CN116384576A - Wind speed prediction method, device, system and storage medium - Google Patents

Wind speed prediction method, device, system and storage medium Download PDF

Info

Publication number
CN116384576A
CN116384576A CN202310361092.5A CN202310361092A CN116384576A CN 116384576 A CN116384576 A CN 116384576A CN 202310361092 A CN202310361092 A CN 202310361092A CN 116384576 A CN116384576 A CN 116384576A
Authority
CN
China
Prior art keywords
wind speed
speed prediction
meteorological data
matrix
layer
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
Application number
CN202310361092.5A
Other languages
Chinese (zh)
Inventor
孙科学
徐俊杰
孙立
王艳
张瑛
成谢锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202310361092.5A priority Critical patent/CN116384576A/en
Publication of CN116384576A publication Critical patent/CN116384576A/en
Pending legal-status Critical Current

Links

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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The invention discloses a wind speed prediction method, a device, a system and a storage medium, which solve the problem that effective characteristic information in wind speed data in a wind power plant cannot be fully excavated in the prior art, and comprise the steps of acquiring meteorological data of adjacent multiple wind power plants; performing error elimination pretreatment on the meteorological data; inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result; according to the wind speed prediction method, the spatial correlation between adjacent wind power plants can be considered in the wind speed prediction process, the capability of extracting effective characteristic information of a wind speed prediction model is improved, and the prediction efficiency and the prediction accuracy are improved.

