CN109086941A - A kind of energy-consuming prediction technique - Google Patents
A kind of energy-consuming prediction technique Download PDFInfo
- Publication number
- CN109086941A CN109086941A CN201810935717.3A CN201810935717A CN109086941A CN 109086941 A CN109086941 A CN 109086941A CN 201810935717 A CN201810935717 A CN 201810935717A CN 109086941 A CN109086941 A CN 109086941A
- Authority
- CN
- China
- Prior art keywords
- energy
- frog
- prediction
- imf
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Operations Research (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to energy forecast technical fields, a kind of more particularly to energy-consuming prediction technique, comprising: sample data of the acquisition including historic energy consumption figure, the size of population, GDP, the industrial structure, energy consumption structure, Energy Intensity, carbon intensity and total import and export value;Nondimensionalization processing is carried out to sample data, and calculates the grey relational grade of each sample data and energy consumption structure, according to the sequence of grey relational grade come the input factor of screening model;It treats forecasting sequence and carries out the sequence noise reduction based on integrated empirical mode decomposition, obtain multiple IMF components;With the parameter of improved shuffled frog leaping algorithm Optimized Least Square Support Vector, and prediction model is established, the result of prediction is reconstructed, obtains final energy-consuming prediction result.Experiments have shown that predicting that energy-consuming, prediction effect is significant with EMD-ISFLA-LSSVM model.
Description
Technical field
The invention belongs to energy forecast technical field more particularly to a kind of energy-consuming prediction techniques.
Background technique
The energy is the important material base of social development, and consumption affects natural environment and economic stabilization is persistently sent out
Exhibition.With the fast development of China's economic, energy consumption is constantly soaring, has become maximum energy-consuming in the world
State.Therefore, by establishing suitable energy-consuming prediction model, Accurate Prediction total energy consumption can be formulated for China and be closed
The production of energy plan of reason and energy-saving and emission-reduction relevant policies provide the foundation of science, to pushing holding for world economy and environment
Supervention exhibition has important practical significance.
Energy resource system is the non-linear system of complexity, and the development and change of consumption figure have growth property and fluctuation.Mesh
Before, scholars propose various models and predict for energy-consuming, such as: gray theory, fuzzy logic method, multivariate regression models
Deng, but method is relatively traditional.
In recent years, artificial intelligence technology continues to develop, and artificial intelligence prediction model is widely used in prediction field, also wraps
Include energy-consuming prediction field.Support vector machines is at present using a kind of extremely wide artificial intelligence prediction model.Support to
The VC dimension that amount machine is built upon Statistical Learning Theory is theoretical on Structural risk minization basis, is believed according to limited sample
Breath seeks best compromise between the complexity and learning ability of model, in the hope of obtaining the supervised learning mould of best Generalization Ability
Type.It shows many distinctive advantages in solution small sample, the identification of non-linear and high dimensional pattern, and can promote the use of
In the Machine Learning Problems such as prediction.Least square method supporting vector machine is using least-square cost function and equality constraint to standard
A kind of innovatory algorithm of support vector machines, compared with standard support vector machines, with faster training speed and better receipts
Hold back precision.Currently, have scholar to predict energy-consuming using least square method supporting vector machine.In order to improve the standard of prediction
True property, researchers optimize the parameter of LSSVM using many algorithms, as ant group algorithm, artificial bee colony algorithm, grey wolf are calculated
A variety of optimization algorithms such as method.
But since energy-consuming prediction is interfered vulnerable to extraneous factor, sequence includes that noise is more, how to be able to achieve to the energy
It is still the technical issues that need to address that consumption, which carries out accurate prediction,.
Summary of the invention
In order to realize the Accurate Prediction to energy-consuming, the invention proposes a kind of energy-consuming prediction techniques, comprising:
Step 1: acquisition includes historic energy consumption figure, the size of population, GDP, the industrial structure, energy consumption structure, the energy
Sample data including intensity, carbon intensity and total import and export value;
Step 2: nondimensionalization processing being carried out to sample data, and calculates the ash of each sample data and energy consumption structure
The color degree of association, according to the sequence of grey relational grade come the input factor of screening model;
Step 3: treating forecasting sequence and obtained multiple based on the sequence noise reduction of integrated empirical mode decomposition (EEMD)
IMF component;
Step 4: optimizing most to by integrating the IMF component that empirical mode decomposition obtains with improved shuffled frog leaping algorithm
Small two multiply the parameter of support vector machines (ISFLA-LSSVM), and establish prediction model, are reconstructed, obtain to the result of prediction
Final energy-consuming prediction result.
