CN111461416B - Wind speed prediction method, system, electronic equipment and storage medium - Google Patents

Wind speed prediction method, system, electronic equipment and storage medium Download PDF

Info

Publication number
CN111461416B
CN111461416B CN202010208630.3A CN202010208630A CN111461416B CN 111461416 B CN111461416 B CN 111461416B CN 202010208630 A CN202010208630 A CN 202010208630A CN 111461416 B CN111461416 B CN 111461416B
Authority
CN
China
Prior art keywords
wind speed
sequence
sub
wind
training
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.)
Active
Application number
CN202010208630.3A
Other languages
Chinese (zh)
Other versions
CN111461416A (en
Inventor
成骁彬
蒋勇
许王建
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.)
Shanghai Electric Wind Power Group Co Ltd
Original Assignee
Shanghai Electric Wind Power Group Co Ltd
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 Shanghai Electric Wind Power Group Co Ltd filed Critical Shanghai Electric Wind Power Group Co Ltd
Priority to CN202010208630.3A priority Critical patent/CN111461416B/en
Publication of CN111461416A publication Critical patent/CN111461416A/en
Application granted granted Critical
Publication of CN111461416B publication Critical patent/CN111461416B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • 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
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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
    • 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)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Economics (AREA)
  • Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Marketing (AREA)
  • Computational Linguistics (AREA)
  • General Business, Economics & Management (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Operations Research (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Quality & Reliability (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Development Economics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Control Of Eletrric Generators (AREA)

Abstract

The invention discloses a wind speed prediction method, a wind speed prediction system, electronic equipment and a storage medium. The wind speed prediction method comprises the following steps: acquiring an initial wind speed sequence comprising wind speeds acquired in N continuous periods; training a wind speed prediction model for outputting a predicted wind speed of the Mth period according to wind speeds acquired from the Mth-i to Mth-1 th periods according to the initial wind speed sequence; outputting the predicted wind speed of the period to be predicted according to the wind speeds acquired by i periods before the period to be predicted and the wind speed prediction model obtained through training; monitoring whether the difference between the predicted wind speed and the actual wind speed of the period to be predicted exceeds the control limit of the period to be predicted based on the EWMA control chart; if yes, retraining the wind speed prediction model. According to the invention, the prediction effect of the wind speed prediction model is monitored based on the EWMA control chart, and the wind speed prediction model is retrained when the prediction effect is monitored to be poor, so that the wind speed prediction model can be maintained at a better prediction effect.