Description

Wind speed prediction method, device, system and storage medium
Technical Field
The invention relates to a wind speed prediction method, a device, a system and a storage medium, and belongs to the technical field of wind speed prediction.
Background
With the increasing prominence of energy situation and environmental problems, the development of clean energy has now become a hotspot problem that is emphasized in various developed countries around the world. Wind power is used as a main part of new energy, and research and application of wind power have become main research targets of scientists. With the continuous development of wind power generation, the negative effects caused by the intermittence and uncertainty of wind speed are gradually revealed, and the fluctuation of wind power has serious influence on the safety of a power grid. The accurate wind power prediction can provide more comprehensive information for the management of the wind power plant, so that the flexibility and the optimality of the control management of the wind power plant are improved, and the method has great strategic significance for the safe operation and the economic benefit of the national power grid.
The prediction of wind power and wind speed is not separated from a scientific prediction model and an effective theoretical method. In recent years, with the rapid development of deep learning technology, data information mining methods have also begun to transition from conventional shallow information layer to deep mining based on information. Scientists at home and abroad have done a lot of work on wind speed prediction and many solutions are proposed for practical problems. In the prior art, there are many methods for predicting wind speed, the prediction accuracy is also gradually improved, and common wind speed predictions are as follows: the univariate model considers characteristics of a single wind farm, and the multivariate model considers time correlation of adjacent multiple wind farms. But there is also a lack of an efficient way to consider both the temporal and spatial correlation between adjacent multiple wind farms.
The current common wind speed predictions suffer from the following disadvantages: the univariate model only considers the characteristics of a single wind power plant, the spatial correlation of adjacent multi-wind power plants is ignored, the multivariate model considers the time correlation of adjacent multi-wind power plants, the spatial correlation inside the multi-wind power plants is ignored, and the problem that effective characteristic information in wind speed data in the wind power plants cannot be fully mined exists.
Disclosure of Invention
The invention aims to provide a wind speed prediction method, a device, a system and a storage medium, which solve the problem that effective characteristic information in wind speed data in a wind power plant cannot be fully mined in the prior art.
In order to achieve the above purpose, the invention is realized by adopting the following technical scheme:
in a first aspect, the present invention provides a wind speed prediction method comprising:
acquiring meteorological data of adjacent multiple wind power plants;
performing error elimination pretreatment on the meteorological data;
and inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
With reference to the first aspect, further, performing error elimination preprocessing on the meteorological data includes:
and cleaning and standardizing the meteorological data to obtain the meteorological data with standard normal distribution.
With reference to the first aspect, further, the wind speed prediction model includes:
the graph convolutional neural network is used for extracting the spatial characteristics of meteorological data of adjacent multi-wind power plants;
the gating circulation unit is used for extracting time characteristics of meteorological data of adjacent multiple wind farms;
the attention mechanism is arranged in the gating circulating unit and is used for improving the capability of the wind speed prediction model for extracting effective information;
and the full connection layer is used for converting the input spatial characteristics and the input time characteristics into wind speed prediction results.
With reference to the first aspect, further, a formula for calculating a spatial feature in the graph convolution neural network is:
Figure BDA0004165178170000021
wherein H is (l+1) Is the output matrix of the layer 1 network, H (l) Is the input matrix of the layer 1 network,
Figure BDA0004165178170000031
represented is a Laplace matrix, +.>
Figure BDA0004165178170000032
A is the adjacency matrix of the undirected graph, I N Is an n-order identity matrix>
Figure BDA0004165178170000033
A degree matrix of the undirected graph, sigma is a ReLU activation function, theta (l) Is the parameter matrix of the first layer and the first layer, and the output matrix of the last layer of network is the space characteristic;
the adjacency matrix of the undirected graph is obtained by the following method: and acquiring historical wind speed measured values of adjacent multiple wind power plants, constructing a complex adjacency matrix according to the historical wind speed measured values, and taking the complex adjacency matrix as an adjacency matrix of the undirected graph.
With reference to the first aspect, further, the calculation process of the gating cycle unit is as follows:
R t ＝σ(W xi X t +W hi h t-1 +b i )
Z t ＝σ(W xf X t +W hi h t-1 +b f )
d t ＝σ(W xc X t +W hc (h t-1 ·R t )+b c )
Figure BDA0004165178170000034
Figure BDA0004165178170000035
wherein R is t Is the output result of the reset gate, Z t Is the output result of the update gate, d t Is the input of the hidden layer, X t Is the input information at the t-th time step, h t-1 Is the memory information at the t-1 time step,
