Wind power prediction method and system based on multi-scale feature fusion

Abstract

The invention relates to a wind power prediction method and system based on multi-scale feature fusion, and the method comprises the steps: obtaining original data in a wind power generation state, carrying out the preprocessing of the original data, obtaining the preprocessing data of different scales, and enabling the preprocessing data to comprise season item data and trend item data, the original data are wind speed, temperature, air pressure, humidity and wind direction; respectively mapping the season item data and the trend item data through linear transformation to generate corresponding feature vectors, and calculating attention scores of different scale sequences based on the feature vectors; and on the basis of different attention scores, calculating fusion features containing multi-scale information, inputting the fusion features into a state space model, and obtaining predicted power of wind power generation through a linear layer. According to the invention, more accurate power can be predicted, and the safety and efficient operation of a power grid can be improved.

Description

Technical Field

[0001] This invention relates to the field of power equipment management technology, and in particular to a wind power prediction method and system based on multi-scale feature fusion. Background Technology

[0002] Wind power is a major form of wind energy utilization. Compared with renewable energy technologies such as solar energy and biomass, wind power is more mature, lower in cost, and less damaging to the environment. With the development of science and technology, wind power technology is also making continuous breakthroughs, and its economies of scale are becoming increasingly apparent.

[0003] Currently, with the increasing proportion of wind power generation in the power system, grid connection and stability issues are becoming increasingly prominent. Due to geographical and climatic factors, wind speed has a high degree of randomness and volatility, which makes the output power of wind power generation significantly volatile. This volatility makes the power output of wind power generation unpredictable and unstable, posing a great challenge to the dispatching and load management of the power system.

[0004] On the one hand, the power uncertainty caused by wind speed fluctuations may lead to fluctuations in grid frequency and voltage, thereby threatening the stability of the grid and the security of power supply. On the other hand, the randomness of wind power generation requires grid operators to configure a large number of backup power sources to balance supply and demand, which increases the operating cost of the grid. Therefore, in order to achieve efficient and reliable wind power grid connection, accurate wind power forecasting is crucial for the safety and efficient operation of the grid. Summary of the Invention

[0005] Therefore, the purpose of this invention is to provide a wind power prediction method and system based on multi-scale feature fusion to address the shortcomings of the prior art.

[0006] To achieve the above objectives, this invention provides a wind power prediction method based on multi-scale feature fusion, the method comprising:

[0007] The raw data under wind power generation conditions are acquired, and the raw data is preprocessed to obtain preprocessed data at different scales. The preprocessed data includes seasonal data and trend data. The raw data includes wind speed, temperature, air pressure, humidity and wind direction.

[0008] The seasonal data and the trend data are mapped to corresponding feature vectors by linear transformation, and attention scores for sequences at different scales are calculated based on the feature vectors.

[0009] Based on the different attention scores, a fusion feature containing multi-scale information is calculated. The fusion feature is then input into the state space model, and the predicted power of wind power generation is obtained through a linear layer.

[0010] The beneficial effects of this invention are as follows: by preprocessing the original data, preprocessed data of different scales are obtained. Then, by linear transformation, the seasonal and trend data in the preprocessed data are mapped to generate corresponding feature vectors. Based on the feature vectors, attention scores of sequences at different scales are calculated. Then, fusion features containing multi-scale information are calculated using different attention scores. The fusion features are input into the state space model. The predicted power of wind power generation is output using the linear layer in the state space model. Unlike existing technologies, this invention can predict more accurate power, which is beneficial to improving the safety and efficient operation of the power grid and also helps to simplify the scheduling and load management of the power system.

[0011] Furthermore, the step of preprocessing the raw data includes:

[0012] The original data is resampled and normalized sequentially to obtain the data to be processed;

[0013] The data to be processed is downsampled by average pooling of different sizes to obtain a multi-scale wind power generation time series set.

[0014] The multi-scale wind power generation time series is mapped to the hidden layer feature space by one-dimensional convolution, and the multi-scale wind power generation time series is decomposed into seasonal data and trend data of different scales by moving average pooling.

[0015] Furthermore, the mathematical expression for the trend data is as follows:

[0016] T = V Pool (P a (x))

[0017] The mathematical expression for the seasonal data is shown below:

[0018] S = xT

[0019] Where T represents the trend data, V Pool P represents average pooling. a This indicates a fill operation, where x represents a single scale sequence in the multi-scale wind power generation time series set, and S represents seasonal data.

[0020] Furthermore, the step of calculating the attention scores for sequences at different scales based on the feature vector includes:

[0021] Attention scores for trend data at different scales are obtained by performing a dot product operation on the query vector corresponding to the trend data and the key vector corresponding to the seasonal data at other scales.

