the prediction model.

[0107] Specifically, the computer device obtains sample monitoring data from the wind turbine, extracts a sample risk sequence from the sample monitoring data, and determines the sample shift based on the target time period and data collection frequency. For example, if the target time period is to predict the wind speed for 10 minutes and the data collection frequency is every 30 seconds, the sample shift is 20.

[0108] Step S704: for each sample moment in the sample wind speed sequence, a sample prediction value corresponding to the corresponding sample moment is determined based on the sample translation amount, and a sample prediction sequence is determined based on the multiple sample prediction values.

[0109] Specifically, for each sample moment in the sample wind speed sequence, the computer device determines the translation moment corresponding to each sample moment based on the sample translation amount; wherein each translation moment is within the sample time period corresponding to the sample wind speed sequence. The computer device determines the sample wind speed corresponding to each sample moment based on the sample wind speed sequence. For each sample moment, the computer device uses the sample wind speed at the translation moment corresponding to the corresponding sample moment as the sample prediction value corresponding to the corresponding sample moment. The computer device determines a sample prediction sequence based on the multiple sample prediction values.

[0110] For example, Figure 8 As shown, 0 to 39 in the sample wind speed sequence (i.e., the original wind speed sequence in the figure) are all represented as sample wind speeds corresponding to 40 sample moments. Among them, the target time period is to predict the wind speed for 10 minutes, and the data collection frequency is once every 30 seconds, then the computer device determines that the sample shift amount is 20. The computer device uses the 20 sample moments after the 20th as shift moments, for example, shift moments 1 to 20. Among them, sample moment 0 corresponds to shift moment 1, and so on. The computer device uses the sample wind speed corresponding to shift moment 1 (i.e., the sample moment is 20) as the predicted value for sample moment 0, and the sample wind speed corresponding to shift moment 2 as the predicted value for sample moment 1. The computer device uses each sample prediction value as a label to form a sample prediction sequence. Among them, the construction process of the sample prediction sequence is as follows. Figure 9 As shown in FIG. , the sample prediction value at each translation moment in the sample prediction sequence corresponds one-to-one to the sample wind speed at the sample moment in the sample wind speed sequence.

[0111] Step S706 : constructing an initial prediction model based on the model parameters, and training and testing the initial prediction model using the sample wind speed sequence and the sample prediction sequence to determine a prediction model for predicting the wind speed of the wind turbine.

[0112] Specifically, an initial prediction model is constructed based on the model parameters. The computer device uses the sample wind speed sequence and the sample prediction sequence as a model feature set, where the model feature set contains the sample prediction value and sample wind speed corresponding to each sample moment. The computer device uses the model feature set to train and test the initial prediction model to determine the prediction model used to predict the wind speed of the wind turbine.

[0113] For example, Figure 10 As shown, the computer device will randomly select 80% of the model feature set as the training set and 20% as the test set. Construct a catboost model, set the learning rate to α, and the L2 regularization coefficient to reg l2 , the maximum depth of each tree is D tree , the maximum number of spanning trees is M tree In this example, α=0.02,reg l2 =3,D tree =8,M tree = 1000. The computer device trains the catboost model using the training set to obtain a trained model. The computer device tests the trained model using the test set to obtain a test prediction value for each test sample. The corresponding label in the test set is recorded as y, and the mean square error (RMS) is used as the performance evaluation indicator. The specific calculation is as follows:

[0114]

[0115] Among them, m is the number of test sets, y i is the label corresponding to the i-th test sample in the test set, is the test prediction value of the i-th test sample in the test set. In this embodiment, the comparison results between the predicted value and the true value are as follows Figure 11 As shown in the figure, due to the large number of samples, only the results of the first 200 samples are shown.

[0116] In this embodiment, by determining the number of sample shifts, the sample prediction value corresponding to each sample moment in the sample wind speed sequence can be determined, and the sample prediction sequence can be determined based on the sample prediction value, thereby establishing a mapping relationship between the sample wind speed and the sample prediction value. In this way, by training and testing the prediction model based on this mapping relationship, a highly accurate prediction model can be obtained, greatly improving the effectiveness of the prediction model.

