CN110264012A - Renewable energy power combination prediction technique and system based on empirical mode decomposition - Google Patents
Renewable energy power combination prediction technique and system based on empirical mode decomposition Download PDFInfo
- Publication number
- CN110264012A CN110264012A CN201910562179.2A CN201910562179A CN110264012A CN 110264012 A CN110264012 A CN 110264012A CN 201910562179 A CN201910562179 A CN 201910562179A CN 110264012 A CN110264012 A CN 110264012A
- Authority
- CN
- China
- Prior art keywords
- renewable energy
- algorithm
- prediction
- vector machine
- predicted value
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Computational Linguistics (AREA)
- Development Economics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of renewable energy power combination prediction technique and system based on empirical mode decomposition form renewable energy time series method includes the following steps: obtaining the renewable energy operation data at multiple moment；Use empirical mode decomposition method by renewable energy Time Series for several intrinsic modal components and a residual components；Least square method supporting vector machine algorithm and radial base neural net algorithm is utilized respectively to predict obtained several intrinsic modal components；Reconstruct is combined using the predicted value that induced ordered weighted averaging operator obtains least square method supporting vector machine algorithm and radial base neural net algorithm, obtains final predicted value.
Description
Technical field
This disclosure relates to energy forecast technical field more particularly to a kind of renewable energy function based on empirical mode decomposition
Rate combination forecasting method and system.
Background technique
In recent years, growing and environmental pollution the constantly aggravation of energy demand causes countries in the world government to energy
The extensive concern of source transition.Significantly to improve air quality, damage cost caused by environmental pollution is preferably minimized, is made simultaneously
It obtains carbon discharge capacity to be remarkably decreased, it is necessary to which the burning for reducing the traditional fossil energies such as coal actively develops clean energy resource, so that energy knot
The mode of structure towards cleaning, low-carbon changes.At the same time, the continuous promotion of the renewable energy technologies based on wind-powered electricity generation, photovoltaic
And it graduallys mature and provides a new road for energy resource structure transition.
Honourable distributed renewable energy itself has certain randomness and fluctuation, by its own energy characteristics
Influence is more obvious, and controllability is poor.Simultaneously as the pattern of the unidirectional trend of traditional power grid is as distributed type renewable energy
The access in source and change, this will affect power supply reliability and power quality, and the effect of should having is not fully played.
Therefore, being effectively predicted for renewable energy source power is of great significance to the reliability that it is powered.
Because renewable energy randomness affects to its power prediction, have at present based on empirical mode decomposition,
The prediction technique for becoming the technologies such as mode decomposition and WAVELET PACKET DECOMPOSITION, although preferably reducing renewable energy randomness to it
The influence of power prediction, but while carrying out power prediction again after data decomposition, is all made of the prediction algorithm of single machine learning, deposits
In certain limitation.
Summary of the invention
In order to overcome the above-mentioned deficiencies of the prior art, present disclose provides a kind of renewable energies based on empirical mode decomposition
Original series are decomposed into several stationary components through EMD first by source power combination forecasting method and system, then using minimum two
Multiply support vector machines (LS-SVM) and radial base (RBF) neural network algorithm individually predicts each component, finally with changing
Weight coefficient is adjusted in real time into IOWA operator, and each component is reconstructed into final prediction result.
On the one hand a kind of renewable energy power combination prediction technique based on empirical mode decomposition that the disclosure provides
Technical solution is:
A kind of renewable energy power combination prediction technique based on empirical mode decomposition, characterized in that including following step
It is rapid:
The renewable energy operation data at multiple moment is obtained, renewable energy time series is formed；
Use empirical mode decomposition method by renewable energy Time Series for several intrinsic modal components and one
A residual components；
Least square method supporting vector machine algorithm and radial base neural net algorithm are utilized respectively to obtained several eigen modes
State component is predicted；
Using induced ordered weighted averaging operator to least square method supporting vector machine algorithm and radial base neural net algorithm
Obtained predicted value is combined reconstruct, obtains final predicted value.
On the other hand a kind of renewable energy power combination forecasting system based on empirical mode decomposition that the disclosure provides
Technical solution be:
A kind of renewable energy power combination forecasting system based on empirical mode decomposition, the system include:
Time series constructs module, for obtaining the renewable energy operation data at multiple moment, forms renewable energy
Time series；
Decomposing module, for using empirical mode decomposition method by renewable energy Time Series for several eigen modes
State component and a residual components；
Monomer prediction module, for being utilized respectively least square method supporting vector machine algorithm and radial base neural net algorithm pair
Obtained several intrinsic modal components are predicted；
Combined prediction module, for using induced ordered weighted averaging operator to least square method supporting vector machine algorithm and diameter
The predicted value obtained to base neural net algorithm is combined reconstruct, obtains final predicted value.