[0022] Attention scores for the seasonal data at different scales are obtained by performing a dot product operation on the query vector corresponding to the seasonal data and the key vector corresponding to the trend data at other scales.

[0023] Furthermore, the step of calculating the fusion feature containing multi-scale information based on different attention scores includes:

[0024] The attention scores are normalized using the softmax function to obtain the corresponding normalized attention weights.

[0025] Based on the normalized attention weights, a fusion feature containing multi-scale information is calculated.

[0026] Furthermore, the formula for calculating the predicted power is as follows:

[0027] h'(t) = exp(ΔA)h(t) + (ΔA) -1 (exp(ΔA)-1)x(t)

[0028] w(t)=Bh'(t)

[0029] Where h'(t) represents the transmitted signal of the state-space model, exp represents the exponential function, h(t) represents the hidden state variables of the system, A represents the system matrix of the state-space model, Δ represents the time step parameter, x(t) represents the input sequence, w(t) represents the predicted power, and B represents the output matrix of the state-space model.

[0030] To achieve the above objectives, the present invention also provides a wind power prediction system based on multi-scale feature fusion, used to implement the provincial wind power generation project management and planning method described above, the system comprising:

[0031] The preprocessing module is used to acquire raw data under wind power generation conditions, preprocess the raw data to obtain preprocessed data at different scales, and the preprocessed data includes seasonal data and trend data. The raw data includes wind speed, temperature, air pressure, humidity and wind direction.

[0032] The mapping and calculation module is used to map the seasonal data and the trend data to generate corresponding feature vectors through linear transformation, and to calculate the attention scores of sequences at different scales based on the feature vectors.

[0033] The prediction module is used to calculate fusion features containing multi-scale information based on different attention scores, input the fusion features into the state space model, and obtain the predicted power of wind power generation through a linear layer.

[0034] Furthermore, the preprocessing module includes:

[0035] The resampling unit is used to resample and normalize the original data sequentially to obtain the data to be processed.

[0036] The downsampling unit is used to downsample the data to be processed by average pooling of different sizes to obtain a multi-scale wind power generation time series set.

[0037] The decomposition unit is used to map the multi-scale wind power generation time series set to the hidden layer feature space through one-dimensional convolution, and to decompose the multi-scale wind power generation time series set into seasonal data and trend data of different scales through moving average pooling.

[0038] Furthermore, the mathematical expression for the trend data is as follows:

[0039] T = V Pool (P a (x))

[0040] The mathematical expression for the seasonal data is shown below:

[0041] S = xT

[0042] Where T represents the trend data, V Pool P represents average pooling. a This indicates a fill operation, where x represents a single scale sequence in the multi-scale wind power generation time series set, and S represents seasonal data. Attached Figure Description

[0043] Figure 1 This is a flowchart of a wind power prediction method based on multi-scale feature fusion according to an embodiment of the present invention;

[0044] Figure 2 This is a structural block diagram of a wind power prediction system based on multi-scale feature fusion according to an embodiment of the present invention.

[0045] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0047] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0048] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0049] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0050] Example 1

[0051] Please see Figure 1The flowchart below shows a wind power prediction method based on multi-scale feature fusion according to the first embodiment of the present invention. The method includes the following steps:

[0052] Step S101: Obtain raw data under wind power generation conditions, preprocess the raw data to obtain preprocessed data at different scales, the preprocessed data including seasonal data and trend data;

[0053] Furthermore, the step of preprocessing the raw data includes:

[0054] The original data is resampled and normalized sequentially to obtain the data to be processed;

[0055] The data to be processed is downsampled by average pooling of different sizes to obtain a multi-scale wind power generation time series set.

[0056] The multi-scale wind power generation time series is mapped to the hidden layer feature space by one-dimensional convolution, and the multi-scale wind power generation time series is decomposed into seasonal data and trend data of different scales by moving average pooling.

[0057] The original data consists of five meteorological elements: wind speed, temperature, air pressure, humidity, and wind direction. The predicted power is the power generation capacity one day later. In the data preprocessing, abnormal and missing values ​​in the original data are first processed. Specifically, after obtaining the original data, the intervals of the original data are first resampled to remove abnormal values, and then the resampled data is normalized to fill in the missing values, thus obtaining the data to be processed for input.

[0058] Step S102: The seasonal data and the trend data are mapped to corresponding feature vectors by linear transformation, and the attention scores of sequences at different scales are calculated based on the feature vectors.

