CN115630731A - Image-based wind power short-term prediction method and system with self-attention mechanism and gating - Google Patents
Image-based wind power short-term prediction method and system with self-attention mechanism and gating Download PDFInfo
- Publication number
- CN115630731A CN115630731A CN202211205360.6A CN202211205360A CN115630731A CN 115630731 A CN115630731 A CN 115630731A CN 202211205360 A CN202211205360 A CN 202211205360A CN 115630731 A CN115630731 A CN 115630731A
- Authority
- CN
- China
- Prior art keywords
- self
- wind power
- data
- neural network
- gating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000007246 mechanism Effects 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000007781 pre-processing Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 18
- 238000012545 processing Methods 0.000 claims abstract description 10
- 238000010606 normalization Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 10
- 238000000354 decomposition reaction Methods 0.000 claims description 10
- 238000007670 refining Methods 0.000 claims description 6
- 238000010248 power generation Methods 0.000 abstract description 5
- 238000005516 engineering process Methods 0.000 abstract description 3
- 239000000284 extract Substances 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 10
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 239000002803 fossil fuel Substances 0.000 description 2
- 230000009021 linear effect Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000002123 temporal effect Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 230000009022 nonlinear effect Effects 0.000 description 1
- 239000003921 oil Substances 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 230000035899 viability Effects 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Operations Research (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Biomedical Technology (AREA)
- Entrepreneurship & Innovation (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a wind power short-term prediction method and system with a self-attention mechanism and a gating image, belonging to the technical field of wind power generation prediction, wherein the method comprises the following steps: acquiring NWP meteorological data and SCADA fan data for preprocessing, and dividing the preprocessed data into a training set and a test set; decomposing the training set into a plurality of continuous historical time subsequences and reconstructing the subsequences into two-dimensional images; establishing a residual error-based deep convolutional neural network, and adding a gating convolutional neural network layer and a self-attention mechanism to obtain the residual error-based deep convolutional neural network with the self-attention mechanism and gating; inputting a two-dimensional image into the neural network for training; inputting the test set into the trained neural network for prediction to obtain the short-term wind power. The method comprehensively uses historical time sequence data and meteorological grid data, extracts features and converts the features into images, and predicts the wind power by using an advanced image processing technology, so that the prediction accuracy is greatly improved.
Description
Technical Field
The invention relates to the technical field of wind power generation prediction, in particular to a method and a system for short-term prediction of wind power based on images with a self-attention mechanism and gating.
Background
With the rapid development of economy and the continuous improvement of living standard, the demand of human for energy is also increasing. However, the use of conventional fossil energy sources, such as coal, oil and natural gas, releases pollution, destroys the environment, and causes global warming. Furthermore, due to the non-renewable nature and limited reserves of fossil fuels, over-mining will result in depletion of energy resources. Therefore, in order to solve the energy crisis and the environmental problem, renewable energy is required. Wind energy is a renewable energy source which is pollution-free and widely distributed, and has attracted global attention.
Unlike the steady supply of fossil fuels, wind power generation systems often exhibit non-stationary and non-linear uncertainties. Meteorological parameters such as temperature, humidity, air pressure, wind direction and wind speed affect wind power generated by the wind turbine. These factors present significant challenges to the management and operation of electrical power systems, including wind energy. Research on this problem has shown that accurate wind power prediction improves the reliability and economic viability of these systems by reducing their integration and operating costs of power generation. However, accurate prediction is a difficult process due to the high uncertainty in wind speed and direction. The uncertainty of the wind data complicates the model learning process, resulting in large prediction errors. Therefore, predicting wind energy is considered a challenging task.
It is well known that wind speed data has non-stationary and non-linear properties. Due to the cubic relationship of wind speed to wind power, the uncertainty level of the power data is higher than the wind speed data. Therefore, the wind power data is decomposed, and deep characteristic information is extracted from the subsequences. VMD is a non-recursive decomposition algorithm with good theoretical basis, and can simultaneously extract modes. In the paper "An improved residual-based connected temporal network for top short-term with power for acquiring", the time series is subjected to VMD feature extraction, and then the features of the historical time series are reconstructed into RGB color space images. Converting the time series into visual patterns provides an advantage that the 2D-CNN visual representation with inherent spatial invariance provides the best input for the convolutional layer, and using the images as input and output, a residual-based deep learning model is built, as shown in fig. 1, which has lower complexity and lower computation cost, can be based on multiple learnable parameters, and can produce more effective prediction results.