A kind of technical solution of on the other hand computer readable storage medium that the disclosure provides is:
A kind of computer readable storage medium, is stored thereon with computer program, realization when which is executed by processor
Step in renewable energy power combination prediction technique based on empirical mode decomposition as described above.
A kind of technical solution of on the other hand computer equipment that the disclosure provides is:
A kind of computer equipment can be run on a memory and on a processor including memory, processor and storage
Computer program, characterized in that the processor is realized when executing described program as described above based on empirical mode decomposition
Step in renewable energy power combination prediction technique.
Through the above technical solutions, the beneficial effect of the disclosure is:
(1) unstable amount is decomposed into after several stable quantities using EMD decomposition method and carries out renewable energy again by the disclosure
The prediction of power effectively reduces randomness, indirect and the fluctuation of renewable energy to prediction bring interference；
(2) disclosure is using improving IOWA operator, by dynamic adjust weight coefficient by the predicted value of LS-SVM algorithm and
The predicted value portfolio restructuring of RBF algorithm obtains final predicted value, avoids the limitation of Individual forecast algorithm, improves prediction essence
Degree.
Detailed description of the invention
The Figure of description for constituting a part of this disclosure is used to provide further understanding of the disclosure, and the disclosure is shown
Meaning property embodiment and its explanation do not constitute the improper restriction to the disclosure for explaining the application.
Fig. 1 is the flow chart of one renewable energy power combination prediction technique of embodiment；
Fig. 2 is the original wind series figure of embodiment one；
Fig. 3 is signal graph of the embodiment after EMD decomposition；
Fig. 4 is each prediction algorithm prediction result figure of embodiment one；
Fig. 5 is each prediction model absolute error absolute value comparison diagram of embodiment one；
Fig. 6 is the structure chart of two renewable energy power combination forecasting system of embodiment.
Specific embodiment
The disclosure is described further with embodiment with reference to the accompanying drawing.
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the disclosure.Unless another
It indicates, all technical and scientific terms that the disclosure uses have logical with disclosure person of an ordinary skill in the technical field
The identical meanings understood.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
Explanation of nouns:
(1) EMD, Empirical Mode Decomposition, empirical mode decomposition is a kind of for non-linear, non-
The adaptive signal decomposition algorithm of stationary signal.
(2) LS-SVM, least square method supporting vector machine algorithm.
(3) RBF neural, radial base neural net are a kind of three-layer neural networks comprising input layer, hidden layer, defeated
Layer out；Transformation from the input space to hidden layer space is nonlinear, and is linear from hidden layer space to output layer spatial alternation
's.
(4) IOWA operator, induced ordered weighted averaging operator.
Embodiment one
The present embodiment provides a kind of renewable energy power combination prediction techniques decomposed based on EMD, firstly, by original sequence
Column are decomposed into several stationary components through EMD, then, use least square method supporting vector machine (LS-SVM) and radial base (RBF)
Neural network algorithm is combined prediction to each component, finally, weight coefficient is adjusted in real time with IOWA operator is improved, by each component
It is reconstructed into final prediction result.
Please refer to attached drawing 1, the renewable energy power combination prediction technique decomposed based on EMD the following steps are included:
S101 obtains the renewable energy operation data at multiple moment, forms renewable energy time series.
In the present embodiment, renewable energy operation data can be wind farm wind velocity data, can also be photovoltaic generation power
Data.
The renewable energy time series that the step 101 obtains is X (t).
S102 uses empirical mode decomposition method by renewable energy Time Series for several stable intrinsic mode
Component and a residual components.
Specifically, in the step 102, renewable energy time series X (t) is decomposed using empirical mode decomposition method
For several intrinsic modal components Ci(t) and a residual components R (t), expression formula are as follows:
In formula, X (t) indicates renewable energy time series, Ci(t) i-th of intrinsic modal components is indicated, R (t) indicates surplus
Remaining component.
Specifically, in the step 102, the specific of renewable energy time series is decomposed using empirical mode decomposition method
Implementation is as follows:
S102-1 finds all minimum points of renewable energy time series and maximum point to be decomposed, passes through three
All minimum points and maximum point fitting are become the envelope up and down of the renewable energy time series by secondary spline function.
Step 102-2 calculates being averaged for renewable energy time series to be decomposed and envelope up and down obtained above
Difference.
Step 102-3, judges whether difference obtained above meets IMF requirement.If the difference is unsatisfactory for requiring, again
The maximum and minimum for finding the renewable energy time series, repeat the above steps, until meeting the requirements；If the difference is full
Foot requires, then enabling the difference is the one-component of renewable energy time series, finds out the difference of original signal Yu this component,
It enables this difference as new signal to be decomposed, repeats the above steps, until the termination rules for meeting previously given terminate to decompose
Process.
In the present embodiment, IMF requires as there is no negative local maximums and positive local minimum.
