CN109146183A - Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm - Google Patents
Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm Download PDFInfo
- Publication number
- CN109146183A CN109146183A CN201810974163.8A CN201810974163A CN109146183A CN 109146183 A CN109146183 A CN 109146183A CN 201810974163 A CN201810974163 A CN 201810974163A CN 109146183 A CN109146183 A CN 109146183A
- Authority
- CN
- China
- Prior art keywords
- particle
- learning machine
- breadth
- elm
- extreme learning
- 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
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 68
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 54
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000005457 optimization Methods 0.000 title claims abstract description 24
- 230000006870 function Effects 0.000 claims abstract description 32
- 230000000694 effects Effects 0.000 claims abstract description 13
- 239000002245 particle Substances 0.000 claims description 49
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000000295 complement effect Effects 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 11
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 7
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 238000012804 iterative process Methods 0.000 claims description 3
- 230000003340 mental effect Effects 0.000 claims description 3
- 230000002028 premature Effects 0.000 claims description 3
- 230000005611 electricity Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000002123 temporal effect Effects 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 238000002922 simulated annealing Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
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]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Water Supply & Treatment (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Operations Research (AREA)
- Public Health (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Entrepreneurship & Innovation (AREA)
Abstract
The invention discloses a kind of short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm, include the following steps: S1, for the non-stationary of load data, using complementation set empirical mode decomposition (CEEMD) by the Time Series of original loads at several intrinsic mode functions (IMFs)；Positive and negative pairs of white noise is added into original time series for complementation set empirical mode decomposition (CEEMD), it not only can guarantee the discomposing effect possessed with set empirical mode decomposition (EEMD) equally in this way, but also can be reduced because of sequence reconstructed error caused by adding white noise；The present invention decomposes a sequence into several modal components using decomposition technique, and combine the parameter of optimization algorithm optimal prediction model, the prediction result of each component is finally superimposed as final predicted value, compared with other models, which can obtain higher precision of prediction in short-term impact load prediction.
Description
Technical field
The present invention relates to Load Prediction In Power Systems technical fields, and in particular to one kind is based on signal decomposition and intelligent optimization
The short-term impact load forecasting model method for building up of algorithm.
Background technique
Load Prediction In Power Systems are the key that unit generation foundation and electricity market adjustment electricity in real time are coordinated in power plant
The accuracy of the main source of valence, prediction will directly affect the cost of electricity-generating of power plant, the electricity consumption of dispatching of power netwoks and Area Inhabitants
Quality.With the growth of urban power consumption, increasing for electricity consumption user causes the complexity of regional load type, single load
Prediction technique (such as fuzzy logic method, time series method, support vector machines, artificial neural network) is easily trapped into part most
Excellent, convergence rate is slower, has been difficult to meet current load prediction precision and generalization demand.Therefore, combination forecasting obtains
To extensive concern and application.Currently, Combination thought mainly has following four kinds: first is that being carried out in advance with several models to original series
It surveys, obtains the higher prediction result of more single model accuracy using the comprehensive each model result of weighting scheme, still, the method needs
Multiple models are predicted simultaneously, and the mode of weighted calculation is complex；Second is that by establishing error correction to prediction model output
Model, still, the method are easily trapped into local optimum in the biggish area of load fluctuation, and generalization is insufficient；Third is that using optimization
Algorithm optimizes fundamental forecasting model parameter；Fourth is that original signal is decomposed by multiple components using signal decomposition technology,
Each component is individually modeled；Latter two built-up pattern embodies better prediction effect than single model, but still can not
Meet the load prediction requirement containing a large amount of impact load areas.
Traditional neural network, such as BP, Elamn and SOM, parameter adjustment is complicated in complication system, and convergence rate is slow,
And extreme learning machine (ELM) this new neural network is adjusted with faster pace of learning because of its less parameter in short term
It has received widespread attention and uses in prediction, but equally exist local optimum problem, its ginseng need to be optimized using intelligent algorithm
Number；For general optimization algorithm, such as genetic algorithm (GA), particle swarm algorithm (PSO), simulated annealing (SA) are optimizing
Later period it is possible that local optimum problem, pertinent literature points out, intersects optimization algorithm (CSO) in length and breadth and possesses the powerful overall situation
Search capability and faster convergence rate, effectively avoiding model parameter, the phase falls into local optimum after optimization, is suitable for extensive
Nonlinear system.
Impact load will lead to the biggish fluctuation of somewhere daily load curve, directly utilize this area's original loads data
It is modeled and is predicted, it is different surely to obtain ideal precision of prediction.Pertinent literature is using empirical mode decomposition (EMD) or gathers
Empirical mode decomposition (EEMD) decomposes wind speed time series, after wind speed time series passes through resolution process, resulting bottle sequence
Column fluctuation reduces, and precision of prediction greatly improves, but empirical mode decomposition (EMD) and set empirical mode decomposition (EEMD) point
Not there is the problems such as mode mixing and sequence reconstructed error.Pertinent literature utilizes improved complementary set empirical mode decomposition
(CEEMD) method handles air speed data, efficiently solves empirical mode decomposition (EMD) and set empirical mode decomposition (EEMD) institute
There are the problem of, further improve the precision of forecasting wind speed.Currently, empirical mode decomposition (CEEMD) method is gathered in complementation
Apply the research in load prediction few, pertinent literature is proposed based on the short-term of complementary set empirical mode decomposition (CEEMD)
Load forecasting model, the model obtain preferable effect in emulation experiment, still, and intelligent algorithm are not used to the ginseng of model
Number optimizes, and when being predicted containing a large amount of impact load areas, the generalization ability of the model needs to be proved.
