CN113779892A - Wind speed and wind direction prediction method - Google Patents

Wind speed and wind direction prediction method Download PDF

Info

Publication number
CN113779892A
CN113779892A CN202111133887.8A CN202111133887A CN113779892A CN 113779892 A CN113779892 A CN 113779892A CN 202111133887 A CN202111133887 A CN 202111133887A CN 113779892 A CN113779892 A CN 113779892A
Authority
CN
China
Prior art keywords
data set
data
network
hidden state
target
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.)
Granted
Application number
CN202111133887.8A
Other languages
Chinese (zh)
Other versions
CN113779892B (en
Inventor
臧增亮
牛丹
陈善龙
陈夕松
李毅
尤伟
潘晓滨
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.)
National University of Defense Technology
Original Assignee
National University of Defense Technology
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 National University of Defense Technology filed Critical National University of Defense Technology
Priority to CN202111133887.8A priority Critical patent/CN113779892B/en
Publication of CN113779892A publication Critical patent/CN113779892A/en
Application granted granted Critical
Publication of CN113779892B publication Critical patent/CN113779892B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/08—Fluids
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a method for predicting wind speed and wind direction, which comprises the steps of acquiring observation data and mode data of a target station, and establishing an initial data set according to the observation data and the mode data; preprocessing the initial data set to obtain a target data set; extracting time characteristic information from the target data set, and extracting and adding two groups of characteristic data from the target data set by using a linear prediction module; in a fusion self-adaptive weight mechanism and an improved long and short memory cell sequence network, cell hidden state information and input information in the long and short memory network are interactively processed; carrying out probability sampling processing on the hidden state information; carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism to generate a network prediction model; and performing ensemble learning on the target data set by using a network prediction model to obtain a final model, and predicting the wind speed and the wind direction by using the final prediction model, so that the wind direction and the wind speed are accurately predicted.

