Model training method and device for wind power climbing scene and wind speed prediction method and device

Abstract

The invention provides a wind power climbing scene-oriented model training method and device, and a wind speed prediction method and device, and belongs to the technical field of wind power. The model training method comprises the following steps: decomposing each piece of first wind speed sequence data in a first wind speed data set into a plurality of target eigencomponents; for any target eigencomponent, determining a target meteorological feature data set corresponding to the target eigencomponent according to a first meteorological data set corresponding to the first wind speed data set; constructing training data according to the first wind speed data set and the plurality of target meteorological feature data sets; the initial wind speed prediction model is trained based on the training data, a trained wind speed prediction model is obtained, the wind speed prediction model is used for obtaining wind speed prediction sequence data, and the climbing event is recognized based on the wind speed prediction sequence data. According to the embodiment of the invention, the model training effect can be improved, and the wind speed prediction model with high prediction accuracy can be obtained.

Description

Technical Field

[0001] The present disclosure relates to the field of wind power technology, and in particular to a model training method, a wind speed prediction method and a device for wind power ramping scenarios. Background Art

[0002] As a renewable energy source, wind energy boasts the advantages of being green, clean, and highly efficient. When utilizing wind energy for power generation, wind turbines convert wind energy into mechanical energy when wind speeds reach a certain level, which in turn drives generators to generate electricity. However, wind power generation is highly susceptible to weather factors, such as wind speed. Weather factors are inherently random and variable, which can impact the quality of power delivered by the power grid. Therefore, accurate wind speed forecasting is crucial for the development of wind power generation. Summary of the Invention

[0003] The present disclosure provides a model training method, a wind speed prediction method and a device for wind power ramping scenarios.

[0004] In a first aspect, the present disclosure provides a model training method for a wind power ramping scenario, the model training method comprising: decomposing each first wind speed sequence data in a first wind speed data set into a plurality of target intrinsic components; for any target intrinsic component, determining a target meteorological feature data set corresponding to the target intrinsic component based on a first meteorological data set corresponding to the first wind speed data set, wherein the first meteorological data set includes a plurality of meteorological feature data, and the target meteorological feature data set includes at least one meteorological feature data screened from the first meteorological data set; constructing training data based on the first wind speed data set and the plurality of target meteorological feature data sets; training an initial wind speed prediction model based on the training data, and obtaining to a trained wind speed prediction model, the wind speed prediction model is used to obtain wind speed prediction sequence data to identify climbing events based on the wind speed prediction sequence data; wherein, for any target meteorological feature data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and one predicted bias component corresponding to the target meteorological feature data set based on the target meteorological feature data set, obtain second wind speed sequence data corresponding to the target meteorological feature data set based on the multiple predicted intrinsic components and one predicted bias component, and adjust the parameters of the wind speed prediction model according to the multiple second wind speed sequence data corresponding to the multiple target meteorological feature data sets and the multiple first wind speed sequence data in the first wind speed data set.

[0005] In the second aspect, the present disclosure provides a wind speed prediction method for a wind power ramping scenario, which includes: obtaining a second meteorological data set for a first preset time period, wherein the second meteorological data set includes multiple meteorological feature data; inputting the second meteorological data set for the first preset time period into a wind speed prediction model to obtain wind speed prediction sequence data for the second preset time period; wherein the wind speed prediction model is obtained using a model training method for a wind power ramping scenario as described in any one of the embodiments of the present disclosure.

[0006] In a third aspect, the present disclosure provides a model training device for wind power ramping scenarios, the model training device comprising: a decomposition module for decomposing each first wind speed sequence data in a first wind speed data set into a plurality of target intrinsic components; a determination module for determining, for any target intrinsic component, a target meteorological feature data set corresponding to the target intrinsic component based on a first meteorological data set corresponding to the first wind speed data set, wherein the first meteorological data set includes a plurality of meteorological feature data, and the target meteorological feature data set includes at least one meteorological feature data screened from the first meteorological data set; a construction module for constructing training data based on the first wind speed data set and a plurality of the target meteorological feature data sets; a training module for determining the target meteorological feature data set corresponding to the target intrinsic component based on the training data; The wind speed prediction model is trained to obtain a trained wind speed prediction model, and the wind speed prediction model is used to obtain wind speed prediction sequence data to identify climbing events based on the wind speed prediction sequence data; wherein, for any target meteorological feature data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and one predicted bias component corresponding to the target meteorological feature data set according to the target meteorological feature data set, obtain second wind speed sequence data corresponding to the target meteorological feature data set based on the multiple predicted intrinsic components and one predicted bias component, and adjust the parameters of the wind speed prediction model according to the multiple second wind speed sequence data corresponding to the multiple target meteorological feature data sets and the multiple first wind speed sequence data in the first wind speed data set.

[0007] In a fourth aspect, the present disclosure provides a wind speed prediction device for a wind power ramping scenario, the wind speed prediction device comprising: an acquisition module for acquiring a second meteorological data set of a first preset time period, wherein the second meteorological data set includes a plurality of meteorological feature data; a prediction module for inputting the second meteorological data set of the first preset time period into a wind speed prediction model to obtain wind speed prediction sequence data of the second preset time period; wherein the wind speed prediction model is obtained using a model training method for a wind power ramping scenario as described in any one of the embodiments of the present disclosure.

[0008] The embodiment provided by the present disclosure decomposes each first wind speed sequence data in a first wind speed data set into multiple target intrinsic components; for any target intrinsic component, a target meteorological feature data set corresponding to the target intrinsic component is determined based on a first meteorological data set corresponding to the first wind speed data set, wherein the first meteorological data set includes multiple meteorological feature data, and the target meteorological feature data set includes at least one meteorological feature data screened from the first meteorological data set; training data is constructed based on the first wind speed data set and the multiple target meteorological feature data sets; an initial wind speed prediction model is trained based on the training data to obtain a trained wind speed prediction model, and the wind speed prediction model is used to obtain wind speed prediction sequence data so as to identify climbing events based on the wind speed prediction sequence data.

[0009] It can be seen that in the embodiment of the present disclosure, by decomposing the first wind speed sequence data, a target intrinsic component with stronger regularity and better characterization of the essence of wind speed can be obtained. For any target intrinsic component, the part of the meteorological characteristic data that has a greater impact on the target intrinsic component is screened out from multiple meteorological characteristic data to obtain a target meteorological characteristic data set corresponding to the target intrinsic component. On this basis, training data is constructed based on the first wind speed data set and multiple target meteorological characteristic data sets, and the wind speed prediction model is trained using the training data. Among them, for any target meteorological characteristic data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and a predicted bias component corresponding to the target meteorological characteristic data set based on the target meteorological characteristic data set, obtain second wind speed sequence data corresponding to the target meteorological characteristic data set based on the multiple predicted intrinsic components and the predicted bias component, and adjust the parameters of the wind speed prediction model based on the multiple second wind speed sequence data corresponding to the multiple target meteorological characteristic data sets and the multiple first wind speed sequence data in the first wind speed data set. Based on this, the model training effect can be improved, and a wind speed prediction model with higher prediction accuracy can be obtained.