Summary of the invention
The purpose of the present invention is to overcome the shortcomings of the existing technology and deficiency, provides a kind of excellent based on signal decomposition and intelligence
Change the short-term impact load forecasting model method for building up of algorithm, which decomposes a sequence into several mode using decomposition technique
Component, and combine the parameter of optimization algorithm optimal prediction model, is finally superimposed the prediction result of each component as final predicted value,
Compared with other models, which can obtain higher precision of prediction in short-term impact load prediction.
The purpose of the invention is achieved by the following technical solution:
A kind of short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm, including it is following
Step:
S1 gathers empirical mode decomposition (CEEMD) for original loads using complementation for the non-stationary of load data
Time Series are at several intrinsic mode functions (IMFs)；
Positive and negative pairs of white noise is added into original time series for complementation set empirical mode decomposition (CEEMD), in this way
Not only can guarantee possess with set empirical mode decomposition (EEMD) the same discomposing effect, but also caused by can be reduced because of addition white noise
Sequence reconstructed error；
Wherein, the processing step of complementary set empirical mode decomposition (CEEMD) is as follows:
Positive and negative pairs of white noise is added into original time series by S1.1, and generates two kinds by additional noise and time
The sequence M1 and M2 that sequence mixes, two kinds of sequence M1 and M2 are obtained by following formula (3):
Wherein NE is the white noise of addition, and X is time series, then M1 is the summation of time series and positive noise, when M2 is
Between sequence and negative noise summation；
S1.2 contains positive and negative white noise for what M1 and M2 were separately disassembled into respective complementation by empirical mode decomposition (EMD)
Several intrinsic mode functions (IMFs) pairs of component；
S1.3 combines each pair of component containing positive and negative white noise as final intrinsic mode function (IMF) point
Amount；
S2, using the parameter of the optimization extreme learning machine of crossover algorithm in length and breadth, and to all intrinsic mode functions (IMF) point
Amount establishes the prediction model of crossover algorithm optimization extreme learning machine (CSO-ELM) in length and breadth respectively；
S2.1, extreme learning machine；
Equipped with N number of by inputting xiWith output yiMutually different sample (the x of compositioni,yi), xi=[xi1,xi2,…xin]T∈
Rn,yi=[yi1,yi2,…yim]T∈Rm, i ∈ [1, N], then the output of a feedforward neural network with L hidden node can
To be indicated by following formula (4):
Wherein αi=[αi1,αi2,…αin]TIt is the input weight for connecting input layer to i-th of hidden layer node, biIt is i-th
The deviation of a hidden layer node, inputs weight and deviation generates at random；βi=[βi1,βi2,…βim]TIt is i-th of hidden layer section
Point arrives the output weight of output layer；αi*xiIndicate vector αiAnd xiInner product；G (x) is excitation function；If the feedforward neural network
N number of sample can be approached with zero error, then there is one group of data αi、biAnd βi, meet following formula (5):
Wherein above-mentioned formula (5) can be reduced to following formula (6):
H β=Y, (6)
Wherein H is the hidden layer input matrix of network, in conjunction with output sample Y, just can be determined by following formula (7) implicit
Layer output matrix β:
β=H-1Y； (7)
S2.2, in length and breadth crossover algorithm；Crossover algorithm (CSO) is by lateral cross and crossed longitudinally two seed nucleus mental arithmetic subgroup in length and breadth
At；In each iterative process, alternately, the filial generation generated after intersection and its parent compete two kinds of operators, preferentially retain；
S2.2.1, lateral cross operation；Lateral cross be in population two mutually different particles between identical dimension
A kind of calculation mechanism；If the d dimension of parent particle X (i) and X (j) carries out lateral cross, they are according to following formula (8) and (9)
Generate filial generation:
MShc(i, d)=r1×X(i,d)+(1-r1)×X(j,d)+c1× (X (i, d)-X (j, d)), (8)
Wherein r1、r2Random number between ∈ [0,1]；c1、c2Random number between ∈ [- 1,1]；M is particle scale；D is
Dimension；X (i, d), X (j, d) indicate that the d of parent particle X (i) and X (j) are tieed up；MShc(i,d)、MShc(j, d) difference table
Show that X (i, d) and X (j, d) tie up filial generation by the d that lateral cross generates；
S2.2.2, crossed longitudinally operation；Crossed longitudinally is a kind of calculation mechanism carried out between same particle different dimensional, often
Secondary crossed longitudinally operation only generates a filial generation；It is assumed that the d of particle X (i)1Peacekeeping d2Dimension progress is crossed longitudinally, according to following
Formula (10) generates filial generation:
Wherein MSvc(i,d1) indicate parent particle X (i) d1Peacekeeping d2The filial generation that dimension passes through crossed longitudinally generation；
In fact, the probability that precocity occurs in dimensionality of particle level is less, therefore crossed longitudinally Probability pvIt is less than lateral friendship
Pitch Probability ph, and iteration is only updated one of particle every time, and effect is equivalent to the convergence direction for making the particle
The change of small probability occurs, to jump out local optimum；
S2.3, crossover algorithm (CSO) optimizes extreme learning machine (ELM) parameter step in length and breadth；
For extreme learning machine (ELM) due to being easy Premature Convergence, generally requiring a large amount of hidden layer node can be only achieved ideal
Precision of prediction；Crossover algorithm (CSO) ability of searching optimum is strong in length and breadth, and fast convergence rate can effectively solve problem above；
Wherein, specific step is as follows for crossover algorithm (CSO) optimization extreme learning machine (ELM) parameter in length and breadth:
S2.3.1, initialization population；Individual particles in population input weight by extreme learning machine (ELM) and hidden layer is inclined
Composition is set, length, that is, dimension D=b (a+1) of particle, wherein a, b are respectively input layer and node in hidden layer；
Wherein θmFor m (1≤m≤popsize) a particle in population；For [- Xmax,Xmax] in
Machine number, XmaxGenerally take 1；
S2.3.2 optimizes extreme learning machine (CSO- using decomposed historical load data as crossover algorithm in length and breadth
ELM) the training sample of prediction model calculates fitness Fitness, the Fitness expression formula of each particle in initial population
(12) as follows:
Wherein yjFor actual negative charge values, TjFor predicted value；
S2.3.3 finds out the smallest optimal solution of fitness in population i.e. optimal particle using crossover algorithm in length and breadth (CSO), kind
Parameter in group's optimal particle is the optimal input weight and deviation of extreme learning machine (ELM) prediction model；Combined training sample
This input value, by Sigmoid activation primitive, i.e.,Find out hidden layer input matrix H；
S2.3.4, the output matrix Y and hidden layer input matrix H of combined training sample, finds out according to above-mentioned formula (10)
Hidden layer output matrix β；
The input value of forecast sample is substituted into all Decomposition Sequences the extreme learning machine for having optimized weight and biasing by S3
(ELM) prediction model, is superimposed the output valve of each series model, to obtain final load prediction results.