Description

Wind speed prediction method, system, electronic equipment and storage medium
Technical Field
The present invention relates to the field of power technologies, and in particular, to a wind speed prediction method, a wind speed prediction system, an electronic device, and a storage medium.
Background
Based on the accurate prediction of the short-time wind speed of the environment where the wind field is located, the fan control system is operated and changed to adjust the specific position (such as direction and angle) of the fan in the wind field, so that the maximum power generation amount of the wind field is improved, however, when the prediction effect is poor, the current wind speed prediction model is difficult to automatically update, so that the current wind speed prediction model is difficult to maintain a good prediction effect and has low automation degree.
Disclosure of Invention
The invention aims to overcome the defects that a wind speed prediction model in the prior art is difficult to maintain a better prediction effect and has low automation degree, and provides a wind speed prediction method, a wind speed prediction system, electronic equipment and a storage medium.
The invention solves the technical problems by the following technical scheme:
a method of wind speed prediction, comprising:
acquiring an initial wind speed sequence, wherein the initial wind speed sequence comprises wind speeds acquired in N continuous periods;
training a wind speed prediction model according to the initial wind speed sequence, wherein the wind speed prediction model is used for outputting the predicted wind speed of the Mth period according to the wind speeds acquired from the Mth period to the Mth period;
Acquiring wind speeds acquired in i periods before a period to be predicted;
outputting the predicted wind speed of the period to be predicted according to the wind speeds acquired in i periods before the period to be predicted and a wind speed prediction model obtained through training;
acquiring the actual wind speed acquired in the period to be predicted;
monitoring whether a difference between a predicted wind speed and an actual wind speed of the period to be predicted exceeds a control limit of an EWMA control chart based on the EWMA (Exponentially Weighted Moving Average) control chart;
if yes, the step of acquiring an initial wind speed sequence and the step of training a wind speed prediction model according to the initial wind speed sequence are executed again so as to retrain the wind speed prediction model;
wherein N is greater than or equal to M > i, and N, M and i are both positive integers.
Preferably, the step of training a wind speed prediction model according to the initial wind speed sequence specifically includes:
respectively constructing a wind speed derivative sequence according to each wind speed in the initial wind speed sequence;
training a wind speed prediction model according to the wind speed derivative sequence; wherein:
the wind speed derivative sequence comprises wind speeds and derivative data corresponding to the wind speeds;
the wind speed prediction model is used for outputting the predicted wind speed of the Mth period according to a wind speed derivative sequence corresponding to the wind speeds acquired from the Mth period to the Mth period.
Preferably, the step of constructing a wind speed derivative sequence from each wind speed in the initial wind speed sequence comprises:
decomposing wind speed to obtain a sub wind speed sequence, wherein the sub wind speed sequence comprises a plurality of sub wind speeds;
and splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence.
Preferably, after the step of decomposing the wind speed to obtain the sub-wind speed sequence, the method further comprises:
training a SVR (Support Vactor Regerssion, support vector regression) model by taking a sub-wind speed sequence as input and the wind speed corresponding to the sub-wind speed sequence as output;
inputting a sub wind speed sequence into a SVR model obtained by training to output SVR wind speed corresponding to the sub wind speed sequence;
the step of splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence comprises the following steps:
splicing wind speeds, sub-wind speed sequences corresponding to the wind speeds and SVR wind speeds corresponding to the sub-wind speed sequences to obtain wind speed derivative sequences;
or alternatively, the first and second heat exchangers may be,
the step of decomposing the wind speed to obtain a sub-wind speed sequence further comprises the following steps:
training an RF (Random Forest) model by taking a sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output;
Inputting a sub wind speed sequence into a trained RF model to output RF wind speed corresponding to the sub wind speed sequence;
the step of splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence comprises the following steps:
splicing wind speeds, sub-wind speed sequences corresponding to the wind speeds and RF wind speeds corresponding to the sub-wind speed sequences to obtain wind speed derivative sequences;
or alternatively, the first and second heat exchangers may be,
the step of decomposing the wind speed to obtain a sub-wind speed sequence further comprises the following steps:
training an SVR model by taking a sub-wind speed sequence as input and the wind speed corresponding to the sub-wind speed sequence as output;
inputting a sub wind speed sequence into a SVR model obtained by training to output SVR wind speed corresponding to the sub wind speed sequence;
training an RF model by taking a sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output;
inputting a sub wind speed sequence into a trained RF model to output RF wind speed corresponding to the sub wind speed sequence;
the step of splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence comprises the following steps:
splicing wind speeds, sub wind speed sequences corresponding to the wind speeds, SVR wind speeds corresponding to the sub wind speed sequences and RF wind speeds corresponding to the sub wind speed sequences to obtain wind speed derivative sequences.
Preferably, the step of training a wind speed prediction model according to the initial wind speed sequence comprises:
training a wind speed prediction model by using an LSTM (Long Short Term Memory, long-term memory network) model based on an attention mechanism according to the initial wind speed sequence;
and/or the number of the groups of groups,
parameters of the wind speed prediction model are optimized based on DE.
A wind speed prediction system, comprising:
the first acquisition module is used for acquiring an initial wind speed sequence, wherein the initial wind speed sequence comprises wind speeds acquired in N continuous periods;
the training module is used for training a wind speed prediction model according to the initial wind speed sequence, and the wind speed prediction model is used for outputting the predicted wind speed of the M th period according to the wind speeds collected from the M-i th period to the M-1 th period;
the second acquisition module is used for acquiring the wind speeds acquired in i periods before the period to be predicted;
the prediction module is used for outputting the predicted wind speed of the period to be predicted according to the wind speeds acquired in i periods before the period to be predicted and the wind speed prediction model obtained through training;
the third acquisition module is used for acquiring the actual wind speed acquired in the period to be predicted;
the monitoring module is used for monitoring whether the difference value between the predicted wind speed and the actual wind speed of the period to be predicted exceeds the control limit of the EWMA control chart or not based on the EWMA control chart;
If yes, the first acquisition module and the training module are called again to retrain the wind speed prediction model;
wherein N is greater than or equal to M > i, and N, M and i are both positive integers.
Preferably, the training module includes:
a construction unit for constructing a wind speed derivative sequence from each wind speed in the initial wind speed sequence, respectively;
the training unit is used for training a wind speed prediction model according to the wind speed derivative sequence; wherein:
the wind speed derivative sequence comprises wind speeds and derivative data corresponding to the wind speeds;
the wind speed prediction model is used for outputting the predicted wind speed of the Mth period according to a wind speed derivative sequence corresponding to the wind speeds acquired from the Mth period to the Mth period.
Preferably, the construction unit includes:
the wind speed decomposing subunit is used for decomposing wind speed to obtain a sub wind speed sequence, and the sub wind speed sequence comprises a plurality of sub wind speeds;
and the splicing subunit is used for splicing the wind speed and the sub wind speed sequence corresponding to the wind speed to obtain a wind speed derivative sequence.
Preferably, the construction unit further comprises:
the first training subunit is used for taking a sub wind speed sequence as input and wind speed corresponding to the sub wind speed sequence as output to train the SVR model;
The first prediction subunit is used for inputting a sub-wind speed sequence into the SVR model obtained through training so as to output SVR wind speed corresponding to the sub-wind speed sequence;
the splicing subunit is specifically configured to splice wind speeds, a sub wind speed sequence corresponding to the wind speeds, and an SVR wind speed corresponding to the sub wind speed sequence to obtain a wind speed derivative sequence;