However, in the model, only one kind of data is used, eight hours of input data is taken as a unit to be extracted and converted into one picture, and the selection of the picture is random when the training set and the test set are constructed, which means that the model ignores the very important seasonality in wind speed and wind power, and does not consider the time correlation between the picture and the picture, so that the final prediction result of the deep learning model is effective, but has a large error.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the related art.
Therefore, the first purpose of the invention is to provide a short-term wind power prediction method with a self-attention mechanism and a gating image, and the method greatly improves the prediction accuracy.
A second object of the invention is to propose a wind power short-term prediction system based on images with self-attentiveness mechanism and gating.
A third object of the present invention is to propose a wind power short term prediction device.
A fourth object of the invention is to propose a non-transitory computer-readable storage medium.
In order to achieve the above object, a first aspect of the present invention provides a method for short-term prediction of wind power based on images with a self-attentive power mechanism and gating, comprising the following steps: the method comprises the following steps of S1, acquiring NWP meteorological data and SCADA fan data for preprocessing, and dividing the preprocessed data into a training set and a test set; s2, decomposing the training set into a plurality of continuous historical time subsequences, and reconstructing the plurality of historical time subsequences into a two-dimensional image; s3, establishing a residual error-based deep convolutional neural network, and adding a gated convolutional neural network layer and a self-attention mechanism to obtain a residual error-based deep convolutional neural network with the self-attention mechanism and a gated mode; s4, inputting the two-dimensional image into the residual error-based depth convolution neural network with the self-attention mechanism and the gating for training; and S5, inputting the test set into a trained residual error-based deep convolution neural network with a self-attention mechanism and gating for prediction to obtain short-term wind power.
The wind power short-term prediction method with the self-attention mechanism and the gating based on the image comprises the steps that in the first stage, the feature extraction based on the variation modal decomposition is carried out, and the features are converted into the image; in the second stage, the wind power is predicted by using a residual error-based deep convolutional neural network with a self-attention mechanism and gating, the model adopts meteorological wind speed, wind direction and wind power data as a data set, and the data come from combination of numerical weather forecast NWP and fan data acquired based on SCADA; the historical time sequence data and meteorological grid data are comprehensively used, the characteristics are extracted and converted into images, advanced image processing technology is used for predicting wind power, and prediction accuracy is greatly improved.
In addition, the wind power short-term prediction method with the self-attention mechanism and the gating image based according to the above embodiment of the invention can also have the following additional technical features:
further, in one embodiment of the invention, the SCADA wind turbine data comprises wind power, 10-minute average wind speed, wind turbine speed and wind turbine state, and the NWP meteorological data comprises 100 meters of transverse and longitudinal wind, temperature and humidity in a 7-by-7 grid centered on the coordinates of the wind field.
Further, in an embodiment of the present invention, the preprocessing procedure of step S1 is:
and refining the time granularity of the NWP meteorological data by using an interpolation method, aligning the NWP meteorological data with the SCADA fan data in a time step, and splicing the characteristics.
Further, in an embodiment of the present invention, the step S2 specifically includes: step S201, decomposing the training set into a plurality of historical time subsequences by adopting a Variational Modal Decomposition (VMD); step S202, wind power data in the plurality of historical time subsequences are normalized by a minimum-maximum normalization method to ensure that all features are distributed according to the same scale, and the plurality of historical time subsequences are updated; step S203, reconstructing the updated plurality of historical time subsequences into the two-dimensional image, wherein the two-dimensional image comprises a hue graph, a saturation graph and a value graph, each two-dimensional graph is composed of a 2 x 4 array in sequence, and the depth is 8 bits.
In order to achieve the above object, a second aspect of the present invention provides a system for short-term prediction of wind power based on images with a self-attention mechanism and gating, comprising: the acquisition and preprocessing module is used for acquiring NWP meteorological data and SCADA fan data for preprocessing, and dividing the preprocessed data into a training set and a test set; the reconstruction module is used for decomposing the training set into a plurality of continuous historical time subsequences and reconstructing the plurality of historical time subsequences into two-dimensional images; the building module is used for building a residual error-based deep convolutional neural network, and adding a gated convolutional neural network layer and a self-attention mechanism to obtain the residual error-based deep convolutional neural network with the self-attention mechanism and the gated convolutional neural network; the training module is used for inputting the two-dimensional image into the residual error-based deep convolution neural network with the self-attention mechanism and the gating for training; and the prediction module is used for inputting the test set into a trained residual error-based deep convolution neural network with a self-attention mechanism and gating for prediction to obtain the short-term wind power.