[0117] In one embodiment, Figure 12 As shown, the computer device constructs a sample prediction sequence (i.e., a label) through historical data. The computer device determines the sample wind speed sequence through historical data, and based on the sample wind speed sequence, the computer device performs modal decomposition of the wind speed sequence through ensemble empirical mode decomposition to obtain sample mode component data. The computer device obtains the sample active power and sample ambient temperature corresponding to the sample wind speed sequence. The computer device combines the sample active power, sample ambient temperature, and sample mode component data to obtain aggregate features, and performs time shift processing on the sample mode component to obtain sample time features. The approximate environment device splices and combines the sample aggregate features and the sample time features to obtain a sample feature subset, thereby completing the feature extraction process. The computer device performs feature cross-processing based on the combined features of each sample in the sample feature subset to obtain a sample cross-feature, and based on the sample cross-feature and the sample feature subset, trains the prediction model in combination with the sample prediction sequence to obtain a trained prediction model.

[0118] The computer device performs collective empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period (i.e., real-time monitoring data). The active power of the power grid and the ambient temperature corresponding to the historical time period are obtained, and the active power of the power grid, the ambient temperature, and the mode component data are combined to obtain aggregate features. The mode component data is time-shifted to obtain time features, and the aggregate features and the time features are spliced ​​and combined to obtain a feature subset. Feature cross-processing is performed based on each combined feature in the feature subset to obtain a cross feature. Wind speed prediction is performed on the cross feature and the feature subset through a prediction model to obtain a predicted wind speed value of the wind turbine in the target time period.

[0119] In this embodiment, ensemble empirical mode decomposition (EEMD) is performed on the wind speed sequence to obtain mode component data corresponding to each mode component. This process, based on EEMD, effectively avoids modal chaos and ensures the validity of the mode component data. The grid active power and ambient temperature corresponding to the historical time period are obtained and combined with the mode component data to generate aggregated features. This aggregated feature captures multi-dimensional information and significantly enhances the validity of the aggregated features. The mode component data is time-shifted to generate temporal features, which are then concatenated with the temporal features to form a feature subset. Feature cross-processing is performed on each combined feature in the feature subset to generate cross-features, effectively expanding the number of features used for wind speed prediction. Wind speed prediction is performed using the cross-features and the feature subset using a prediction model to obtain a predicted wind speed value for the wind turbine within the target time period. This, based on a sufficient number of highly effective cross-features and feature subsets, significantly improves the quality of features used for wind speed prediction. Therefore, based on the cross-features and feature subsets with high feature quality, a high-precision wind speed prediction value can be obtained through the prediction model, that is, the accuracy of wind speed prediction for wind turbines is greatly increased.

[0120] In order to facilitate a clearer understanding of the technical solution of this application, a more detailed embodiment is provided for explanation. Figure 13 As shown, the computer device obtains the active power, external temperature and original wind speed sequence of the transformer grid side from the real-time monitoring data, wherein the original wind speed sequence contains the wind speed of the wind turbine at each historical moment in the historical time period. The computer device combines the original wind speed sequence with white noise to obtain the wind speed sequence to be processed. The computer device performs collective empirical mode decomposition on the wind speed sequence to be processed to obtain the mode component data corresponding to each modal component. Specifically, initialize l = 1, r l (t) = X(t), and set the iteration stop threshold s, s takes a value between 0.2 and 0.3, and initialize j = 1, h j-1 (t) = r0(t). The computer equipment finds h j-1 All local maximum and minimum points in (t) are fitted with cubic spline interpolation for all maximum points to obtain the upper envelope; all local minimum points are fitted with cubic spline interpolation for all minimum points to obtain the lower envelope. The computer equipment calculates the mean m of the upper and lower envelopes j-1 (t). The computer equipment performs standard deviation S on the mean d Calculation, that is:

[0121]

[0122] Among them, T is the number of sequence h0(t) (i.e. the number of sample points), h j (t) = h j-1 (t)-m j-1 (t). Computer equipment determines S d Is it less than the threshold s? If so, the calculation of the lth IMF component is completed. l (t) = h j (t). If not, set j = j + 1 and return to find h j-1 The steps of all local maximum and minimum points in (t) are continued until it is less than the threshold s. The computer device passes r0(t)=r0(t)-IMF l (t) Update the sequence and determine whether r0(t) contains more than two extreme values. If so, set p = p + 1 and return to the step of initializing l = 1 to continue execution. If not, determine that the ensemble empirical mode decomposition is completed and obtain p IMF components. The computer device repeatedly performs ensemble empirical mode decomposition on the wind speed sequence to be processed M times to obtain M groups of model wind volume vectors (i.e., IMF components). Let the jth IMF component obtained by the mth empirical mode decomposition be C j,m The computer is set to average the IMF components corresponding to each stage obtained by M-times empirical mode decomposition. The final mode component data (i.e., multiple IMF components, i.e., IMF component 1, IMF component 2, ..., IMF component p) are obtained. Among them, the mode component data IMF of the kth IMF component is k The calculation is as follows:

[0123]

[0124] The computer device groups and aggregates the grid active power, ambient temperature, and mode component data corresponding to the historical time period by hour, i.e., groups data with the same year, month, and day averages into one group, to obtain multiple sub-feature sets. For each sub-feature set, multiple statistical calculations are performed on the individual grid active powers corresponding to the corresponding sub-feature set to obtain multiple power statistics corresponding to each sub-feature set. For each sub-feature set, multiple statistical calculations are performed on the individual ambient temperatures corresponding to the corresponding sub-feature set to obtain multiple temperature statistics corresponding to each sub-feature set. For each sub-feature set, multiple statistical calculations are performed on the individual mode component data corresponding to the corresponding sub-feature set to obtain multiple mode component statistics corresponding to each sub-feature set; the statistical calculations include at least one of median calculation, mean calculation, standard deviation calculation, maximum calculation, minimum calculation, and range calculation. For each sub-feature set, the power statistics, temperature statistics, and mode component statistics corresponding to the corresponding sub-feature set are used as sub-aggregate features corresponding to the corresponding sub-feature set. The sub-aggregate features corresponding to each sub-feature set are combined to obtain an aggregate feature. The computer device shifts multiple pattern component data (i.e., multiple IMF classifications) backward by T, 2T, …, 10T time units, respectively. Each pattern component data then generates 10 sub-time features (i.e., historical time features), where T represents the temporal resolution of sample acquisition. The computer device concatenates the aggregated features with the time features to obtain a feature subset to be processed, where the feature subset to be processed contains multiple combined features. The computer device removes missing values ​​from the feature subset to obtain a processed feature subset.

[0125] The computer device initializes the outer loop number i and the inner loop number j, and the number of features of the feature subset N fea , and number each combination feature, i.e. 1 to 10. Under the current outer loop number, the current inner loop number is obtained by initializing the inner loop number, and it is determined whether the current outer loop number is equal to the current inner loop number. If the current outer loop number is not equal to the inner loop number, then based on the current outer loop number, the current inner loop number, and the sequence number of each combination feature, the first combination feature fea corresponding to the current outer loop number is determined. i , and the second combined feature fea corresponding to the current number of inner loops j Divide the first feature by the second combined feature to obtain the sub-cross processing result Fea corresponding to the current inner loop number p. If the current number of inner loops is less than the number of features, the current number of inner loops is incremented by one to obtain the next number of inner loops, and the next number of inner loops is used as the current number of inner loops corresponding to the next inner loop iteration, and the next inner loop iteration is entered, and the step of determining whether the current number of outer loops is equal to the current number of inner loops is returned to continue the execution until the current number of inner loops is equal to the number of features. If the current number of inner loops is equal to the number of features, the sub-cross processing results obtained after multiple inner loop iterations under the current number of outer loops are obtained. Determine whether the current number of outer loops is less than the number of features. If the current number of outer loops is less than the number of features, the current number of outer loops is incremented by one to obtain the next number of outer loops, and the next number of outer loops is used as the current number of outer loops corresponding to the next outer loop iteration, and the next outer loop iteration is entered, and the step of obtaining the current number of inner loops by initializing the number of inner loops under the current number of outer loops is returned to continue the execution until the current number of outer loops is equal to the number of features. If the current number of outer loops is equal to the number of features, the sub-cross processing results obtained after multiple outer loop iterations are obtained, and the cross features are determined by each sub-cross processing result. The computer device filters out missing values ​​or infinite sub-cross processing results from each sub-cross processing result to obtain sub-cross features. The computer device determines a cross feature based on the multiple sub-cross features corresponding to each combined feature. The computer device uses the cross feature and the feature subset as a prediction feature. The computer device inputs the prediction feature into a prediction model to obtain a predicted wind speed value for the wind turbine within a target time period.