The integrated empirical mode decomposition is that Gauss is added in original signal on the basis of Conventional wisdom mode decomposition
White noise, the statistical property being distributed using white noise frequency-flat eliminate the intermittency in original signal, to effectively inhibit
Modal overlap problem.
Specific step is as follows for the integrated empirical mode decomposition:
Step 301: random Gaussian white noise sequence is added in echo signal,
Xm(t)=X (t)+knm(t) (1)
In formula, k is the amplitude coefficient for the white noise being added, nmIt (t) is white noise sequence, X (t) is echo signal, Xm(t)
For the signal that white noise is added;
Step 302: the signal that white noise is added is decomposed into one group of IMF using EMD;
Step 303: different white noise sequences being added every time, repeats the above steps;
Step 304: calculate the mean value of IMF after decomposing, the mean value of each IMF that decomposition is obtained as finally as a result,
In formula, m is the integrated number of EMD, m=1,2 ..., N;ci,mGenerated i-th of IMF is decomposed for the m times EMD,For the mean value for decomposing obtained each IMF.
The improved shuffled frog leaping algorithm is the inertia power that decreases in non-linear is introduced in traditional shuffled frog leaping algorithm
Value, avoids algorithm from falling into locally optimal solution.
The improved shuffled frog leaping algorithm formula is as follows:
Di=rand () (Xbi-Xwi) (8)
X′w=ω Xw+Di(-Dmax≤Di≤Dmax) (10)
In formula, DiFor the step-length of frog, rand () is distributed across the random number between [0,1], Xbi、XwiRespectively adapt to
Spend i-th optimal and worst of frog;X′wFitness for the frog obtained after update, XwFor the fitness of original frog,
ω is inertia weight, DmaxIt is the maximum step-length of frog, ωmaxFor maximum inertia weight;ωminFor minimum inertia weight;T is to work as
The product of the number of iterations of preceding sub- population and current total mixed iteration number;K is the total the number of iterations of sub- population.
The parameter with improved shuffled frog leaping algorithm Optimized Least Square Support Vector specifically includes:
Step 401: the parameter of improved shuffled frog leaping algorithm being set, frog population is initialized;
Step 402: calculating the fitness value of each frog individual and be ranked up;
Step 403: carrying out sub-group division, and determine the optimal solution in each sub-group, worst solution and group's overall situation most
Excellent solution;
Step 404: local search being done to the worst frog individual in each sub-group and is updated operation, until part
Search terminates;
Step 405: updated sub-group is mixed;
Step 406: judging whether to reach maximum number of iterations, if reached, stop optimization, export optimal solution, otherwise turn
To step 402;
Step 407: the parameter value after optimization being assigned to least square method supporting vector machine, building prediction model is predicted.
Beneficial effects of the present invention:
The present invention realizes the optimization to LSSVM using shuffled frog leaping algorithm, and shuffled frog leaping algorithm simulates frog
The behavior of information exchange is unfolded when each sub-group looks for food in group, is a kind of completely new heuristic Swarm Evolution algorithm,
With efficient calculated performance and excellent ability of searching optimum.Noise reduction is carried out to energy-consuming sequence using EEMD algorithm.It is real
Verify it is bright energy-consuming is predicted with EMD-ISFLA-LSSVM model, the fitting degree of prediction curve and actual curve
Very good, prediction effect is significant.
Detailed description of the invention
Fig. 1 is the integrated empirical mode decomposition result in embodiment.
Fig. 2 a~2d is respectively the prediction result and residue signal of IMF1, IMF2, IMF3 in embodiment.
Fig. 3 is final prediction result and residual error in embodiment.
Specific embodiment
With reference to the accompanying drawing, it elaborates to embodiment.