Description

Wind speed and wind direction prediction method
Technical Field
The application relates to the field of renewable energy sources, in particular to a method for predicting wind speed and wind direction.
Background
Wind speed and wind direction are important data in the field of meteorology, generally have the property of being unadjustable and uncontrollable, and influence the normal life of people all the time. Accurate prediction of wind speed and wind direction is an important measure for further improving the utilization rate of wind energy, and wind energy has great potential as an environment-friendly renewable energy source, and plays an important role in improving people's life and promoting resource conservation and utilization.
Currently, in wind speed and wind direction detection, wind energy related data are measured mainly by means of radar equipment and a wind measuring tower; in the field of forecasting wind speed and wind direction, numerical weather forecasting and time series forecasting are the mainstream methods at present. The accuracy of numerical weather forecast in short-term prediction and long-term prediction is low, and the accuracy of time series prediction of a single model in short-term prediction and long-term prediction is also low, so that how to accurately predict wind speed and wind direction becomes a technical problem to be solved urgently.
The above is only for the purpose of assisting understanding of the technical aspects of the present invention, and does not represent an admission that the above is prior art.
Disclosure of Invention
The invention mainly aims to provide a method for predicting wind speed and wind direction, and aims to solve the technical problem that the wind speed and the wind direction cannot be accurately predicted in the prior art.
To achieve the above object, the present invention provides a method for predicting wind speed and wind direction, the method comprising:
acquiring observation data and mode data of a target site, and establishing an initial data set according to the observation data and the mode data;
preprocessing the initial data set to obtain a target data set;
extracting time characteristic information from the target data set, and extracting and adding two groups of characteristic data from the target data set by using a linear prediction module;
in a fusion self-adaptive weight mechanism and an improved long and short memory cell sequence network, carrying out interactive processing on cell hidden state information and input information in the long and short memory network;
carrying out probability sampling processing on the hidden state information;
carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism to generate a network prediction model;
and performing ensemble learning on the target data set by using the network prediction model to obtain a final model, and predicting the wind speed and the wind direction by using the final prediction model.
Optionally, the step of preprocessing the initial data set to obtain a target data set includes:
rejecting data units which do not meet preset conditions in the initial data set;
when the data unit is detected to be missing or wrong, the average value of the data in the corresponding threshold value range is used for supplementing;
and obtaining a target data set according to the processing result.
Optionally, the step of extracting temporal feature information from the target data set and extracting and adding two sets of feature data from the target data set by using a linear prediction module includes:
carrying out specification matching on the mode data in the target data set and the real data, wherein the mode data M1,…,MmAnd observation data O1,…,OnWherein M isiIth feature data representing mode prediction at time tI is more than or equal to 1 and less than or equal to m, m represents the total number of features of the pattern data at the time t, OjJ is more than or equal to 1 and less than or equal to n, and n represents the total number of features of the actual observed data at the time t;
and adding two groups of trend sequences for the observation data of wind speed and wind direction from the target data set by using a trend item model of a linear prediction module and the data tuples of the observation data, arranging tuples for the sequence of the observation data, and adding the tuples into the target data set as two groups of characteristic data.
Optionally, after the step of adding two trend sequences to the wind speed and wind direction observation data from the target data set by using the trend item model of the data tuple using the linear prediction module and the observation data, and arranging tuples of the observation data in order as two sets of feature data into the target data set, the method further includes:
and recombining the newly generated data module and the original data set, aggregating the newly generated data module and the original data set into a new data set tuple, and dividing the new data set tuple into a training set and a test set.
Optionally, the step of performing interactive processing on the hidden state information and the input information of the cell in the long and short memory networks in the fusion adaptive weight mechanism and the improved long and short memory cell sequence network includes:
in the fusion of adaptive weight mechanism and improved long-short memory cell sequence network, the improved cell unit outputs h to the last unitprevAlternately processing with the current cell input x, and defining as Inter (x, h)prev,cprev) The interaction process is (° denotes hadmard product):
Figure BDA0003281347320000031
for odd i e [1 … r ]]
Figure BDA0003281347320000032
for even i e [1 … r ]]
Q and R are extra setting matrixes, and the parameter i controls x and h to carry out interactive calculation;
the values after the extraction operation are input into a sequence unit network, and training is spliced into four states (degree represents hadmard product):
It＝σ(Wixx+Wihhprev+bi)
Ft＝σ(Wfxx+Wfhhprev+bf)
Ot＝σ(Woxx+Wohhprev+bo)
Figure BDA0003281347320000034
Figure BDA0003281347320000035
wherein It，Ft，Ot，CtH respectively represents an input gate, a forgetting gate, an output gate, a memory cell state and a next round of candidate hidden states;
the current memory cell state (c) and the next candidate hidden state (h) are obtained and transmitted to the next network unit.
Optionally, the step of performing probability sampling processing on the hidden state information includes:
setting probability p as a selection real mark aiming at the hidden state (h) of each cell unit in training, and selecting the output of the model by the probability of 1-p;
with the training, the value of p is reduced, that is, the output of the model is selected as much as possible, so that the model training and the prediction are consistent, and the probability p is redefined for each epoch training:
Figure BDA0003281347320000033
and integrating the probability selection learning method into a training network.
Optionally, the step of performing a weighted summation on the hidden state information of the encoding module and the decoding module by using an adaptive weighting mechanism to generate the network prediction model includes:
the method for carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism comprises the following steps: information X ═ X for input1,x2,…,xt]Performing attention-stepping calculation to define AtteniThe steps are for attention:
Atteni＝softmax(score(Xi,q))
Figure BDA0003281347320000041
wherein score (X)iQ) a scoring mechanism for bilinear models;
the specific way of performing weighted average on the processed information is as follows:
Figure BDA0003281347320000042
wherein q is context information;
and generating a network prediction model according to the calculation steps.
According to the method, observation data and mode data of a target site are obtained, and an initial data set is established according to the observation data and the mode data; preprocessing the initial data set to obtain a target data set; extracting time characteristic information from the target data set, and extracting and adding two groups of characteristic data from the target data set by using a linear prediction module; in a fusion self-adaptive weight mechanism and an improved long and short memory cell sequence network, carrying out interactive processing on cell hidden state information and input information in the long and short memory network; carrying out probability sampling processing on the hidden state information; carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism to generate a network prediction model; and performing ensemble learning on the target data set by using the network prediction model to obtain a final model, and predicting the wind speed and the wind direction by using the final prediction model, so that the wind direction and the wind speed are accurately predicted.
Drawings
FIG. 1 is a schematic flow chart illustrating a method for predicting wind speed and wind direction according to a first embodiment of the present invention;
FIG. 2 is a diagram of a network structure based on a fusion adaptive weighting mechanism and an improved memory unit sequence according to a first embodiment of a method for predicting wind speed and wind direction of the present invention;
FIG. 3 is a diagram of a network structure of an improved cell unit according to a first embodiment of the method for predicting wind speed and wind direction of the present invention;
FIG. 4 is a block diagram of an adaptive weighting mechanism module according to a first embodiment of a method for predicting wind speed and wind direction of the present invention.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
An embodiment of the present invention provides a method for predicting a wind speed and a wind direction, and referring to fig. 1, fig. 1 is a schematic flow chart of a first embodiment of the method for predicting a wind speed and a wind direction according to the present invention.
In this embodiment, the method for predicting the wind speed and the wind direction includes the following steps:
step S10: the method comprises the steps of obtaining observation data and mode data of a target site, and establishing an initial data set according to the observation data and the mode data.
The observation data related to the wind speed and the wind direction are from observation data of a Chinese meteorological high-altitude station, the mode data are from Nanjing information engineering university, the data are 5-minute grid data, and the time range is from 1 month in 2020 to 12 months in 2020.
In this embodiment, a network structure diagram based on a fusion adaptive weight mechanism and an improved memory unit sequence as shown in fig. 2 is built.
Step S20: and preprocessing the initial data set to obtain a target data set.
It can be understood that the preprocessing of the initial data set refers to data combination and elimination of data units (e.g., 0, 999999, etc.) of existing abnormal data, and deletion of multiple times of fused repeated data units; for individual dirty or missing data, the mean of the data within the threshold range is used for supplementation;
step S30: temporal feature information is extracted from the target dataset and two sets of feature data are extracted and added from the target dataset using a linear prediction module.
Further, the step of extracting temporal feature information from the target data set and using a linear prediction module to extract and add two sets of feature data from the target data set comprises: carrying out specification matching on the mode data in the target data set and the real data, wherein the mode data M1,…,MmAnd observation data O1,…,OnWherein M isiI is more than or equal to 1 and less than or equal to m, m represents the total number of features of the pattern data at the time t, OjJ is more than or equal to 1 and less than or equal to n, and n represents the total number of features of the actual observed data at the time t; and adding two groups of trend sequences for the observation data of wind speed and wind direction from the target data set by using a trend item model of a linear prediction module and the data tuples of the observation data, arranging tuples for the sequence of the observation data, and adding the tuples into the target data set as two groups of characteristic data.
Further, the step of adding two sets of trend sequences to the wind speed and wind direction observation data by using the trend term model of the data tuples of the linear prediction module and the observation data from the target data set, and adding the sequentially arranged tuples of the observation data as two sets of feature data into the target data set further includes: and recombining the newly generated data module and the original data set, aggregating the newly generated data module and the original data set into a new data set tuple, and dividing the new data set tuple into a training set and a test set.