Figure BDA0004165178170000036
is the memory information at the t-th time step, h t Is the final memory information at the t-th time step, W xi 、W xf And W is xc The weights of the input information in the reset gate, update gate and hidden layer, b i 、b f And b c The method comprises the steps of resetting bias items in a gate, updating the gate and hiding a layer respectively, wherein sigma is a ReLU activation function, tanh is a tanh activation function, and final memory information corresponding to the last time step is a time feature.
With reference to the first aspect, further, the calculation process of the attention mechanism includes:
u t ＝tanh(W h h t +b h )
Figure BDA0004165178170000037
Figure BDA0004165178170000041
wherein u is t Is the t-th moment hidden state h t Corresponding attention score value, tanh is tanh activation function, W h Is h t Weights of (h), h t Is the final memory information at the t-th time step, b h Is h t Bias term, a t Is performed on all u by the Softmax function t Normalized weighting coefficients, exp, represent an exponential function based on a natural constantT represents a transpose, v is u t S is the feature vector obtained by multiplying the attention probability weight with the hidden layer state of the history input node and accumulating.
With reference to the first aspect, further, the meteorological data includes an ambient temperature, an ambient humidity, and an illumination intensity.
In a second aspect, the present invention also provides a wind speed prediction apparatus comprising:
and a data acquisition module: the method comprises the steps of acquiring meteorological data of adjacent multiple wind farms;
and a data preprocessing module: the method comprises the steps of performing error elimination pretreatment on meteorological data;
wind speed prediction module: the method is used for inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
In a third aspect, the present invention also provides a wind speed prediction system comprising:
a memory for storing instructions;
a processor for executing the instructions to cause the system to perform operations implementing the wind speed prediction method according to any one of the first aspects.
In a fourth aspect, the present invention also provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a wind speed prediction method as in any of the first aspects.
Compared with the prior art, the invention has the following beneficial effects:
according to the wind speed prediction method, the device, the system and the storage medium, the acquired meteorological data comprise the meteorological data of the adjacent multiple wind power plants, the spatial correlation between the adjacent wind power plants is considered when the spatial characteristics are extracted from the trained wind speed prediction model, the capability of the wind speed prediction model for extracting effective characteristic information is improved, and the prediction efficiency and the prediction accuracy are improved;
the graph convolution neural network is used as a spatial feature extraction module to extract complex spatial features, the gating circulation unit is used as a time feature extraction module to extract complex time features, and attention mechanisms are used to distinguish the importance degrees of the features at different time points, so that the wind speed prediction of the multi-wind power plant is finally realized, and the accuracy of the prediction is improved.
Drawings
FIG. 1 is a flowchart of a method for predicting wind speed according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a second method for predicting wind speed according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of extracting spatial features according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of extracting temporal features according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of the attention mechanism provided by an embodiment of the present invention;
fig. 6 is a schematic structural diagram of temporal feature extraction and attention mechanism fusion provided by an embodiment of the present invention.
Detailed Description
The present invention will be further described with reference to the accompanying drawings, and the following examples are only for more clearly illustrating the technical aspects of the present invention, and are not to be construed as limiting the scope of the present invention.
Example 1
As shown in fig. 1, an embodiment of the present invention provides a wind speed prediction method, including the steps of:
s1, acquiring meteorological data of adjacent multiple wind farms.
The embodiment also obtains the historical wind speed data of the adjacent multi-wind power plant, wherein the historical wind speed data comprises the wind speed and the wind speed direction, and the meteorological data comprises the ambient temperature, the ambient humidity and the illumination intensity.
S2, performing error elimination pretreatment on the meteorological data.
And cleaning and standardizing the acquired meteorological data, converting each data into standard normal distribution, and eliminating errors caused by different dimensions.
S3, inputting the meteorological data subjected to the error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
Because of the geographic difference between adjacent wind power plants, the wind speed correlation between the adjacent wind power plants also has the difference, the embodiment constructs a complex adjacency matrix which can simultaneously represent the space-time correlation between the adjacent wind power plants, and takes the complex adjacency matrix as an adjacency matrix required by a graph convolution neural network, and the characteristics between the adjacent multiple wind power plants are extracted through the complex adjacency matrix; the complex adjacency matrix designed by carrying out correlation modeling according to the wind speed sequence can simultaneously represent the time and space correlation of wind speeds among a plurality of wind power stations, so that the characteristic extraction of the graph convolution neural network is not limited to the same moment any more, and the space-time difference existing among adjacent wind power stations can be described.