1, empirical mode decomposition
(1)EMD
Sophisticated signal is decomposed into several IMF (Intrinsic by screening mode by EMD (empirical mode decomposition) algorithm
Mode Function), IMF illustrates the internal characteristics vibration mode of signal.IMF component must satisfy following two condition:
1) number of extreme point and the number of zero crossing it is identical or at most differ 1;
2) upper and lower envelope answers Local Symmetric.
For signal X (t), the process of EMD algorithm is as follows:
1) all maximum of signal X (t) and minimum point are determined;
2) according to the maximum of signal and minimum point, using the method for cubic spline interpolation construct respectively X (t) it is upper,
The f of lower envelope linea(t) and fb(t);
3) local mean value of signal: f is acquiredm(t)=(fa(t)+fb(t))/2;
4) X (t) and f are calculatedm(t) difference, E (t)=X (t)-fm(t);
5) judge whether E (t) meets IMF condition, obtain first IMF component imf if meeting1(t), it otherwise repeats
Above-mentioned steps are until signal meets IMF condition.
6) r (t)=X (t)-imf is enabled1(t), judge whether r (t) needs to continue to decompose, replace X with r (t) if needing
(t) above step is repeated, otherwise decomposing terminates.
So far, signal EMD decomposition terminates, and last signal X (t) is decomposed into several IMF components imfi(t) and residual components
The sum of r (t).I.e.
(2) EEMD (integrated empirical mode decomposition)
EMD algorithm can introduce modal overlap problem, and to solve this problem, Huang and Wu propose EEMD method.It is passing
It unites on the basis of EMD, white Gaussian noise is added in EEMD method in original signal, the statistics being distributed using white noise frequency-flat
Characteristic eliminates the intermittency in original signal, to effectively inhibit modal overlap problem.
The specific decomposition step of EEMD is as follows:
1) random Gaussian white noise sequence is added in echo signal.
Xm(t)=X (t)+knm(t) (1)
In formula, k is the amplitude coefficient for the white noise being added, nmIt (t) is white noise sequence.
2) signal that white noise is added is decomposed into one group of IMF using EMD;
3) different white noise sequences is added every time, repeats the above steps;
4) mean value of IMF after decomposing is calculated, the mean value for each IMF that decomposition is obtained is as final result.
In formula, N is the integrated number of EMD;ci,mGenerated i-th of IMF is decomposed for the m times EMD.2,LSSVM
Least square method supporting vector machine (LSSVM) is a kind of pair of support vector machines (SVM) proposed by Suykens et al.
Innovatory algorithm, the algorithm convert nonlinear problem to the Linear Estimation problem of high-dimensional feature space, improve solving speed and
Generalization ability.
If training set is (xi,yi), i=1,2 ..., N, N are training sample sum, xi∈RmFor input sample, yiFor output
Sample, linear function of the LSSVM in high-dimensional feature space are as follows:
In formula, ω is weight coefficient vector, and b is biasing constant.The regression problem considers function complexity and error of fitting, can
It is denoted as the optimization problem of an equality constraint, it may be assumed that
If αiFor Lagrange multiplier, by establishing Lagrange's equation, wushu (2) is converted into unconstrained optimization problem,
As shown in formula (3):
Parameter alpha is acquired according to the Karush-Kuhn-Tucker condition of non-linear optimum programmingi, b, substitute into formula (1), obtain
To the output of support vector machines:
Radial basis kernel function RBF is selected to obtain the anticipation function of LSSVM as the kernel function of LSSVM are as follows:
3, improved shuffled frog leaping algorithm
(1)SFLA
Shuffled frog leaping algorithm SFLA was proposed by Eusuff and Lansey in 2003 at first.As a kind of novel bionical object
Intelligent optimization algorithm is learned, SFLA combines mould because of algorithm (MA, memetic algorithm) and particle swarm optimization algorithm
The advantages of (PSO, particle swarm optimization) two population intelligent optimization algorithm.The algorithm has concept simple,
The parameter of adjustment is few, and calculating speed is fast, and global search optimizing ability is strong, it is easy to accomplish the characteristics of.