[0010] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. The above and other features and advantages will become more apparent to those skilled in the art by describing detailed example embodiments with reference to the accompanying drawings. In the accompanying drawings:

[0012] Figure 1 A flowchart of a model training method for wind power ramping scenarios provided in an embodiment of the present disclosure.

[0013] Figure 2 A schematic diagram of a wind farm provided in an embodiment of the present disclosure.

[0014] Figure 3 A schematic diagram of sorting the importance assessment results of meteorological characteristic data provided in an embodiment of the present disclosure.

[0015] Figure 4 A schematic diagram of sorting the importance assessment results of meteorological characteristic data provided in an embodiment of the present disclosure.

[0016] Figure 5 A schematic diagram of the processing process of a model training method for wind power ramping scenarios provided in an embodiment of the present disclosure.

[0017] Figure 6 A flow chart of a wind speed prediction method for wind power ramping scenarios provided by an embodiment of the present disclosure.

[0018] Figure 7 A schematic diagram of a wind speed prediction method for wind power ramping scenarios provided by an embodiment of the present disclosure.

[0019] Figure 8 A schematic diagram of a model training device for wind power ramping scenarios provided by an embodiment of the present disclosure.

[0020] Figure 9 A schematic diagram of a wind speed prediction device for wind power ramping scenarios provided by an embodiment of the present disclosure.

[0021] Figure 10 A block diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0022] To enable those skilled in the art to better understand the technical solutions of the present disclosure, exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0023] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0024] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0025] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded. Similar words such as "connected" or "connected" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0026] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless expressly defined as such herein.

[0027] Wind power is intermittent, random, and uncertain, which can lead to frequent, short-term, and dramatic changes in wind power output, known as power ramp events (referred to as ramp events). The large-scale integration of wind power has brought increasingly significant challenges to the safe and stable operation, scheduling, and real-time control of the power grid. Therefore, high-precision and timely wind speed prediction technology has become a core requirement for wind power integration and grid resilience improvement.

[0028] Furthermore, during the climbing process, the difficulty in wind speed prediction mainly stems from the fact that the wind speed sequence is affected by the coupling of multi-dimensional environmental factors such as terrain, temperature, and air pressure, and has significant non-stationary and nonlinear characteristics. The relevant technologies are less sensitive to the high-frequency mutation components in extreme events, and it is difficult to distinguish between noise and effective signals. In addition, the climbing event response time window is short, and balancing long-term trend prediction and short-term detail capture is also one of the prediction difficulties.

[0029] In view of this, the embodiments of the present disclosure provide a model training method for wind power ramping scenarios, a wind speed prediction method for wind power ramping scenarios, a model training device for wind power ramping scenarios, a wind speed prediction device for wind power ramping scenarios, an electronic device, a computer-readable storage medium, and a computer program product.

[0030] A first aspect of the embodiments of the present disclosure provides a model training method for wind power ramping scenarios.

[0031] Figure 1 This is a flow chart of a model training method for wind power ramping scenarios provided by an embodiment of the present disclosure. Figure 1, the model training method for wind power ramping scenarios may include the following steps.

[0032] Step S101: decompose each first wind speed sequence data in the first wind speed data set into a plurality of target eigencomponents.

[0033] In some optional embodiments, the first wind speed data set includes a plurality of first wind speed sequence data, each of which is sequence data about wind speed and corresponds to a certain time period.

[0034] In other words, the first wind speed sequence data may be historical wind speed real-time sequence data, and a plurality of first wind speed sequence data may correspond to the same time period but to different locations.

[0035] Figure 2 A wind farm schematic diagram provided in an embodiment of the present disclosure. Figure 2 As shown, four grid points are set around the wind farm, namely grid point 1, grid point 2, grid point 3, and grid point 4. For example, for this wind farm, the corresponding first wind speed data set includes N first wind speed sequence data, which include wind speed data of the four grid points in the past 24 hours, and the time resolution is 15 minutes (i.e., wind speed sampling is performed every 15 minutes).

[0036] It should be noted that the above description of the first wind speed data set is merely an example, and the embodiments of the present disclosure do not limit this.

[0037] Furthermore, considering that the wind speed sequence data is a complex nonlinear and non-stationary signal, in order to better perform data processing, the first wind speed sequence data in the aforementioned first wind speed data set can be decomposed to perform corresponding data processing using the decomposed signal components.

[0038] In some optional embodiments, each first wind speed sequence data in the first wind speed data set is decomposed into multiple target intrinsic components, including: decomposing the first wind speed sequence data into multiple target intrinsic components based on a preset decomposition method; wherein the preset decomposition method includes at least one of the following: empirical mode decomposition (EMD), local mean decomposition, and wavelet decomposition.

[0039] Exemplarily, any first wind speed sequence data in the first wind speed data set can be decomposed into multiple intrinsic mode functions (IMFs) and a residual component based on the EMD decomposition method, where the multiple intrinsic mode functions are multiple target intrinsic components.

[0040] In this decomposition method, the obtained multiple target intrinsic components (i.e., IMF functions) can reflect the essential characteristics of wind speed at different time scales and frequencies. The different scales or trend components that actually exist in the first wind speed series data can be decomposed step by step to generate a series of data sequences with different characteristic scales. The decomposed sequences have stronger regularity than the original first wind speed series data, which can improve the prediction accuracy.

[0041] For example, the first wind speed sequence data I can be decomposed into:

[0042]

[0043] Where IMF is the target intrinsic component, i is the serial number of the target intrinsic component, and 1≤i≤M, M is the number of target intrinsic components, and M is an integer greater than 1; Res M is the residual component.

[0044] Exemplarily, any first wind speed sequence data in the first wind speed data set can be decomposed into several low-frequency components and several high-frequency components based on wavelet decomposition. These low-frequency components and high-frequency components are multiple target eigencomponents.

[0045] The low-frequency component retains the slowly varying, relatively smooth portions of the signal, reflecting the signal's main characteristics and outline. In other words, the low-frequency component primarily characterizes the overall trend of wind speed and provides a rough approximation of the wind speed sequence. The high-frequency component, representing the rapidly varying portions of wind speed, reflects detailed information and high-frequency components at different scales. This signal decomposition method enables analysis of wind speed at various scales.

[0046] It should be noted that the above preset decomposition method is only an example, and the embodiments of the present disclosure do not limit this.

[0047] Furthermore, in some optional embodiments, multiple decomposition methods may be combined to achieve decomposition of the first wind speed sequence data.

[0048] For example, for any first wind speed sequence data in the first wind speed data set, the first wind speed sequence data can be first subjected to a short-time Fourier transform (STFT) to obtain local information of the wind speed in time and frequency, so as to capture the transient changes and time-varying characteristics of the wind speed; on this basis, the STFT decomposition result is subjected to an EMD decomposition to obtain multiple target intrinsic components, which reflect the intrinsic time-frequency characteristics of the local information, so as to achieve more detailed analysis and processing of the wind speed sequence data.