The wind speed prediction model takes a complex adjacent matrix as an adjacent matrix required by the graph convolution neural network, and uses a two-layer graph convolution neural network to perform feature extraction on the preprocessed meteorological data of the adjacent multi-wind fields to obtain the spatial features among the adjacent multi-wind fields. The calculation formula is as follows:
Figure BDA0004165178170000061
wherein f (X, A) is the output result of the graph roll-up neural network,
Figure BDA0004165178170000062
represented is a laplace matrix;
Figure BDA0004165178170000063
a is an adjacency matrix of the undirected graph; i N Is an n-order identity matrix. />
Figure BDA0004165178170000064
A degree matrix for the undirected graph; sigma is a ReLU activation function, θ (0) 、θ (1) Respectively a two-layer convolved weight matrix.
The method comprises the step of cutting a K-order polynomial into a 1-order polynomial by using a Chebyshev polynomial approximation and a first-order approximation in a graph convolution neural network. In matrix multiplication convolution operation, the complexity of matrix calculation and feature decomposition is too large, and the spatial locality is ignored, so that the convolution kernel is approximated by using the weight coefficient of a filter and using chebyshev inequality, and the formula is as follows:
Figure BDA0004165178170000071
Figure BDA0004165178170000072
wherein g θ (Λ) is a eigenvalue function of the Laplace matrix L, K is the graph convolution kernel size, θ k Is a polynomial coefficient, Λ k Is a corresponding eigenvalue matrix, X is an input matrix, U is an eigenvector matrix, T is a transpose, L is a Laplacian matrix, L k Is the K-th power of the laplace matrix L.
To reduce complexity, g θ The parameterization of (Λ) is chebyshev polynomials of the formula:
Figure BDA0004165178170000073
Figure BDA0004165178170000074
wherein g θ (L) is a eigenvalue function parameterized as a Chebyshev polynomial, T k Is the K-th order of the chebyshev polynomial,
Figure BDA0004165178170000079
is a K-order approximation function in Laplace polynomial, the time complexity is linear with the edge, lambda max Refers to the maximum value of the laplace matrix L. L (L) sym Is a symmetric normalization of the laplace matrix L.
To reduce the operation cost, let k=1, λ max =2, the resulting convolution expression is as follows:
Figure BDA0004165178170000075
wherein H is (l+1) Is the output matrix of the layer 1 network, H (l) Is the input matrix of the layer 1 network,
Figure BDA0004165178170000076
represented is a Laplace matrix, +.>
Figure BDA0004165178170000077
A is the adjacency matrix of the undirected graph, I N Is an n-order identity matrix>
Figure BDA0004165178170000078
A degree matrix of the undirected graph, sigma is a ReLU activation function, theta (l) The parameter matrix is a parameter matrix of the first layer and the parameter matrix of the last layer is a spatial characteristic, and after being approximated by a Chebyshev polynomial, the node characteristic in the 1 st-order neighborhood is reserved, and the extraction capacity of the spatial characteristic is improved.
In fig. 3, the GCN (graph convolutional neural network) of the first layer can be used to extract information on the central point 1 near the vertices 2, 3, 4, 5, and by multi-layer stacking, the receptive field of the convolutional layer becomes larger with the increase of the convolutional layer and a more abstract representation of the information is obtained. After passing through the two layers of GCN, the center point 1 obtains vertex information on adjacent vertices 2, 3, 4, 5, 6, 7.
As shown in fig. 2, the spatial features extracted from the graph convolutional neural network are input into a gating cycle unit, and the temporal features are extracted.
The calculation process of the gating cycle unit is as follows:
R t ＝σ(W xi X t +W hi h t-1 +b i )
Z t ＝σ(W xf X t +W hi h t-1 +b f )
d t ＝σ(W xc X t +W hc (h t-1 ·R t )+b c )
Figure BDA0004165178170000081
Figure BDA0004165178170000082
wherein R is t Is the output result of the reset gate, Z t Is the output result of the update gate, d t Is the input of the hidden layer, X t Is the input information at the t-th time step, h t-1 Is the memory information at the t-1 time step,
Figure BDA0004165178170000083
is the memory information at the t-th time step, h t Is the final memory information at the t-th time step, W xi 、W xf And W is xc The weights of the input information in the reset gate, update gate and hidden layer, b i 、b f And b c The method comprises the steps of resetting bias items in a gate, updating the gate and hiding a layer respectively, wherein sigma is a ReLU activation function, tanh is a tanh activation function, and final memory information corresponding to the last time step is a time feature.
And the wind speed prediction model inputs the spatial characteristics into a gating circulation unit to obtain the time characteristics between adjacent multi-wind fields. Reset gate R in FIG. 4 t Results of (a) and previous time step information h t-1 Multiplying. h is a t Representing the final memory information on the current time step, which transfers the information retaining the current time step to the next cell by updating the gate Z t Determining the memory content of the current time step
Figure BDA0004165178170000084