The present invention have compared with prior art it is below the utility model has the advantages that
(1) the present invention is based on complementations to gather empirical mode decomposition (Complementary Ensemble Empirical
Mode Decomposition, CEEMD) and crossover algorithm (Crisscross Optimization, CSO) optimizes the limit in length and breadth
The method for building up of the short-term impact load forecasting model of learning machine (Extreme Learning Machine, ELM), uses first
Load Time Series are resolved into multiple mode being distributed from high frequency to low frequency by complementation set empirical mode decomposition (CEEMD) technology
Then component is established extreme learning machine model to whole components respectively and is predicted, and optimized using crossover algorithm (CSO) in length and breadth
The input weight and hidden layer of extreme learning machine (ELM) bias, and are finally superimposed the prediction result of each component.Simulation result table
Bright, complementation set empirical mode decomposition (CEEMD) solves the problems, such as set empirical mode decomposition (EEMD), indulges
Horizontal crossover algorithm (CSO) makes extreme learning machine (ELM) neural network obtain more preferably parameter, compared with other models, the group
Molding type can obtain higher precision of prediction in short-term impact load prediction；
(2) positive and negative pairs of white noise is added complementary set empirical mode decomposition (CEEMD) of the invention into original series
Sound not only can guarantee the discomposing effect possessed with set empirical mode decomposition (EEMD) equally in this way, but also can be reduced because adding white noise
Sequence reconstructed error caused by sound further reduced the non-stationary of sample data；Crossover algorithm (CSO) has powerful in length and breadth
Ability of searching optimum, the weight for making extreme learning machine (ELM), straggling parameter more preferably, improve the model precision of prediction and
Generalization；
(3) simulation result of the invention shows that CEEMD-CSO-ELM stability is high, generalization ability is strong, with other models pair
Than obtaining highest precision of prediction, prediction curve more approaches actual value, basic load is relatively small, impact load institute accounting
It is worthy of popularization in the short-term load forecasting of example larger area.
Detailed description of the invention
Fig. 1 is CSO-ELM flow chart of the invention；
Fig. 2 is that the present invention is based on the load forecasting model structure charts of CEEMD-CSO-ELM；
Fig. 3 is that prediction error of the invention changes schematic diagram with IMF number；
Fig. 4 is Load Time Series CEEMD decomposition result schematic diagram of the invention.
Specific embodiment
Present invention will now be described in further detail with reference to the embodiments and the accompanying drawings, but embodiments of the present invention are unlimited
In this.
As shown in figures 1-4, a kind of to be established based on signal decomposition and the short-term impact load forecasting model of intelligent optimization algorithm
Method includes the following steps:
S1 gathers empirical mode decomposition (CEEMD) for original loads using complementation for the non-stationary of load data
Time Series are at several intrinsic mode functions (IMFs)；
Set empirical mode decomposition (EEMD) method solves mode present in empirical mode decomposition (EMD) in itself
Aliasing Problem, but the white noise in initial data addition cannot be completely eliminated；Complementation gathers empirical mode decomposition (CEEMD) to original
Positive and negative pairs of white noise is added in beginning time series, both can guarantee possess with set empirical mode decomposition (EEMD) equally in this way
Discomposing effect, and can be reduced because addition white noise caused by sequence reconstructed error；
Wherein, the decomposition principle of complementary set empirical mode decomposition (CEEMD) is as follows:
Empirical mode decomposition (EMD) hypothesis collectively constitutes time series by different oscillation modes, these complex characteristics are hidden
It ensconces in time series, several intrinsic mode functions (IMFs) of this adjoint oscillating function will be by empirical mode decomposition (EMD)
It is extracted from initial data, therefore, the time series x (t) of initial data can resolve into several natural mode of vibration components and remaining
Amount, such as shown in following formula (1):
M is the sum of intrinsic mode function (IMF), c in above formulai(t) component of i-th of intrinsic mode function (IMF), rm
It (t) is surplus；Several intrinsic mode functions (IMFs) should meet following two requirement: 1) number (including the maximum value of extreme value
And minimum value) and zero cross point quantity answer it is equal, or at most difference one；2) the envelope average value of maximum value and minimum value
It should be zero；These steps are repeated, to the last a data series r (t) cannot be decomposed again, and screening process terminates.To overcome
Defect is mixed by the mode that empirical mode decomposition (EMD) is generated, gathers empirical mode decomposition (EEMD) into original time series
White noise signal is added, the white noise of addition has to comply with statistical law below, such as shown in following formula (2):
N is the set number for adding white noise in above formula, and ε is the standard deviation of additional noise, εmIt is ultimate criterion deviation, usually
It is defined as former time series and decomposes the difference of obtained several intrinsic mode functions (IMFs) component；Complementation set empirical modal
It decomposes (CEEMD) and then overcomes set empirical mode decomposition (EEMD) by the way that positive and negative pairs of white noise is added into time series
The shortcomings that, obtain complementary several intrinsic mode functions (IMFs)；
Wherein, the processing step of complementary set empirical mode decomposition (CEEMD) is as follows:
Positive and negative pairs of white noise is added into original time series by S1.1, and generates two kinds by additional noise and time
The sequence M1 and M2 that sequence mixes, two kinds of sequence M1 and M2 are obtained by following formula (3):