or alternatively, the first and second heat exchangers may be,
the construction unit further includes:
the second training subunit is used for taking the sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output to train the RF model;
the second prediction subunit is used for inputting the sub wind speed sequence into the trained RF model so as to output the RF wind speed corresponding to the sub wind speed sequence;
the splicing subunit is specifically configured to splice wind speeds, a sub wind speed sequence corresponding to the wind speeds, and RF wind speeds corresponding to the sub wind speed sequence to obtain a wind speed derivative sequence;
or alternatively, the first and second heat exchangers may be,
the construction unit further includes:
the first training subunit is used for taking a sub wind speed sequence as input and wind speed corresponding to the sub wind speed sequence as output to train the SVR model;
the first prediction subunit is used for inputting a sub-wind speed sequence into the SVR model obtained through training so as to output SVR wind speed corresponding to the sub-wind speed sequence;
The second training subunit is used for taking the sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output to train the RF model;
the second prediction subunit is used for inputting the sub wind speed sequence into the trained RF model so as to output the RF wind speed corresponding to the sub wind speed sequence;
the splicing subunit is specifically configured to splice a wind speed, a sub-wind speed sequence corresponding to the wind speed, an SVR wind speed corresponding to the sub-wind speed sequence, and an RF wind speed corresponding to the sub-wind speed sequence to obtain a wind speed derivative sequence.
Preferably, the training module is specifically configured to train a wind speed prediction model according to the initial wind speed sequence by using an LSTM model based on an attention mechanism;
and/or the number of the groups of groups,
the training module further comprises:
and the optimizing unit is used for optimizing parameters of the wind speed prediction model based on the DE.
An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing any of the wind speed prediction methods described above when executing the computer program.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of any of the wind speed prediction methods described above.
The invention has the positive progress effects that: according to the wind speed prediction model based on the EWMA control graph, the prediction effect of the wind speed prediction model is monitored, and when the wind speed prediction model is monitored to be poor, the wind speed prediction model can be maintained to be good in prediction effect and high in automation degree, accurate prediction of short-time wind speed is facilitated, and further control of a fan control system is facilitated to improve maximum power generation of a wind field.
Drawings
FIG. 1 is a flowchart of a wind speed prediction method according to embodiment 1 of the present invention.
Fig. 2 is a flowchart of step S2 in the wind speed prediction method according to embodiment 1 of the present invention.
Fig. 3 is a flowchart of step S21 in the wind speed prediction method according to embodiment 1 of the present invention.
FIG. 4 is a logic diagram of the implementation of the LSTM model based on the attention mechanism in the wind speed prediction method according to embodiment 1 of the present invention.
Fig. 5 is a schematic structural diagram of LSTM model based on attention mechanism in the wind speed prediction method according to embodiment 1 of the present invention.
FIG. 6 is a block diagram of a wind speed prediction system according to embodiment 2 of the present invention.
Fig. 7 is a schematic structural diagram of an electronic device according to embodiment 3 of the present invention.
Detailed Description
The invention is further illustrated by means of the following examples, which are not intended to limit the scope of the invention.
Example 1
The present embodiment provides a wind speed prediction method, referring to fig. 1, the wind speed prediction method of the present embodiment includes:
s1, acquiring an initial wind speed sequence;
s2, training a wind speed prediction model according to the initial wind speed sequence;
s3, acquiring wind speeds acquired in i periods before a period to be predicted;
s4, outputting a predicted wind speed of the period to be predicted according to the wind speeds acquired in i periods before the period to be predicted and a wind speed prediction model obtained through training;
s5, acquiring an actual wind speed acquired in a period to be predicted;
s6, monitoring whether the difference value between the predicted wind speed and the actual wind speed of the period to be predicted exceeds the control limit of the EWMA control chart or not based on the EWMA control chart;
if yes, return to step S1.
In this embodiment, the initial wind speed sequence may include wind speeds acquired in N consecutive periods, where N is a positive integer that may be set in a user-defined manner according to practical applications, for example, N may take on a value of 200, and the initial wind speed sequence may be represented as s= [ V ] 1 ，V 2 ，…，V N-1 ，V N ]. The wind speed prediction model is used for sampling according to the M-i th cycle to the M-1 th cycle Outputting the predicted wind speed of the M th period by the collected wind speed, wherein N is more than or equal to M>i and M and i are both positive integers, and the acquisition period can be set in a customized manner according to practical applications, for example, the acquisition period can be set to 30 minutes.
Referring to fig. 2, step S2 in this embodiment may specifically include:
s21, respectively constructing a wind speed derivative sequence according to each wind speed in the initial wind speed sequence;
s22, training a wind speed prediction model according to the wind speed derivative sequence.
In this embodiment, the wind speed derivative sequence may include a wind speed and derivative data obtained by processing the wind speed, and accordingly, the wind speed prediction model may be specifically configured to output a predicted wind speed of the mth cycle according to a wind speed derivative sequence corresponding to wind speeds collected from the mth-i cycle to the mth-1 cycle.
Referring to fig. 3, step S21 in this embodiment may specifically include:
s211, decomposing the wind speed to obtain a sub wind speed sequence.
Specifically, in this embodiment, the sub-wind speed sequence is derived data of wind speed, where the sub-wind speed sequence may include a plurality of sub-wind speeds, and generation of the sub-wind speed sequence may implement expansion of input parameters required for deep learning, and it is assumed that the K-dimensional sub-wind speed sequence is obtained after wind speed decomposition, and wind speed V T The corresponding sub-wind speed sequence may be denoted as V TS ＝[V T1 ,V T2 ,…,V T(K-1) ,V TK ]Wherein T is 1.ltoreq.T.ltoreq.N and T is a positive integer. Further, in step S211, the sub-wind speed sequence may be obtained specifically based on the VMD (Variational Mode Decomposition, variational modal decomposition) decomposed wind speed, or may include obtaining the sub-wind speed sequence based on the EMD (Empirical Mode Decomposition, empirical modal decomposition) decomposed wind speed, where the VMD has an advantage over the EMD in that the dimension of the sub-wind speed sequence, that is, the number of included sub-wind speeds, may be customized according to the actual application.
In the present embodiment, it is preferable to obtain the sub wind speed sequence based on the VMD decomposed wind speed. In particular, the VMD may convert the true signal f (t) (i.e., the present realityWind speed in the example) into a series of modes u k (i.e., the sub-wind speeds in this embodiment), the VMD assumes each u k Can have an intermediate pulse signal w in the frequency domain k And a range defined by the upper and lower sidebands, wherein the implementation logic of the VMD may include:
u using Hilbert transform k Converting to a frequency domain;
removing the spectrum of the mode;
the demodulated sidebands are estimated by the gaussian equation:
s.t.∑ k u k ＝f(t)
where f (t) represents the main path of the decomposition, δ (t) represents the Dirac distribution, x represents the convolution, and in order to introduce the constraint, the lagrangian penalty factors λ and α are taken into account, so the optimization equation is modified as follows, and then the optimal solution calculation is performed:
Referring to fig. 3, step S21 in this embodiment may further include:
s212, training an SVR model by taking a sub-wind speed sequence as input and wind speed corresponding to the sub-wind speed sequence as output;
s213, inputting the sub-wind speed sequence into the SVR model obtained through training to output SVR wind speed corresponding to the sub-wind speed sequence.
Specifically, in the present embodiment, the training data of the SVR model may include [ V T ,V TS ]. After SVR model training is completed, the wind speed sequence V is used as a sub-wind speed sequence TS Output for input SVR wind speed V predicted by training derived SVR model T-SVR . Wherein, SVR wind speed V predicted by SVR model T-SVR And is also derived data of wind speed.
Referring to fig. 3, step S21 in this embodiment may further include:
s214, training an RF model by taking a sub wind speed sequence as input and wind speed corresponding to the sub wind speed sequence as output;
s215, inputting the sub wind speed sequence into the trained RF model to output the RF wind speed corresponding to the sub wind speed sequence.
Specifically, in the present embodiment, the training data of the RF model may include [ V T ,V TS ]. After the RF model training is completed, the wind speed sequence V is used as a sub-wind speed sequence TS Outputting for input the RF wind speed V predicted by the trained RF model T-RF . Wherein the RF wind speed V predicted by the RF model T-RF And is also derived data of wind speed.
Referring to fig. 3, step S21 in this embodiment may further include:
s216, splicing wind speeds, sub-wind speed sequences corresponding to the wind speeds, SVR wind speeds corresponding to the sub-wind speed sequences and RF wind speeds corresponding to the sub-wind speed sequences to obtain wind speed derivative sequences.
In the present embodiment, the spliced wind speed V is preferable T Wind speed V T Corresponding sub wind speed sequence V TS Sequence of sub wind speeds V TS Corresponding SVR wind speed V T-SVR Sub-wind speed sequence V TS Corresponding RF wind speed V T-RF To construct wind speed derivative sequences to achieve further expansion of the input parameters required for deep learning, where S can be used T ＝[V T ,V T1 ,V T2 ,…,V T(K-1) ,V TK ,V T-SVR ,V T-RF ]To represent wind speed V T A corresponding sequence of sub-wind speeds. In other embodiments of the invention, the wind speed derivative sequences may also incorporate SVR model predicted wind speeds V based on wind speeds and their corresponding sub-sequences T-SVR Or the wind speed V predicted by the RF model T-RF The structure is obtained.
In the present embodiment, training data [ V ] can be constructed M ,S M-i ,S M-i+1 ,…,S M-2 ,S M-1 ]For training a wind speed prediction model for deriving the sequence from i wind speeds S M-i ,S M-i+1 ,…,S M-2 ,S M-1 ]To output the predicted wind speed V of the Mth period M . Specifically, it can be based on the input S 1 ,S 2 ,…,S i-1 ,S i ]To predict wind speed V i+1 According to the input S 2 ,S 3 ,…,S i ,S i+1 ]To predict V i+2 … … according to the input [ S ] N-i ,S N-i+1 ,…,S N-2 ,S N-1 ]To predict wind speed V N 。
In step S4, i wind speed derivative sequences may be constructed according to the wind speeds acquired i periods before the period to be predicted, and then the constructed i wind speed derivative sequences may be input into a wind speed prediction model obtained by training to output the predicted wind speed of the period to be predicted.
In this embodiment, step S2 may specifically include a step of training a wind speed prediction model using an attention-based LSTM model according to an initial wind speed sequence, and step S22 may specifically include a step of training a wind speed prediction model using an attention-based LSTM model according to a wind speed derivative sequence. Referring to FIG. 4, the implementation logic of the LSTM model based on the attention mechanism may include:
1) Forgetting gate, which is used to decide whether to discard the current cell state, the mathematical formula is: f (f) t ＝σ(W f X t +R f h t-1 +b f ) Wherein σ represents a sigmoid equation, h t-1 And X t Data respectively representing the previous layer and the current input layer, (W) f ,R f ,b f ) Representing the input weight, recursive weight, and bias of the forgetting gate.
2) An input gate for determining whether to store new data into the cell, the mathematical formula of which is: i.e t ＝σ(W i X t +R i h t-1 +b i ) Wherein, (W) i ,R i ,b i ) Representing the input weight, recursive weight, and bias of the input gate. In addition, a tanh layer is used to form new memories: g t ＝tanh(W g X t +R g h t-1 +b g ) Wherein, (W) g ,R g ,b g ) Input weights, recursive weights, and bias representing new memoriesAnd (5) placing.
3) The state of the cell will be updated as: c (C) t ＝C t-1 ×f f +g t ×i t 。
4) The output gate is used for deciding whether to transfer the information in the cell to the current hidden layer, and the mathematical formula is as follows: o (o) t ＝σ(W o X t +R o h t-1 +b o ) Wherein, (W) o ,R o ,b o ) Representing the output gate input weight, recursive weight, and bias. Finally, multiplying the output gate with the sigmoid gate: h is a t ＝o t ×tanh(C t )。
In this example, the softmax conversion is performed at the last output Layer LSTM Layer: p is p t ＝softx max(o t ) Finally, the LSTM is convolved with softmax, e.g. y t ＝p t *o t ,. Referring to fig. 5, in the structure of the LSTM model based on the attention mechanism: input_1 represents the input layer, lstm_1 represents the last output layer, dense_1 represents the weight layer of softmax, dense_1 represents the inner product of input_1 and dense_1, and dense_2 represents the predicted wind speed of the output.
In this embodiment, compared with the normal LSTM model, the LSTM model based on the attention mechanism can avoid overlengthy memory chains, and the attention mechanism gives a weight to each neuron, specifically, gives a larger weight to neurons with accurate prediction results and gives smaller weight to neurons with inaccurate prediction results, so that the embodiment can analyze neurons with high weight without analyzing each neuron, and in addition, the weights can be automatically obtained without manual intervention when the model iterates.
Step S2 in this embodiment may further include a step of optimizing parameters of the wind speed prediction model based on DE, and for optimizing parameters of the LSTM model based on the attention mechanism, implementation logic may include:
Initializing a search cluster, wherein the search cluster is an upper limit range and a lower limit range of a needed solving parameter, and can be set in a self-defining way according to practical application;
setting the MAE as the minimized objective function:
wherein A is t F for the actual wind speed at time t t For the predicted wind speed at this point, n is the calculated array;
parameters that determine DE, for example, optimized parameters may include, but are not limited to, batch_ size, neuro, look _back and dropout;
transferring the parameters to an LSTM model based on an attention mechanism to calculate MAE;
and if the result is converged, the optimization parameters are obtained.
Further, in this embodiment, parameters of the SVR model may be optimized based on the DE, and the optimized parameters may include, but are not limited to, gamma and cost, and parameters of the RF model may be optimized based on the DE. The prediction effect of the model can be further improved through optimizing the model parameters.
In step S6 of the present embodiment, the logic for determining the control limit of the EWMA control map may include:
let the residual error of the model predicted value and the actual value in the i time period be X i ；
Constructing statistics of EWMA: z is Z i ＝λX i +(1-λ)Z i-1 Wherein λ represents the weight value of EWMA, Z i EWMA statistics representing i time periods, Z i-1 A value of EWMA representing the i-1 time period;
Calculating an upper control limit:wherein L is a parameter of EWMA;
calculating a lower control limit:
the values of (λ, L) are looked up, and since this embodiment is suitable for industrial applications, λ=0.1, l=2.7 can be chosen and corresponds to the industry's conventional 3-sigma criterion.
In this embodiment, when the difference between the predicted wind speed and the actual wind speed in the period to be predicted exceeds the range defined by the upper control limit and the lower control limit, it may be determined that the difference between the predicted wind speed output by the wind speed prediction model and the actual wind speed is large, that is, the prediction capability of the wind speed prediction model currently used for outputting the predicted wind speed is reduced, and it is necessary to retrain the wind speed prediction model again to ensure that the wind speed prediction model has a better prediction effect. Specifically, when the step S6 determines that it is yes, the step returns to the step S1 to re-acquire the initial wind speed sequence, where the re-acquired initial wind speed sequence may include wind speeds acquired in a plurality of recent consecutive periods, and then step S2 is performed again, and the wind speed prediction model is retrained according to the re-acquired initial wind speed sequence.
In this embodiment, the initial wind speed sequence may be normalized, or training data for training a model may be divided into a training set and a test set.
According to the embodiment, the prediction effect of the wind speed prediction model is monitored based on the EWMA control chart, and when the wind speed prediction model is monitored to be poor in prediction effect, the wind speed prediction model can be maintained to be good in prediction effect and high in automation degree, accurate prediction of short-time wind speed is facilitated, and further control of a fan control system is facilitated to improve maximum power generation of a wind field.
Example 2
The present embodiment provides a wind speed prediction system, referring to fig. 6, the wind speed prediction system of the present embodiment includes:
the first acquisition module 1 is used for acquiring an initial wind speed sequence;
the training module 2 is used for training a wind speed prediction model according to the initial wind speed sequence;
the second acquisition module 3 is used for acquiring the wind speeds acquired by i periods before the period to be predicted;