Step 102-4 obtains final decomposition result by above step are as follows:
In formula, X (t) indicates original renewable energy time series, Ci(t) i-th of intrinsic modal components, R (t) table are indicated
Show residual components.
The present embodiment uses EMD decomposition method, and unstable amount is decomposed into several stable quantities and carries out renewable energy function again
The prediction of rate effectively reduces randomness, indirect and the fluctuation of renewable energy to prediction bring interference.
S103, several intrinsic mode point that step 102 is obtained using least square method supporting vector machine algorithm (LS-SVM)
Amount is predicted.
Specifically, in the step 103, several intrinsic modal components are predicted using LS-SVM algorithm specific reality
Existing mode is as follows:
Given prediction collection data acquisition system (xi,yi), i=1,2 ..., l, xi∈Rd, yi∈ R, corresponding regression function are as follows:
In formula, xiFor the intrinsic modal components of input, yiFor the predicted value of output, l is number of samples in training set, and w is power
Weight coefficient, b is deviation,For nonlinear mapping function.
Optimize weight coefficient w, optimization object function and constraint may be expressed as:
In formula, eiFor slack variable error term, C is punishment regularization parameter, indicates the punishment degree to error.
The optimization object function and constraint condition introduce Lagrange multiplier α and obtain unrestricted function are as follows:
In formula, xiFor input vector, yiFor scalar output, l is number of samples in training set, and w is weight coefficient, and b is inclined
Difference,For nonlinear mapping function, eiFor slack variable error term.
It can be obtained by KKT condition:
Lagrangian is collated to be obtained:
In formula, G=(1,1 ...)T, α=(α1,α2,…,αl)T,Ω∈Rl×lFor l × l dimensional vector, andK indicates the kernel function for meeting Mercer requirement.
Kernel function K is due to mapping functionIt can during input vector is mapped to higher dimensional space from the input space
Can meeting so that dimension generation it is explosive so thatOperation it is considerably complicated and introduce, the kernel function makes
With gaussian radial basis function:
Wherein, xiFor input vector, σ is the width parameter of function, controls the radial effect range of function.
It is optimized by parameter σ of the genetic algorithm to kernel function.
S104 carries out several intrinsic modal components that step 102 obtains using radial base (RBF) neural network algorithm pre-
It surveys.
Specifically, in the step 104, RBF neural has three layers, including input layer, hidden layer, output layer, input
Layer is first layer, is signal source node；Hidden layer is the second layer, for the input space to be mapped to higher dimensional space, hidden layer list
Depending on the number of member depends on actual need, transforming function transformation function is radial basis function；Output layer is third layer, for making sound to input
It answers.
The basic function of the hidden layer is Gaussian function, is indicated are as follows:
In formula, ciIndicate the center of i-th of basic function, σiConstant is extended for i-th of hidden layer node, which determines
The shape of radial basis function.
Using unsupervised learning method come Selecting All Parameters, m are chosen from training sample set using k-means clustering method and is gathered
Class center is regarded to m center of radial basis function, the selection formula of the extension constant by these points are as follows:
In formula, cmaxFor the maximum distance between the center of selection, m is the number by clustering obtained implicit node.
Hidden layer neuron number and extension constant are optimized by genetic algorithm.
S105 is combined reconstruct using the predicted value that improved IOWA operator obtains step 103 and step 104, obtains
To final predicted value.
Specifically, in the step 105, improved IOWA operator is as follows:
In formula, xitPredicted value for i-th kind of prediction technique in t moment, i=1,2 ..., N, t=1,2 ..., N, aitIt is
The precision of prediction of i kind prediction technique t moment, xa-index(it)For certain individual event prediction algorithm in t moment according to precision of prediction
The predicted value of sequence, kiFor the corresponding weight coefficient of i-th kind of prediction technique predicted value of each moment point.
The solution formula of the combining weights coefficient are as follows:
In formula, xitPredicted value for i-th kind of prediction technique in t moment, i=1,2 ..., N, t=1,2 ..., N,
xa-index(it)For the predicted value that certain individual event prediction algorithm sorts in t moment according to precision of prediction, kiIt is i-th kind of each moment point
The corresponding weight coefficient of prediction technique predicted value.
In the present embodiment, the prediction technique of use includes LS-SVM algorithm and RBF neural network algorithm, and use is improved
The corresponding weight coefficient of predicted value for the LS-SVM algorithm that IOWA operator is adjusted and the predicted value of RBF neural network algorithm are corresponding
Weight coefficient.
Specifically, in the step 105, step 103 and step 104 are obtained using improved IOWA operator predicted value
The specific implementation being combined is as follows:
Step 105-1 obtains LS-SVM algorithm and RBF neural network algorithm to the predicted value of intrinsic modal components, setting
The precision of prediction of LS-SVM algorithm and RBF neural network algorithm each moment.