[0059] Each seasonal item data and each trend item data will be mapped to generate three feature vectors, namely, each seasonal item data can generate a corresponding query vector, key vector and value vector, and similarly, each trend item data can generate a corresponding query vector, key vector and value vector.

[0060] Furthermore, the step of calculating the attention scores for sequences at different scales based on the feature vector includes:

[0061] Attention scores for trend data at different scales are obtained by performing a dot product operation on the query vector corresponding to the trend data and the key vector corresponding to the seasonal data at other scales.

[0062] Attention scores for the seasonal data at different scales are obtained by performing a dot product operation on the query vector corresponding to the seasonal data and the key vector corresponding to the trend data at other scales.

[0063] Understandably, the purpose of this step is to measure the similarity between sequences at different scales.

[0064] Step S103: Calculate the fusion features containing multi-scale information based on the different attention scores, input the fusion features into the state space model, and obtain the predicted power of wind power generation through a linear layer.

[0065] Specifically, based on the attention scores corresponding to different seasonal data items, a fusion feature containing multi-scale information is calculated for the seasonal data items. Similarly, based on the attention scores corresponding to different trend data items, a fusion feature containing multi-scale information is calculated for the trend data items.

[0066] Through the above steps, the original data is preprocessed to obtain preprocessed data at different scales. Then, through linear transformation, the seasonal and trend data in the preprocessed data are mapped to generate corresponding feature vectors. Based on the feature vectors, attention scores for sequences at different scales are calculated. Then, fusion features containing multi-scale information are calculated through different attention scores. The fusion features are input into the state space model, and the predicted power of wind power generation is output using the linear layer in the state space model. Unlike existing technologies, this method can predict more accurate power, which is beneficial to improving the safety and efficient operation of the power grid, and also helps to simplify the dispatching and load management of the power system.

[0067] Furthermore, the mathematical expression for the trend data is as follows:

[0068] T = V Pool (P a (x))

[0069] The mathematical expression for the seasonal data is shown below:

[0070] S = xT

[0071] Where T represents the trend data, V Pool P represents average pooling. a This indicates a fill operation, where x represents a single scale sequence in the multi-scale wind power generation time series set, and S represents seasonal data.

[0072] Furthermore, the expression for calculating the attention score is as follows:

[0073]

[0074] in, M represents the scaling factor, used to balance the range of the inner product result and prevent excessively large or small values ​​from affecting subsequent normalization processing. ij Q represents the attention score. i Represents the query vector. This represents the key vector.

[0075] Furthermore, the step of calculating the fusion feature containing multi-scale information based on different attention scores includes:

[0076] The attention scores are normalized using the softmax function to obtain the corresponding normalized attention weights.

[0077] Based on the normalized attention weights, a fusion feature containing multi-scale information is calculated.

[0078] Specifically, the attention scores corresponding to the trend item data at different scales are normalized one by one using the softmax function. Then, the normalized attention weights corresponding to the different trend item data obtained after normalization are multiplied and weighted summed to obtain the fusion feature containing multi-scale information corresponding to the trend item data.

[0079] Similarly, the attention scores corresponding to the seasonal data at different scales are normalized one by one using the softmax function. Then, the normalized attention weights corresponding to the different seasonal data obtained after normalization are multiplied and weighted summed to obtain the fusion features containing multi-scale information corresponding to the seasonal data.

[0080] Furthermore, the expressions for generating corresponding feature vectors for the trend data and seasonal data images at different scales are as follows:

[0081] Q i =P q ·T i ,K i =P k ·T i V i =P s ·T i

[0082] Q' i =P' q ·S i ,K' i =P' k ·S i ,V' i =P' s ·S i

[0083] Among them, T i P represents the trend term data at different scales. q P represents the query vector transformation coefficient. k Q represents the key vector transformation coefficient, P represents the value vector transformation coefficient, and Q represents the key vector transformation coefficient. i K represents the query vector corresponding to the trend item data. i V represents the key vector corresponding to the trend item data. i S represents the value vector corresponding to the trend item data. i P' represents the seasonal data at different scales. q P' represents the query vector transformation coefficient. k P represents the key vector transformation coefficient. ' s Q represents the value vector transformation coefficient. ' i K represents the query vector corresponding to the seasonal data. ' i V represents the key vector corresponding to the seasonal data. ' i This represents the value vector corresponding to the seasonal data.

[0084] Furthermore, the formula for calculating the predicted power is as follows:

[0085] h'(t) = exp(ΔA)h(t) + (ΔA) -1 (exp(ΔA)-1)x(t)

[0086] w(t)=Bh'(t)

[0087] Where h'(t) represents the transmitted signal of the state-space model, exp represents the exponential function, h(t) represents the hidden state variables of the system, A represents the system matrix of the state-space model, Δ represents the time step parameter, x(t) represents the input sequence, w(t) represents the predicted power, and B represents the output matrix of the state-space model.