[0126] In this embodiment, ensemble empirical mode decomposition (EEMD) is performed on a wind speed sequence to be processed, obtaining mode component data corresponding to each mode component. The wind speed sequence to be processed is obtained by collecting wind speed data from wind turbines during a historical time period. Processing the wind speed sequence based on EEMD effectively avoids modal chaos and ensures the validity of the mode component data. The grid active power and ambient temperature corresponding to the historical time period are obtained, and the grid active power, ambient temperature, and mode component data are combined to obtain aggregated features. The aggregated features obtained by combining information on the three variables of power, temperature, and wind speed reflect multi-dimensional information, significantly enhancing the validity of the aggregated features. The mode component data is time-shifted to obtain time features, and the aggregated features are concatenated with the time features to obtain a feature subset. Feature cross-processing is performed on each combined feature in the feature subset to obtain cross-features, thereby effectively expanding the number of features used for wind speed prediction. Wind speed prediction is performed using the cross-features and the feature subset using a prediction model to obtain a predicted wind speed value for the wind turbine during the target time period. In this way, based on a sufficient number of highly effective cross-features and feature subsets, the quality of features used for wind speed prediction is greatly improved. Therefore, based on high-quality cross-features and feature subsets, a multi-step wind speed prediction is directly performed through the prediction model to obtain a highly accurate wind speed prediction value, which greatly increases the accuracy of wind turbine wind speed prediction. Furthermore, based on this prediction value, the power grid can be assisted in grid scheduling and grid frequency regulation, ensuring the stability of the power grid and power supply instructions. Furthermore, it can also help the wind turbine control system to perform real-time control adjustments on the wind turbine based on the wind speed prediction value, ensuring that the wind turbine does not experience excessive instantaneous loads, which can easily cause damage to the wind turbine.

[0127] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0128] Based on the same inventive concept, embodiments of the present application also provide a device for predicting wind turbine wind speed for implementing the aforementioned method for predicting wind turbine wind speed. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the device for predicting wind turbine wind speed provided below can be found in the aforementioned definition of the method for predicting wind turbine wind speed, and will not be further elaborated here.

[0129] In one embodiment, Figure 14 As shown, a device for predicting wind speed of a wind turbine is provided, comprising: a decomposition module 1402, a combination module 1404, a translation module 1406, a cross module 1408 and a prediction module 1410, wherein:

[0130] The decomposition module 1402 is used to perform collective empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period.

[0131] The combining module 1404 is configured to obtain the grid active power and the ambient temperature corresponding to the historical time period, and combine the grid active power, the ambient temperature, and the mode component data to obtain an aggregated feature.

[0132] The translation module 1406 is used to perform time translation processing on the pattern component data to obtain a time feature, and to combine the aggregate feature and the time feature to obtain a feature subset.

[0133] The cross module 1408 is configured to perform feature cross processing based on each combined feature in the feature subset to obtain cross features.

[0134] The prediction module 1410 is configured to perform wind speed prediction on the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within a target time period.

[0135] In one embodiment, the decomposition module 1402 is configured to perform multiple empirical mode decompositions on the wind speed sequence to be processed to obtain multiple groups of mode component vectors; each group of mode component vectors includes multiple mode component values. For each mode component, the corresponding mode component value is obtained from the multiple groups of mode component vectors, and the mean of the multiple mode component values ​​corresponding to the corresponding mode component is calculated to obtain the mean corresponding to the corresponding mode component. The mean corresponding to the mode component is used as the mode component data corresponding to the corresponding mode component to obtain multiple mode component data.

[0136] In one embodiment, the combining module 1404 is configured to divide the historical time period into multiple sub-historical time periods, and to divide the grid active power, the ambient temperature, and the mode component data corresponding to the historical time period into each sub-historical time period to obtain sub-feature sets. Each sub-feature set corresponds to a sub-historical time period, and each sub-feature set includes the grid active power, ambient temperature, and mode component data corresponding to the corresponding sub-historical time period. For each sub-feature set, multiple statistical calculations are performed on the grid active power corresponding to the sub-feature set to obtain multiple power statistics corresponding to each sub-feature set. For each sub-feature set, multiple statistical calculations are performed on the ambient temperature corresponding to the sub-feature set to obtain multiple temperature statistics corresponding to each sub-feature set. For each sub-feature set, multiple statistical calculations are performed on the mode component data corresponding to the sub-feature set to obtain multiple mode component statistics corresponding to each sub-feature set. The statistical calculations include at least one of median calculation, mean calculation, standard deviation calculation, maximum calculation, minimum calculation, and range calculation. For each sub-feature set, the power statistics, temperature statistics, and mode component statistics corresponding to the corresponding sub-feature set are used as the sub-aggregate features corresponding to the corresponding sub-feature set. The sub-aggregate features corresponding to each sub-feature set are combined to obtain the aggregate feature.