the prediction module 4 is used for outputting the predicted wind speed of the period to be predicted according to the wind speeds acquired by i periods before the period to be predicted and the wind speed prediction model obtained by training;
the third acquisition module 5 is used for acquiring the actual wind speed acquired in the period to be predicted;
the monitoring module 6 monitors whether the difference between the predicted wind speed and the actual wind speed of the period to be predicted exceeds the control limit of the EWMA control chart based on the EWMA control chart;
If yes, the first acquisition module 1 and the training module 2 are called again.
In this embodiment, the initial wind speed sequence may include wind speeds acquired in N consecutive periods, where N is a positive integer that may be set in a user-defined manner according to practical applications, for example, N may take on a value of 200, and the initial wind speed sequence may be represented as s= [ V ] 1 ,V 2 ,…,V N-1 ,V N ]. The wind speed prediction model is used for outputting the predicted wind speed of the Mth period according to the wind speeds acquired from the Mth period to the Mth period, wherein N is more than or equal to M>i and M and i are both positive integers, and the acquisition period can be set in a customized manner according to practical applications, for example, the acquisition period can be set to 30 minutes.
Referring to fig. 6, the training module 2 in this embodiment may specifically include:
a construction unit 21 for constructing a wind speed derivative sequence from each wind speed in the initial wind speed sequence, respectively;
a training unit 22 for training a wind speed prediction model according to the wind speed derivative sequence.
In this embodiment, the wind speed derivative sequence may include a wind speed and derivative data obtained by processing the wind speed, and accordingly, the wind speed prediction model may be specifically configured to output a predicted wind speed of the mth cycle according to a wind speed derivative sequence corresponding to wind speeds collected from the mth-i cycle to the mth-1 cycle.
Referring to fig. 6, the construction unit 21 in the present embodiment may specifically include:
a decomposition subunit 211, configured to decompose wind speeds to obtain a sub-wind speed sequence.
Concrete embodimentsIn this embodiment, the sub-wind speed sequence is derived data of wind speed, where the sub-wind speed sequence may include a plurality of sub-wind speeds, and generation of the sub-wind speed sequence may implement expansion of input parameters required for deep learning, and it is assumed that the K-dimensional sub-wind speed sequence is obtained after wind speed decomposition, and wind speed V T The corresponding sub-wind speed sequence may be denoted as V TS ＝[V T1 ,V T2 ,…,V T(K-1) ,V TK ]Wherein T is 1.ltoreq.T.ltoreq.N and T is a positive integer. Further, the decomposition subunit 211 may specifically obtain the sub-wind speed sequence based on the VMD (Variational Mode Decomposition, variational modal decomposition) decomposition wind speed, or may include obtaining the sub-wind speed sequence based on the EMD (Empirical Mode Decomposition, empirical modal decomposition) decomposition wind speed, where the VMD has an advantage over the EMD in that the dimension of the sub-wind speed sequence, that is, the number of included sub-wind speeds, may be customized according to the actual application. In the present embodiment, it is preferable to obtain the sub wind speed sequence based on the VMD decomposed wind speed. In particular, the VMD may decompose the true signal f (t) (i.e., wind speed in this embodiment) into a series of modes u k (i.e., the sub-wind speeds in this embodiment), the VMD assumes each u k Can have an intermediate pulse signal w in the frequency domain k And a range defined by the upper and lower sidebands, wherein the implementation logic of the VMD may include:
u using Hilbert transform k Converting to a frequency domain;
removing the spectrum of the mode;
the demodulated sidebands are estimated by the gaussian equation:
s.t.∑ k u k ＝f(t)
where f (t) represents the main path of the decomposition, δ (t) represents the Dirac distribution, x represents the convolution, and in order to introduce the constraint, the lagrangian penalty factors λ and α are taken into account, so the optimization equation is modified as follows, and then the optimal solution calculation is performed:
referring to fig. 6, the construction unit 21 in the present embodiment may further include:
a first training subunit 212, configured to train the SVR model with the sub-wind speed sequence as input and the wind speed corresponding to the sub-wind speed sequence as output;
the first prediction subunit 213 is configured to input the sub-wind speed sequence into the trained SVR model, so as to output an SVR wind speed corresponding to the sub-wind speed sequence.
Specifically, in the present embodiment, the training data of the SVR model may include [ V T ,V TS ]. After SVR model training is completed, the wind speed sequence V is used as a sub-wind speed sequence TS Output for input SVR wind speed V predicted by training derived SVR model T-SVR . Wherein, SVR wind speed V predicted by SVR model T-SVR And is also derived data of wind speed.
Referring to fig. 6, the construction unit 21 in the present embodiment may further include:
a second training subunit 214, configured to train the RF model with the sub-wind speed sequence as input and the wind speed corresponding to the sub-wind speed sequence as output;
the second prediction subunit 215 is configured to input the sub-wind speed sequence into the trained RF model, so as to output an RF wind speed corresponding to the sub-wind speed sequence.
Specifically, in the present embodiment, the training data of the RF model may include [ V T ，V TS ]. After the RF model training is completed, the wind speed sequence V is used as a sub-wind speed sequence TS Outputting for input the RF wind speed V predicted by the trained RF model T-RF . Wherein the RF wind speed V predicted by the RF model T-RF And is also derived data of wind speed.
Referring to fig. 6, the construction unit 21 in the present embodiment may further include:
the splicing subunit 216 is configured to splice the wind speed, the sub-wind speed sequence corresponding to the wind speed, the SVR wind speed corresponding to the sub-wind speed sequence, and the RF wind speed corresponding to the sub-wind speed sequence to obtain a wind speed derivative sequence.
In the present practiceIn the embodiment, the spliced wind speed V is preferable T Wind speed V T Corresponding sub wind speed sequence V TS Sequence of sub wind speeds V TS Corresponding SVR wind speed V T-SVR Sub-wind speed sequence V TS Corresponding RF wind speed V T-RF To construct wind speed derivative sequences to achieve further expansion of the input parameters required for deep learning, where S can be used T ＝[V T ,V T1 ,V T2 ,…,V T(K-1) ,V TK ,V T-SVR ,V T-RF ]To represent wind speed V T A corresponding sequence of sub-wind speeds. In other embodiments of the present invention, the splicing subunit 216 may also combine the SVR model predicted wind speed V based on the spliced wind speed and the corresponding subsequence T-SVR Or the wind speed V predicted by the RF model T-RF Constructing a wind speed derivative sequence.
In the present embodiment, training data [ V ] can be constructed M ,S M-i ,S M-i+1 ,…,S M-2 ,S M-1 ]For training a wind speed prediction model for deriving the sequence from i wind speeds S M-i ，S M-i+1 ，…，S M-2 ，S M-1 ]To output the predicted wind speed V of the Mth period M . Specifically, it can be based on the input S 1 ,S 2 ,…,S i-1 ,S i ]To predict wind speed V i+1 According to the input S 2 ，S 3 ,…,S i ,S i+1 ]To predict V i+2 … … according to the input [ S ] N-i ,S N-i+1 ,…,S N-2 ,S N-1 ]To predict wind speed V N 。
In this embodiment, the prediction module 4 may construct i wind speed derivative sequences according to the wind speeds acquired i periods before the period to be predicted, and then may input the constructed i wind speed derivative sequences into the wind speed prediction model obtained by training to output the predicted wind speed of the period to be predicted.
In this embodiment, the training module 2 may be specifically configured to train the wind speed prediction model with the attention-based LSTM model according to the initial wind speed sequence, and the training unit 22 may be specifically configured to train the wind speed prediction model with the attention-based LSTM model according to the wind speed derivative sequence. Referring to FIG. 4, the implementation logic of the LSTM model based on the attention mechanism may include:
1) Forgetting gate, which is used to decide whether to discard the current cell state, the mathematical formula is: f (f) t ＝σ(W f X t +R f h t-1 +b f ) Wherein σ represents a sigmoid equation, h t-1 And X t Data respectively representing the previous layer and the current input layer, (W) f ,R f ,b f ) Representing the input weight, recursive weight, and bias of the forgetting gate.
2) An input gate for determining whether to store new data into the cell, the mathematical formula of which is: i.e t ＝σ(W i X t +R i h t-1 +b i ) Wherein, (W) i ,R i ,b i ) Representing the input weight, recursive weight, and bias of the input gate. In addition, a tanh layer is used to form new memories: g t ＝tanh(W g X t +R g h t-1 +b g ) Wherein, (W) g ,R g ,b g ) Representing the newly memorized input weights, recursive weights and biases.