And memorizing the content h in the previous time step t-1 The information that needs to be acquired. Z is Z t Is the result of the activation output of the update gate, which is equal to h t-1 Representing the memory content stored from the previous time step to the current time step and then associated with the current time stepMemory content->
Figure BDA0004165178170000091
And the matrix products of (2) are added to obtain final output information.
The attention mechanism can extract importance levels of different feature states based on past time, weight past feature states in the prediction model, and focus key information. The gating circulation unit network is used as a time correlation basic network structure, and the integration attention mechanism can better enable the model to extract effective information in input data, so that the accuracy and efficiency of a prediction task are improved. S in FIG. 5 1 ，s 2 ，...，s t For the current input sequence, h 1 ，h 2 ，...，h t Maintaining hidden layer state values, a, corresponding to the current input sequence t Attention weight to the current input sequence for hidden state of historical input. The attention mechanism can extract importance levels of different feature states based on the past time, weighting the past feature states in the predictive model.
In FIG. 6, for each hidden state output by the gating cyclic element, the attention mechanism unit scores each hidden state according to its operation logic to obtain the importance ratio of each hidden state, and further obtains the weight coefficient { α } normalized by the Softmax function 1 ，α 2 ，...，α t This weight coefficient corresponds to each hidden state. And finally, multiplying and adding the weight coefficients of all the hidden states to obtain the final output of the states.
After the spatial features and the time features are extracted, the spatial features and the time features are input into a full-connection layer, and a final wind speed prediction result is obtained.
In this embodiment, the training process of the wind speed prediction model is as follows: the training effect of the neural network has a larger relationship with the number of hidden layer nodes of the neural network and the number of hidden layer nodes of the neural network, and the number of hidden layer nodes is related to the input and output dimensions. In the model of this example, the numbers of hidden layer nodes in the GCN layer and the GRU layer are both designed to be equal in size. The optimal model parameters are found through multiple experiments, the number of GCN hidden layer nodes of the GCN-GRUA model is set to 64, the batch size is set to 64, the learning rate is set to 0.001, and the iteration times are set to 1000 times.
An algorithm model with a certain representativeness in wind speed prediction is selected as a comparison model, wherein the algorithm model comprises the following components:
(1) MLP: the multilayer perceptron is a simple multilayer fully-connected neural network.
(2) SVR: the support vector machine is a classical machine-learned prediction algorithm.
(3) GRU: and (5) gating the circulating neural network, and acquiring time related information by using the GRU. The importance of the time information in the wind speed prediction is checked by comparing it.
(4) CNN: convolutional neural networks, which employ CNNs to extract spatial correlation, are compared to check the importance of spatial information in wind speed predictions.
(5) CNN-GRU: and acquiring information of the spatial correlation through the CNN, and acquiring a model of the information related to time through a gating circulation unit.
(6) GCN-GRU: and extracting information of the spatial correlation through the GCN, and acquiring a model of the information related to time through a gating circulation unit.
In order to evaluate the Accuracy of the prediction method given in this embodiment, three evaluation indexes of Root Mean Square Error (RMSE), mean Absolute Error (MAE), and Accuracy (Accuracy) are selected for the prediction error.
TABLE 1 evaluation index values of different prediction methods
Figure BDA0004165178170000101
Through comprehensive comparison with other models, the wind speed prediction method provided by the embodiment has obviously better performance than a comparison algorithm and better prediction effect.
Examples
The embodiment of the invention also provides a wind speed prediction device, which comprises:
and a data acquisition module: the method comprises the steps of acquiring meteorological data of adjacent multiple wind farms;
and a data preprocessing module: the method comprises the steps of performing error elimination pretreatment on meteorological data;
wind speed prediction module: the method is used for inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
Example 3
The embodiment of the invention provides a wind speed prediction system, which comprises:
a memory for storing instructions;
a processor for executing the instructions to cause the system to perform operations implementing a wind speed prediction method as follows:
acquiring meteorological data of adjacent multiple wind power plants;
performing error elimination pretreatment on the meteorological data;
and inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
Example 4
Embodiments of the present invention provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a wind speed prediction method of:
acquiring meteorological data of adjacent multiple wind power plants;
performing error elimination pretreatment on the meteorological data;
and inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that modifications and variations could be made by those skilled in the art without departing from the technical principles of the present invention, and such modifications and variations should also be regarded as being within the scope of the invention.