[0137] In one embodiment, the translation module 1406 is configured to increase the translation amount based on the translation amount of the previous translation to obtain the translation amount corresponding to the current translation. The pattern component data is translated according to the translation amount corresponding to the current translation to obtain each sub-time feature corresponding to the current translation. The translation amount is increased based on the translation amount of the current translation to continue the next translation operation until a preset number of translations is reached, thereby obtaining each sub-time feature corresponding to each translation. The sub-time features obtained by multiple translation operations are combined to obtain a time feature.

[0138] In one embodiment, the intersection module 1408 is configured to compare the current combined feature in the feature subset with the other combined features in the feature subset except the current combined feature to obtain a sub-intersection feature corresponding to the current combined feature. The intersection feature is determined based on the sub-intersection features corresponding to each combined feature in the feature subset.

[0139] In one embodiment, the prediction module 1410 is used to obtain a sample wind speed sequence corresponding to a wind turbine and determine the number of sample shifts based on a target time period and a data acquisition frequency; wherein the sample wind speed sequence includes sample wind speeds at each sample moment, and the target time period is used to represent the prediction time period of the prediction model. For each sample moment in the sample wind speed sequence, a sample prediction value corresponding to the corresponding sample moment is determined based on the number of sample shifts, and a sample prediction sequence is determined based on the multiple sample prediction values. An initial prediction model is constructed based on the model parameters, and the initial prediction model is trained and tested using the sample wind speed sequence and the sample prediction sequence to determine a prediction model for predicting the wind speed of the wind turbine.

[0140] Each module in the aforementioned wind turbine wind speed prediction device may be implemented in whole or in part via software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor within a computer device in the form of hardware, or may be stored in a memory within the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0141] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 15 As shown. The computer device includes a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store predicted data on wind turbine wind speed. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for predicting wind turbine wind speed is implemented.

[0142] Those skilled in the art will understand that Figure 15 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0143] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0144] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0145] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0146] 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, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0147] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may 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 may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0148] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, 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.

[0149] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for predicting wind speed of a wind turbine, characterized in that: The method comprises: Performing ensemble empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine in a historical time period; Obtaining grid active power and ambient temperature corresponding to the historical time period, and dividing the grid active power, ambient temperature, and mode component data based on time information of the historical time period to obtain sub-feature sets; the time information of each sub-feature set is different; Perform statistical calculations on each sub-feature set to obtain the statistical results of the corresponding sub-feature set, and combine the statistical results corresponding to each sub-feature set to obtain the aggregated feature; Performing time translation processing on the pattern component data to obtain time features, and combining the aggregate features and the time features to obtain a feature subset; For a current combination feature in the feature subset, performing division calculation on the current combination feature and other combination features in the feature subset except the current combination feature, to obtain a sub-cross processing result corresponding to the current combination feature; Filtering each sub-cross processing result to obtain a sub-cross feature, and determining a cross feature based on multiple sub-cross features corresponding to each combined feature; The wind speed is predicted for the cross feature and the feature subset using a prediction model to obtain a predicted wind speed value of the wind turbine within a target time period.

2. The method according to claim 1, characterized in that The process of performing ensemble empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component includes: Perform multiple empirical mode decompositions on the wind speed sequence to be processed to obtain multiple groups of mode component vectors; each group of mode component vectors contains the values ​​of multiple mode components; For each mode component, the value of the corresponding mode component is obtained from multiple groups of mode component vectors, and the values ​​of multiple mode components corresponding to the corresponding mode component are averaged to obtain the mean value corresponding to the corresponding mode component; By taking the mean values ​​corresponding to the mode components as the mode component data corresponding to the corresponding mode components, a plurality of mode component data are obtained.