Wherein NE is the white noise of addition, and X is time series, then M1 is the summation of time series and positive noise, when M2 is
Between sequence and negative noise summation；
S1.2 contains positive and negative white noise for what M1 and M2 were separately disassembled into respective complementation by empirical mode decomposition (EMD)
Several intrinsic mode functions (IMFs) pairs of component；
S1.3 combines each pair of component containing positive and negative white noise as final intrinsic mode function (IMF) point
Amount；
S2, it is insufficient for the generalization ability of extreme learning machine prediction model, it is easily trapped into local optimum, using intersecting in length and breadth
The parameter of algorithm optimization extreme learning machine, and it is excellent to establish crossover algorithm in length and breadth respectively to all intrinsic mode functions (IMF) component
Change the prediction model of extreme learning machine (CSO-ELM)；
S2.1, extreme learning machine；
Equipped with N number of by inputting xiWith output yiMutually different sample (the x of compositioni,yi), xi=[xi1,xi2,…xin]T∈
Rn,yi=[yi1,yi2,…yim]T∈Rm, i ∈ [1, N], then the output of a feedforward neural network with L hidden node can
To be indicated by following formula (4):
Wherein αi=[αi1,αi2,…αin]TIt is the input weight for connecting input layer to i-th of hidden layer node, biIt is i-th
The deviation (bias) of a hidden layer node, inputs weight and deviation generates at random；βi=[βi1,βi2,…βim]TIt is hidden i-th
Output weight of the hiding node layer to output layer；αi*xiIndicate vector αiAnd xiInner product；G (x) is excitation function, if the feedforward is refreshing
N number of sample can be approached with zero error through network, then there is one group of data αi、biAnd βi, meet following formula (5):
Wherein above-mentioned formula (5) can be reduced to following formula (6):
H β=Y, (6)
Wherein H is the hidden layer input matrix of network, in conjunction with output sample Y, just can be determined by following formula (7) implicit
Layer output matrix β:
β=H-1Y； (7)
S2.2 crossover algorithm in length and breadth；Crossover algorithm (CSO) is by lateral cross and crossed longitudinally two seed nucleus mental arithmetic subgroup in length and breadth
At；In each iterative process, alternately, the filial generation generated after intersection and its parent compete two kinds of operators, preferentially retain；
The operation of S2.2.1 lateral cross；Lateral cross be in population two mutually different particles between identical dimension one
Kind calculation mechanism；If the d dimension of parent particle X (i) and X (j) carries out lateral cross, they are produced according to following formula (8) and (9)
Raw filial generation:
MShc(i, d)=r1×X(i,d)+(1-r1)×X(j,d)+c1× (X (i, d)-X (j, d)), (8)
Wherein r1、r2Random number between ∈ [0,1]；c1、c2Random number between ∈ [- 1,1]；M is particle scale；D is
Dimension；X (i, d), X (j, d) indicate that the d of parent particle X (i) and X (j) are tieed up；MShc(i,d)、MShc(j, d) difference table
Show that X (i, d) and X (j, d) tie up filial generation by the d that lateral cross generates；
The crossed longitudinally operation of S2.2.2；Crossed longitudinally is a kind of calculation mechanism carried out between same particle different dimensional, often
Secondary crossed longitudinally operation only generates a filial generation；It is assumed that the d of particle X (i)1Peacekeeping d2Dimension progress is crossed longitudinally, according to following
Formula (10) generates filial generation:
Wherein MSvc(i,d1) indicate parent particle X (i) d1Peacekeeping d2The filial generation that dimension passes through crossed longitudinally generation；
In fact, the probability that precocity occurs in dimensionality of particle level is less, therefore crossed longitudinally Probability pvIt is less than lateral friendship
Pitch Probability ph, and iteration is only updated one of particle every time, and effect is equivalent to the convergence direction for making the particle
The change of small probability occurs, to jump out local optimum；
S2.3, crossover algorithm (CSO) optimizes extreme learning machine (ELM) parameter step in length and breadth；
For extreme learning machine (ELM) due to being easy Premature Convergence, generally requiring a large amount of hidden layer node can be only achieved ideal
Precision of prediction；Crossover algorithm (CSO) ability of searching optimum is strong in length and breadth, and fast convergence rate can effectively solve problem above；Such as figure
Shown in 1, specific step is as follows for crossover algorithm (CSO) optimization extreme learning machine (ELM) parameter in length and breadth:
S2.3.1, initialization population；Individual particles in population input weight by extreme learning machine (ELM) and hidden layer is inclined
Composition is set, length, that is, dimension D=b (a+1) of particle, wherein a, b are respectively input layer and node in hidden layer；
Wherein θmFor m (1≤m≤popsize) a particle in population；For [- Xmax,Xmax] in
Machine number, XmaxGenerally take 1；
S2.3.2 optimizes extreme learning machine (CSO- using decomposed historical load data as crossover algorithm in length and breadth
ELM) the training sample of prediction model calculates fitness Fitness, the Fitness expression formula of each particle in initial population
(12) as follows:
Wherein yjFor actual negative charge values, TjFor predicted value；
S2.3.3 finds out the smallest optimal solution of fitness in population i.e. optimal particle using crossover algorithm in length and breadth (CSO), kind
Parameter in group's optimal particle is the optimal input weight and deviation of extreme learning machine (ELM) prediction model；Combined training sample
This input value, by Sigmoid activation primitive, i.e.,Find out hidden layer input matrix H；
S2.3.4, the output matrix Y and hidden layer input matrix H of combined training sample, finds out according to above-mentioned formula (10)
Hidden layer output matrix β；
The input value of forecast sample is substituted into all Decomposition Sequences the crossover algorithm in length and breadth for having optimized weight and biasing by S3
Optimize extreme learning machine (CSO-ELM) model, is superimposed the output valve of each sequence prediction sample, obtains final load prediction knot
Fruit.