[0049] Step S102 : for any target intrinsic component, a target meteorological characteristic dataset corresponding to the target intrinsic component is determined based on a first meteorological dataset corresponding to a first wind speed dataset.

[0050] The first meteorological data set includes a plurality of meteorological characteristic data, and the target meteorological characteristic data set includes at least one meteorological characteristic data selected from the first meteorological data set.

[0051] For example, the first meteorological data set corresponding to the first wind speed data set refers to a set of meteorological characteristic data that is temporally correlated with the first wind speed data set (e.g., corresponding to the same time period or two adjacent time periods, etc.) and corresponds to the same observation location (e.g., a grid point). The meteorological characteristic data may include temperature, relative humidity, sea level pressure, precipitation, low cloud cover, high cloud cover, zonal wind, meridional wind, wind speed, divergence, potential vorticity, atmospheric precipitable water, convective effective potential energy, etc., and the embodiments of the present disclosure are not limited thereto.

[0052] Since the first meteorological data set contains a large amount of meteorological characteristic data, and some of the meteorological characteristic data may have little effect on wind speed, this part of the data can be filtered out to reduce the amount of data processing.

[0053] In some optional embodiments, a target meteorological characteristic data set corresponding to a target intrinsic component is determined based on a first meteorological data set corresponding to a first wind speed data set, including: for any target intrinsic component, determining a first meteorological characteristic model corresponding to the target intrinsic component based on the first meteorological data set, the first meteorological characteristic model including at least one meteorological decision tree; determining an importance evaluation result of each meteorological characteristic data based on the first meteorological characteristic model; screening out at least one meteorological characteristic data from a plurality of meteorological characteristic data based on the importance evaluation result, and obtaining a target meteorological characteristic data set corresponding to the target intrinsic component based on the at least one screened meteorological characteristic data; or, determining an importance evaluation result of each meteorological characteristic data based on a preset second meteorological characteristic model, the second meteorological characteristic model being a model determined based on a third wind speed data set; screening out at least one meteorological characteristic data from a plurality of meteorological characteristic data based on the importance evaluation result, and obtaining a target meteorological characteristic data set corresponding to the target intrinsic component based on the at least one screened meteorological characteristic data.

[0054] From this, it can be seen that for each target intrinsic component, the importance evaluation results of each meteorological characteristic data can be determined based on the first meteorological characteristic model or the second meteorological characteristic model, and then a part of the meteorological characteristic data can be screened out according to the importance evaluation results to obtain the target meteorological characteristic data set corresponding to the target intrinsic component.

[0055] In other words, in this way, the local correlation characteristics between the wind speed decomposition signal and the space-time meteorological field can be mined to achieve effective extraction and utilization of meteorological characteristic data, while also reducing the amount of data processing and shortening the data processing time.

[0056] Exemplarily, for any target intrinsic component, a first meteorological feature model corresponding to the target intrinsic component can be determined based on a first meteorological data set by using the eXtreme Gradient Boosting (XGBoost) method, and the first meteorological feature model includes multiple meteorological decision trees. For example, a first meteorological feature model can be initialized, and in each round of model iteration, the residual of the current first meteorological feature model (i.e., the difference between the true value and the predicted value) is calculated, and then a new meteorological decision tree is trained with the residual as the target, and the prediction result of the new meteorological decision tree is weightedly combined with the previous model through the learning rate to update the first meteorological feature model. After multiple rounds of iterations, if the preset stopping condition is met (such as the number of iterations is greater than a preset threshold, etc.), the trained first meteorological feature model is obtained. On this basis, according to the first meteorological characteristic model, the index values ​​such as the number of splitting times and information gain of each meteorological characteristic data are determined, and the importance evaluation results of each meteorological characteristic data are determined based on these index values. Then, according to the importance evaluation results, at least one meteorological characteristic data with greater importance is screened out from multiple meteorological characteristic data, and the target meteorological characteristic data set corresponding to the target intrinsic component is obtained based on these screened meteorological characteristic data.

[0057] For example, considering that the distribution of meteorological characteristic data has certain similarities, the second meteorological characteristic model can be trained in advance using the third wind speed data set, and based on a method similar to the first meteorological characteristic model, the index values ​​such as the number of splitting times and information gain of each meteorological characteristic data can be determined, and the importance evaluation results of each meteorological characteristic data can be determined based on these index values. Then, based on the importance evaluation results, at least one meteorological characteristic data with greater importance is screened out from multiple meteorological characteristic data, and the target meteorological characteristic data set corresponding to the target intrinsic component is obtained based on these screened meteorological characteristic data.

[0058] For example, for a certain target intrinsic component, it corresponds to 228 meteorological characteristic data, its time range is 12 months (including January to December), and corresponds to 4 grid points (including grid point 1, grid point 2, grid point 3 and grid point 4).

[0059] Figure 3 A schematic diagram of the ranking of importance evaluation results of meteorological characteristic data provided by the embodiment of the present disclosure, corresponding to Figure 2 wind farms, in Figure 3Only the meteorological characteristic data of January, April, July and October are shown as examples. Moreover, by performing an importance analysis on these meteorological characteristic data, it can be determined that the meteorological characteristic data ranked in the top 20 in January, April, July and October have a significant impact on the wind speed, and the remaining meteorological characteristic data (not shown in the figure) have little impact on the wind speed. Based on this, the top 20 meteorological characteristic data can be used to construct a target meteorological characteristic data set corresponding to the target intrinsic component. Among them, the first meteorological characteristic data of January is "200m zonal wind-2", which represents the 200m zonal wind of grid point 2 in January. Other meteorological characteristic data are similar to it, and the "-" is the meteorological characteristic before the "-", and the "-" is the identifier or serial number of the corresponding grid point.

[0060] Furthermore, the target meteorological characteristic dataset is analyzed in the location dimension. Figure 3 As shown in the figure, in January and April, the number of grid points corresponding to the 20 meteorological characteristic data points is the largest, while in July and October, the number of grid points corresponding to the 20 meteorological characteristic data points is the largest. The reason for this is that grid point 4 is located downwind of the dominant wind direction and is second only to grid point 2 in distance from the wind farm. Therefore, the meteorological characteristic data of grid point 4 has a greater impact on the wind speed of the wind farm. For July and October, the dominant wind direction is not prominent, tending towards the southeast, while grid point 2 is upwind and closest to the wind farm. Its location factor has a greater impact on wind speed than the wind direction factor. In other words, the ranking of the meteorological characteristic data of each month and its corresponding position further illustrate that the contribution of meteorological characteristic data (or its importance, or its impact on wind speed) is closely related to the geographical location, and the meteorological characteristic data of grid points located downwind and close to the wind farm have the strongest correlation with the actual wind speed value.

[0061] Figure 4 A schematic diagram of the ranking of importance evaluation results of meteorological characteristic data provided by the embodiment of the present disclosure, corresponding to Figure 2 wind farms, in Figure 4 The figure only shows the meteorological characteristic data of five time periods, namely the current time, 1 hour before the current time, 2 hours before the current time, 3 hours before the current time and 4 hours before the current time in January, April, July and October. Based on this, the importance of the meteorological characteristic data is analyzed from the time dimension.