The basic thought of shuffled frog leaping algorithm is as follows:
For a D-dimensional target search space, initial population is generated by way of generating at random, and entire group is by F
Frog is constituted.Then, then by population be divided into S sub-group, n frog contained in each sub-group, and meet F=S ×
N, and frog individual is ranked up according to fitness.The division mode of frog is executed by following rule: the 1st frog is divided
Enter the 1st subgroup, the 2nd frog is subdivided into the 2nd subgroup, and jth frog is subdivided into j-th of subgroup, then+1 frog of jth
It is subdivided into+1 subgroup of jth.And so on, terminate until all frogs divide.For each sub-group, if fitness it is optimal and
Worst frog is respectively XbAnd Xw, in addition, there is the frog of adaptive optimal control degree to be expressed as X in populationg.In each of sub-group
In secondary evolutionary process, to XwIt does local search operation and updates it, the strategy taken is as follows:
Di=rand () * (Xbi-Xwi) (8)
X′w=Xw+Di(-Dmax≤Di≤Dmax) (9)
In formula, rand () is distributed across the random number between [0,1];DmaxIt is the maximum step-length of frog.If being obtained after updating
Frog X 'wFitness than original frog XwIt is more excellent, just use X 'wReplace Xw;Otherwise, X is just usedgReplace Xb, then by formula (8) and
(9) local search procedure is executed;If the new explanation obtained at this time still cannot be than original frog XwIt is more excellent, it just takes and is randomly generated
The method of one new position replaces original Xw.Aforesaid operations are constantly repeated, until terminating local search.Later, it repartitions
Population, and local search is unfolded again.It constantly repeats the above steps, until reaching termination condition.
(2)ISFLA
In traditional SFLA algorithm, influence of the feature of frog individual to be updated to updated value be it is invariable,
This way can not only be such that convergence speed of the algorithm declines, and be also easy to that it is caused to fall into local optimum.Therefore, to make searching for algorithm
Rope efficiency is improved, and is introduced the Inertia Weight of decreases in non-linear herein, is improved to traditional SFLA algorithm, and algorithm is avoided to fall into
Enter locally optimal solution.
Inertia Weight is introduced, shown in more new strategy such as formula (3) at this time:
X′w=ω Xw+Di(-Dmax≤Di≤Dmax) (10)
Inertia Weight ω reflects the movement tendency of frog, and the value of traditional Inertia Weight is taken with population iteration time
The strategy of several increase and linear decrease is difficult to reflect the local search of population complexity although the strategy is simple and directly
Process, and be easy that algorithm is made to fall into locally optimal solution.Therefore, following nonlinear decreasing strategy is used herein:
In formula, ωmaxFor maximum inertia weight;ωminFor minimum inertia weight;T be current sub- population the number of iterations with
The product of current total mixed iteration number;K is the total the number of iterations of sub- population.Above-mentioned decrementing procedure makes ω in kind of a group hunting
Numerical value early period it is larger, become smaller to the later period, reduce the probability that algorithm falls into local optimum to a certain extent.
4, ISFLA-LSSVM (improved shuffled frog leaping algorithm Optimized Least Square Support Vector)
It is predicted using the LSSVM model of Radial basis kernel function (RBF), it is thus necessary to determine that regularization parameter is wide with kernel function
Two parameters are spent, they will directly affect the performance of model, and then influence the precision of prediction.Therefore, it selects herein improved mixed
The algorithm that leapfrogs is closed to optimize the parameter of least square method supporting vector machine.The step of optimization, is as follows:
Step 1: the parameter of setting ISFLA model initializes frog population;
Step 2: it calculates the fitness value of each frog individual and is ranked up;
Step 3: carrying out sub-group division, and determines the optimal solution in each sub-group, worst solution and global optimum of group
Solution;
Step 4: doing local search to the worst frog individual in each sub-group and be updated operation, until part
Search terminates;
Step 5: updated sub-group is mixed;
Step 6: judging whether to reach maximum number of iterations, if reached, stops optimization, exports optimal solution, otherwise go to
Step 2;
Step 7: the parameter value after optimization is assigned to LSSVM, building prediction model is predicted.