The intrinsic modal components actual value that certain prediction technique is predicted is xt, t=1,2 ..., N, if using altogether
M kind prediction technique predicts it, then xitPredicted value for i-th kind of prediction technique in t moment, i=1,2 ..., N, t=1,
2 ..., N, the present embodiment is predicted using two kinds of prediction techniques of LS-SVM algorithm and RBF neural network algorithm respectively, if aitFor
The precision of prediction of i-th kind of prediction technique t moment is expressed as follows shown:
If precision of prediction aitMeetAnd using the precision of prediction at each moment as m kind prediction technique every
The induction value at one moment.
The precision of prediction of all prediction techniques and predicted value are constituted a two-dimensional array (< a by step 105-21t,x1t>,<
a2t,x2t>,…,<amt,xmt>)。
Step 105-3, by m kind prediction technique each moment predicted value according to precision of prediction aitSize sequence,
xa-index(it)Indicate the predicted value that certain individual event prediction algorithm sorts in t moment according to precision of prediction, kiIndicate each moment point
The corresponding weight coefficient of i-th kind of prediction technique predicted value.
Step 105-4 solves the combining weights coefficient k of LS-SVM algorithm and RBF neural network algorithm1、k2, solution formula
Are as follows:
In formula, xitPredicted value for i-th kind of prediction technique in t moment, i=1,2 ..., N, t=1,2 ..., N,
xa-index(it)For the predicted value that certain individual event prediction algorithm sorts in t moment according to precision of prediction, kiIt is i-th kind of each moment point
The corresponding weight coefficient of prediction technique predicted value.
S105-5 establishes weighted array prediction model, solves IOWA operator combination predicted value.
In the present embodiment, the weighted array prediction model that the step 105 is established are as follows:
f1(t)=k1x1(t)+k2x2(t)
Wherein, k1For a kind of higher predicted value of algorithm of precision of prediction in LS-SVM algorithm and RBF neural network algorithm
Corresponding weight coefficient；k2For a kind of corresponding weight coefficient of predicted value of algorithm of residue；x1It (t) is LS-SVM algorithm and RBF mind
Through a kind of higher predicted value of algorithm of precision of prediction in network algorithm；x2It (t) is a kind of remaining predicted value of algorithm；f1(t) it is
IOWA operator combination predicted value.
The disclosure adjusts weight coefficient for the predicted value and RBF of LS-SVM algorithm using IOWA operator is improved, by dynamic
The predicted value portfolio restructuring of algorithm obtains final predicted value, avoids the problem of Individual forecast algorithm is brought.
Experimental verification
In order to prove the present embodiment propose prediction technique validity, by taking the wind power plant of Jinan as an example, using Rstudio
Language writes algorithm routine, constructs three kinds of prediction models: LS-SVM prediction model respectively, and RBF neural prediction model changes
Into IOWA operator combination prediction model, predictive simulation is carried out to the wind speed in the one section time.Fig. 2 is that wind power plant surveys wind speed number
According to sample, which takes 500 hours to sample a point per hour per hour for interval altogether, and preceding 450 hours are used to instruct
Practice, 1h prediction in advance is carried out to rear 50 hours.Mean square error RMSE is respectively adopted to the performance evaluation of prediction result and is put down
Equal absolute percent error MAPE.Wherein, mean square error RMSE and mean absolute percentage error MAPE expression formula are as follows:
In formula, xiIndicate actual value, x 'iIndicate the predicted value at the moment, n indicates the number of future position.
To time series of the original air speed data after EMD is decomposed, arranged altogether according to from high frequency to low frequency
6 IMF components and a residual components are decomposed into, as shown in Figure 3.
Fig. 4 is each prediction algorithm prediction result figure, can obtain each prediction model absolute error absolute value in Fig. 5 by calculating
And 1 three kinds of respective prediction evaluation indexes of prediction model of table.
1 three kinds of respective prediction evaluation indexes of prediction model of table
It is computed verifying, is decomposed without using EMD, mean square error is 0.41 when LS-SVM prediction is used only, average absolute hundred
Dividing ratio error is 7.19%, is individually 0.47 with RBF neural prediction mean square error, average absolute value percentage error is
7.03%, it the use of the prediction technique mean square error that the present embodiment proposes is 6.22, average absolute value percentage error is 5.96%,
Mean square error reduces 14.6% and 25.5% compared to LS-SVM and RBF neural respectively.Simulation result shows EMD
The randomness that can reduce time series is decomposed, while combination forecasting has merged each individual event prediction model advantage, can effectively mention
High precision of prediction.Forecasting wind speed result be can be calculated into prediction wind power by wind-powered electricity generation model.