[0062] like Figure 4As shown, for January and October, the meteorological characteristic data corresponding to the current hour accounts for the largest proportion of the 20 meteorological characteristic data, for April it is the previous hour, and for July it is the previous two hours. In other words, in April, July, and October, the meteorological characteristic data corresponding to the current hour and the previous hour are most highly correlated with the actual wind speed, while in January, the meteorological characteristic data corresponding to the previous two hours and the previous three hours are most highly correlated. This suggests a general pattern: the closer the time, the greater the proportion of the corresponding meteorological characteristic data among the 20 meteorological characteristic data.

[0063] In summary, the meteorological feature data contained in the target meteorological feature data set determined based on the importance assessment result are the meteorological feature data that have a greater impact on wind speed prediction.

[0064] Step S103: constructing training data based on the first wind speed data set and multiple target meteorological feature data sets.

[0065] In some optional embodiments, the first wind speed data set and multiple target meteorological feature data sets can be directly used as training data, so that the wind speed prediction model obtains multiple predicted second wind speed sequence data based on the multiple target meteorological feature data sets, and calculates the loss value based on the first wind speed data set and the multiple second wind speed sequence data, and then updates the parameters of the wind speed prediction model based on the loss value, and repeats the above process until a trained wind speed prediction model is obtained.

[0066] In some optional embodiments, in order to improve the wind speed prediction model's ability to identify hill climbing events, some hill climbing features may be input into the model so that the wind speed prediction model can learn the ability to identify hill climbing events based on these hill climbing features.

[0067] In some optional embodiments, constructing training data based on a first wind speed dataset and multiple target meteorological feature datasets includes: obtaining a first slope climbing feature based on the first wind speed dataset; constructing training data based on the first slope climbing feature, the first wind speed dataset, and multiple target meteorological feature datasets; wherein the wind speed prediction model is further used to generate a second slope climbing feature, and adjusting parameters of the wind speed prediction model based on the second slope climbing feature, multiple second wind speed series data, and multiple first wind speed series data. Thus, compared to training data constructed directly from the first wind speed dataset and multiple target meteorological feature datasets, the addition of the first slope climbing feature results in expanded training data.

[0068] In some optional embodiments, the first climbing feature includes a climbing event identification feature or a climbing event identifier; accordingly, based on the first wind speed data set, the first climbing feature is obtained, including: for at least part of the first wind speed sequence data in the first wind speed data set, obtaining electric energy sequence data corresponding to at least part of the first wind speed sequence data; performing feature extraction based on at least part of the first wind speed sequence data and the corresponding electric energy sequence data to obtain a climbing event identification feature; or, marking at least part of the first wind speed sequence data with a climbing event identifier.

[0069] In other words, the first climbing feature can have multiple forms. It can be feature data extracted from the first wind speed sequence data and the corresponding electric energy sequence data, or it can be a climbing event identifier marked in the first wind speed sequence data. The embodiment of the present disclosure does not limit this.

[0070] Step S104 : training the initial wind speed prediction model based on the training data to obtain a trained wind speed prediction model. The wind speed prediction model is used to obtain wind speed prediction sequence data, so as to identify a hill climbing event based on the wind speed prediction sequence data.

[0071] In some optional embodiments, for any target meteorological characteristic data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and one predicted bias component corresponding to the target meteorological characteristic data set based on the target meteorological characteristic data set, obtain second wind speed sequence data corresponding to the target meteorological characteristic data set based on the multiple predicted intrinsic components and one predicted bias component, and adjust the parameters of the wind speed prediction model based on the multiple second wind speed sequence data corresponding to the multiple target meteorological characteristic data sets and the multiple first wind speed sequence data in the first wind speed data set.

[0072] In some optional embodiments, the wind speed prediction model is constructed based on a deep neural network (DNN), which includes an input layer, a hidden layer, and an output layer. Each node in a network layer has an operational relationship with all nodes in the next layer. Furthermore, there can be multiple hidden layers. Increasing the number of hidden layers can better separate the features of the data. However, if there are too many hidden layers, the training time may increase and overfitting may also occur. Furthermore, the DNN neural network can be expressed as follows:

[0073]

[0074] Among them, X i (i=1,2,…,n) represents the n-dimensional wind speed feature vector (corresponding to the input data of the network layer), x i ∈X represents the i-th element of X; W ij Indicates that from X i to jth Adaptive weight matrix between layer product neurons; W jk Indicates that from j th Layer product neurons up to k th Adaptive weight matrix between layer neurons; Y N is the wind speed output by the neural network; m represents the number of layers of the corresponding neural network. When m = 1, the hidden layer contains only one layer of product neurons; β represents the basis function vector; θ ij represents the adaptive threshold; σ(x) is the nonlinear activation function, h km represents the response of the corresponding network layer, ξ jk Represents the connection weight.

[0075] In some optional embodiments, if the training data does not include the first slope characteristic, a first loss function can be set, and the target meteorological characteristic dataset can be input into the wind speed prediction model to obtain multiple second wind speed series data. A first loss value of the first loss function is calculated based on the multiple second wind speed series data and the multiple first wind speed series data in the first wind speed dataset, and the model parameters are updated based on the first loss value. Similarly, when it is determined that the preset stopping condition is met, the training is stopped, and a trained wind speed prediction model is obtained.

[0076] In some optional embodiments, if the training data includes a first slope characteristic, a second loss function and a third loss function can be set, and the target meteorological feature data set is input into the wind speed prediction model to obtain multiple second wind speed sequence data and a second slope characteristic. On the one hand, a second loss value of the second loss function is calculated based on the multiple second wind speed sequence data and the multiple first wind speed sequence data in the first wind speed data set. On the other hand, a third loss value of the third loss function is calculated based on the second slope characteristic and the first slope characteristic, and the model parameters are jointly updated based on the second loss value and the third loss value. Similarly, when it is determined that the preset stop condition is met, the training is stopped to obtain a trained wind speed prediction model.

[0077] In some optional embodiments, test data may also be provided. For example, 80% of the training data constructed based on the first wind speed dataset and multiple target meteorological characteristic datasets may be used as a training set, and the remaining 20% ​​may be used as a test set.

[0078] In some optional embodiments, the wind speed prediction model can also be evaluated. For example, the wind speed prediction model of the embodiment of the present disclosure can be compared with the prediction results of other models. The test indicators used may include mean deviation (ME), mean absolute error (MAE), root mean square error (RMSE), and correlation coefficient (r), which respectively describe the deviation size, degree of deviation, degree of dispersion, and correlation between the predicted wind speed and the observed wind speed.

[0079] Figure 5 This is a schematic diagram of the processing process of a model training method for wind power ramping scenarios provided by an embodiment of the present disclosure. Figure 5 As shown, the first wind speed data set includes K pieces of first wind speed sequence data (first wind speed sequence data S11, first wind speed sequence data S21, ..., first wind speed sequence data SK1), where K is an integer greater than 1.