[0088] Example 2

[0089] Please see Figure 2 The diagram below shows the structural block diagram of a wind power prediction system based on multi-scale feature fusion according to the second embodiment of the present invention. The system includes:

[0090] The preprocessing module is used to acquire raw data under wind power generation conditions, preprocess the raw data to obtain preprocessed data at different scales, and the preprocessed data includes seasonal data and trend data.

[0091] The mapping and calculation module is used to map the seasonal data and the trend data to generate corresponding feature vectors through linear transformation, and to calculate the attention scores of sequences at different scales based on the feature vectors.

[0092] The prediction module is used to calculate fusion features containing multi-scale information based on different attention scores, input the fusion features into the state space model, and obtain the predicted power of wind power generation through a linear layer.

[0093] In practical implementation, the original data is preprocessed to obtain preprocessed data at different scales. Then, through linear transformation, the seasonal and trend data in the preprocessed data are mapped to generate corresponding feature vectors. Based on the feature vectors, attention scores for sequences at different scales are calculated. Then, fusion features containing multi-scale information are calculated through different attention scores. The fusion features are input into the state space model, and the predicted power of wind power generation is output using the linear layer in the state space model. Unlike existing technologies, this method can predict more accurate power, which is beneficial to improving the safety and efficient operation of the power grid, and also helps to simplify the dispatching and load management of the power system.

[0094] Furthermore, the preprocessing module 10 includes:

[0095] The resampling unit is used to resample and normalize the original data sequentially to obtain the data to be processed.

[0096] The downsampling unit is used to downsample the data to be processed by average pooling of different sizes to obtain a multi-scale wind power generation time series set.

[0097] The decomposition unit is used to map the multi-scale wind power generation time series set to the hidden layer feature space through one-dimensional convolution, and to decompose the multi-scale wind power generation time series set into seasonal data and trend data of different scales through moving average pooling.

[0098] Furthermore, the mathematical expression for the trend data is as follows:

[0099] T = V Pool (P a (x))

[0100] The mathematical expression for the seasonal data is shown below:

[0101] S = xT

[0102] Where T represents the trend data, V Pool P represents average pooling. aThis indicates a fill operation, where x represents a single scale sequence in the multi-scale wind power generation time series set, and S represents seasonal data.

[0103] Furthermore, the mapping and calculation module 20 includes:

[0104] The first calculation unit is used to obtain the attention scores corresponding to the trend item data at different scales by performing a dot product operation on the query vector corresponding to the trend item data and the key vector corresponding to the seasonal item data at other scales.

[0105] The second calculation unit is used to obtain the attention scores corresponding to the seasonal data at different scales by performing a dot product operation on the query vector corresponding to the seasonal data and the key vector corresponding to the trend data at other scales.

[0106] Furthermore, the prediction module 30 includes:

[0107] The normalization unit is used to normalize each attention score using the softmax function to obtain the corresponding normalized attention weights;

[0108] The third calculation unit is used to calculate the fusion features containing multi-scale information based on the normalized attention weights.

[0109] Furthermore, the formula for calculating the predicted power is as follows:

[0110] h'(t) = exp(ΔA)h(t) + (ΔA) -1 (exp(ΔA)-1)x(t)

[0111] w(t)=Bh'(t)

[0112] Where h'(t) represents the transmitted signal of the state-space model, exp represents the exponential function, h(t) represents the hidden state variables of the system, A represents the system matrix of the state-space model, Δ represents the time step parameter, x(t) represents the input sequence, w(t) represents the predicted power, and B represents the output matrix of the state-space model.

[0113] Example 3

[0114] In the third embodiment of the present invention, based on the same inventive concept, the present invention proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the wind power prediction method based on multi-scale feature fusion of the above embodiments.

[0115] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that contains storage, communication, propagation, or transmission programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0116] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0117] The memory may include a large-capacity storage device for data or instructions. For example, and not limitingly, the memory may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory may include removable or non-removable (or fixed) media. Where appropriate, the memory may be internal or external to the data processing device. In a particular embodiment, the memory is non-volatile memory. In a particular embodiment, the memory includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0118] Example 4

[0119] In the fourth embodiment of the present invention, based on the same inventive concept, the present invention proposes a terminal, the terminal comprising: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to execute the wind power prediction method based on multi-scale feature fusion of the above embodiments.

[0120] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0121] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0122] Without causing conflict, those skilled in the art can freely combine and use the above-mentioned additional technical features.

[0123] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A wind power prediction method based on multi-scale feature fusion, characterized in that, The method includes: The raw data under wind power generation conditions are acquired, and the raw data is preprocessed to obtain preprocessed data at different scales. The preprocessed data includes seasonal data and trend data. The raw data includes wind speed, temperature, air pressure, humidity and wind direction. The seasonal data and the trend data are mapped to corresponding feature vectors by linear transformation, and attention scores for sequences at different scales are calculated based on the feature vectors. Based on the different attention scores, a fusion feature containing multi-scale information is calculated. The fusion feature is then input into the state space model, and the predicted power of wind power generation is obtained through a linear layer.

2. The wind power prediction method based on multi-scale feature fusion according to claim 1, characterized in that, The preprocessing steps for the raw data include: The original data is resampled and normalized sequentially to obtain the data to be processed; The data to be processed is downsampled by average pooling of different sizes to obtain a multi-scale wind power generation time series set. The multi-scale wind power generation time series is mapped to the hidden layer feature space by one-dimensional convolution, and the multi-scale wind power generation time series is decomposed into seasonal data and trend data of different scales by moving average pooling.

3. The wind power prediction method based on multi-scale feature fusion according to claim 1, characterized in that, The mathematical expression for the trend data is shown below: T=V Pool (P a (x)) The mathematical expression for the seasonal data is shown below: S = xT Where T represents the trend data, V Pool P represents average pooling. a This indicates a fill operation, where x represents a single scale sequence in the multi-scale wind power generation time series set, and S represents seasonal data.

4. The wind power prediction method based on multi-scale feature fusion according to claim 1, characterized in that, The step of calculating the attention scores of sequences at different scales based on the feature vector includes: Attention scores for trend data at different scales are obtained by performing a dot product operation on the query vector corresponding to the trend data and the key vector corresponding to the seasonal data at other scales. Attention scores for the seasonal data at different scales are obtained by performing a dot product operation on the query vector corresponding to the seasonal data and the key vector corresponding to the trend data at other scales.

5. The wind power prediction method based on multi-scale feature fusion according to claim 1, characterized in that, The step of calculating the fusion feature containing multi-scale information based on different attention scores includes: The attention scores are normalized using the softmax function to obtain the corresponding normalized attention weights. Based on the normalized attention weights, a fusion feature containing multi-scale information is calculated.

6. The wind power prediction method based on multi-scale feature fusion according to claim 1, characterized in that, The formula for calculating the predicted power is as follows: h'(t)=exp(ΔA)h(t)+(ΔA) -1 (exp(ΔA)-1)x(t) w(t)=Bh′(t) Where h'(t) represents the transmitted signal of the state-space model, exp represents the exponential function, h(t) represents the hidden state variables of the system, A represents the system matrix of the state-space model, Δ represents the time step parameter, x(t) represents the input sequence, w(t) represents the predicted power, and B represents the output matrix of the state-space model.

7. A wind power prediction system based on multi-scale feature fusion, used to implement the wind power prediction method based on multi-scale feature fusion as described in any one of claims 1-6, characterized in that, The system includes: The preprocessing module is used to acquire raw data under wind power generation conditions, preprocess the raw data to obtain preprocessed data at different scales, and the preprocessed data includes seasonal data and trend data. The raw data includes wind speed, temperature, air pressure, humidity and wind direction. The mapping and calculation module is used to map the seasonal data and the trend data to generate corresponding feature vectors through linear transformation, and to calculate the attention scores of sequences at different scales based on the feature vectors. The prediction module is used to calculate fusion features containing multi-scale information based on different attention scores, input the fusion features into the state space model, and obtain the predicted power of wind power generation through a linear layer.

8. The wind power prediction system based on multi-scale feature fusion according to claim 7, characterized in that, The preprocessing module includes: The resampling unit is used to resample and normalize the original data sequentially to obtain the data to be processed. The downsampling unit is used to downsample the data to be processed by average pooling of different sizes to obtain a multi-scale wind power generation time series set. The decomposition unit is used to map the multi-scale wind power generation time series set to the hidden layer feature space through one-dimensional convolution, and to decompose the multi-scale wind power generation time series set into seasonal data and trend data of different scales through moving average pooling.

9. The wind power prediction system based on multi-scale feature fusion according to claim 8, characterized in that, The mathematical expression for the trend data is shown below: T=V Pool (P a (x)) The mathematical expression for the seasonal data is shown below: S = xT Where T represents the trend data, V Pool P represents average pooling. a This indicates a fill operation, where x represents a single scale sequence in the multi-scale wind power generation time series set, and S represents seasonal data.

Images