Claims (10)

1. A method of predicting wind speed, comprising:
acquiring meteorological data of adjacent multiple wind power plants;
performing error elimination pretreatment on the meteorological data;
and inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
2. The method of claim 1, wherein the error-canceling pre-processing of the meteorological data comprises:
and cleaning and standardizing the meteorological data to obtain the meteorological data with standard normal distribution.
3. The method of claim 1, wherein the wind speed prediction model comprises:
the graph convolutional neural network is used for extracting the spatial characteristics of meteorological data of adjacent multi-wind power plants;
the gating circulation unit is used for extracting time characteristics of meteorological data of adjacent multiple wind farms;
the attention mechanism is arranged in the gating circulating unit and is used for improving the capability of the wind speed prediction model for extracting effective information;
and the full connection layer is used for converting the input spatial characteristics and the input time characteristics into wind speed prediction results.
4. A method of predicting wind speed according to claim 3, wherein the formula for calculating spatial features in the graph convolution neural network is:
Figure FDA0004165178160000011
wherein H is (l+1) Is the output matrix of the layer 1 network, H (l) Is the input matrix of the layer 1 network,
Figure FDA0004165178160000012
represented is a Laplace matrix, +.>
Figure FDA0004165178160000013
A is the adjacency matrix of the undirected graph, I N Is an n-order identity matrix>
Figure FDA0004165178160000014
A degree matrix of the undirected graph, sigma is a ReLU activation function, theta (l) Is the parameter matrix of the first layer and the first layer, and the output matrix of the last layer of network is the space characteristic;
the adjacency matrix of the undirected graph is obtained by the following method: and acquiring historical wind speed measured values of adjacent multiple wind power plants, constructing a complex adjacency matrix according to the historical wind speed measured values, and taking the complex adjacency matrix as an adjacency matrix of the undirected graph.
5. A method of predicting wind speed according to claim 3, wherein the gating cycle unit is calculated as follows:
R t ＝σ(W xi X t +W hi h t-1 +b i )
Z t ＝σ(W xf X t +W hi h t-1 +b f )
d t ＝σ(W xc X t +W hc (h t-1 ·R t )+b c )
Figure FDA0004165178160000021
Figure FDA0004165178160000022
wherein R is t Is the output result of the reset gate, Z t Is the output result of the update gate, d t Is the input of the hidden layer, X t Is the input information at the t-th time step, h t-1 Is the memory information at the t-1 time step,
Figure FDA0004165178160000023
is the memory information at the t-th time step, h t Is the final memory information at the t-th time step, W xi 、W xf And W is xc The weights of the input information in the reset gate, update gate and hidden layer, b i 、b f And b c The method comprises the steps of resetting bias items in a gate, updating the gate and hiding a layer respectively, wherein sigma is a ReLU activation function, tanh is a tanh activation function, and final memory information corresponding to the last time step is a time feature.
6. The method of claim 5, wherein the calculation of the attention mechanism comprises:
u t ＝tanh(W h h t +b h )
Figure FDA0004165178160000024
Figure FDA0004165178160000031
wherein u is t Is the t-th moment hidden state h t Corresponding attention score value, tanh is tanh activation function, W h Is h t Weights of (h), h t Is the final memory information at the t-th time step, b h Is h t Bias term, a t Is performed on all u by the Softmax function t The normalized weight coefficient exp represents an exponential function based on a natural constant, T represents a transpose, and v is u t S is the feature vector obtained by multiplying the attention probability weight with the hidden layer state of the history input node and accumulating.
7. The method of claim 1, wherein the meteorological data comprises ambient temperature, ambient humidity, and illumination intensity.
8. Wind speed prediction device, characterized by comprising:
and a data acquisition module: the method comprises the steps of acquiring meteorological data of adjacent multiple wind farms;
and a data preprocessing module: the method comprises the steps of performing error elimination pretreatment on meteorological data;
wind speed prediction module: the method is used for inputting the meteorological data subjected to error elimination pretreatment into a trained wind speed prediction model to obtain a wind speed prediction result.
9. A wind speed prediction system, comprising:
a memory for storing instructions;
a processor for executing the instructions to cause the system to perform operations implementing a wind speed prediction method as claimed in any one of claims 1 to 7.
10. Computer readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements a wind speed prediction method according to any of claims 1-7.
CN202310361092.5A 2023-04-04 2023-04-04 Wind speed prediction method, device, system and storage medium Pending CN116384576A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310361092.5A CN116384576A (en) 2023-04-04 2023-04-04 Wind speed prediction method, device, system and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310361092.5A CN116384576A (en) 2023-04-04 2023-04-04 Wind speed prediction method, device, system and storage medium