[0080] The first wind speed sequence data S11 is decomposed to obtain N11 target eigencomponents. For any of these target eigencomponents, the XGBoost method is used to filter out the meteorological feature data of high importance from the first meteorological dataset corresponding to the first wind speed dataset to obtain the target meteorological feature dataset corresponding to the target eigencomponent. Ultimately, N11 target meteorological feature datasets are obtained, where N11 is an integer greater than 1.

[0081] Similarly, for the first wind speed sequence data S21, N21 target intrinsic components and N21 target meteorological characteristic data sets can be obtained, ..., for the first wind speed sequence data SK1, NK1 target intrinsic components and NK1 target meteorological characteristic data sets can be obtained, where N21 and NK1 are both integers greater than 1.

[0082] Furthermore, based on the first wind speed sequence data S11, N11 target meteorological feature data sets, the first wind speed sequence data S21, N21 target meteorological feature data sets, ..., the first wind speed sequence data SK1, NK1 target meteorological feature data sets, training data is constructed and the training data is input into the wind speed prediction model for model training.

[0083] During the training process, the wind speed prediction model generates N12 groups of predicted eigenvalues ​​and predicted bias components based on the N11 target meteorological feature data sets (each group includes multiple predicted eigenvalues ​​and one predicted bias component), and obtains the second wind speed sequence data S12 based on this prediction (N12 and N11 can be the same or different).

[0084] Similarly, the wind speed prediction model generates N22 groups of predicted intrinsic components and predicted bias components based on the N21 target meteorological feature data sets, and obtains the second wind speed sequence data S22 based on this prediction;...; the wind speed prediction model generates NK2 groups of predicted intrinsic components and predicted bias components based on the NK1 target meteorological feature data sets, and obtains the second wind speed sequence data SK2 based on this prediction.

[0085] On this basis, the parameters of the wind speed prediction model can be adjusted according to the first wind speed sequence data S11 to the first wind speed sequence data SK1 and the second wind speed sequence data S11 to the second wind speed sequence data SK2 to achieve a round of iteration of the wind speed prediction model.

[0086] The above process is repeated. If it is determined that the preset stop condition is met, the currently obtained wind speed prediction model is determined as the final wind speed prediction model, and the wind speed prediction model can be used to perform tasks such as wind speed prediction.

[0087] It should be noted that in some optional embodiments, slope characteristics can also be input into the wind speed prediction model as training data, allowing the wind speed prediction model to learn the ability to identify slope events through the slope characteristics. In other words, the output of the wind speed prediction model can include slope event recognition results in addition to the second wind speed sequence data. Accordingly, when adjusting parameters, in addition to the first and second wind speed sequence data, it is also necessary to adjust parameters based on the slope characteristics and the slope event recognition results. Based on these two dimensions, the wind speed prediction model can be more capable of identifying slope events.

[0088] To summarize, in an embodiment of the present disclosure, each first wind speed sequence data in the first wind speed data set is decomposed into multiple target intrinsic components; for any target intrinsic component, a target meteorological feature data set corresponding to the target intrinsic component is determined based on the first meteorological data set corresponding to the first wind speed data set, the first meteorological data set includes multiple meteorological feature data, and the target meteorological feature data set includes at least one meteorological feature data screened out from the first meteorological data set; training data is constructed based on the first wind speed data set and the multiple target meteorological feature data sets; the initial wind speed prediction model is trained based on the training data to obtain a trained wind speed prediction model, and the wind speed prediction model is used to obtain wind speed prediction sequence data so as to identify climbing events based on the wind speed prediction sequence data.

[0089] It can be seen that in the embodiment of the present disclosure, by decomposing the first wind speed sequence data, a target intrinsic component with stronger regularity and better characterization of the essence of wind speed can be obtained. For any target intrinsic component, the part of the meteorological characteristic data that has a greater impact on the target intrinsic component is screened out from multiple meteorological characteristic data to obtain a target meteorological characteristic data set corresponding to the target intrinsic component. On this basis, training data is constructed based on the first wind speed data set and multiple target meteorological characteristic data sets, and the wind speed prediction model is trained using the training data. Among them, for any target meteorological characteristic data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and a predicted bias component corresponding to the target meteorological characteristic data set based on the target meteorological characteristic data set, obtain second wind speed sequence data corresponding to the target meteorological characteristic data set based on the multiple predicted intrinsic components and the predicted bias component, and adjust the parameters of the wind speed prediction model based on the multiple second wind speed sequence data corresponding to the multiple target meteorological characteristic data sets and the multiple first wind speed sequence data in the first wind speed data set. Based on this, the model training effect can be improved, and a wind speed prediction model with higher prediction accuracy can be obtained.

[0090] A second aspect of the embodiments of the present disclosure provides a wind speed prediction method for wind power ramping scenarios.

[0091] Figure 6 This is a flow chart of a wind speed prediction method for a wind power ramping scenario provided by an embodiment of the present disclosure. Figure 6 The wind speed prediction method for wind power ramping scenarios may include the following steps.

[0092] Step S601: Acquire a second meteorological data set of a first preset time period, where the second meteorological data set includes a plurality of meteorological characteristic data.

[0093] Step S602: Input the second meteorological data set of the first preset period into the wind speed prediction model to obtain wind speed prediction sequence data of the second preset period. The wind speed prediction model is obtained using the model training method for wind power ramping scenarios of any embodiment of the present disclosure.

[0094] It can be seen from this that after obtaining the second meteorological data set of the first preset time period, it is input into the wind speed prediction model to obtain the wind speed prediction sequence data of the second preset time period.

[0095] In some optional embodiments, the first preset period and the second preset period may correspond to different time periods.

[0096] For example, the first preset period corresponds to XXXX year X month 3, including 24 hours on the 3rd, and the second preset period corresponds to XXXX year X month 4, including 24 hours on the 4th. It can be seen that the wind speed data on the 4th is predicted based on the meteorological data on the 3rd.

[0097] In some optional embodiments, the first preset period and the second preset period may correspond to the same time period; accordingly, the method may further include: correcting the third wind speed sequence data according to the wind speed prediction sequence data to obtain the target wind speed sequence data for the second preset period; wherein, the third wind speed sequence data is the wind speed prediction data for the second preset period obtained based on a preset prediction method, and the preset prediction method includes a wind speed prediction method other than the wind speed prediction method of the embodiment of the present disclosure, and the embodiment of the present disclosure does not limit this.

[0098] In other words, the wind speed data predicted by the wind speed prediction model (ie, wind speed prediction sequence data) can be used to correct the third wind speed sequence data obtained based on the preset prediction method to obtain more accurate wind speed prediction data.

[0099] For example, the first preset time period corresponds to XXXX year X month 3, including the 24 hours of the 3rd day, and the second preset time period also corresponds to XXXX year X month 3, including the 24 hours of the 3rd day. Based on this, the third wind speed sequence data of the 3rd day can be corrected based on the wind speed forecast sequence data of the 3rd day, thereby obtaining a more accurate corrected wind speed forecast sequence.