Embodiment two
The present embodiment provides a kind of renewable energy power combination forecasting system based on empirical mode decomposition, the system packet
It includes:
Time series constructs module 201, for obtaining the renewable energy operation data at multiple moment, forms renewable energy
Source time sequence；
Decomposing module 202, for using empirical mode decomposition method by renewable energy Time Series for several
Levy modal components and a residual components；
Monomer prediction module 203 is calculated for being utilized respectively least square method supporting vector machine algorithm and radial base neural net
Method predicts obtained several intrinsic modal components；
Combined prediction module 204, for using induced ordered weighted averaging operator to least square method supporting vector machine algorithm
The predicted value obtained with radial base neural net algorithm is combined reconstruct, obtains final predicted value.
Specifically, the decomposing module 202 is specifically used for:
All minimum points of renewable energy time series and maximum point to be decomposed are searched, cubic spline letter is passed through
All minimum points and maximum point fitting are become the envelope up and down of the renewable energy time series by number；
Calculate renewable energy time series and the mean difference of obtained envelope up and down to be decomposed；
Judge whether obtained mean difference meets IMF requirement, i.e., there is no negative local maximum and positive local poles
Small value；If not satisfied, then finding the maximum and minimum of the renewable energy time series again, repeat the above steps, directly
To meeting the requirements；If meeting the requirements, enabling the mean difference is first intrinsic modal components of renewable energy time series,
The difference for finding out former data Yu this intrinsic modal components enables this difference as new data to be decomposed, repeats the above steps,
Termination rules until meeting setting terminate decomposable process.
Specifically, the monomer prediction module 203 is specifically used for:
The weight coefficient of Optimized Least Square Support Vector algorithm, using weight coefficient, nonlinear mapping function and partially
Difference building regression function；
Several intrinsic modal components composing training collection input regression functions that will be obtained, export predicted value.
The optimization object function of the weight coefficient of the least square method supporting vector machine algorithm and constraint may be expressed as:
In formula, eiFor slack variable error term, C is punishment regularization parameter, indicates the punishment degree to error；xiIt is defeated
Incoming vector, l are number of samples in training set, and w is weight coefficient, and b is deviation,For nonlinear mapping function.
The optimization object function of the LS-SVM and constraint introduce Lagrange multiplier α and obtain unrestricted function are as follows:
It can be obtained by KKT condition
The Lagrangian is collated to be obtained:
In formula, G=(1,1 ...)T, α=(α1,α2,…,αl)T,Ω∈Rl×lFor l × l dimensional vector, andK indicates the kernel function for meeting Mercer.
The kernel function uses gaussian radial basis function:
Specifically, the monomer prediction module 203 also particularly useful for:
Each hidden layer node extension constant is chosen, radial basis function of the Gaussian function as hidden layer is constructed；
Several intrinsic modal components are sent into the input layer of radial base neural net, use radial basis function by hidden layer
Intrinsic modal components are transformed into higher dimensional space, the predicted value of intrinsic modal components is exported by output layer.
The radial basis function of the hidden layer is Gaussian function, is indicated are as follows:
In formula, ciIndicate the center of i-th of basic function, σiConstant is extended for i-th of hidden layer node.
Extend the selection formula of constant are as follows:
In formula, cmaxFor the maximum distance between the center of selection, m is the number for passing through the implicit node that cluster obtains.
Specifically, the combined prediction module 204 is specifically used for:
Least square method supporting vector machine algorithm and radial base neural net algorithm are obtained to the predicted value of intrinsic modal components,
Least square method supporting vector machine algorithm and radial base neural net algorithm are set in the precision of prediction at per moment；
Using least square method supporting vector machine algorithm and radial base neural net algorithm to the predicted value of intrinsic modal components
And the precision of prediction of least square method supporting vector machine algorithm and radial base neural net algorithm at per moment constitutes two-dimensional array；
By least square method supporting vector machine algorithm and radial base neural net algorithm at per moment to intrinsic modal components
Predicted value is ranked up according to corresponding precision of prediction size；
Solve the combining weights coefficient of least square method supporting vector machine algorithm and radial base neural net algorithm；
Combining weights coefficient based on least square method supporting vector machine algorithm and radial base neural net algorithm establishes weighting
Combination forecasting；
Least square method supporting vector machine algorithm and radial base neural net algorithm is defeated to the predicted value of intrinsic modal components
Enter weighted array prediction model and be combined prediction, obtains the combined prediction value of induced ordered weighted averaging operator.
The IOWA combined prediction value formula are as follows:
In formula, xitPredicted value for i-th kind of prediction technique in t moment, i=1,2 ..., N, t=1,2 ..., N, aitIt is
The precision of prediction of i kind prediction technique t moment, xa-index(it)For certain individual event prediction algorithm in t moment according to precision of prediction
The predicted value of sequence, kiFor the corresponding weight coefficient of i-th kind of prediction technique predicted value of each time point.
The solution formula of the combining weights coefficient are as follows:
In formula, xitPredicted value for i-th kind of prediction technique in t moment, i=1,2 ..., N, t=1,2 ..., N,
xa-index(it)For the predicted value that certain individual event prediction algorithm sorts in t moment according to precision of prediction, kiIt is i-th kind of each moment point
The corresponding weight coefficient of prediction technique predicted value.