Specifically, based on the prediction model of CEEMD-CSO-ELM, include the following steps:
1, data processing；
Using the historical load in Guangdong somewhere on December 31st, 1 day 1 June in 2010, temporal resolution is
5min, i.e., have 1 data point for every 5 minutes, and load prediction temporal resolution is 1 hour.It will be hourly negative on the day before prediction day
Day horal load is predicted in trained extreme learning machine (ELM) model, output for charge values input；In view of load variations
With the relationship of weather, input quantity also includes the highest temperature for predicting the same day and the previous day, minimum other than the previous day load data
Temperature, rainfall, day type (working day, weekend, festivals or holidays), and the sample of CEEMD-CSO-ELM model is trained then to choose prediction
5 months a few days ago data.
2, modeling process；
As shown in Fig. 2, wherein IMF1~IMFn is with oscillating function component, r (n) is surplus；
Specific step is as follows:
1) sequence samples are obtained according to specific prediction day.
2) using complementary set empirical mode decomposition (CEEMD) by the Time Series in sample at it is multiple from high frequency to
The intrinsic mode function (IMF) and surplus r (n) of low frequency distribution, decompose impact load high fdrequency component；If intrinsic mode function
(IMF) insufficient, then prediction effect is not achieved, if intrinsic mode function (IMF) is excessively, on the one hand predicted time can be made too long,
On the other hand then can failure law sequence, make data distortion；By taking summer predicts day as an example, precision of prediction is with intrinsic mode function
(IMF) changing rule of component increase and decrease is more obvious.
As shown in figure 3, prediction error is smaller and smaller, and natural mode of vibration with the increase of intrinsic mode function (IMF) number
When function (IMF) number is 8, there is an inflection point, continue growing component, error increases instead；Spring, the autumn, the winter prediction day
There is identical changing rule, therefore selecting intrinsic mode function (IMF) number is 8；If gathering number N is 500, in each set member
The standard deviation of middle addition white noise is set as 0.2；Summer continuous 30 days load datas are taken, temporal resolution is 1 hour, complementary
It is as shown in Figure 4 to gather empirical mode decomposition (CEEMD) decomposition result.
3) crossover algorithm in length and breadth is established to whole subsequence IMF1, IMF2 ... IMF8 and r8 respectively and optimizes extreme learning machine
(CSO-ELM) model obtains the optimal solution of model parameter using training sample training pattern.
4) prediction day horal load is predicted using above-mentioned optimal models.
5) superposition all sequences obtain final predicted value, and carry out error analysis with actual value；In order to fully and objectively
The accuracy of evaluation model, prediction error assessment function use mean absolute percentage error (Mean Absolute
Percentage Error, MAPE), as shown in following formula:
In above formula, ytFor actual value, TjFor predicted value.
The present invention is based on complementations to gather empirical mode decomposition (Complementary Ensemble Empirical Mode
Decomposition, CEEMD) and crossover algorithm (Crisscross Optimization, CSO) optimizes extreme learning machine in length and breadth
The method for building up of the short-term impact load forecasting model of (Extreme Learning Machine, ELM) uses complementary sets first
It closes empirical mode decomposition (CEEMD) technology and Load Time Series is resolved into multiple modal components being distributed from high frequency to low frequency,
Then extreme learning machine model is established respectively to whole components to predict, and the limit is optimized using crossover algorithm (CSO) in length and breadth
The input weight and hidden layer of learning machine (ELM) bias, and are finally superimposed the prediction result of each component；Simulation result shows mutually
Supplementary set closes empirical mode decomposition (CEEMD) and solves the problems, such as set empirical mode decomposition (EEMD), hands in length and breadth
Fork algorithm (CSO) makes extreme learning machine (ELM) obtain more preferably parameter, and compared with other models, the built-up pattern is short-term
Higher precision of prediction can be obtained in impact load prediction；Complementation set empirical mode decomposition (CEEMD) adds into original series
Enter positive and negative pairs of white noise, not only can guarantee the discomposing effect possessed with set empirical mode decomposition (EEMD) equally in this way, but also
Sequence reconstructed error caused by can be reduced because of addition white noise, further reduced the non-stationary of sample data；Intersect in length and breadth
Algorithm (CSO) has powerful ability of searching optimum, and the weight for making extreme learning machine (ELM), straggling parameter more preferably, improve this
The precision of prediction and generalization of model；Simulation result shows that CEEMD-CSO-ELM stability is high, generalization ability is strong, with other moulds
Type comparison obtains highest precision of prediction, and prediction curve more approaches actual value, basic load is relatively small, impact load institute
It is worthy of popularization in the short-term load forecasting of accounting example larger area.
Above-mentioned is the preferable embodiment of the present invention, but embodiments of the present invention are not limited by the foregoing content,
His any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention, should be
The substitute mode of effect, is included within the scope of the present invention.