3) The state of the cell will be updated as: c (C) t ＝C t-1 ×f f +g t ×i t 。
4) The output gate is used for deciding whether to transfer the information in the cell to the current hidden layer, and the mathematical formula is as follows: o (o) t ＝σ(W o X t +R o h t-1 +b o ) Wherein, (W) o ,R o ,b o ) Representing the output gate input weight, recursive weight, and bias. Finally, multiplying the output gate with the sigmoid gate: h is a t ＝o t ×tanh(C t )。
In this example, the softmax conversion is performed at the last output Layer LSTM Layer: p is p t ＝softx max(o t ) Finally, the LSTM is convolved with softmax, e.g. y t ＝p t * o t. Referring to fig. 5, in the structure of the LSTM model based on the attention mechanism:input_1 represents the input layer, lstm_1 represents the last output layer, dense_1 represents the weight layer of softmax, dense_1 represents the inner product of input_1 and dense_1, and dense_2 represents the predicted wind speed of the output.
In this embodiment, compared with the normal LSTM model, the LSTM model based on the attention mechanism can avoid overlengthy memory chains, and the attention mechanism gives a weight to each neuron, specifically, gives a larger weight to neurons with accurate prediction results and gives smaller weight to neurons with inaccurate prediction results, so that the embodiment can analyze neurons with high weight without analyzing each neuron, and in addition, the weights can be automatically obtained without manual intervention when the model iterates.
The training module 2 in this embodiment may further include a first optimizing unit for optimizing parameters of the wind speed prediction model based on DE, and in particular, the first optimizing unit may be configured to optimize parameters of the LSTM model based on the attention mechanism, and the implementation logic may include:
initializing a search cluster, wherein the search cluster is an upper limit range and a lower limit range of a needed solving parameter, and can be set in a self-defining way according to practical application;
setting the MAE as the minimized objective function:
wherein A is t F for the actual wind speed at time t t For the predicted wind speed at this point, n is the calculated array;
parameters that determine DE, for example, optimized parameters may include, but are not limited to, batch_ size, neuro, look _back and dropout;
Transferring the parameters to an LSTM model based on an attention mechanism to calculate MAE;
and if the result is converged, the optimization parameters are obtained.
Further, in this embodiment, the training module 2 may further include a second optimizing unit for optimizing parameters of the SVR model based on the DE, where the optimized parameters may include, but are not limited to, gamma and cost, and may further include a third optimizing unit for optimizing parameters of the RF model based on the DE. The prediction effect of the model can be further improved through optimizing the model parameters.
In this embodiment, the logic for determining the control limits of the EWMA control map may include:
let the residual error of the model predicted value and the actual value in the i time period be X i ；
Constructing statistics of EWMA: z is Z i ＝λX i +(1-λ)Z i-1 Wherein λ represents the weight value of EWMA, Z i EWMA statistics representing i time periods, Z i-1 A value of EWMA representing the i-1 time period;
calculating an upper control limit:wherein L is a parameter of EWMA;
calculating a lower control limit:
the values of (λ, L) are looked up, and since this embodiment is suitable for industrial applications, λ=0.1, l=2.7 can be chosen and corresponds to the industry's conventional 3-sigma criterion.
In this embodiment, when the difference between the predicted wind speed and the actual wind speed in the period to be predicted exceeds the range defined by the upper control limit and the lower control limit, it may be determined that the difference between the predicted wind speed output by the wind speed prediction model and the actual wind speed is large, that is, the prediction capability of the wind speed prediction model currently used for outputting the predicted wind speed is reduced, and it is necessary to retrain the wind speed prediction model again to ensure that the wind speed prediction model has a better prediction effect. Specifically, when the monitoring module 6 determines that it is yes, the first obtaining module 1 is called again to re-obtain the initial wind speed sequence, where the re-obtained initial wind speed sequence may include wind speeds collected in a plurality of recent consecutive periods, and then the training module 2 is called again to re-train the wind speed prediction model according to the re-obtained initial wind speed sequence.
In this embodiment, the wind speed training device may further include a preprocessing module for normalizing the initial wind speed sequence, and may further include a dividing module for dividing training data for training the model into a training set and a test set.
According to the embodiment, the prediction effect of the wind speed prediction model is monitored based on the EWMA control chart, and when the wind speed prediction model is monitored to be poor in prediction effect, the wind speed prediction model can be maintained to be good in prediction effect and high in automation degree, accurate prediction of short-time wind speed is facilitated, and further control of a fan control system is facilitated to improve maximum power generation of a wind field.
Example 3
The present embodiment provides an electronic device, which may be expressed in the form of a computing device (for example, may be a server device), including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor may implement the wind speed prediction method provided in embodiment 1 when executing the computer program.
Fig. 7 shows a schematic diagram of the hardware structure of the present embodiment, and as shown in fig. 7, the electronic device 9 specifically includes:
at least one processor 91, at least one memory 92, and a bus 93 for connecting the different system components (including the processor 91 and the memory 92), wherein:
The bus 93 includes a data bus, an address bus, and a control bus.
The memory 92 includes volatile memory such as Random Access Memory (RAM) 921 and/or cache memory 922, and may further include Read Only Memory (ROM) 923.
Memory 92 also includes a program/utility 925 having a set (at least one) of program modules 924, such program modules 924 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.
The processor 91 executes various functional applications and data processing, such as the wind speed prediction method provided in embodiment 1 of the present invention, by running a computer program stored in the memory 92.
The electronic device 9 may further communicate with one or more external devices 94 (e.g., keyboard, pointing device, etc.). Such communication may occur through an input/output (I/O) interface 95. Also, the electronic device 9 may communicate with one or more networks such as a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, such as the Internet, through a network adapter 96. The network adapter 96 communicates with other modules of the electronic device 9 via the bus 93. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in connection with the electronic device 9, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, data backup storage systems, and the like.
It should be noted that although several units/modules or sub-units/modules of an electronic device are mentioned in the above detailed description, such a division is merely exemplary and not mandatory. Indeed, the features and functionality of two or more units/modules described above may be embodied in one unit/module according to embodiments of the present application. Conversely, the features and functions of one unit/module described above may be further divided into ones that are embodied by a plurality of units/modules.
Example 4
The present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the wind speed prediction method provided by embodiment 1.
More specifically, among others, readable storage media may be employed including, but not limited to: portable disk, hard disk, random access memory, read only memory, erasable programmable read only memory, optical storage device, magnetic storage device, or any suitable combination of the foregoing.
In a possible embodiment, the invention may also be implemented in the form of a program product comprising program code for causing a terminal device to carry out the steps of implementing the wind speed prediction method as described in example 1, when said program product is run on the terminal device.
Wherein the program code for carrying out the invention may be written in any combination of one or more programming languages, which program code may execute entirely on the user device, partly on the user device, as a stand-alone software package, partly on the user device and partly on the remote device or entirely on the remote device.
While specific embodiments of the invention have been described above, it will be appreciated by those skilled in the art that this is by way of example only, and the scope of the invention is defined by the appended claims. Various changes and modifications to these embodiments may be made by those skilled in the art without departing from the principles and spirit of the invention, but such changes and modifications fall within the scope of the invention.