The renewable energy power combination forecasting system that the present embodiment proposes uses empirical mode decomposition by decomposing module
Unstable amount is decomposed into the prediction for carrying out renewable energy source power after several stable quantities again by method, effectively reduces renewable energy
Randomness, indirect and the fluctuation in source are to prediction bring interference；By combined prediction module using IOWA operator is improved, lead to
It crosses dynamic adjustment weight coefficient and the predicted value portfolio restructuring of the predicted value of LS-SVM algorithm and RBF algorithm is obtained into final prediction
Value avoids the limitation of Individual forecast algorithm, improves precision of prediction.
Embodiment three
The present embodiment provides a kind of computer readable storage mediums, are stored thereon with computer program, and the program is processed
The step in the renewable energy power combination prediction technique based on empirical mode decomposition as shown in Figure 1 is realized when device executes.
Example IV
The present embodiment provides a kind of computer equipment, including memory, processor and storage on a memory and can located
The computer program run on reason device, realization is as shown in Figure 1 when the processor executes described program is divided based on empirical modal
Step in the renewable energy power combination prediction technique of solution.
It should be understood by those skilled in the art that, embodiment of the disclosure can provide as method, system or computer program
Product.Therefore, the shape of hardware embodiment, software implementation or embodiment combining software and hardware aspects can be used in the disclosure
Formula.Moreover, the disclosure, which can be used, can use storage in the computer that one or more wherein includes computer usable program code
The form for the computer program product implemented on medium (including but not limited to magnetic disk storage and optical memory etc.).
Although above-mentioned be described in conjunction with specific embodiment of the attached drawing to the disclosure, model not is protected to the disclosure
The limitation enclosed, those skilled in the art should understand that, on the basis of the technical solution of the disclosure, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within the protection scope of the disclosure.
Claims (10)
1. a kind of renewable energy power combination prediction technique based on empirical mode decomposition, characterized in that the following steps are included:
The renewable energy operation data at multiple moment is obtained, renewable energy time series is formed；
Empirical mode decomposition method is used to remain renewable energy Time Series for several intrinsic modal components and one
Remaining component；
Least square method supporting vector machine algorithm and radial base neural net algorithm are utilized respectively to obtained several intrinsic mode point
Amount is predicted；
Least square method supporting vector machine algorithm and radial base neural net algorithm are obtained using induced ordered weighted averaging operator
Predicted value be combined reconstruct, obtain final predicted value.
2. the renewable energy power combination prediction technique according to claim 1 based on empirical mode decomposition, feature
It is that the renewable energy operation data is wind farm wind velocity data or photovoltaic generation power data.
3. the renewable energy power combination prediction technique according to claim 1 based on empirical mode decomposition, feature
It is, it is described to include: by the step of renewable energy Time Series using empirical mode decomposition method
All minimum points of renewable energy time series and maximum point to be decomposed are searched, it will by cubic spline function
All minimum points and maximum point fitting become the envelope up and down of the renewable energy time series；
Calculate renewable energy time series and the mean difference of obtained envelope up and down to be decomposed；
Judge whether obtained mean difference meets IMF requirement, i.e., there is no negative local maximums and positive local minimum；
If not satisfied, then finding the maximum and minimum of the renewable energy time series again, repeat the above steps, until meeting
It is required that；If meeting the requirements, enabling the mean difference is first intrinsic modal components of renewable energy time series, finds out original
The difference of data and this intrinsic modal components enables this difference as new data to be decomposed, repeats the above steps, Zhi Daoman
The termination condition set enough terminates decomposable process.
4. the renewable energy power combination prediction technique according to claim 1 based on empirical mode decomposition, feature
It is that described the step of being predicted using least square method supporting vector machine algorithm obtained several intrinsic modal components includes:
The weight coefficient of Optimized Least Square Support Vector algorithm utilizes weight coefficient, nonlinear mapping function and deviation structure
Build regression function；
It by obtained several intrinsic modal components composing training collection and inputs regression function and handles, export intrinsic modal components
Predicted value.
5. the renewable energy power combination prediction technique according to claim 1 based on empirical mode decomposition, feature
It is that the optimization object function of the weight coefficient of the least square method supporting vector machine algorithm and constraint may be expressed as:
In formula, eiFor slack variable error term, C is punishment regularization parameter, indicates the punishment degree to error；xiFor input to
Amount, l are number of samples in training set, and w is weight coefficient, and b is deviation,For nonlinear mapping function.
6. the renewable energy power combination prediction technique according to claim 1 based on empirical mode decomposition, feature
It is that described the step of being predicted using radial base neural net algorithm obtained several intrinsic modal components includes:
Each hidden layer node extension constant is chosen, radial basis function of the Gaussian function as hidden layer is constructed；
Several intrinsic modal components are sent into the input layer of radial base neural net, incite somebody to action this using radial basis function by hidden layer
Sign modal components transform to higher dimensional space, and the predicted value of intrinsic modal components is exported by output layer.