5. pre- flow gauge
The prediction accuracy of energy-consuming will receive the influence of factors, in order to realize to the accurate pre- of energy-consuming
It surveys, is considering population, GDP, the industrial structure (value of secondary industry accounting), energy consumption structure, Energy Intensity, carbon row herein
On the basis of putting the energy-consumings predicted impact factor such as intensity and total import and export value, the EEMD-ISFLA-LSSVM energy is proposed
Consumption predictions model.The specific prediction steps of the prediction model are as follows:
(1) data acquisition is screened with influence factor
Collecting sample data, including historic energy consumption, population, GDP, the industrial structure, energy consumption structure, energy source strength
The data such as degree, carbon intensity and total import and export value.Then to data carry out nondimensionalization processing, and calculate each influence because
The grey relational grade of element and energy-consuming, according to the sequence of grey relational grade come the input factor of screening model.
(2) the sequence noise reduction based on EEMD
It treats forecasting sequence and carries out integrated empirical mode decomposition, obtain multiple IMF components.
(3) the energy-consuming prediction based on ISFLA-LSSVM
On the basis of considering energy-consuming predicted impact factor, divide by integrating the IMF that empirical mode decomposition obtains
Amount is predicted respectively with ISFLA-LSSVM model, and the result of prediction is reconstructed, and it is pre- to obtain final energy-consuming
Survey result.
In the present invention, we have collected population, GDP, the industrial structure (the secondary industry increase of -2016 years Chinese nineteen nineties
Be worth accounting), energy consumption structure, Energy Intensity, the energy-consumings influence factor data such as carbon intensity and total import and export value.
(data source is in World Bank's database and China Statistical Yearbook).
In order to determine the input of prediction model, we screen energy-consuming influence factor using grey relational grade index,
Realize Feature Dimension Reduction.The grey relational grade calculated result of each influence factor and energy-consuming is as follows:
1 grey relational grade calculated result of table
According to table 1, we choose five factors of the grey relational grade greater than 0.7 as the defeated of energy-consuming prediction model
Enter, respectively population, GDP, the industrial structure, energy consumption structure and total import and export value.Model output is energy-consuming.
In order to realize the Accurate Prediction to energy-consuming, we carry out integrated Empirical Mode to prime energy consumption sequence first
State is decomposed, and to realize Noise reducing of data, by integrating empirical mode decomposition, is obtained three IMF components and a residue signal, is such as schemed
Shown in 1:
We are using the data of Chinese 1990-2009 as training sample set, using the data of 2010-2016 as test
Sample set predicts IMF component obtained above and residue signal with ISFLA-LSSVM model respectively.Model ginseng
Number set as follows: for frog population quantity as 500, subgroup quantity is 50, and subgroup inner search number is 10 times;Least square is supported
The search range of the regularization parameter of vector machine is [0.1,150], the search range of Radial basis kernel function parameter be [0.01,
30], global maximum number of iterations is 150.Prediction result and residual error are as shown in Figure 2:
The prediction result of IMF component and residue signal is reconstructed, obtains the final prediction result of energy-consuming, such as
Shown in Fig. 3:
The relative error of each future position is as shown in table 2:
The relative error of each future position of table 2
According to Fig. 3 and table 2, energy-consuming is predicted with EEMD-ISFLA-LSSVM model, prediction curve and reality
The fitting degree of border curve is very good, and prediction effect is significant, and the relative error of all future positions is no more than 5%.
This embodiment is merely preferred embodiments of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of protection of the claims
Subject to.
Claims (6)
1. a kind of energy-consuming prediction technique characterized by comprising
Step 1: acquisition include historic energy consumption figure, the size of population, GDP, the industrial structure, energy consumption structure, Energy Intensity,
Sample data including carbon intensity and total import and export value;
Step 2: nondimensionalization processing being carried out to sample data, and the grey for calculating each sample data and energy consumption structure is closed
Connection degree, according to the sequence of grey relational grade come the input factor of screening model;
Step 3: treating forecasting sequence and carry out the sequence noise reduction based on integrated empirical mode decomposition, obtain multiple IMF components;
Step 4: optimizing minimum two with improved shuffled frog leaping algorithm to by integrating the IMF component that empirical mode decomposition obtains
Multiply the parameter of support vector machines, and establish prediction model, the result of prediction is reconstructed, obtains final energy-consuming prediction
As a result.