In the specific implementation, the mode data M is subjected to specification matching with the real data1,…,MmAnd observation data O1,…,On，MiIndicating mode prediction at time ti characteristic data, i is more than or equal to 1 and less than or equal to m, m represents the total number of characteristics of the pattern data at the moment t, OjJ is more than or equal to 1 and less than or equal to n, n represents the total number of features of the actual observed data at the time t, and the value of n is 2, and the value of m is 24.
In specific implementation, after a data set with a uniform specification is obtained, a timestamp of each piece of data is supplemented, detailed time characteristic information (Date, Hour) of the data is extracted and supplemented to each data list, the timeliness of the data set is enhanced, and a new sample data set is formed; adding two groups of trend sequences to the observation data of wind speed and wind direction by using a linear prediction module and a data tuple of the observation data and a trend item model, and arranging tuples of the observation data in sequence to be used as two groups of characteristic data to be added into a data set; and recombining the newly generated data module and the original data set, aggregating the newly generated data module and the original data set into a new data set tuple, and dividing the new data set tuple into a training set and a test set.
Step S40: and in the fusion of an adaptive weight mechanism and the improved long and short memory cell sequence network, carrying out interactive processing on the hidden state information and the input information of the cell in the long and short memory network.
It should be noted that the embodiment shown in fig. 3 proposes an improved cell network structure diagram.
Further, the step of performing interactive processing on the hidden state information and the input information of the cell in the long and short memory network in the fusion adaptive weight mechanism and the improved long and short memory cell sequence network includes: in the fusion of adaptive weight mechanism and improved long-short memory cell sequence network, the improved cell unit outputs h to the last unitprevAlternately processing with the current cell input x, and defining as Inter (x, h)prev,cprev) The interaction process is (° denotes hadmard product):
Figure BDA0003281347320000061
for odd i e [1 … r ]]
Figure BDA0003281347320000062
for even i e [1 … r ]]
Q and R are extra setting matrixes, and the parameter i controls x and h to carry out interactive calculation; the values after the extraction operation are input into a sequence unit network, and training is spliced into four states (degree represents hadmard product):
It＝σ(Wixx+Wihhprev+bi)
Ft＝σ(Wfxx+Wfhhprev+bf)
Ot＝σ(Woxx+Wohhprev+bo)
Figure BDA0003281347320000073
Figure BDA0003281347320000074
wherein It，Ft，Ot，CtH respectively represents an input gate, a forgetting gate, an output gate, a memory cell state and a next round of candidate hidden states; the current memory cell state (c) and the next candidate hidden state (h) are obtained and transmitted to the next network unit.
Step S50: and carrying out probability sampling processing on the hidden state information.
Step S60: and carrying out weighted summation on the hidden state information of the coding module and the decoding module by using an adaptive weight mechanism to generate a network prediction model.
It should be noted that the structure of the adaptive weight mechanism module used in this embodiment is shown in fig. 4.
Further, the step of performing a weighted summation of the hidden state information of the encoding module and the decoding module by using an adaptive weighting mechanism to generate the network prediction model includes: the method for carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism comprises the following steps: for inputInformation X ═ X1,x2,…,xt]Performing attention-stepping calculation to define AtteniThe steps are for attention:
Atteni＝softmax(score(Xi,q))
Figure BDA0003281347320000071
wherein score (X)iQ) a scoring mechanism for bilinear models; the specific way of performing weighted average on the processed information is as follows:
Figure BDA0003281347320000072
wherein q is context information; and generating a network prediction model according to the calculation steps.
Step S70: and performing ensemble learning on the target data set by using the network prediction model to obtain a final model, and predicting the wind speed and the wind direction by using the final prediction model.
In specific implementation, the improved data set is put into a network, and hyper-parameter information (table 2) is defined and trained; correcting and optimizing the problems occurring in the training process to finally obtain a prediction model; and finally, after real data testing, the Root Mean Square Error (RMSE) result is 0.376, which is superior to the plum blossom morning equal prediction model result 0.385 and Liu Yongqian equal prediction model result 0.71.
Table 1 shows the configuration information of the computer platform for operating the wind speed and direction prediction model according to the present invention.
Figure BDA0003281347320000081
Table 1 table 2 shows configuration information of relevant parameters of the model in the present invention.
Name (R) Parameter information
Learning rate 0.0001
epoch 800
Weight attenuation 0.0005
Dimension of input 28
Output dimension 2
TABLE 2
The method comprises the steps of acquiring observation data and mode data of a target site, and establishing an initial data set according to the observation data and the mode data; preprocessing the initial data set to obtain a target data set; extracting time characteristic information from the target data set, and extracting and adding two groups of characteristic data from the target data set by using a linear prediction module; in a fusion self-adaptive weight mechanism and an improved long and short memory cell sequence network, carrying out interactive processing on cell hidden state information and input information in the long and short memory network; carrying out probability sampling processing on the hidden state information; carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism to generate a network prediction model; and performing ensemble learning on the target data set by using the network prediction model to obtain a final model, and predicting the wind speed and the wind direction by using the final prediction model, so that the wind direction and the wind speed are accurately predicted.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., a rom/ram, a magnetic disk, an optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims (7)