According to the wind power short-term prediction system with the self-attention mechanism and the gating based on the images, in the first stage, feature extraction based on variational modal decomposition is carried out, and the features are converted into the images; in the second stage, the wind power is predicted by using a residual error-based deep convolutional neural network with a self-attention mechanism and gating, the model adopts meteorological wind speed, wind direction and wind power data as a data set, and the data come from combination of numerical weather forecast NWP and fan data acquired based on SCADA; the historical time sequence data and the meteorological grid data are comprehensively used, the characteristics are extracted and converted into images, advanced image processing technology is used for predicting wind power, and the prediction accuracy is greatly improved.
In addition, the wind power short-term prediction system with the self-attention mechanism and the gating image based according to the above embodiment of the invention can also have the following additional technical features:
further, in one embodiment of the present invention, the SCADA wind turbine data includes wind power, 10 minute average wind speed, wind rotor speed, and wind turbine status, and the NWP meteorological data includes 100 meters of transverse and longitudinal wind, temperature, and humidity in a 7 x 7 grid centered on the coordinates of the wind field.
Further, in an embodiment of the present invention, the preprocessing process of the acquiring and preprocessing module is: and refining the time granularity of the NWP meteorological data by using an interpolation method, aligning the NWP meteorological data with the SCADA fan data in a time step, and splicing the characteristics.
Further, in an embodiment of the present invention, the reconstruction module is specifically configured to: decomposing the training set into the plurality of historical time subsequences by using a Variational Modal Decomposition (VMD); carrying out normalization processing on the wind power data in the plurality of historical time subsequences by adopting a minimum-maximum normalization method so as to ensure that all the characteristics are distributed according to the same scale, and updating the plurality of historical time subsequences; and reconstructing the updated plurality of historical time subsequences into the two-dimensional image, wherein the two-dimensional image comprises a hue graph, a saturation graph and a value graph, each two-dimensional graph sequentially consists of a 2 x 4 array, and the depth of each two-dimensional graph is 8 bits.
In order to achieve the above object, a wind power short-term prediction device according to a third aspect of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the wind power short-term prediction device implements the wind power short-term prediction method with a self-attention mechanism and a gating image based as described in the above embodiments.
To achieve the above object, a fourth aspect of the present invention provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for short-term prediction of wind power with a self-attentive power mechanism and gating an image based as described in the above embodiments.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The foregoing and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a schematic diagram of a prior art residual-based deep learning model according to an embodiment of the present invention;
FIG. 2 is a flow diagram of a method for short term prediction of wind power with a self-attentive mechanism and gating image based, according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a deep convolutional neural network based on residual error according to an embodiment of the present invention;
FIG. 4 is a diagram illustrating an overall implementation of gated convolutional neural network layers in accordance with an embodiment of the present invention;
FIG. 5 is a schematic diagram of a self-attention mechanism implementation of one embodiment of the present invention;
FIG. 6 is a flow chart of predicting short term wind power according to one embodiment of the present invention;
FIG. 7 is a schematic diagram of a system for short-term prediction of wind power based on images with self-attentiveness and gating, according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are illustrative and intended to explain the present invention and should not be construed as limiting the present invention.
The method and system for short-term wind power prediction based on self-attentive power and gating provided by the embodiment of the invention are described below with reference to the accompanying drawings, and firstly, the method for short-term wind power prediction based on self-attentive power and gating provided by the embodiment of the invention is described with reference to the accompanying drawings.
FIG. 2 is a flow chart of a method for short-term prediction of wind power with a self-attentive mechanism and gating image based, according to an embodiment of the invention.
As shown in fig. 2, the method for short-term prediction of wind power based on images with self-attention mechanism and gating comprises the following steps:
in step S1, the NWP meteorological data and the SCADA fan data are obtained for preprocessing, and the preprocessed data are divided into a training set and a testing set.