2. method according to claim 1, which is characterized in that the integrated empirical mode decomposition is in Conventional wisdom mode point
On the basis of solution, white Gaussian noise is added in original signal, the statistical property being distributed using white noise frequency-flat is eliminated former
Intermittency in signal, to effectively inhibit modal overlap problem.
3. method according to claim 1 or claim 2, which is characterized in that specific step is as follows for the integrated empirical mode decomposition:
Step 301: random Gaussian white noise sequence is added in echo signal,
Xm(t)=X (t)+knm(t) (1)
In formula, k is the amplitude coefficient for the white noise being added, nmIt (t) is white noise sequence, X (t) is echo signal, XmIt (t) is to add
Enter the signal of white noise;
Step 302: the signal that white noise is added is decomposed into one group of IMF using EMD;
Step 303: different white noise sequences being added every time, repeats the above steps;
Step 304: calculate the mean value of IMF after decomposing, the mean value of each IMF that decomposition is obtained as finally as a result,
In formula, m is the integrated number of EMD, m=1,2 ..., N;ci,mGenerated i-th of IMF is decomposed for the m times EMD,To divide
The mean value for each IMF that solution obtains.
4. method according to claim 1 or claim 2, which is characterized in that the improved shuffled frog leaping algorithm is to mix in traditional
Conjunction leapfrogs in algorithm, introduces the Inertia Weight of decreases in non-linear, algorithm is avoided to fall into locally optimal solution.
5. method according to claim 1 or claim 2, which is characterized in that the improved shuffled frog leaping algorithm formula is as follows:
Di=rand () (Xbi-Xwi) (8)
X′w=ω Xw+Di(-Dmax≤Di≤Dmax) (10)
In formula, DiFor the step-length of frog, rand () is distributed across the random number between [0,1], Xbi、XwiRespectively fitness is optimal
With i-th worst of frog;X′wFitness for the frog obtained after update, XwFor the fitness of original frog, ω is used
Property weight, DmaxIt is the maximum step-length of frog, ωmaxFor maximum inertia weight;ωminFor minimum inertia weight;T is current son kind
The product of the number of iterations of group and current total mixed iteration number;K is the total the number of iterations of sub- population.
6. method according to claim 1 or claim 2, which is characterized in that described minimum with the optimization of improved shuffled frog leaping algorithm
Two parameters for multiplying support vector machines specifically include:
Step 401: the parameter of improved shuffled frog leaping algorithm being set, frog population is initialized;
Step 402: calculating the fitness value of each frog individual and be ranked up;
Step 403: carrying out sub-group division, and determine the optimal solution in each sub-group, worst solution and group's globally optimal solution;
Step 404: local search being done to the worst frog individual in each sub-group and is updated operation, until local search
Terminate;
Step 405: updated sub-group is mixed;
Step 406: judging whether to reach maximum number of iterations, if reached, stop optimization, export optimal solution, otherwise go to step
Rapid 402;
Step 407: the parameter value after optimization being assigned to least square method supporting vector machine, building prediction model is predicted.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810935717.3A CN109086941A (en) | 2018-08-16 | 2018-08-16 | A kind of energy-consuming prediction technique |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810935717.3A CN109086941A (en) | 2018-08-16 | 2018-08-16 | A kind of energy-consuming prediction technique |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109086941A true CN109086941A (en) | 2018-12-25 |
Family
ID=64793569
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810935717.3A Pending CN109086941A (en) | 2018-08-16 | 2018-08-16 | A kind of energy-consuming prediction technique |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109086941A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112990826A (en) * | 2021-04-15 | 2021-06-18 | 西南交通大学 | Short-time logistics demand prediction method, device, equipment and readable storage medium |