1. A method of predicting wind speed and direction, the method comprising:
acquiring observation data and mode data of a target site, and establishing an initial data set according to the observation data and the mode data;
preprocessing the initial data set to obtain a target data set;
extracting time characteristic information from the target data set, and extracting and adding two groups of characteristic data from the target data set by using a linear prediction module;
in a fusion self-adaptive weight mechanism and an improved long and short memory cell sequence network, carrying out interactive processing on cell hidden state information and input information in the long and short memory network;
carrying out probability sampling processing on the hidden state information;
carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism to generate a network prediction model;
and performing ensemble learning on the target data set by using the network prediction model to obtain a final model, and predicting the wind speed and the wind direction by using the final prediction model.
2. The method of claim 1, wherein the step of preprocessing the initial data set to obtain a target data set comprises:
rejecting data units which do not meet preset conditions in the initial data set;
when the data unit is detected to be missing or wrong, the average value of the data in the corresponding threshold value range is used for supplementing;
and obtaining a target data set according to the processing result.
3. The method of claim 1, wherein the step of extracting temporal feature information from the target dataset and using a linear prediction module to extract and add two sets of feature data from the target dataset comprises:
carrying out specification matching on the mode data in the target data set and the real data, wherein the mode data M1，…，MmAnd observation data O1，…，OnWherein M isiI is more than or equal to 1 and less than or equal to m, m represents the total number of features of the pattern data at the time t, OjJ is more than or equal to 1 and less than or equal to n, and n represents the total number of features of the actual observed data at the time t;
and adding two groups of trend sequences for the observation data of wind speed and wind direction from the target data set by using a trend item model of a linear prediction module and the data tuples of the observation data, arranging tuples for the sequence of the observation data, and adding the tuples into the target data set as two groups of characteristic data.
4. The method of claim 2, wherein the step of adding two sets of trend sequences to the wind speed and direction observations using their trend term models from the target data set using the data tuples of linear prediction modules and observations, and arranging the tuples for observation in order, as two sets of feature data to be added to the target data set, further comprises:
and recombining the newly generated data module and the original data set, aggregating the newly generated data module and the original data set into a new data set tuple, and dividing the new data set tuple into a training set and a test set.
5. The method according to claim 1, wherein the step of interactively processing the hidden state information and the input information of the cell in the long and short memory network in the network of fusing the adaptive weight mechanism and the improved long and short memory cell sequence comprises:
in the fusion of adaptive weight mechanism and improved long-short memory cell sequence network, the improved cell unit outputs h to the last unitprevAlternately processing with the current cell input x, and defining as Inter (x, h)prev，cprev) The interaction process is (
Figure FDA0003281347310000023
Representing the hadmard product):
Figure FDA0003281347310000021
odd i ∈ [1.. r ]]
Figure FDA0003281347310000022
Even number i ∈ [1.. r]
Q and R are extra setting matrixes, and the parameter i controls x and h to carry out interactive calculation;
extracting the calculated values as input into a sequence unit network, training and splicing into four states (
Figure FDA0003281347310000024
Representing the hadmard product):
It＝σ(Wixx+Wihhprev+bi)
Ft＝σ(Wfxx+Wfhhprev+bf)
Ot＝σ(Woxx+Wohhprev+bo)
Figure FDA0003281347310000025
Figure FDA0003281347310000026
wherein It，Ft，Ot，CtH respectively represents an input gate, a forgetting gate, an output gate, a memory cell state and a next round of candidate hidden states;
the current memory cell state (c) and the next candidate hidden state (h) are obtained and transmitted to the next network unit.
6. The method of claim 1, wherein the step of probabilistically sampling the hidden state information comprises:
setting probability p as a selection real mark aiming at the hidden state (h) of each cell unit in training, and selecting the output of the model by the probability of 1-p;
with the training, the value of p is reduced, that is, the output of the model is selected as much as possible, so that the model training and the prediction are consistent, and the probability p is redefined for each epoch training:
Figure FDA0003281347310000031
and integrating the probability selection learning method into a training network.
7. The method of claim 1, wherein the step of performing a weighted summation of the hidden state information of the encoding module and the decoding module using an adaptive weighting mechanism to generate the network prediction model comprises:
the method for carrying out weighted summation on the hidden state information of the coding module and the decoding module by utilizing a self-adaptive weight mechanism comprises the following steps: information X ═ X for input1，x2，...，xt]Performing attention-stepping calculation to define AtteniThe steps are for attention:
Atteni＝softmax(score(Xi，q))
Figure FDA0003281347310000032
wherein score (X)iQ) a scoring mechanism for bilinear models;
the specific way of performing weighted average on the processed information is as follows:
Figure FDA0003281347310000033
wherein q is context information;
and generating a network prediction model according to the calculation steps.
CN202111133887.8A 2021-09-27 2021-09-27 Method for predicting wind speed and wind direction Active CN113779892B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111133887.8A CN113779892B (en) 2021-09-27 2021-09-27 Method for predicting wind speed and wind direction