7. the renewable energy power combination prediction technique according to claim 1 based on empirical mode decomposition, feature
It is, it is described that least square method supporting vector machine algorithm and radial base neural net algorithm are obtained using induced ordered weighted averaging operator
To predicted value be combined reconstruct the step of include:
Least square method supporting vector machine algorithm and radial base neural net algorithm are obtained to the predicted value of intrinsic modal components, setting
The precision of prediction of least square method supporting vector machine algorithm and radial base neural net algorithm at per moment；
Using least square method supporting vector machine algorithm and radial base neural net algorithm to the predicted value of intrinsic modal components and
The precision of prediction of least square method supporting vector machine algorithm and radial base neural net algorithm at per moment constitutes two-dimensional array；
Prediction by least square method supporting vector machine algorithm and radial base neural net algorithm at per moment to intrinsic modal components
Value is ranked up according to corresponding precision of prediction size；
Solve the combining weights coefficient of least square method supporting vector machine algorithm and radial base neural net algorithm；
Combining weights coefficient based on least square method supporting vector machine algorithm and radial base neural net algorithm, establishes weighted array
Prediction model,
Least square method supporting vector machine algorithm and radial base neural net algorithm add the predicted value input of intrinsic modal components
Power combination forecasting is combined prediction, obtains the combined prediction value of induced ordered weighted averaging operator.
8. a kind of renewable energy power combination forecasting system based on empirical mode decomposition, characterized in that include:
Time series constructs module, for obtaining the renewable energy operation data at multiple moment, forms renewable energy source time
Sequence；
Decomposing module, for using empirical mode decomposition method by renewable energy Time Series for several intrinsic mode point
Amount and a residual components；
Monomer prediction module, for being utilized respectively least square method supporting vector machine algorithm and radial base neural net algorithm to obtaining
Several intrinsic modal components predicted；
Combined prediction module, for using induced ordered weighted averaging operator to least square method supporting vector machine algorithm and radial base
The predicted value that neural network algorithm obtains is combined reconstruct, obtains final predicted value.
9. a kind of computer readable storage medium, is stored thereon with computer program, characterized in that the program is executed by processor
The Shi Shixian renewable energy power combination prediction technique for example of any of claims 1-7 based on empirical mode decomposition
In step.
10. a kind of computer equipment including memory, processor and stores the meter that can be run on a memory and on a processor
Calculation machine program, characterized in that realize when the processor executes described program and be based on as of any of claims 1-7
Step in the renewable energy power combination prediction technique of empirical mode decomposition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910562179.2A CN110264012B (en) | 2019-06-26 | 2019-06-26 | Renewable energy power combination prediction method and system based on empirical mode decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910562179.2A CN110264012B (en) | 2019-06-26 | 2019-06-26 | Renewable energy power combination prediction method and system based on empirical mode decomposition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110264012A true CN110264012A (en) | 2019-09-20 |
| CN110264012B CN110264012B (en) | 2021-05-07 |
Family
ID=67921858
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910562179.2A Active CN110264012B (en) | 2019-06-26 | 2019-06-26 | Renewable energy power combination prediction method and system based on empirical mode decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110264012B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111049134A (en) * | 2019-12-26 | 2020-04-21 | 清华大学 | Calculation method for peak clipping requirements of multi-energy complementary park response power system |
| CN111258980A (en) * | 2020-01-18 | 2020-06-09 | 重庆邮电大学 | Dynamic file placement method based on combination prediction in cloud storage system |
| CN111276983A (en) * | 2019-12-27 | 2020-06-12 | 西华大学 | Photovoltaic smoothing method and system based on fuzzy self-adaptive complete empirical mode decomposition algorithm |
| CN111327052A (en) * | 2020-03-12 | 2020-06-23 | 清华大学 | Method and device for accelerating and optimizing combination of power system units |
| CN111985711A (en) * | 2020-08-19 | 2020-11-24 | 华北电力大学（保定） | Wind power probability prediction model establishing method |
| CN116565898A (en) * | 2023-07-04 | 2023-08-08 | 昆明理工大学 | Wind-storage combined primary frequency modulation self-adaptive control method based on wind speed prediction |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108233417A (en) * | 2018-01-22 | 2018-06-29 | 华北电力大学（保定） | A kind of regenerative resource of more prediction model cascade compensation parallel coupleds, which is contributed, predicts computational methods |