Claims (1)
1. a kind of short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm, feature exist
In including the following steps:
S1 gathers empirical mode decomposition (CEEMD) for the time of original loads using complementation for the non-stationary of load data
Sequence resolves into several intrinsic mode functions (IMFs)；
Positive and negative pairs of white noise is added into original time series for complementation set empirical mode decomposition (CEEMD), in this way can
Guarantee to possess the discomposing effect with set empirical mode decomposition (EEMD) equally, and can be reduced because of sequence caused by adding white noise
Reconstructed error；
Wherein, the processing step of complementary set empirical mode decomposition (CEEMD) is as follows:
Positive and negative pairs of white noise is added into original time series by S1.1, and generates two kinds by additional noise and time series
The sequence M1 and M2 mixed, two kinds of sequence M1 and M2 are obtained by following formula (3):
Wherein NE is the white noise of addition, and X is time series, then M1 is the summation of time series and positive noise, and M2 is time sequence
The summation of column and negative noise；
S1.2, by empirical mode decomposition (EMD) if by M1 and M2 be separately disassembled into respective complementation containing positive and negative white noise
The pairs of component of dry intrinsic mode function (IMFs)；
S1.3 combines each pair of component containing positive and negative white noise as final intrinsic mode function (IMF) component；
S2, using the parameter of the optimization extreme learning machine of crossover algorithm in length and breadth, and to all intrinsic mode function (IMF) components point
The prediction model of crossover algorithm optimization extreme learning machine (CSO-ELM) in length and breadth is not established；
S2.1, extreme learning machine；
Equipped with N number of by inputting xiWith output yiMutually different sample (the x of compositioni,yi), xi=[xi1,xi2,…xin]T∈Rn,yi
=[yi1,yi2,…yim]T∈Rm, i ∈ [1, N], then the output of a feedforward neural network with L hidden node can be by
Following formula (4) indicate:
xi∈Rn,αi∈Rn,βi∈Rm, (4)
Wherein αi=[αi1,αi2,…αin]TIt is the input weight for connecting input layer to i-th of hidden layer node, biIt is hidden i-th
Deviation containing node layer, inputs weight and deviation generates at random；βi=[βi1,βi2,…βim]TIt is that i-th of hiding node layer arrives
The output weight of output layer；αi*xiIndicate vector αiAnd xiInner product；G (x) is excitation function；If the feedforward neural network can be with
Zero error approaches N number of sample, then there is one group of data αi、biAnd βi, meet following formula (5):
Wherein above-mentioned formula (5) can be reduced to following formula (6):
H β=Y, (6)
Wherein H is the hidden layer input matrix of network, in conjunction with output sample Y, just can determine that hidden layer is defeated by following formula (7)
Matrix β out:
β=H-1Y； (7)
S2.2, in length and breadth crossover algorithm；Crossover algorithm (CSO) is made of lateral cross and crossed longitudinally two seed nucleus mental arithmetic in length and breadth；
In each iterative process, alternately, the filial generation generated after intersection and its parent compete two kinds of operators, preferentially retain；
S2.2.1, lateral cross operation；Lateral cross is the one kind of two mutually different particles between identical dimension in population
Calculation mechanism；If the d dimension of parent particle X (i) and X (j) carries out lateral cross, they are generated according to following formula (8) and (9)
Filial generation:
MShc(i, d)=r1×X(i,d)+(1-r1)×X(j,d)+c1× (X (i, d)-X (j, d)), (8)
MShc(j, d)=r2×X(j,d)+(1-r2)×X(i,d)+c2× (X (j, d)-X (i, d)), (9)
I, j ∈ N (1, M), d ∈ N (1, D),
Wherein r1、r2Random number between ∈ [0,1]；c1、c2Random number between ∈ [- 1,1]；M is particle scale；D is variable
Dimension；X (i, d), X (j, d) indicate that the d of parent particle X (i) and X (j) are tieed up；MShc(i,d)、MShc(j, d) respectively indicates X
(i, d) and X (j, d) tie up filial generation by the d that lateral cross generates；
S2.2.2, crossed longitudinally operation；Crossed longitudinally is a kind of calculation mechanism carried out between same particle different dimensional, vertical every time
A filial generation is only generated to crossover operation；It is assumed that the d of particle X (i)1Peacekeeping d2Dimension progress is crossed longitudinally, according to following formula
(10) filial generation is generated:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2), (10)
I ∈ N (1, M), d1,d2∈ N (1, D), r ∈ [0,1],
Wherein MSvc(i,d1) indicate parent particle X (i) d1Peacekeeping d2The filial generation that dimension passes through crossed longitudinally generation；
In fact, the probability that precocity occurs in dimensionality of particle level is less, therefore crossed longitudinally Probability pvIt is less than lateral cross probability
ph, and iteration is only updated one of particle every time, and effect, which is equivalent to, keeps the convergence direction generation of the particle small
The change of probability, to jump out local optimum；
S2.3, crossover algorithm (CSO) optimizes extreme learning machine (ELM) parameter step in length and breadth；
Due to being easy Premature Convergence, generally require a large amount of hidden layer node can be only achieved preferably in advance extreme learning machine (ELM)
Survey precision；Crossover algorithm (CSO) ability of searching optimum is strong in length and breadth, and fast convergence rate can effectively solve problem above；
Wherein, specific step is as follows for crossover algorithm (CSO) optimization extreme learning machine (ELM) parameter in length and breadth:
S2.3.1, initialization population；Individual particles in population input weight by extreme learning machine (ELM) and hidden layer biases structure