3. The method according to claim 1, characterized in that The specific steps of determining the aggregation features include: Dividing the historical time period into a plurality of sub-historical time periods, and dividing the grid active power, the ambient temperature, and the mode component data corresponding to the historical time period according to each sub-historical time period to obtain each sub-feature set; wherein each sub-feature set corresponds to a sub-historical time period, and each sub-feature set includes the grid active power, the ambient temperature, and the mode component data corresponding to the corresponding sub-historical time period; For each sub-feature set, multiple statistical calculations are performed on the active power of each power grid corresponding to the corresponding sub-feature set to obtain multiple power statistical results corresponding to each sub-feature set; For each sub-feature set, performing multiple statistical calculations on each ambient temperature corresponding to the corresponding sub-feature set to obtain multiple temperature statistical results corresponding to each sub-feature set; For each sub-feature set, performing multiple statistical calculations on each pattern component data corresponding to the corresponding sub-feature set to obtain multiple pattern component statistical results corresponding to each sub-feature set; wherein the statistical calculations include at least one of median calculation, mean calculation, standard deviation calculation, maximum calculation, minimum calculation, and range calculation; For each sub-feature set, the power statistics, temperature statistics, and mode component statistics corresponding to the corresponding sub-feature set are used as sub-aggregate features corresponding to the corresponding sub-feature set; The sub-aggregate features corresponding to each sub-feature set are combined to obtain the aggregate feature.

4. The method according to claim 1, wherein The performing time shift processing on the pattern component data to obtain a time feature includes: The translation amount is increased based on the translation amount of the previous translation to obtain the translation amount corresponding to the current translation; Shifting the pattern component data according to the shift amount corresponding to the current shift to obtain each sub-time feature corresponding to the current shift; The translation amount is increased based on the translation amount of the current translation to continue the next translation operation until the preset number of translations is reached and the sub-time features corresponding to each translation are obtained; The sub-time features obtained by multiple translation operations are combined to obtain the time feature.

5. The method according to any one of claims 1 to 4, characterized in that The step of determining the prediction model includes: Obtain a sample wind speed sequence corresponding to the wind turbine, and determine the number of sample shifts based on a target time period and a data acquisition frequency; wherein the sample wind speed sequence includes sample wind speeds at each sample moment, and the target time period is used to represent the prediction time period of the prediction model; For each sample moment in the sample wind speed sequence, a sample prediction value corresponding to the corresponding sample moment is determined based on the sample translation amount, and a sample prediction sequence is determined based on the multiple sample prediction values; An initial prediction model is constructed based on the model parameters, and the initial prediction model is trained and tested using the sample wind speed sequence and the sample prediction sequence to determine a prediction model for predicting the wind speed of the wind turbine.

6. A wind speed prediction device for a wind turbine, characterized in that: The device comprises: A decomposition module is used to perform collective empirical mode decomposition on the wind speed sequence to be processed to obtain mode component data corresponding to each mode component; wherein the wind speed sequence to be processed is obtained by collecting the wind speed of the wind turbine within a historical time period; a combination module for obtaining the grid active power and ambient temperature corresponding to the historical time period, and dividing the grid active power, ambient temperature, and mode component data based on the time information of the historical time period to obtain sub-feature sets; the time information of each sub-feature set is different; performing statistical calculations on each sub-feature set to obtain statistical results of the corresponding sub-feature set, and splicing and combining the statistical results corresponding to each sub-feature set to obtain an aggregate feature; A translation module, configured to perform time translation processing on the pattern component data to obtain a time feature, and to combine the aggregate feature and the time feature to obtain a feature subset; a cross module, configured to perform division calculations on a current combined feature in the feature subset by each of the other combined features in the feature subset except the current combined feature, to obtain a sub-cross processing result corresponding to the current combined feature; filter each sub-cross processing result to obtain a sub-cross feature, and determine a cross feature based on the multiple sub-cross features corresponding to each combined feature; The prediction module is used to perform wind speed prediction on the cross-features and the feature subsets through a prediction model to obtain a wind speed prediction value of the wind turbine within a target time period.

7. The device according to claim 6, characterized in that The decomposition module is used to perform multiple empirical mode decompositions on the wind speed sequence to be processed to obtain multiple groups of mode component vectors; wherein each group of mode component vectors contains multiple mode component values; for each mode component, the value of the corresponding mode component is obtained from the multiple groups of mode component vectors, and the mean of the values ​​of the multiple mode components corresponding to the corresponding mode component is calculated to obtain the mean corresponding to the corresponding mode component; by using the mean corresponding to the mode component as the mode component data corresponding to the corresponding mode component, multiple mode component data are obtained.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images