Further, in one embodiment of the present invention, the SCADA wind turbine data includes wind power, 10 minute average wind speed, wind turbine speed, and wind turbine status, and the NWP meteorological data includes 100 meters of cross-wind, longitudinal wind, temperature, and humidity in a 7 x 7 grid centered on the coordinates of the wind field.
Specifically, the wind turbine data collected by the main manual SCADA system and NWP numerical meteorological data provided by the EC are provided, the SCADA data use fields mainly comprise wind power, 10-minute average wind speed, wind wheel rotating speed, wind turbine state and the like, the EC data use 7 × 7 grid data with the coordinates of the wind field as the center, the grid size is 0.125 ℃, and the fields mainly use 100-meter transverse and longitudinal wind, temperature and humidity. For SCADA data, firstly, the SCADA data with the resolution of 0.5s are divided according to the time span of 10 minutes, averaging processing is carried out on the data in each 10-minute span to obtain the SCADA data of each 10-minute integral point, so that an averaging method is adopted instead of selecting a specific time point, the influence of wind fluctuation on the data is avoided, and the wind speed condition in the time period can be better reflected by using the average data. And for the data in the missing time period, filling is not carried out so as to prevent the data noise from influencing the prediction result. For NWP data, the NWP data downloaded from the EC are nc-format data, and need to be read by using a python library function provided by the EC official website, and after the NWP data are read according to coordinate classification, the NWP data are reintegrated into table data sorted according to time. And finally, transforming the NWP data into three-dimensional grid data with 7 × feature numbers.
Further, the pretreatment process in the embodiment of the invention is as follows: and (3) refining the time granularity of the NWP meteorological data by using an interpolation method, aligning the NWP meteorological data with the SCADA fan data in a time step, and splicing the characteristics.
In step S2, the training set is decomposed into a plurality of consecutive historical time subsequences, and the plurality of historical time subsequences are reconstructed into two-dimensional images.
Further, in an embodiment of the present invention, step S2 specifically includes:
step S201, decomposing a training set into a plurality of historical time subsequences by adopting variational modal decomposition VMD;
step S202, carrying out normalization processing on the wind power data in the plurality of historical time subsequences by adopting a minimum-maximum normalization method to ensure that all features are distributed according to the same scale, and updating the plurality of historical time subsequences;
step S203, reconstructing the updated plurality of historical time subsequences into a two-dimensional image, wherein the two-dimensional image comprises a hue map, a saturation map and a value map, each two-dimensional image is composed of 2 x 4 arrays in sequence, and the depth is 8 bits.
Specifically, a time series p (t) of the training set is decomposed into a series of K eigenmode functions, i.e., a plurality of historical time subsequences, by using a variational mode decomposition VMD, as shown in the following formula:
wherein each eigenmode function is represented as u k (t) having the following characteristics in amplitude and frequency:
it should be noted that the VMD algorithm adopted in the present invention is a constrained variation optimization algorithm, and can be represented by the following formula:
wherein, w k As the center frequency, δ (t) is a pulse signal,
is u k (t) a hilbert transform;
that is, the data of the VMD training set decomposed by the variational modality is divided into small lots, and the data in each lot needs to be data of a continuous period of time.
And then, reconstructing the obtained historical time subsequence into a two-dimensional image, so as to facilitate the learning by using a deep neural network in the subsequent steps. Reconstructing the features of the historical temporal sub-sequence into visual patterns provides an advantage of providing the best input for 2D CNN visual representation convolutional layers with inherent spatial invariance.
It should be noted that the two-dimensional image includes the hourly input correlations for the plant. Before the reconstruction process, the wind power data are normalized by a minimum-maximum normalization method so as to ensure that all the characteristics are distributed according to the same scale. Image format selection for the last 8 hours, each set of images was composed of a 2 x 4 array containing 8 data points. Subsequently to be used as input parameters, the reconstructed feature can be defined as f i 1 ,f i 2 ,L,f i N Where N is the total data point, i is the number of input features, and i =1,2,l,8.
For each historical time series, e.g., decomposed IMF components, the data points are rearranged into an RGB image, 8 bits deep, comprising a map of hue, saturation, and value. Each graph is generated from a different historical time series for each input parameter.
In step S3, a residual error-based deep convolutional neural network is established, and a gated convolutional neural network layer and a self-attention mechanism are added to obtain a residual error-based deep convolutional neural network with a self-attention mechanism and a gate.