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111133887.8A CN113779892B (en) 2021-09-27 2021-09-27 Method for predicting wind speed and wind direction

Publications (2)

Publication Number Publication Date
CN113779892A true CN113779892A (en) 2021-12-10
CN113779892B CN113779892B (en) 2022-09-02

Family

ID=78853749

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111133887.8A Active CN113779892B (en) 2021-09-27 2021-09-27 Method for predicting wind speed and wind direction

Country Status (1)

Country Link
CN (1) CN113779892B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117022032A (en) * 2023-10-10 2023-11-10 深圳市瑞凯诺科技有限公司 New energy automobile electricity taking method and system based on display screen

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112257850A (en) * 2020-10-26 2021-01-22 河南大学 Vehicle track prediction method based on generation countermeasure network
CN112949945A (en) * 2021-04-15 2021-06-11 河海大学 Wind power ultra-short-term prediction method for improving bidirectional long-short term memory network

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112257850A (en) * 2020-10-26 2021-01-22 河南大学 Vehicle track prediction method based on generation countermeasure network
CN112949945A (en) * 2021-04-15 2021-06-11 河海大学 Wind power ultra-short-term prediction method for improving bidirectional long-short term memory network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
谢林枫等: "基于一种新技术的风电功率短期预测", 《计算机仿真》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117022032A (en) * 2023-10-10 2023-11-10 深圳市瑞凯诺科技有限公司 New energy automobile electricity taking method and system based on display screen
CN117022032B (en) * 2023-10-10 2023-12-29 深圳市瑞凯诺科技有限公司 New energy automobile electricity taking method and system based on display screen

Also Published As

Publication number Publication date
CN113779892B (en) 2022-09-02

Similar Documents

Publication Publication Date Title
CN103996084B (en) Wind power probability Forecasting Methodology based on longitudinal moment Markov chain model
CN112285807A (en) Meteorological information prediction method and device
CN109472283B (en) Dangerous weather prediction method and device based on multiple incremental regression tree model
CN114493052B (en) Multi-model fusion self-adaptive new energy power prediction method and system
CN111178585A (en) Fault reporting amount prediction method based on multi-algorithm model fusion
CN114462718A (en) CNN-GRU wind power prediction method based on time sliding window
CN117335411B (en) Medium-and-long-term power generation capacity prediction method for photovoltaic power station group
CN114330934A (en) Model parameter self-adaptive GRU new energy short-term power generation power prediction method
CN117909888B (en) Intelligent artificial intelligence climate prediction method
CN113779892B (en) Method for predicting wind speed and wind direction
CN115032720A (en) Application of multi-mode integrated forecast based on random forest in ground air temperature forecast
CN117117819A (en) Photovoltaic power generation short-term power prediction method, system, equipment and medium
CN114819385A (en) Wind power prediction method and device, electronic equipment and storage medium
CN116663393A (en) Random forest-based power distribution network continuous high-temperature fault risk level prediction method
CN117033923A (en) Method and system for predicting crime quantity based on interpretable machine learning
CN112434750B (en) Dam monitoring data development mode identification method based on convolutional neural network
CN112580899A (en) Medium-and-long-term wind power generation prediction method and system fused with machine learning model
CN115209441A (en) Method, device, equipment and storage medium for predicting base station out-of-service alarm
CN117688367B (en) Wind power generation ultra-short term power prediction method and device based on instant learning
CN113919533A (en) Air quality tracing and forecasting method based on artificial intelligence
CN117437230B (en) Photovoltaic power station power prediction method and system based on image restoration
KR102685543B1 (en) A system for determining the optimal bid amount of a stochastic scenario technique using weather forecast data
CN114612147A (en) Power price prediction method for subgraph learning
Liu et al. Predicting Power Outage at Low-Lying Area Substations During Storm Surge Disasters Using Multi-Grained Cascaded Forest
CN115983434A (en) Short-term wind field wind speed prediction method and system based on neural network

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
GR01 Patent grant
GR01 Patent grant