CN114692959A (en) * | 2022-03-16 | 2022-07-01 | 国网福建省电力有限公司经济技术研究院 | Urban energy data fusion system based on gray level correlation analysis method |
CN115829081A (en) * | 2022-10-19 | 2023-03-21 | 华中科技大学 | Urban traffic carbon emission prediction method based on support vector regression model |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106779157A (en) * | 2016-11-18 | 2017-05-31 | 中铁第勘察设计院集团有限公司 | The Forecasting Methodology of Regional Energy consumption demand |
-
2018
- 2018-08-16 CN CN201810935717.3A patent/CN109086941A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106779157A (en) * | 2016-11-18 | 2017-05-31 | 中铁第勘察设计院集团有限公司 | The Forecasting Methodology of Regional Energy consumption demand |
Non-Patent Citations (2)
Title |
---|
王宏力 等: "《惯性测量组合智能故障诊断及预测技术》", 31 May 2017 * |
索春梅: ""基于灰色关联理论与经验模态分解的光伏处理超短期预测", 《中国电力》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112990826A (en) * | 2021-04-15 | 2021-06-18 | 西南交通大学 | Short-time logistics demand prediction method, device, equipment and readable storage medium |
CN112990826B (en) * | 2021-04-15 | 2021-11-02 | 西南交通大学 | Short-time logistics demand prediction method, device, equipment and readable storage medium |
CN114692959A (en) * | 2022-03-16 | 2022-07-01 | 国网福建省电力有限公司经济技术研究院 | Urban energy data fusion system based on gray level correlation analysis method |
CN115829081A (en) * | 2022-10-19 | 2023-03-21 | 华中科技大学 | Urban traffic carbon emission prediction method based on support vector regression model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Improved tunicate swarm algorithm: Solving the dynamic economic emission dispatch problems | |
Wang et al. | An improved grey model optimized by multi-objective ant lion optimization algorithm for annual electricity consumption forecasting | |
Reagen et al. | A case for efficient accelerator design space exploration via bayesian optimization | |
Hassan et al. | Operational framework for recent advances in backtracking search optimisation algorithm: A systematic review and performance evaluation | |
Cheng et al. | A new combined model based on multi-objective salp swarm optimization for wind speed forecasting | |
Wang et al. | Stochastic economic emission load dispatch through a modified particle swarm optimization algorithm | |
Ewees et al. | Enhanced salp swarm algorithm based on firefly algorithm for unrelated parallel machine scheduling with setup times | |
CN109492814A (en) | A kind of Forecast of Urban Traffic Flow prediction technique, system and electronic equipment | |
CN109685277A (en) | Electricity demand forecasting method and device | |
He et al. | Optimising the job-shop scheduling problem using a multi-objective Jaya algorithm | |
CN108133258A (en) | A kind of mixing global optimization method | |
Orouskhani et al. | Evolutionary dynamic multi-objective optimization algorithm based on Borda count method | |
CN109086941A (en) | A kind of energy-consuming prediction technique | |
Xuemei et al. | Building cooling load forecasting using fuzzy support vector machine and fuzzy C-mean clustering | |
Zhao et al. | A decomposition-based many-objective ant colony optimization algorithm with adaptive reference points | |
CN106855957A (en) | Factory's bus load prediction based on similar day and least square method supporting vector machine | |
Iranfar et al. | Multiagent reinforcement learning for hyperparameter optimization of convolutional neural networks | |
Jamous et al. | Modifications of particle swarm optimization techniques and its application on stock market: A survey | |
CN106228235B (en) | A kind of land utilization space Optimal Configuration Method for taking pattern Yu PROCESS COUPLING effect into account | |
Liu et al. | A survey of artificial bee colony algorithm | |
Xi et al. | Whale optimization algorithm based on nonlinear adjustment and random walk strategy | |
CN108805254A (en) | Parameter Optimization System for time series forecasting | |
Zhang et al. | High dimensional missing data imputation for classification problems: A hybrid model based on k-nearest neighbor and genetic algorithm | |
Zivkovic et al. | Optimized recurrent neural networks with attention for wind farm energy generation forecasting | |
Yang et al. | Hybrid Taguchi-based particle swarm optimization for flowshop scheduling problem |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181225 |
|
RJ01 | Rejection of invention patent application after publication |