Specifically, a residual-based deep convolutional neural network is established first, and as shown in fig. 3, the nonlinear wind power needs to be predicted through extensive depth feature extraction at any wind power frequency. Therefore, the embodiment of the invention constructs a robust and effective depth residual error CNN model, and fully utilizes image recognition to extract the advanced features of the convolutional layer. To achieve efficient aggregation of feature maps, the extracted salient features are brought from the initial layer to deeper layers using residual concatenation, the remaining concatenation also including skip concatenation, providing additional feature information for more accurate prediction.
Furthermore, a gate control convolution neural network layer for extracting seasonal characteristics is added on the basis of the deep convolution neural network, the change characteristics of wind power in different dates are learned, the input of the model needs to be noticed, and the pictures are sequentially input according to the time sequence. The training set needs to correspond to the test set: if the wind power of a certain day in March needs to be predicted, the training set should also have wind power of March and before and after March in the last years.
Therefore, the embodiment of the invention uses a gated convolutional neural network layer to learn the seasonality of wind power, and as shown in fig. 4, the gated convolutional neural network layer consists of a one-dimensional convolutional layer and a gated linear unit. This particular design allows for a parallel and controllable training process. In this process, let the convolution kernel be Γ, input be Y, output be Z, and divide the output into two parts on average, which can be represented as Z = [ P, Q ], where [ P, Q ] is also used as the input of the gate control. The computation of the entire gated convolutional neural network layer can be expressed as:
Γ*Y＝Peσ(Q)
where σ (Q) is the compositional structure and dynamic changes in the time series between learning inputs P.
Further, as shown in FIG. 5, a self-attention mechanism is added for the same input X t By the method, sensitive relations can be established between any time point in the long sequence and other time points, the problem of dependence failure caused by overlong time window is solved, the dependence capability of long-time window input is improved, namely the robustness and generalization capability of the model are improved, and the method is more suitable for variable wind power generation prediction problems in actual application. The above calculation is formulated as follows:
head＝Attention(QW Q ,KW K ,VW V )。
in step S4, the two-dimensional image is input into a residual-based deep convolutional neural network with a self-attention mechanism and gating for training.
In step S5, the test set is input into a trained deep convolutional neural network with a self-attention mechanism and gating and based on residual errors for prediction, so as to obtain short-term wind power.
It should be noted that, the data in the test set is sent to the network to evaluate the result of the network training, and the data in the test set is complete and does not need to be divided.
Specifically, as shown in fig. 6, the test set data is input into the trained residual error-based deep convolutional neural network with the self-attention mechanism and the gate control, the data is input into the residual error-based deep convolutional neural network, the result of the residual error-based deep convolutional neural network is sent into the gate control neural network layer, the model learns the change characteristics (seasonality) of the wind power at different dates, the result of the gate control neural network layer is sent into the self-attention layer, the result is added to the result of the gate control neural network layer in a weighted manner, and the calculation results of the two are subjected to regression prediction to obtain the final wind power prediction result, that is, the short-term wind power, and the wind power 4 hours after the point of the prediction is started.
Then, real-time SCADA data are acquired from a fan manufacturer every day, real-time NWP data are acquired from an EC, the data are sent into a model for prediction, prediction data after a starting point is four hours are generated, meanwhile, the data in the previous day are sent into the model for incremental training, and the performance of the model is enhanced.
In conclusion, the method for predicting the wind power in the short term based on the image with the self-attention mechanism and the gating, which is provided by the embodiment of the invention, has the following beneficial effects:
decomposing and reconstructing the NWP data and the SCADA data into a 2D image by comprehensively using the NWP data and the SCADA data, and providing the optimal input for the convolutional layer by comprehensively reconstructing the characteristics of the historical time sequence to visually represent the 2D CNN with inherent spatial invariance;
the gated convolution neural network layer is used for learning the relation between wind power time sequence images on different dates, namely the seasonality of wind power, so that the prediction capability of the model is improved;
the initial training result of the 2D image formed by reconstructing the fan data is sent to an attention mechanism, the prediction results of the two parts are integrated, the problem of dependence failure caused by overlong time window is solved, and the robustness and the generalization capability of the model are improved.