At length, that is, dimension D=b (a+1) of particle, wherein a, b are respectively input layer and node in hidden layer；
Wherein θmFor m (1≤m≤popsize) a particle in population；For [- Xmax,Xmax] in random number,
XmaxGenerally take 1；
S2.3.2, it is pre- using decomposed historical load data as the optimization of crossover algorithm in length and breadth extreme learning machine (CSO-ELM)
The training sample for surveying model, calculates fitness Fitness, the Fitness expression formula (12) of each particle in initial population such as
Under:
Wherein yjFor actual negative charge values, TjFor predicted value；
S2.3.3 finds out the smallest optimal solution of fitness in population i.e. optimal particle using crossover algorithm in length and breadth (CSO), and population is most
Parameter in excellent particle is the optimal input weight and deviation of extreme learning machine (ELM) prediction model；Combined training sample
Input value, by Sigmoid activation primitive, i.e.,Find out hidden layer input matrix H；
S2.3.4, the output matrix Y and hidden layer input matrix H of combined training sample, finds out implicit according to above-mentioned formula (10)
Layer output matrix β；
The input value of forecast sample is substituted into all Decomposition Sequences the extreme learning machine (ELM) for having optimized weight and biasing by S3
Prediction model, is superimposed the output valve of each series model, to obtain final load prediction results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810974163.8A CN109146183A (en) | 2018-08-24 | 2018-08-24 | Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810974163.8A CN109146183A (en) | 2018-08-24 | 2018-08-24 | Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109146183A true CN109146183A (en) | 2019-01-04 |
Family
ID=64827866
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810974163.8A Pending CN109146183A (en) | 2018-08-24 | 2018-08-24 | Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109146183A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110059844A (en) * | 2019-02-01 | 2019-07-26 | 东华大学 | Energy storage device control method based on set empirical mode decomposition and LSTM |
| CN110648017A (en) * | 2019-08-30 | 2020-01-03 | 广东工业大学 | Short-term impact load prediction method based on two-layer decomposition technology |
| CN111062538A (en) * | 2019-12-21 | 2020-04-24 | 东北电力大学 | CEEMD method for determining reasonable maximum value of cellular load in space load prediction |
| CN111126659A (en) * | 2019-11-18 | 2020-05-08 | 国网安徽省电力有限公司六安供电公司 | Power load prediction method and system |
| CN111539558A (en) * | 2020-04-03 | 2020-08-14 | 东南大学 | Power load prediction method adopting optimized extreme learning machine |
| CN111882106A (en) * | 2020-06-16 | 2020-11-03 | 江苏大学 | Short-term power load prediction method based on comprehensive factors and CEEMD-IGWO-GRNN |
| CN112115754A (en) * | 2019-07-30 | 2020-12-22 | 嘉兴学院 | Short-term traffic flow prediction model based on firework differential evolution hybrid algorithm-extreme learning machine |
| CN112417762A (en) * | 2020-11-25 | 2021-02-26 | 中国民航大学 | Sector flow short-term prediction method based on decomposition integration methodology |
| CN112613650A (en) * | 2020-12-14 | 2021-04-06 | 燕山大学 | Novel three-stage short-term power load prediction method |
| CN112766076A (en) * | 2020-12-31 | 2021-05-07 | 上海电机学院 | Power load ultra-short-term prediction method, system, equipment and storage medium |
| CN113239624A (en) * | 2021-05-21 | 2021-08-10 | 长沙理工大学 | Short-term load prediction method, equipment and medium based on neural network combination model |
| CN113378387A (en) * | 2021-06-10 | 2021-09-10 | 大连海事大学 | Ship motion forecasting method based on improved EMD-AR model |
| CN114638552A (en) * | 2022-04-22 | 2022-06-17 | 国网浙江省电力有限公司营销服务中心 | Impact load identification method based on reference day-minute-scale curve and sliding window |
| CN114692692A (en) * | 2022-04-02 | 2022-07-01 | 河海大学 | Snowfall identification method based on microwave attenuation signal fusion kernel extreme learning machine |
| CN117635245A (en) * | 2023-11-30 | 2024-03-01 | 广东电力交易中心有限责任公司 | Power price prediction method and system based on multilevel frequency domain decomposition and IBM H optimization DELM |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104408529A (en) * | 2014-11-21 | 2015-03-11 | 广东工业大学 | Short-term load predicting method of power grid |
| CN107292462A (en) * | 2017-08-25 | 2017-10-24 | 广东工业大学 | A kind of short-term load forecasting method, apparatus and system |
| CN107516150A (en) * | 2017-08-25 | 2017-12-26 | 广东工业大学 | A kind of Forecasting Methodology of short-term wind-electricity power, apparatus and system |
-
2018
- 2018-08-24 CN CN201810974163.8A patent/CN109146183A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104408529A (en) * | 2014-11-21 | 2015-03-11 | 广东工业大学 | Short-term load predicting method of power grid |
| CN107292462A (en) * | 2017-08-25 | 2017-10-24 | 广东工业大学 | A kind of short-term load forecasting method, apparatus and system |
| CN107516150A (en) * | 2017-08-25 | 2017-12-26 | 广东工业大学 | A kind of Forecasting Methodology of short-term wind-electricity power, apparatus and system |
Non-Patent Citations (2)
| Title |
|---|
| 殷豪等: "基于结合混沌纵横交叉的粒子群算法优化极限学习机的短期负荷预测", 《计算机应用研究》 * |
| 董朕等: "基于混合算优化的短期的风功率预测", 《智慧电力》 * |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110059844A (en) * | 2019-02-01 | 2019-07-26 | 东华大学 | Energy storage device control method based on set empirical mode decomposition and LSTM |
| CN110059844B (en) * | 2019-02-01 | 2021-10-08 | 东华大学 | Energy storage device control method based on ensemble empirical mode decomposition and LSTM |
| CN112115754A (en) * | 2019-07-30 | 2020-12-22 | 嘉兴学院 | Short-term traffic flow prediction model based on firework differential evolution hybrid algorithm-extreme learning machine |