Claims (8)

1. A method of wind speed prediction, comprising:
acquiring an initial wind speed sequence, wherein the initial wind speed sequence comprises wind speeds acquired in N continuous periods;
training a wind speed prediction model according to the initial wind speed sequence, wherein the wind speed prediction model is used for outputting the predicted wind speed of the Mth period according to the wind speeds acquired from the Mth period to the Mth period;
Acquiring wind speeds acquired in i periods before a period to be predicted;
outputting the predicted wind speed of the period to be predicted according to the wind speeds acquired in i periods before the period to be predicted and a wind speed prediction model obtained through training;
acquiring the actual wind speed acquired in the period to be predicted;
monitoring whether the difference between the predicted wind speed and the actual wind speed of the period to be predicted exceeds the control limit of the EWMA control chart or not based on the EWMA control chart;
if yes, the step of acquiring an initial wind speed sequence and the step of training a wind speed prediction model according to the initial wind speed sequence are executed again so as to retrain the wind speed prediction model;
wherein N is more than or equal to M > i, and N, M and i are both positive integers;
the step of training a wind speed prediction model according to the initial wind speed sequence specifically comprises the following steps:
respectively constructing a wind speed derivative sequence according to each wind speed in the initial wind speed sequence;
training a wind speed prediction model according to the wind speed derivative sequence; wherein:
the wind speed derivative sequence comprises wind speeds and derivative data corresponding to the wind speeds;
the wind speed prediction model is used for outputting the predicted wind speed of the Mth period according to a wind speed derivative sequence corresponding to the wind speed collected from the Mth period to the Mth period;
The step of constructing a wind speed derivative sequence from each wind speed in the initial wind speed sequence comprises:
decomposing wind speed to obtain a sub wind speed sequence, wherein the sub wind speed sequence comprises a plurality of sub wind speeds;
and splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence.
2. The method of predicting wind speed of claim 1, further comprising, after the step of decomposing the wind speed to obtain a sequence of sub-wind speeds:
training an SVR model by taking a sub-wind speed sequence as input and the wind speed corresponding to the sub-wind speed sequence as output;
inputting a sub wind speed sequence into a SVR model obtained by training to output SVR wind speed corresponding to the sub wind speed sequence;
the step of splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence comprises the following steps:
splicing wind speeds, sub-wind speed sequences corresponding to the wind speeds and SVR wind speeds corresponding to the sub-wind speed sequences to obtain wind speed derivative sequences;
or alternatively, the first and second heat exchangers may be,
the step of decomposing the wind speed to obtain a sub-wind speed sequence further comprises the following steps:
training an RF model by taking a sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output;
inputting a sub wind speed sequence into a trained RF model to output RF wind speed corresponding to the sub wind speed sequence;
The step of splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence comprises the following steps:
splicing wind speeds, sub-wind speed sequences corresponding to the wind speeds and RF wind speeds corresponding to the sub-wind speed sequences to obtain wind speed derivative sequences;
or alternatively, the first and second heat exchangers may be,
the step of decomposing the wind speed to obtain a sub-wind speed sequence further comprises the following steps:
training an SVR model by taking a sub-wind speed sequence as input and the wind speed corresponding to the sub-wind speed sequence as output;
inputting a sub wind speed sequence into a SVR model obtained by training to output SVR wind speed corresponding to the sub wind speed sequence;
training an RF model by taking a sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output;
inputting a sub wind speed sequence into a trained RF model to output RF wind speed corresponding to the sub wind speed sequence;
the step of splicing the sub-wind speed sequences corresponding to the wind speeds to obtain a wind speed derivative sequence comprises the following steps:
splicing wind speeds, sub wind speed sequences corresponding to the wind speeds, SVR wind speeds corresponding to the sub wind speed sequences and RF wind speeds corresponding to the sub wind speed sequences to obtain wind speed derivative sequences.
3. The method of wind speed prediction according to claim 1, wherein the step of training a wind speed prediction model from the initial wind speed sequence comprises:
Training a wind speed prediction model by using an LSTM model based on an attention mechanism according to the initial wind speed sequence;
and/or the number of the groups of groups,
parameters of the wind speed prediction model are optimized based on DE.
4. A wind speed prediction system, comprising:
the first acquisition module is used for acquiring an initial wind speed sequence, wherein the initial wind speed sequence comprises wind speeds acquired in N continuous periods;
the training module is used for training a wind speed prediction model according to the initial wind speed sequence, and the wind speed prediction model is used for outputting the predicted wind speed of the M th period according to the wind speeds collected from the M-i th period to the M-1 th period;
the second acquisition module is used for acquiring the wind speeds acquired in i periods before the period to be predicted;
the prediction module is used for outputting the predicted wind speed of the period to be predicted according to the wind speeds acquired in i periods before the period to be predicted and the wind speed prediction model obtained through training;
the third acquisition module is used for acquiring the actual wind speed acquired in the period to be predicted;
the monitoring module is used for monitoring whether the difference value between the predicted wind speed and the actual wind speed of the period to be predicted exceeds the control limit of the EWMA control chart or not based on the EWMA control chart;
If yes, the first acquisition module and the training module are called again to retrain the wind speed prediction model;
wherein N is more than or equal to M > i, and N, M and i are both positive integers;
the training module comprises:
a construction unit for constructing a wind speed derivative sequence from each wind speed in the initial wind speed sequence, respectively;
the training unit is used for training a wind speed prediction model according to the wind speed derivative sequence; wherein:
the wind speed derivative sequence comprises wind speeds and derivative data corresponding to the wind speeds;
the wind speed prediction model is used for outputting the predicted wind speed of the Mth period according to a wind speed derivative sequence corresponding to the wind speed collected from the Mth period to the Mth period;
the construction unit includes:
the wind speed decomposing subunit is used for decomposing wind speed to obtain a sub wind speed sequence, and the sub wind speed sequence comprises a plurality of sub wind speeds;
and the splicing subunit is used for splicing the wind speed and the sub wind speed sequence corresponding to the wind speed to obtain a wind speed derivative sequence.
5. The wind speed prediction system according to claim 4, wherein the construction unit further comprises:
the first training subunit is used for taking a sub wind speed sequence as input and wind speed corresponding to the sub wind speed sequence as output to train the SVR model;
The first prediction subunit is used for inputting a sub-wind speed sequence into the SVR model obtained through training so as to output SVR wind speed corresponding to the sub-wind speed sequence;
the splicing subunit is specifically configured to splice wind speeds, a sub wind speed sequence corresponding to the wind speeds, and an SVR wind speed corresponding to the sub wind speed sequence to obtain a wind speed derivative sequence;
or alternatively, the first and second heat exchangers may be,
the construction unit further includes:
the second training subunit is used for taking the sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output to train the RF model;
the second prediction subunit is used for inputting the sub wind speed sequence into the trained RF model so as to output the RF wind speed corresponding to the sub wind speed sequence;
the splicing subunit is specifically configured to splice wind speeds, a sub wind speed sequence corresponding to the wind speeds, and RF wind speeds corresponding to the sub wind speed sequence to obtain a wind speed derivative sequence;
or alternatively, the first and second heat exchangers may be,
the construction unit further includes:
the first training subunit is used for taking a sub wind speed sequence as input and wind speed corresponding to the sub wind speed sequence as output to train the SVR model;
the first prediction subunit is used for inputting a sub-wind speed sequence into the SVR model obtained through training so as to output SVR wind speed corresponding to the sub-wind speed sequence;
The second training subunit is used for taking the sub wind speed sequence as input and the wind speed corresponding to the sub wind speed sequence as output to train the RF model;
the second prediction subunit is used for inputting the sub wind speed sequence into the trained RF model so as to output the RF wind speed corresponding to the sub wind speed sequence;
the splicing subunit is specifically configured to splice a wind speed, a sub-wind speed sequence corresponding to the wind speed, an SVR wind speed corresponding to the sub-wind speed sequence, and an RF wind speed corresponding to the sub-wind speed sequence to obtain a wind speed derivative sequence.
6. The wind speed prediction system of claim 4, wherein the training module is specifically configured to train a wind speed prediction model from the initial wind speed sequence using an attention-based LSTM model;
and/or the number of the groups of groups,
the training module further comprises:
and the optimizing unit is used for optimizing parameters of the wind speed prediction model based on the DE.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the wind speed prediction method of any one of claims 1 to 5 when the computer program is executed by the processor.
8. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the wind speed prediction method according to any one of claims 1 to 5.
CN202010208630.3A 2020-03-23 2020-03-23 Wind speed prediction method, system, electronic equipment and storage medium Active CN111461416B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010208630.3A CN111461416B (en) 2020-03-23 2020-03-23 Wind speed prediction method, system, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010208630.3A CN111461416B (en) 2020-03-23 2020-03-23 Wind speed prediction method, system, electronic equipment and storage medium