[0100] In some optional embodiments, wind speed correction may be performed only for specific situations to reduce data processing volume, wherein the specific situation may include the presence of a hill climbing event, the receipt of a preset indication message (indicating the need for wind speed correction), etc.

[0101] In some optional embodiments, the first preset period and the second preset period correspond to the same time period; accordingly, the method may further include: obtaining a climbing event identification result for the second preset period, the climbing event identification result being output by a wind speed prediction model, or being obtained based on wind speed prediction sequence data; when the climbing event identification result indicates the presence of a climbing event, correcting the third wind speed sequence data according to the wind speed prediction sequence data to obtain target wind speed sequence data for the second preset period; wherein the third wind speed sequence data is wind speed prediction data for the second preset period obtained based on a preset prediction method.

[0102] In some optional embodiments, obtaining the second meteorological data set for the first preset time period includes obtaining a plurality of initial meteorological characteristic data for the first preset time period, determining the degree of influence of each initial meteorological characteristic data on wind speed, and then selecting a plurality of meteorological characteristic data with a greater degree of influence from the plurality of initial meteorological characteristic data to form the second meteorological data set. The degree of influence refers to the magnitude of the influence of the corresponding meteorological characteristic data on the predicted wind speed.

[0103] For example, the influence of each initial meteorological feature data on wind speed can be determined using an XGBoost approach. Similar to determining the importance assessment results of meteorological feature data, multiple decision trees can be determined based on the multiple initial meteorological feature data. Furthermore, the influence of each initial meteorological feature data can be determined based on indicators such as the number of splits and information gain of the initial meteorological feature data. Based on this, a second meteorological dataset is determined.

[0104] For example, based on the target meteorological feature data set determined during the training process of the wind speed prediction model, the degree of influence of each initial meteorological feature data on the wind speed can be determined, and then multiple meteorological feature data can be screened out from the multiple initial meteorological feature data to form a second meteorological data set. In other words, in this way, the importance assessment results of the aforementioned first meteorological data set can be reused, which can reduce the amount of data processing. Moreover, since the distribution of meteorological feature data and the like have certain similarities, the degree of influence determined based on the importance assessment results of the first meteorological data set is also relatively accurate.

[0105] In some optional embodiments, after obtaining the wind speed prediction sequence data, some strategies in the wind power scenario may be formulated based on the wind speed prediction sequence data.

[0106] In some optional embodiments, after inputting the second meteorological data set of the first preset time period into the wind speed prediction model to obtain the wind speed prediction sequence data of the second preset time period, the method may also include: determining the wind power operation strategy based on the target wind speed sequence data of the second preset time period, and the wind power operation strategy is used to indicate the power generation behavior and power supply behavior under the wind power scenario.

[0107] As can be seen, wind speed prediction models can achieve accurate and reliable wind speed forecasts, capable of predicting events such as drastic short-term wind speed fluctuations. This can help optimize wind farm power output control, improve grid stability, and mitigate operational risks. In other words, wind speed prediction can closely integrate meteorological science with power system needs, bridging the gap between the uncertainty of natural wind resources and the rigid demands of the grid. Specifically, first, it can accurately identify the trend and magnitude of wind speed fluctuations in advance, enabling grid operators to formulate more reasonable power generation plans and allocate reserve capacity. Second, high-precision wind speed predictions can reduce the impact of wind power uncertainty on grid frequency and voltage, effectively alleviating the risk of power supply and demand imbalances or load shedding caused by forecast errors. Third, by improving the reliability of wind speed predictions, the grid's adaptability to wind power fluctuations can be enhanced, reducing wind curtailment and improving power generation efficiency, thereby improving wind energy utilization. In short, improving the accuracy and reliability of wind speed predictions can effectively promote the transformation of wind power from a "fluctuating power source" to a "controllable resource," enabling more efficient and convenient use of wind energy.

[0108] In summary, the disclosed embodiment provides a technical concept of "feature screening + machine learning" based on the EMD-XGBoost-DNN model. By matching meteorological feature data, the climbing features of similar samples are obtained. By expanding the training set and updating the machine learning method, it is applicable to a wider range of application scenarios and can further improve performance. Among them, when constructing the target meteorological feature data set, the average level and fluctuation characteristics of meteorological feature data near the wind farm can be characterized by the mean and standard deviation of each grid data; for each meteorological feature data sequence, rolling features can be calculated in a specified time window (1 hour, 3 hours, 6 hours and 12 hours) to capture the changing trend of short-term forecast data; in addition, the slope can also be calculated to characterize the instantaneous change trend; time-based sine and cosine features can also be added to capture the periodic change pattern in the data, especially the influence of diurnal changes. In addition, by eliminating low-information features based on variance and feature importance analysis, about several features that have important contributions to wind speed can be screened out from the initial multiple meteorological feature data. The final target meteorological feature data set will contain richer information and can better characterize the forecast quantity near the wind farm and its changing characteristics.

[0109] Figure 7 This is a schematic diagram of a wind speed prediction method for wind power ramping scenarios provided by an embodiment of the present disclosure. Figure 7 First, historical weather forecasts are obtained, including multiple historical meteorological feature data, and equal-length historical weather forecast sequences are cyclically extracted. The extracted sequences are matched with the 24-hour weather forecast for similarity, and similar sample clusters are obtained based on several sequences with high similarity. Historical wind speed data are obtained, including multiple wind speed sequence data. The historical wind speed data are time-matched with similar sample clusters, and a training sample set is constructed for model training to obtain a trained wind speed prediction model. Furthermore, the wind speed prediction model can be used to predict wind speed, and the prediction results can be judged for a climbing event. If there is no climbing event, the wind speed prediction sequence data is directly output. If there is a climbing event, the 24-hour wind speed forecast can be corrected and the corrected wind speed prediction sequence data can be output. For example, the wind speed-related data in the 24-hour weather forecast is corrected based on the wind speed prediction sequence data, thereby obtaining the corrected wind speed prediction sequence data.

[0110] A third aspect of the embodiments of the present disclosure provides a model training device for wind power ramp-up scenarios.

[0111] Figure 8 This is a schematic diagram of a model training device for wind power ramping scenarios provided by an embodiment of the present disclosure. Figure 8The model training device 800 for wind power ramping scenarios may include the following modules.

[0112] A decomposition module 801 is configured to decompose each first wind speed sequence data in the first wind speed data set into a plurality of target eigencomponents;

[0113] A determination module 802 is configured to determine, for any target intrinsic component, a target meteorological characteristic dataset corresponding to the target intrinsic component based on a first meteorological dataset corresponding to the first wind speed dataset, where the first meteorological dataset includes a plurality of meteorological characteristic data, and the target meteorological characteristic dataset includes at least one meteorological characteristic data selected from the first meteorological dataset;

[0114] A construction module 803 is configured to construct training data based on the first wind speed data set and multiple target meteorological characteristic data sets;

[0115] A training module 804 is configured to train an initial wind speed prediction model based on the training data to obtain a trained wind speed prediction model. The wind speed prediction model is configured to obtain wind speed prediction sequence data, so as to identify a slope climbing event based on the wind speed prediction sequence data.