| CN112115754B (en) * | 2019-07-30 | 2023-05-26 | 嘉兴学院 | Short-time traffic flow prediction model based on firework differential evolution hybrid algorithm-extreme learning machine |
| CN110648017A (en) * | 2019-08-30 | 2020-01-03 | 广东工业大学 | Short-term impact load prediction method based on two-layer decomposition technology |
| CN111126659A (en) * | 2019-11-18 | 2020-05-08 | 国网安徽省电力有限公司六安供电公司 | Power load prediction method and system |
| CN111062538B (en) * | 2019-12-21 | 2022-09-20 | 东北电力大学 | CEEMD method for determining reasonable maximum value of cellular load in space load prediction |
| CN111062538A (en) * | 2019-12-21 | 2020-04-24 | 东北电力大学 | CEEMD method for determining reasonable maximum value of cellular load in space load prediction |
| CN111539558B (en) * | 2020-04-03 | 2023-05-16 | 东南大学 | Power load prediction method adopting optimization extreme learning machine |
| CN111539558A (en) * | 2020-04-03 | 2020-08-14 | 东南大学 | Power load prediction method adopting optimized extreme learning machine |
| CN111882106A (en) * | 2020-06-16 | 2020-11-03 | 江苏大学 | Short-term power load prediction method based on comprehensive factors and CEEMD-IGWO-GRNN |
| CN111882106B (en) * | 2020-06-16 | 2024-06-11 | 江苏大学 | Short-term power load prediction method based on comprehensive factors and CEEMD-IGWO-GRNN |
| CN112417762B (en) * | 2020-11-25 | 2022-04-29 | 中国民航大学 | Sector flow short-term prediction method based on decomposition integration methodology |
| CN112417762A (en) * | 2020-11-25 | 2021-02-26 | 中国民航大学 | Sector flow short-term prediction method based on decomposition integration methodology |
| CN112613650A (en) * | 2020-12-14 | 2021-04-06 | 燕山大学 | Novel three-stage short-term power load prediction method |
| CN112766076B (en) * | 2020-12-31 | 2023-05-12 | 上海电机学院 | Ultra-short-term prediction method, system, equipment and storage medium for power load |
| CN112766076A (en) * | 2020-12-31 | 2021-05-07 | 上海电机学院 | Power load ultra-short-term prediction method, system, equipment and storage medium |
| CN113239624A (en) * | 2021-05-21 | 2021-08-10 | 长沙理工大学 | Short-term load prediction method, equipment and medium based on neural network combination model |
| CN113378387B (en) * | 2021-06-10 | 2022-11-01 | 大连海事大学 | Ship motion forecasting method based on improved EMD-AR model |
| CN113378387A (en) * | 2021-06-10 | 2021-09-10 | 大连海事大学 | Ship motion forecasting method based on improved EMD-AR model |
| CN114692692A (en) * | 2022-04-02 | 2022-07-01 | 河海大学 | Snowfall identification method based on microwave attenuation signal fusion kernel extreme learning machine |
| CN114638552B (en) * | 2022-04-22 | 2022-09-02 | 国网浙江省电力有限公司营销服务中心 | Impact load identification method based on reference day-minute-scale curve and sliding window |
| CN114638552A (en) * | 2022-04-22 | 2022-06-17 | 国网浙江省电力有限公司营销服务中心 | Impact load identification method based on reference day-minute-scale curve and sliding window |
| CN117635245A (en) * | 2023-11-30 | 2024-03-01 | 广东电力交易中心有限责任公司 | Power price prediction method and system based on multilevel frequency domain decomposition and IBM H optimization DELM |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109146183A (en) | Short-term impact load forecasting model method for building up based on signal decomposition and intelligent optimization algorithm | |
| Bento et al. | Ocean wave energy forecasting using optimised deep learning neural networks | |
| CN111210082B (en) | Precipitation prediction method based on optimized BP neural network algorithm | |
| Cinar et al. | Development of future energy scenarios with intelligent algorithms: case of hydro in Turkey | |
| CN110414719A (en) | A kind of vehicle flowrate prediction technique based on Multi-variable Grey Model time series | |
| CN110491146A (en) | A kind of traffic signal control scheme real-time recommendation method based on deep learning | |
| CN104715292A (en) | City short-term water consumption prediction method based on least square support vector machine model | |
| CN106650784A (en) | Feature clustering comparison-based power prediction method and device for photovoltaic power station | |
| CN112415924A (en) | Energy-saving optimization method and system for air conditioning system | |
| CN108074004A (en) | A kind of GIS-Geographic Information System short-term load forecasting method based on gridding method | |
| CN101824807A (en) | Method for short-term prediction of salt tide at water intake in tidal river region | |
| CN110648017A (en) | Short-term impact load prediction method based on two-layer decomposition technology | |
| CN105574586B (en) | General-purpose aircraft boat material needing forecasting method based on MPSO-BP networks | |
| CN104123595A (en) | Power distribution network load prediction method and system | |
| CN108520345A (en) | Evaluation for cultivated-land method and system based on GA-BP neural network models | |
| CN107466816A (en) | A kind of irrigation method based on dynamic multilayer extreme learning machine | |
| Luo et al. | Prediction of vegetable price based on neural network and genetic algorithm | |
| CN110472840A (en) | A kind of agricultural water conservancy dispatching method and system based on nerual network technique | |
| CN113554466A (en) | Short-term power consumption prediction model construction method, prediction method and device | |
| CN110163254A (en) | A kind of cucumber green house yield intelligent Forecasting device based on recurrent neural network | |
| CN110070228A (en) | BP neural network wind speed prediction method for neuron branch evolution | |
| CN112182951B (en) | Land utilization planning method considering multiple uncertainties | |
| CN110119086A (en) | A kind of tomato greenhouse environmental parameter intelligent monitoring device based on ANFIS neural network | |
| Liu et al. | Evaluation and analysis of irrigation water use efficiency based on an extreme learning machine model optimized by the spider monkey optimization algorithm | |
| CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190104 |