| CN108520269A (en) * | 2018-03-10 | 2018-09-11 | 华北电力大学（保定） | A kind of wind speed forecasting method and forecasting wind speed system |
| KR20190034917A (en) * | 2017-09-25 | 2019-04-03 | 엘에스웨어(주) | Blockchain Data analyzing Method based on Neural Network for Electric Power Trading |
-
2019
- 2019-06-26 CN CN201910562179.2A patent/CN110264012B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20190034917A (en) * | 2017-09-25 | 2019-04-03 | 엘에스웨어(주) | Blockchain Data analyzing Method based on Neural Network for Electric Power Trading |
| CN108233417A (en) * | 2018-01-22 | 2018-06-29 | 华北电力大学（保定） | A kind of regenerative resource of more prediction model cascade compensation parallel coupleds, which is contributed, predicts computational methods |
| CN108520269A (en) * | 2018-03-10 | 2018-09-11 | 华北电力大学（保定） | A kind of wind speed forecasting method and forecasting wind speed system |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111049134A (en) * | 2019-12-26 | 2020-04-21 | 清华大学 | Calculation method for peak clipping requirements of multi-energy complementary park response power system |
| CN111276983A (en) * | 2019-12-27 | 2020-06-12 | 西华大学 | Photovoltaic smoothing method and system based on fuzzy self-adaptive complete empirical mode decomposition algorithm |
| CN111258980A (en) * | 2020-01-18 | 2020-06-09 | 重庆邮电大学 | Dynamic file placement method based on combination prediction in cloud storage system |
| CN111258980B (en) * | 2020-01-18 | 2024-02-27 | 广州大鱼创福科技有限公司 | Dynamic file placement method based on combined prediction in cloud storage system |
| CN111327052A (en) * | 2020-03-12 | 2020-06-23 | 清华大学 | Method and device for accelerating and optimizing combination of power system units |
| CN111327052B (en) * | 2020-03-12 | 2021-01-29 | 清华大学 | Method and device for accelerating and optimizing combination of power system units |
| CN111985711A (en) * | 2020-08-19 | 2020-11-24 | 华北电力大学（保定） | Wind power probability prediction model establishing method |
| CN111985711B (en) * | 2020-08-19 | 2024-02-02 | 华北电力大学（保定） | Wind power probability prediction model building method |
| CN116565898A (en) * | 2023-07-04 | 2023-08-08 | 昆明理工大学 | Wind-storage combined primary frequency modulation self-adaptive control method based on wind speed prediction |
| CN116565898B (en) * | 2023-07-04 | 2023-09-15 | 昆明理工大学 | Wind-storage combined primary frequency modulation self-adaptive control method based on wind speed prediction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110264012B (en) | 2021-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110264012A (en) | Renewable energy power combination prediction technique and system based on empirical mode decomposition | |
| CN110059878B (en) | Photovoltaic power generation power prediction model based on CNN LSTM and construction method thereof | |
| Wang et al. | A novel combined model based on hybrid optimization algorithm for electrical load forecasting | |
| Wang et al. | Forecasting the seasonal natural gas consumption in the US using a gray model with dummy variables | |
| CN105512745A (en) | Wind power section prediction method based on particle swarm-BP neural network | |
| CN104636985A (en) | Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network | |
| CN103440370A (en) | Transmission and transformation project construction cost assessment method and device | |
| CN105447509A (en) | Short-term power prediction method for photovoltaic power generation system | |
| CN109193756B (en) | Scene decoupling dynamic economic dispatching model solving method of wind power access system | |
| CN106779177A (en) | Multiresolution wavelet neutral net electricity demand forecasting method based on particle group optimizing | |
| Tsang et al. | Autonomous household energy management using deep reinforcement learning | |
| Li et al. | Short term prediction of photovoltaic power based on FCM and CG-DBN combination | |
| CN113255138A (en) | Load distribution optimization method for power system | |
| CN114626573A (en) | Load prediction method for optimizing extreme learning machine based on improved multivariate universe algorithm | |
| CN112966883A (en) | Comprehensive energy system operation optimization method and terminal equipment | |
| CN117077960A (en) | Day-ahead scheduling optimization method for regional comprehensive energy system | |
| CN111401659A (en) | Ultra-short-term or short-term photovoltaic power generation power prediction method based on case reasoning | |
| CN114117925A (en) | Comprehensive energy system planning method considering high-proportion distributed photovoltaic consumption | |
| Zhang et al. | Joint optimization of the number, type and layout of wind turbines for a new offshore wind farm | |
| Zhao et al. | Super short term combined power prediction for wind power hydrogen production | |
| Zhu | Research on adaptive combined wind speed prediction for each season based on improved gray relational analysis | |
| Hussain et al. | Energy management of buildings with energy storage and solar photovoltaic: A diversity in experience approach for deep reinforcement learning agents | |
| CN109615142A (en) | A kind of wind farm wind velocity combination forecasting method based on wavelet analysis | |
| CN109980695A (en) | A kind of distributed energy and user's behavior prediction method of distribution system | |
| Yin et al. | Quantum-inspired distributed policy-value optimization learning with advanced environmental forecasting for real-time generation control in novel power systems |
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 |