The proposed wind power short-term prediction system with self-attentiveness mechanism and gating image-based will be described next with reference to the accompanying drawings.
FIG. 7 is a flow diagram of a system for short term prediction of wind power with a self-attentive mechanism and gating image based, according to an embodiment of the present invention.
As shown in fig. 7, the system 10 includes: an acquisition and preprocessing module 100, a reconstruction module 200, a construction module 300, a training module 400, and a prediction module 500.
The acquisition and preprocessing module 100 is used for acquiring NWP meteorological data and SCADA wind turbine data for preprocessing, and dividing the preprocessed data into a training set and a test set. The reconstruction module 200 is configured to decompose the training set into a plurality of continuous historical time subsequences, and reconstruct the plurality of historical time subsequences into two-dimensional images. The building module 300 is configured to build a residual error-based deep convolutional neural network, and add a gated convolutional neural network layer and a self-attention mechanism to obtain a residual error-based deep convolutional neural network with a self-attention mechanism and a gated self-attention mechanism. The training module 400 is used for training two-dimensional image input into a residual error-based deep convolutional neural network with a self-attention mechanism and gating. The prediction module 500 is used for inputting the test set into a trained deep convolutional neural network based on residual errors with a self-attention mechanism and gating for prediction to obtain short-term wind power.
Further, in one embodiment of the invention, the SCADA wind turbine data includes wind power, 10 minute average wind speed, rotor speed, and turbine status, and the NWP meteorological data includes 100 meters of cross-wind and longitudinal wind, temperature, and humidity in a 7 by 7 grid centered on the coordinates of the wind field.
Further, in an embodiment of the present invention, the preprocessing process of the acquiring and preprocessing module is: and (3) refining the time granularity of the NWP meteorological data by using an interpolation method, aligning the NWP meteorological data with the SCADA fan data in a time step, and splicing the characteristics.
Further, in an embodiment of the present invention, the reconfiguration module is specifically configured to: decomposing the training set into a plurality of historical time subsequences by adopting Variational Modal Decomposition (VMD); carrying out normalization processing on the wind power data in the plurality of historical time subsequences by adopting a minimum-maximum normalization method so as to ensure that all the characteristics are distributed according to the same scale, and updating the plurality of historical time subsequences; and reconstructing the plurality of updated historical time subsequences into a two-dimensional image, wherein the two-dimensional image comprises a hue graph, a saturation graph and a value graph, each two-dimensional graph is composed of a 2 x 4 array in sequence, and the depth is 8 bits.
It should be noted that the foregoing explanation of the embodiment of the method for short-term wind power prediction based on images with self-attention mechanism and gating is also applicable to the system of the embodiment, and will not be described herein again.
The system for short-term prediction of wind power based on images and provided with the self-attention mechanism and the gating has the following beneficial effects:
decomposing and reconstructing the NWP data and the SCADA data into a 2D image by comprehensively using the NWP data and the SCADA data, and providing the optimal input for the convolutional layer by comprehensively reconstructing the characteristics of the historical time sequence to visually represent the 2D CNN with inherent spatial invariance;
the gated convolution neural network layer is used for learning the relation between wind power time sequence images on different dates, namely the seasonality of wind power, so that the prediction capability of the model is improved;
the initial training result of the 2D image formed by reconstructing the fan data is sent to an automatic attention mechanism, the prediction results of the two parts are integrated, the problem of dependence failure caused by overlong time window is avoided, and the robustness and the generalization capability of the model are improved.
In order to implement the foregoing embodiments, the present invention further provides a wind power short-term prediction apparatus, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the wind power short-term prediction apparatus implements the wind power short-term prediction method with the self-attention power mechanism and the gating image based as described in the foregoing embodiments.
In order to achieve the above embodiments, the present invention further proposes a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for short-term prediction of wind power with a self-attentive power mechanism and gating of images as described in the previous embodiments.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or N embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "N" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more N executable instructions for implementing steps of a custom logic function or process, and alternate implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the embodiments of the present invention.
The logic and/or steps represented in the flowcharts or otherwise described herein, e.g., an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or N wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Further, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are well known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
It will be understood by those skilled in the art that all or part of the steps carried out in the method of implementing the above embodiments may be implemented by hardware related to instructions of a program, which may be stored in a computer readable storage medium, and the program, when executed, includes one or a combination of the steps of the method embodiments.