[0116] Among them, for any target meteorological characteristic data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and one predicted bias component corresponding to the target meteorological characteristic data set based on the target meteorological characteristic data set, obtain second wind speed sequence data corresponding to the target meteorological characteristic data set based on the multiple predicted intrinsic components and one predicted bias component, and adjust the parameters of the wind speed prediction model according to the multiple second wind speed sequence data corresponding to the multiple target meteorological characteristic data sets and the multiple first wind speed sequence data in the first wind speed data set.

[0117] In the embodiment provided by the present disclosure, each first wind speed sequence data in the first wind speed data set is decomposed into multiple target intrinsic components through a decomposition module; a target meteorological feature data set corresponding to the target intrinsic component is determined according to a first meteorological data set corresponding to the first wind speed data set through a determination module, wherein the first meteorological data set includes multiple meteorological feature data, and the target meteorological feature data set includes at least one meteorological feature data screened from the first meteorological data set; training data is constructed according to the first wind speed data set and multiple target meteorological feature data sets through a construction module; an initial wind speed prediction model is trained based on the training data through a training module to obtain to a trained wind speed prediction model, the wind speed prediction model is used to obtain wind speed prediction sequence data to identify climbing events based on the wind speed prediction sequence data; wherein, for any target meteorological feature data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and one predicted bias component corresponding to the target meteorological feature data set based on the target meteorological feature data set, obtain second wind speed sequence data corresponding to the target meteorological feature data set based on the multiple predicted intrinsic components and one predicted bias component, and adjust the parameters of the wind speed prediction model according to the multiple second wind speed sequence data corresponding to the multiple target meteorological feature data sets and the multiple first wind speed sequence data in the first wind speed data set.

[0118] It can be seen that in the embodiment of the present disclosure, by decomposing the first wind speed sequence data, a target intrinsic component with stronger regularity and better characterization of the essence of wind speed can be obtained. For any target intrinsic component, the part of the meteorological characteristic data that has a greater impact on the target intrinsic component is screened out from multiple meteorological characteristic data to obtain a target meteorological characteristic data set corresponding to the target intrinsic component. On this basis, training data is constructed based on the first wind speed data set and multiple target meteorological characteristic data sets, and the wind speed prediction model is trained using the training data. Among them, for any target meteorological characteristic data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and a predicted bias component corresponding to the target meteorological characteristic data set based on the target meteorological characteristic data set, obtain second wind speed sequence data corresponding to the target meteorological characteristic data set based on the multiple predicted intrinsic components and the predicted bias component, and adjust the parameters of the wind speed prediction model based on the multiple second wind speed sequence data corresponding to the multiple target meteorological characteristic data sets and the multiple first wind speed sequence data in the first wind speed data set. Based on this, the model training effect can be improved, and a wind speed prediction model with higher prediction accuracy can be obtained.

[0119] A fourth aspect of the embodiments of the present disclosure provides a wind speed prediction device for wind power ramping scenarios.

[0120] Figure 9This is a schematic diagram of a wind speed prediction device for wind power ramping scenarios provided by an embodiment of the present disclosure. Figure 9 The wind speed prediction device 900 for wind power ramping scenarios may include the following modules.

[0121] An acquisition module 901 is configured to acquire a second meteorological data set for a first preset period of time, wherein the second meteorological data set includes a plurality of meteorological characteristic data;

[0122] Prediction module 902, configured to input the second meteorological data set of the first preset period into the wind speed prediction model to obtain wind speed prediction sequence data of the second preset period;

[0123] The wind speed prediction model is obtained by using the model training method for wind power ramping scenarios as described in any one of the embodiments of the present disclosure.

[0124] In the embodiments provided herein, an acquisition module is used to acquire a second meteorological data set for a first preset time period, wherein the second meteorological data set includes multiple meteorological characteristic data. A prediction module is used to input the second meteorological data set for the first preset time period into a wind speed prediction model to obtain a series of wind speed prediction data for the second preset time period. As can be seen, the embodiments provided herein can obtain highly accurate wind speed prediction results, thereby enabling guidance on power generation and supply behaviors based on the wind speed prediction results, thereby improving wind energy utilization.

[0125] It is understood that the above-mentioned embodiments mentioned in this disclosure can be combined with each other to form a combined embodiment without violating the principle logic. Due to space limitations, this disclosure will not go into details. It is understood by those skilled in the art that in the above-mentioned method of the specific implementation method, the specific execution order of each step and the setting of the functional module should be determined by its function and possible internal logic.

[0126] In addition, the present disclosure also provides an electronic device and a computer-readable storage medium.

[0127] Figure 10 A block diagram of an electronic device provided in an embodiment of the present disclosure.

[0128] Reference Figure 10An embodiment of the present disclosure provides an electronic device, which includes: at least one processor 1001; at least one memory 1002, and one or more I / O interfaces 1003 connected between the processor 1001 and the memory 1002; wherein the memory 1002 stores one or more computer programs that can be executed by the at least one processor 1001, and the one or more computer programs are executed by the at least one processor 1001 so that the at least one processor 1001 can execute the model training method for wind power ramping scenarios or the wind speed prediction method for wind power ramping scenarios described in any one of the embodiments of the present disclosure.

[0129] The present disclosure also provides a computer-readable storage medium having a computer program stored thereon. When executed by a processor / processing core, the computer program implements the model training method for wind power ramping scenarios or the wind speed prediction method for wind power ramping scenarios described in any one of the embodiments of the present disclosure. The computer-readable storage medium may be volatile or non-volatile.

[0130] An embodiment of the present disclosure also provides a computer program product, including a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the model training method for wind power ramping scenarios or the wind speed prediction method for wind power ramping scenarios described in any one of the embodiments of the present disclosure.

[0131] It will be understood by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable storage medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium).

[0132] As is well known to those skilled in the art, the term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information (such as computer-readable program instructions, data structures, program modules or other data). Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technology, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those skilled in the art, communication media typically contains computer-readable program instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0133] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0134] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., utilizing an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions. The electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0135] The computer program product described herein may be implemented in hardware, software, or a combination thereof. In one embodiment, the computer program product is implemented as a computer storage medium. In another embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0136] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0137] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0138] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0139] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and the part of the module, program segment or instruction contains one or more executable instructions for realizing the prescribed logical function. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the prescribed function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0140] Example embodiments have been disclosed herein, and although specific terms are employed, they are used and should be interpreted only in a general illustrative sense and not for purposes of limitation. In some instances, it will be apparent to those skilled in the art that, unless otherwise expressly indicated, features, characteristics, and / or elements described in conjunction with a particular embodiment may be used alone or in combination with features, characteristics, and / or elements described in conjunction with other embodiments. Therefore, it will be understood by those skilled in the art that various changes in form and detail may be made without departing from the scope of the present disclosure as set forth in the appended claims.