Publications (1)

Publication Number Publication Date
CN116384576A true CN116384576A (en) 2023-07-04

Family

ID=86965229

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310361092.5A Pending CN116384576A (en) 2023-04-04 2023-04-04 Wind speed prediction method, device, system and storage medium

Country Status (1)

Country Link
CN (1) CN116384576A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116894384A (en) * 2023-07-11 2023-10-17 湖北工业大学 Multi-fan wind speed space-time prediction method and system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116894384A (en) * 2023-07-11 2023-10-17 湖北工业大学 Multi-fan wind speed space-time prediction method and system
CN116894384B (en) * 2023-07-11 2024-05-07 湖北工业大学 Multi-fan wind speed space-time prediction method and system

Similar Documents

Publication Publication Date Title
Wu et al. A novel hybrid system based on multi-objective optimization for wind speed forecasting
CN108448610B (en) Short-term wind power prediction method based on deep learning
Ma et al. A hybrid attention-based deep learning approach for wind power prediction
CN115018021B (en) Machine room abnormity detection method and device based on graph structure and abnormity attention mechanism
CN111815033A (en) Offshore wind power prediction method based on RCNN and meteorological time sequence characteristics
CN113011570B (en) Facial expression recognition method adopting neural network compression system
CN111027772A (en) Multi-factor short-term load prediction method based on PCA-DBILSTM
CN112100911B (en) Solar radiation prediction method based on depth BILSTM
CN115275991A (en) Active power distribution network operation situation prediction method based on IEMD-TA-LSTM model
CN114781744A (en) Deep learning multi-step long radiance prediction method based on codec
CN116384576A (en) Wind speed prediction method, device, system and storage medium
CN110738363B (en) Photovoltaic power generation power prediction method
CN112906935A (en) Method for predicting ultra-short-term power of wind power plant
CN116960978A (en) Offshore wind power prediction method based on wind speed-power combination decomposition reconstruction
CN116894504A (en) Wind power cluster power ultra-short-term prediction model establishment method
CN114118401A (en) Neural network-based power distribution network flow prediction method, system, device and storage medium
Chen et al. Short-term Wind Speed Forecasting Based on Singular Spectrum Analysis, Fuzzy C-Means Clustering, and Improved POABP
CN116628488A (en) Training method of wind power prediction model, wind power prediction method and device
CN114997360B (en) Evolution parameter optimization method, system and storage medium of neural architecture search algorithm
KR102583178B1 (en) Method and apparatus for predicting power generation
CN115713144A (en) Short-term wind speed multi-step prediction method based on combined CGRU model
CN114139783A (en) Wind power short-term power prediction method and device based on nonlinear weighted combination
CN114239945A (en) Short-term power load prediction method, device, equipment and storage medium
Teslyuk et al. Synthesis of Neurocontroller for Intellectualization Tasks of Process Control Systems
Silva et al. Application of a hybrid neural fuzzy inference system to forecast solar intensity

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