Publications (2)

Publication Number Publication Date
CN111461416A CN111461416A (en) 2020-07-28
CN111461416B true CN111461416B (en) 2023-07-18

Family

ID=71679191

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010208630.3A Active CN111461416B (en) 2020-03-23 2020-03-23 Wind speed prediction method, system, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN111461416B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113033094B (en) * 2021-03-24 2024-02-09 上海海洋大学 Sea wave height prediction method

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102542133A (en) * 2010-12-10 2012-07-04 中国科学院深圳先进技术研究院 Short-time wind speed forecasting method and system for wind power plant
CN202735528U (en) * 2012-08-13 2013-02-13 国电南京自动化股份有限公司 Intelligent meteorological station system capable of predicting meteorological data
CN107909209A (en) * 2017-11-16 2018-04-13 东华大学 The wind speed forecasting method of complete overall experience mode decomposition and depth belief network
CN110083940A (en) * 2019-04-28 2019-08-02 东华大学 A kind of short-term wind speed forecasting method based on SSA-HMD-CNNSVM model
CN110555515A (en) * 2019-08-22 2019-12-10 南京信大气象科学技术研究院有限公司 Short-term wind speed prediction method based on EEMD and LSTM
CN110738010A (en) * 2019-10-17 2020-01-31 湖南科技大学 Wind power plant short-term wind speed prediction method integrated with deep learning model

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102542133A (en) * 2010-12-10 2012-07-04 中国科学院深圳先进技术研究院 Short-time wind speed forecasting method and system for wind power plant
CN202735528U (en) * 2012-08-13 2013-02-13 国电南京自动化股份有限公司 Intelligent meteorological station system capable of predicting meteorological data
CN107909209A (en) * 2017-11-16 2018-04-13 东华大学 The wind speed forecasting method of complete overall experience mode decomposition and depth belief network
CN110083940A (en) * 2019-04-28 2019-08-02 东华大学 A kind of short-term wind speed forecasting method based on SSA-HMD-CNNSVM model
CN110555515A (en) * 2019-08-22 2019-12-10 南京信大气象科学技术研究院有限公司 Short-term wind speed prediction method based on EEMD and LSTM
CN110738010A (en) * 2019-10-17 2020-01-31 湖南科技大学 Wind power plant short-term wind speed prediction method integrated with deep learning model

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Khuram Hayat.《基于融合LSTM的超短期风速概率预测方法研究》.《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》.2020,全文. *
Tascikaraoglu A et al.《A review of combined approaches for prediction of short-term wind speed and power》.《Renewable & Sustainable Energy Reviews》.2014,全文. *

Also Published As

Publication number Publication date
CN111461416A (en) 2020-07-28

Similar Documents

Publication Publication Date Title
Li et al. Smart wind speed forecasting approach using various boosting algorithms, big multi-step forecasting strategy
Wei et al. Ultra-short-term forecasting of wind power based on multi-task learning and LSTM
JP5888640B2 (en) Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program
CN109242212A (en) A kind of wind-powered electricity generation prediction technique based on change Mode Decomposition and length memory network
CN108428023B (en) Trend prediction method based on quantum weighted threshold repetitive unit neural network
CN111340282A (en) DA-TCN-based method and system for estimating residual service life of equipment
CN111832825A (en) Wind power prediction method and system integrating long-term and short-term memory network and extreme learning machine
Zhao et al. Probabilistic remaining useful life prediction based on deep convolutional neural network
CN116502774B (en) Time sequence prediction method based on time sequence decomposition and Legend projection
CN113111572B (en) Method and system for predicting residual life of aircraft engine
CN114139783A (en) Wind power short-term power prediction method and device based on nonlinear weighted combination
CN115564155A (en) Distributed wind turbine generator power prediction method and related equipment
CN113128666A (en) Mo-S-LSTMs model-based time series multi-step prediction method
CN116665798A (en) Air pollution trend early warning method and related device
Zaman et al. Wind speed forecasting using ARMA and neural network models
CN116432812A (en) New energy power prediction method for optimizing LSTM (least squares) by using Zun sea squirt algorithm
CN111461416B (en) Wind speed prediction method, system, electronic equipment and storage medium
CN118040678A (en) Short-term offshore wind power combination prediction method
CN118017482A (en) Flexible climbing capacity demand analysis method based on prediction error feature extraction
CN116960975A (en) Photovoltaic power generation amount prediction method and device
CN116722529A (en) Short-term photovoltaic power prediction method and system
CN111461418B (en) Wind speed prediction method, system, electronic equipment and storage medium
CN117154680A (en) Wind power prediction method based on non-stationary transducer model
US20210232917A1 (en) Estimating useful life
CN113011674A (en) Photovoltaic power generation prediction method and device, electronic equipment and storage medium

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