Claims

1. A model training method for wind power ramping scenarios, characterized by: include: Decomposing each first wind speed sequence data in the first wind speed data set into a plurality of target eigencomponents; For any target intrinsic component, determining a target meteorological characteristic dataset corresponding to the target intrinsic component based on a first meteorological dataset corresponding to the first wind speed dataset, wherein the first meteorological dataset includes a plurality of meteorological characteristic data, and the target meteorological characteristic dataset includes at least one meteorological characteristic data selected from the first meteorological dataset; constructing training data based on the first wind speed dataset and the plurality of target meteorological characteristic datasets; Training an initial wind speed prediction model based on the training data to obtain a trained wind speed prediction model, wherein the wind speed prediction model is used to obtain wind speed prediction sequence data, so as to identify a climbing event based on the wind speed prediction sequence data; Among them, for any target meteorological characteristic data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and one predicted bias component corresponding to the target meteorological characteristic data set based on the target meteorological characteristic data set, obtain second wind speed sequence data corresponding to the target meteorological characteristic data set based on the multiple predicted intrinsic components and one predicted bias component, and adjust the parameters of the wind speed prediction model according to the multiple second wind speed sequence data corresponding to the multiple target meteorological characteristic data sets and the multiple first wind speed sequence data in the first wind speed data set.

2. The method according to claim 1, characterized in that The step of decomposing each piece of first wind speed sequence data in the first wind speed data set into a plurality of target intrinsic components includes: Decomposing the first wind speed sequence data into a plurality of target eigencomponents based on a preset decomposition method; The preset decomposition method includes at least one of the following: empirical mode decomposition, local mean decomposition, and wavelet decomposition.

3. The method according to claim 1, characterized in that The determining, based on the first meteorological dataset corresponding to the first wind speed dataset, a target meteorological characteristic dataset corresponding to the target intrinsic component, includes: For any target intrinsic component, determining a first meteorological characteristic model corresponding to the target intrinsic component based on the first meteorological data set, wherein the first meteorological characteristic model includes at least one meteorological decision tree; Determining, based on the first meteorological characteristic model, an importance assessment result of each meteorological characteristic data; screening out at least one meteorological characteristic data from a plurality of meteorological characteristic data according to the importance evaluation result, and obtaining a target meteorological characteristic data set corresponding to the target intrinsic component according to the at least one screened meteorological characteristic data; or Determining an importance assessment result of each meteorological characteristic data based on a preset second meteorological characteristic model, where the second meteorological characteristic model is a model determined based on the third wind speed data set; At least one meteorological characteristic data is screened out from a plurality of meteorological characteristic data according to the importance evaluation result, and a target meteorological characteristic data set corresponding to the target intrinsic component is obtained according to the at least one screened meteorological characteristic data.

4. The method according to claim 1, wherein The constructing of training data according to the first wind speed dataset and the plurality of target meteorological characteristic datasets includes: Acquire a first climbing characteristic based on the first wind speed dataset; constructing the training data according to the first climbing characteristic, the first wind speed dataset, and a plurality of target meteorological characteristic datasets; The wind speed prediction model is further used to generate a second climbing characteristic, and adjust the parameters of the wind speed prediction model according to the second climbing characteristic, a plurality of the second wind speed sequence data and a plurality of the first wind speed sequence data.

5. The method according to claim 4, characterized in that The first climbing feature includes a climbing event identification feature or a climbing event identifier; The obtaining of a first climbing characteristic based on the first wind speed dataset includes: For at least a portion of the first wind speed sequence data in the first wind speed data set, obtaining electric energy sequence data corresponding to the at least a portion of the first wind speed sequence data; Feature extraction is performed based on at least part of the first wind speed sequence data and the corresponding electric energy sequence data to obtain the hill climbing event identification feature; or the hill climbing event identifier is marked on at least part of the first wind speed sequence data.

6. A wind speed prediction method for wind power ramping scenarios, characterized in that: include: Acquire a second meteorological data set for a first preset time period, wherein the second meteorological data set includes a plurality of meteorological characteristic data; Inputting the second meteorological data set of the first preset time period into the wind speed prediction model to obtain wind speed prediction sequence data of the second preset time period; The wind speed prediction model is obtained by using the model training method for wind power ramping scenarios as described in any one of claims 1 to 5.

7. The method according to claim 6, characterized in that The first preset time period and the second preset time period correspond to the same time period; the method further includes: Obtaining a hill climbing event recognition result for the second preset time period, where the hill climbing event recognition result is output by the wind speed prediction model or obtained based on the wind speed prediction sequence data; When the climbing event identification result indicates that a climbing event exists, the third wind speed sequence data is corrected according to the wind speed prediction sequence data to obtain the target wind speed sequence data for the second preset time period; The third wind speed sequence data is wind speed forecast data for the second preset time period obtained based on a preset forecast method.

8. The method according to claim 7, characterized in that After inputting the second meteorological data set of the first preset time period into the wind speed prediction model to obtain the wind speed prediction sequence data of the second preset time period, the method further includes: A wind power operation strategy is determined according to the target wind speed sequence data of the second preset time period, where the wind power operation strategy is used to indicate power generation behavior and power supply behavior in a wind power scenario.

9. A model training device for wind power ramping scenarios, characterized in that: include: a decomposition module, configured to decompose each first wind speed sequence data in the first wind speed data set into a plurality of target intrinsic components; a determination module, configured to determine, for any target intrinsic component, a target meteorological characteristic dataset corresponding to the target intrinsic component based on a first meteorological dataset corresponding to the first wind speed dataset, wherein the first meteorological dataset includes a plurality of meteorological characteristic data, and the target meteorological characteristic dataset includes at least one meteorological characteristic data selected from the first meteorological dataset; A construction module, configured to construct training data based on the first wind speed dataset and a plurality of target meteorological characteristic datasets; A training module, configured to train an initial wind speed prediction model based on the training data to obtain a trained wind speed prediction model, wherein the wind speed prediction model is configured to obtain wind speed prediction sequence data, so as to identify a hill climbing event based on the wind speed prediction sequence data; Among them, for any target meteorological characteristic data set, the wind speed prediction model is used to obtain multiple predicted intrinsic components and one predicted bias component corresponding to the target meteorological characteristic data set based on the target meteorological characteristic data set, obtain second wind speed sequence data corresponding to the target meteorological characteristic data set based on the multiple predicted intrinsic components and one predicted bias component, and adjust the parameters of the wind speed prediction model according to the multiple second wind speed sequence data corresponding to the multiple target meteorological characteristic data sets and the multiple first wind speed sequence data in the first wind speed data set.

10. A wind speed prediction device for wind power ramp-up scenarios, characterized in that: include: An acquisition module, configured to acquire a second meteorological data set for a first preset time period, wherein the second meteorological data set includes a plurality of meteorological characteristic data; A prediction module, configured to input the second meteorological data set of the first preset time period into a wind speed prediction model to obtain wind speed prediction sequence data of the second preset time period; The wind speed prediction model is obtained by using the model training method for wind power ramping scenarios as described in any one of claims 1 to 5.

Images