In addition, functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may exist alone physically, or two or more units are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a software functional module mode. The integrated module, if implemented in the form of a software functional module and sold or used as a stand-alone product, may also be stored in a computer readable storage medium.
The storage medium mentioned above may be a read-only memory, a magnetic or optical disk, etc. Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are exemplary and not to be construed as limiting the present invention, and that changes, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. A wind power short-term prediction method based on images and provided with a self-attention mechanism and gating is characterized by comprising the following steps:
the method comprises the following steps of S1, acquiring NWP meteorological data and SCADA fan data for preprocessing, and dividing the preprocessed data into a training set and a test set;
s2, decomposing the training set into a plurality of continuous historical time subsequences, and reconstructing the plurality of historical time subsequences into a two-dimensional image;
s3, establishing a residual error-based deep convolutional neural network, and adding a gated convolutional neural network layer and a self-attention mechanism to obtain a residual error-based deep convolutional neural network with the self-attention mechanism and a gated mode;
s4, inputting the two-dimensional image into the residual error-based deep convolution neural network with the self-attention mechanism and the gating for training;
and S5, inputting the test set into a trained residual error-based deep convolution neural network with a self-attention mechanism and gating for prediction to obtain short-term wind power.
2. The method of short-term wind power prediction with self-attentive power mechanism and gating image based on according to claim 1, characterized in that the SCADA wind turbine data comprises wind power, 10 minute average wind speed, wind turbine speed and wind turbine status, and the NWP meteorological data comprises 100 meters of transverse and longitudinal wind, temperature and humidity in a 7 x 7 grid centered on the coordinates of the wind field.
3. The method for short-term prediction of wind power with self-attention mechanism and gating based on image according to claim 1, characterized in that the preprocessing procedure of step S1 is:
and refining the time granularity of the NWP meteorological data by using an interpolation method, aligning the NWP meteorological data with the SCADA fan data in time step, and splicing the characteristics.
4. The method for short-term prediction of wind power with self-attentive power mechanism and gating based on images according to claim 1, wherein said step S2 comprises in particular:
step S201, decomposing the training set into a plurality of historical time subsequences by adopting a Variational Modal Decomposition (VMD);
step S202, carrying out normalization processing on the wind power data in the plurality of historical time subsequences by adopting a minimum-maximum normalization method to ensure that all features are distributed according to the same scale, and updating the plurality of historical time subsequences;
step S203, reconstructing the updated plurality of historical time subsequences into the two-dimensional image, wherein the two-dimensional image comprises a hue graph, a saturation graph and a value graph, each two-dimensional graph is composed of a 2 x 4 array in sequence, and the depth is 8 bits.
5. A system for short-term prediction of wind power with a self-attentive mechanism and gated image-based, comprising:
the acquisition and preprocessing module is used for acquiring NWP meteorological data and SCADA fan data for preprocessing, and dividing the preprocessed data into a training set and a test set;
the reconstruction module is used for decomposing the training set into a plurality of continuous historical time subsequences and reconstructing the plurality of historical time subsequences into two-dimensional images;
the building module is used for building a residual error-based deep convolutional neural network, and adding a gate control convolutional neural network layer and a self-attention mechanism to obtain the residual error-based deep convolutional neural network with the self-attention mechanism and gate control;
the training module is used for inputting the two-dimensional image into the residual error-based deep convolution neural network with the self-attention mechanism and the gating for training;
and the prediction module is used for inputting the test set into a trained residual error-based deep convolution neural network with a self-attention mechanism and gating for prediction to obtain the short-term wind power.
6. The system of claim 5, wherein the SCADA wind turbine data comprises wind power, 10 minute average wind speed, wind turbine speed, and turbine state, and the NWP meteorological data comprises 100 meters of cross-wind, longitudinal wind, temperature, and humidity in a 7 by 7 grid centered on wind farm coordinates.
7. The system of claim 5, wherein the pre-processing of the acquisition and pre-processing module is:
and refining the time granularity of the NWP meteorological data by using an interpolation method, aligning the NWP meteorological data with the SCADA fan data in a time step, and splicing the characteristics.
8. The system of claim 5, wherein the reconstruction module is specifically configured to:
decomposing the training set into the plurality of historical time subsequences by using a Variational Modal Decomposition (VMD);
carrying out normalization processing on the wind power data in the plurality of historical time subsequences by adopting a minimum-maximum normalization method so as to ensure that all the characteristics are distributed according to the same scale, and updating the plurality of historical time subsequences;
and reconstructing the updated plurality of historical time subsequences into the two-dimensional image, wherein the two-dimensional image comprises a hue graph, a saturation graph and a value graph, each two-dimensional graph sequentially consists of a 2 x 4 array, and the depth of each two-dimensional graph is 8 bits.
9. A wind power short term prediction device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor when executing the computer program implementing the method of wind power short term prediction with self-attentive power control and gating of images as claimed in any of claims 1-4.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the method for short-term prediction of wind power with self-attentive power and gating image based as claimed in any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211205360.6A CN115630731A (en) | 2022-09-30 | 2022-09-30 | Image-based wind power short-term prediction method and system with self-attention mechanism and gating |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211205360.6A CN115630731A (en) | 2022-09-30 | 2022-09-30 | Image-based wind power short-term prediction method and system with self-attention mechanism and gating |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115630731A true CN115630731A (en) | 2023-01-20 |
Family
ID=84903984
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211205360.6A Pending CN115630731A (en) | 2022-09-30 | 2022-09-30 | Image-based wind power short-term prediction method and system with self-attention mechanism and gating |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115630731A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117094452A (en) * | 2023-10-20 | 2023-11-21 | 浙江天演维真网络科技股份有限公司 | Drought state prediction method, and training method and device of drought state prediction model |
-
2022
- 2022-09-30 CN CN202211205360.6A patent/CN115630731A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117094452A (en) * | 2023-10-20 | 2023-11-21 | 浙江天演维真网络科技股份有限公司 | Drought state prediction method, and training method and device of drought state prediction model |
| CN117094452B (en) * | 2023-10-20 | 2024-02-06 | 浙江天演维真网络科技股份有限公司 | Drought state prediction method, and training method and device of drought state prediction model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | Model-free renewable scenario generation using generative adversarial networks | |
| Zeng et al. | Short-term wind power prediction using a wavelet support vector machine | |
| CN116070799B (en) | Photovoltaic power generation amount prediction system and method based on attention and deep learning | |
| Khodayar et al. | Robust deep neural network for wind speed prediction | |
| CN114429153A (en) | Lifetime learning-based gearbox increment fault diagnosis method and system | |
| CN115630731A (en) | Image-based wind power short-term prediction method and system with self-attention mechanism and gating | |
| CN117096867A (en) | Short-term power load prediction method, device, system and storage medium | |
| CN115859816A (en) | Wind power minute-level prediction method and system based on CNN-LSTM algorithm | |
| CN113642716B (en) | Depth variation self-encoder model training method, device, equipment and storage medium | |
| Li et al. | Temporal attention based tcn-bigru model for energy time series forecasting | |
| CN116523145B (en) | Photovoltaic power ultra-short-term prediction method and device, computer equipment and storage medium | |
| CN116860824A (en) | Meteorological data knowledge discovery method and system | |
| CN115392525A (en) | Distributed wind power prediction method, device and equipment | |
| CN111932007B (en) | Power prediction method and device for photovoltaic power station and storage medium | |
| Wei et al. | Using a model structure selection technique to forecast short-term wind speed for a wind power plant in North China | |
| Gawali et al. | Fault prediction model in wind turbines using deep learning structure with enhanced optimisation algorithm | |
| CN114118611A (en) | Wind power prediction method and system based on prediction algorithm pool | |
| Wang et al. | Short‐term prediction of wind power based on temporal convolutional network and the informer model | |
| Dhakal et al. | Deep Learning Model with Probability Density Function and Feature Engineering for Short Term Wind Speed Prediction | |
| Asvestopoulos | Explainable Renewable Energy Forecasting | |
| CN117151303B (en) | Ultra-short-term solar irradiance prediction method and system based on hybrid model | |
| Che et al. | Analysis of data mining method for short-term wind measurement of wind farm based on multi-technology fusion | |
| Jiang et al. | Research on time-series based and similarity search based methods for PV power prediction | |
| Zhang et al. | Hybrid Model Based on Isolated Forest and WT-Catboost for Wind Power Forecasting | |
| CN117148325A (en) | Method, system and device for temporarily replacing fault laser wind-finding radar |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |