CN109242532A - The Short-term electricity price forecasting method of RBF neural is decomposed and optimized based on local mean value - Google Patents
The Short-term electricity price forecasting method of RBF neural is decomposed and optimized based on local mean value Download PDFInfo
- Publication number
- CN109242532A CN109242532A CN201810876676.5A CN201810876676A CN109242532A CN 109242532 A CN109242532 A CN 109242532A CN 201810876676 A CN201810876676 A CN 201810876676A CN 109242532 A CN109242532 A CN 109242532A
- Authority
- CN
- China
- Prior art keywords
- electricity price
- mean value
- local mean
- rbf neural
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000005611 electricity Effects 0.000 title claims abstract description 86
- 230000001537 neural effect Effects 0.000 title claims abstract description 37
- 238000013277 forecasting method Methods 0.000 title claims abstract description 17
- 238000013528 artificial neural network Methods 0.000 claims abstract description 24
- 238000005457 optimization Methods 0.000 claims abstract description 16
- 239000002245 particle Substances 0.000 claims description 30
- 239000011159 matrix material Substances 0.000 claims description 15
- 230000003044 adaptive effect Effects 0.000 claims description 12
- 238000012549 training Methods 0.000 claims description 12
- 238000000034 method Methods 0.000 claims description 9
- 239000008186 active pharmaceutical agent Substances 0.000 claims description 6
- 230000002123 temporal effect Effects 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000012804 iterative process Methods 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 230000006978 adaptation Effects 0.000 claims 1
- 238000011160 research Methods 0.000 abstract description 12
- 230000007547 defect Effects 0.000 abstract description 5
- 239000000203 mixture Substances 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000011295 pitch Substances 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000001568 sexual effect Effects 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0206—Price or cost determination based on market factors
-
- 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
-
- 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
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Finance (AREA)
- Marketing (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Primary Health Care (AREA)
- Tourism & Hospitality (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The technical field of Research on electricity price prediction of the present invention is obtained electricity price historical data first and pre-processed to data more particularly, to a kind of Short-term electricity price forecasting method decomposed based on local mean value and optimize RBF neural;Then it is decomposed using local mean value and history electricity price data is resolved into a series of PF components for possessing physical significance;Then 0.5h prediction in advance is carried out to all PF components using the prediction model of the optimization radial base neural net of crossover algorithm in length and breadth;It is finally superimposed the predicted value of all PF components, obtains actual prediction result.The present invention decomposes to reduce the non-stationary and nonlinear problem of electricity price sequence using local mean value, look-ahead is carried out using the optimization radial base neural net composition mixed model of crossover algorithm in length and breadth, compensate for the defect that neural network easily falls into local optimum, improve the generalization ability of neural network, influence of the electricity price sequence complex characteristics to prediction result is reduced, the precision of Electric Price Forecasting is improved.
Description
Technical field
The present invention relates to the technical fields of Research on electricity price prediction, are decomposed and are optimized based on local mean value more particularly, to one kind
The Short-term electricity price forecasting method of RBF neural.
Background technique
With the reform of electricity market, electricity price can trade in the market as general goods, and Research on electricity price prediction is
Fully considering relation between market supply and demand, participant in the market implements the influence factors such as electricity market size and social activities, passes through
It establishes correlation model to study history electricity price data, the changing rule of electricity price itself is analyzed, to future electrical energy market market
Marginal price is predicted.From the point of view of power generation side, accurate Research on electricity price prediction is conducive to it and holds marketplace trend, and it is first to grasp market
Machine, so that optimal generated energy and electricity price bidding strategy are constructed, to obtain maximum profit;From the point of view of electricity consumption side, electricity price is formd
Its power purchase expense, accurate Research on electricity price prediction can allow user to control its electricity consumption according to demand, formulate reasonable electricity consumption plan, and one
A little common users can automatically control the time by purchase, and the higher household electrical appliances of intelligence degree enjoy valley power consumption
Material benefit to reduce living cost, while also functioning to the effect of peak load shifting.Therefore, accurate Research on electricity price prediction is to electric system
It plays an important role with electricity market normal operation.
Currently, most models all concentrate on Electric Price Forecasting in document.These methods include time series models, people
Work model of mind and mixed model.Compared with conventional neural network model, using the neural network mould after intersection optimization in length and breadth
Type compensates for many deficiencies, and the parameter that it avoids neural network falls into the defect of local optimum, improves the general of neural network
Change ability, so can be used for Electric Price Forecasting, however, single prediction model is to be unable to Accurate Prediction electricity price, due to electricity price
Sequence has non-stationary and nonlinear complex characteristics, and original electricity price is resolved into a series of moulds by variation mode decomposition technology
State, then predicted using prediction model, the raising Research on electricity price prediction precision of the big degree of energy.But electricity price not only has with history electricity price
It closes, also offers with weather, Generation Side, the indefinite sexual factor such as human activity is related, the essence of these factors affects to Research on electricity price prediction
Degree.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide one kind to decompose based on local mean value and optimize RBF
The Short-term electricity price forecasting method of neural network can be applied to the scientific research and work of electricity market and electric system related fields
Cheng Yingyong, and the generalization ability and precision of prediction of prediction model can be improved.
In order to solve the above technical problems, the technical solution adopted by the present invention is that:
A kind of Short-term electricity price forecasting method decomposed based on local mean value and optimize RBF neural is provided, including following
Step:
S1. history electricity price data are obtained and history electricity price data are pre-processed;
S2. it is decomposed using local mean value and history electricity price data described in step S1 is decomposed into several PF components;
S3. training sample is selected, the prediction model of crossover algorithm optimization RBF neural in length and breadth is established;
S4. crossover algorithm in length and breadth is all made of described in step S3 to PF components all in step S2 and optimizes RBF neural
Prediction model predicted to obtain the predicted value of PF component;
S5. it is superimposed the predicted value of PF component all in step S4, obtains actual prediction result.
The Short-term electricity price forecasting method decomposed based on local mean value and optimize RBF neural of the invention, using part
Mean value decomposes the non-stationary and nonlinear problem for reducing electricity price sequence, optimizes radial base nerve net using crossover algorithm in length and breadth
Network forms mixed model and carries out look-ahead, reduces influence of the electricity price sequence complex characteristics to prediction result, improves short-term
The precision of Research on electricity price prediction.
Preferably, in step S1, the history electricity price data include continuous 1 week~4 weeks electricity price data, temporal resolution
For 0.25h~1h, one day includes 24~96 data points.
Preferably, in step S1, the history electricity price data include continuous 2 weeks electricity price data, and temporal resolution is
0.5h, one day includes 48 data points.
Preferably, local mean value described in step S2 decompose the following steps are included:
S21. using history electricity price data as original signal x (t), all Local Extremum n in original signal x (t)i(i=1,
2 ..., M), M indicates extreme value points, if any two adjacent extreme point is niAnd ni+1, then have:
mi=(ni+ni+1)/2
ai=| ni+ni+1|/2
In formula, miFor two neighboring extreme point niAnd ni+1Average value, aiFor envelope estimated value;
S22. by miAnd aiDiscrete point is wired to broken line with straight line respectively, using moving average method to miAnd aiLocated
Reason, obtains local mean value function m11(t) and local envelope function a11(t);
S23. by local mean value function m11(t) it separates, obtains from original signal x (t):
h11(t)=x (t)-m11(t)
And to h11(t) it is demodulated, is obtained:
s11(t)=h11(t)/a11(t)
If S24. s11It (t) is pure FM signal, then s11(t) local envelope function meets a12(t)=1;If a12(t)≠
1, then as the following formula iteration to s1n(t) become pure FM Function, i.e. s1n(t) envelope estimation function a1(n+1)(t)=1;Obtain with
Lower formula:
h11(t)=x (t)-m11(t)
h12(t)=s11(t)-m12(t)
h1n(t)=s1(n-1)(t)-m1n(t)
In formula,
s11(t)=h11(t)/a11(t)
s12(t)=h12(t)/a12(t)
s1n(t)=h1n(t)-a1n(t)
Stopping criterion for iteration are as follows:
S25. all envelope estimation functions generated in an iterative process are multiplied to obtain envelope signal a1(t),
S26. by envelope signal a1(t) and pure FM signal s1n(t) it is multiplied and obtains first PF component in original signal
PF1(t), PF1(t)=a1(t)s1n(t);
S27. by PF1(t) it is separated from original signal x (t), obtains a new signal u1(t), u1(t)=x
(t)-PF1(t);
S28. by u1(t) step S21~S27 is repeated as original signal, recycled k times, until ukIt (t) is a dull letter
Until number:
u1(t)=x (t)-PF1(t)
u2(t)=u1(t)-PF2(t)
uk(t)=uk-1(t)-PFk(t)
S29. original signal x (t) is by k PF component and uk(t) it reconstructs, it may be assumed that
It is LMD () that the present invention, which carries out the function that local mean value is decomposed in MATLAB platform, the local mean value in the present invention
It decomposes and original electricity price sequence is decomposed into 6 PF components.
Preferably, in step S24, the range a of an end is taken1n(t) the termination condition of ± Δ x as iteration;Take step
A in S251(t)≈1.Being arranged in this way can be improved calculating speed using following approximate algorithm, and not influence computational accuracy.
Preferably, in step S26, PF1Amplitude be a1(t), instantaneous frequency f1It (t) can be directly by s1n(t) it acquires, i.e.,
Are as follows:
Preferably, in step S3, the prediction model for establishing crossover algorithm optimization RBF neural in length and breadth includes following step
It is rapid:
S31. it according to given training sample, determines the neuron number of neural network topology structure and each layer, and determines
Crossed longitudinally probability P v, population scale M, the maximum number of iterations T of crossover algorithm in length and breadthmaxgen;
S32. the particle to be optimized is encoded, in the solution space of coding, initial population X=[X is randomly generated1,
X2..., XM]T;
S33. fitness evaluation is carried out to particle each in initial population described in step S32 using following formula:
In formula, ptIndicate the reality output of neural network,Indicate the target output of neural network, N indicates training sample
Number;
S34. lateral cross operation is carried out according to the following formula, and the filial generation that lateral cross operation obtains is stored in matrix MShc, then count
Calculate matrix MShcIn all particles adaptive value, obtained adaptive value and the adaptive value of parent population X are compared, selected suitable
The better particle of response is stored in matrix D ShcIn, it is expressed as following formula:
MShc(i, d)=r1× X (i, d)+(1-r1) × X (j, d)+c1× (X (i, d)-X (j, d))
MShc(j, d)=r2× X (j, d)+(1-r2) × X (i, d)+c2× (X (j, d)-X (i, d))
I, j ∈ N (1, M), d ∈ N (1, D)
In formula, r1、r2It is the random number between [0,1];c1、c2It is the random number between [- 1,1];M is the model of population
It encloses;D is the dimension of variable;The d that X (i, d), X (j, d) respectively indicate parent particle X (i) and X (j) is tieed up;MShc(i, d), MShc
(j, d) respectively indicates X (i, d) and X (j, d) and ties up generation filial generation in d by lateral cross.Lateral cross probability of the invention takes
1, lateral cross is crossover operation of doing sums in two particles, and two particles are randomly generated with one-dimensional.
S35. crossed longitudinally operation is carried out according to the following formula, and the solution that crossed longitudinally operation obtains is stored in matrix MSvc, then calculate
Matrix MSvcIn all particles adaptive value, be compared with its parent population X, select the better particle of fitness be retained in square
Battle array DShcIn, it is expressed as following formula:
MSvc(i, d1)=rX (i, d1)+(1-r) X (i, d2)
I ∈ N (1, M), d1, d2∈ N (1, D), r ∈ [0,1]
In formula, MSvc(i, d1) be particle X (i) d1Peacekeeping d2Tie up the filial generation by generating after crossed longitudinally operation.
Crossed longitudinally is a kind of arithmetic crossover that all particles carry out between different dimensional, and bidimensional be random combine together.
S36. judge whether current iteration number k is greater than Tmaxgen, if so, terminate optimization, iteration ends, and by DSvcIn
One group of best solution of fitness is set as weight corresponding to neural network and threshold value;Otherwise, step S34 is gone to be changed again
Generation.
The present invention optimizes the weight and threshold value of radial base neural net input layer and hidden layer using crossover algorithm in length and breadth, more
The defect that neural network easily falls into local optimum has been mended, the generalization ability of neural network is improved.
Preferably, in step S31, the training sample is made of 400~800 history electricity price data.
Preferably, in step S4,0.2h~0.75h prediction in advance is carried out to all PF components.
Preferably, in step S4,0.5h prediction in advance is carried out to all PF components.
Compared with prior art, the beneficial effects of the present invention are:
(1) present invention is decomposed using local mean value to reduce the non-stationary and nonlinear problem of electricity price sequence, using vertical
Traversed by pitches algorithm optimization radial base neural net composition mixed model and carries out look-ahead, reduces electricity price sequence complex characteristics pair
The influence of prediction result improves the precision of Electric Price Forecasting;
(2) present invention optimizes the weight and valve of radial base neural net input layer and hidden layer using crossover algorithm in length and breadth
Value, compensates for the defect that neural network easily falls into local optimum, improves the generalization ability of neural network.
Detailed description of the invention
Fig. 1 is the stream of the Short-term electricity price forecasting method of the invention for being decomposed based on local mean value and optimizing RBF neural
Cheng Tu.
Fig. 2 is the flow chart established crossover algorithm in length and breadth and optimize the prediction model of RBF neural of the invention.
Fig. 3 is the prediction effect figure of LMD-CSO-RBF model of the invention.
Specific embodiment
The present invention is illustrated With reference to embodiment, however protection scope of the present invention is not tightly limited to
Method and its core concept of the invention are merely used to help understand in the explanation of following embodiment, embodiment.It is all in this patent
Spirit and principle within made any modifications, equivalent replacements, and improvements etc., should be included in the present invention claims protection model
Within enclosing.
Embodiment one
It is as shown in Figure 1 to Figure 2 the short-term electricity price decomposed based on local mean value and optimize RBF neural of the present embodiment
The embodiment of prediction technique, comprising the following steps:
S1. history electricity price data are obtained and history electricity price data are pre-processed;
In the present embodiment, history electricity price data include continuous 2 weeks electricity price data, temporal resolution 0.5h, i.e., one day
Include 48 data points.
S2. it is decomposed using local mean value and history electricity price data described in step S1 is decomposed into several PF components;
Local mean value described in step S2 decompose the following steps are included:
S21. using history electricity price data as original signal x (t), all Local Extremum n in original signal x (t)i(i=1,
2 ..., M), M indicates extreme value points, if any two adjacent extreme point is niAnd ni+1, then have:
mi=(ni+ni+1)/2
ai=| ni+ni+1|/2
In formula, miFor two neighboring extreme point niAnd ni+1Average value, aiFor envelope estimated value;
S22. by miAnd aiDiscrete point is wired to broken line with straight line respectively, using moving average method to miAnd aiLocated
Reason, obtains local mean value function m11(t) and local envelope function a11(t);
S23. by local mean value function m11(t) it separates, obtains from original signal x (t):
h11(t)=x (t)-m11(t)
And to h11(t) it is demodulated, is obtained:
s11(t)=h11(t)/a11(t)
If S24. s11It (t) is pure FM signal, then s11(t) local envelope function meets a12(t)=1;If a12(t)≠
1, then as the following formula iteration to s1n(t) become pure FM Function, i.e. s1n(t) envelope estimation function a1(n+1)(t)=1;Obtain with
Lower formula:
h11(t)=x (t)-m11(t)
h12(t)=s11(t)-m12(t)
h1n(t)=s1(n-1)(t)-m1n(t)
In formula,
s11(t)=h11(t)/a11(t)
s12(t)=h12(t)/a12(t)
s1n(t)=h1n(t)-a1n(t)
Stopping criterion for iteration are as follows:
S25. all envelope estimation functions generated in an iterative process are multiplied to obtain envelope signal a1(t),
S26. by envelope signal a1(t) and pure FM signal s1n(t) it is multiplied and obtains first PF component in original signal
PF1(t), PF1(t)=a1(t)s1n(t);
S27. by PF1(t) it is separated from original signal x (t), obtains a new signal u1(t), u1(t)=x
(t)-PF1(t);
S28. by u1(t) step S21~S27 is repeated as original signal, recycled k times, until ukIt (t) is a dull letter
Until number:
u1(t)=x (t)-PF1(t)
u2(t)=u1(t)-PF2(t)
uk(t)=uk-1(t)-PFk(t)
S29. original signal x (t) is by k PF component and uk(t) it reconstructs, it may be assumed that
It is LMD () that the present embodiment, which carries out the function that local mean value is decomposed in MATLAB platform, and the local in the present invention is equal
Value, which is decomposed, is decomposed into 6 PF components for original electricity price sequence.To improve calculating speed and not influencing computational accuracy, the present embodiment benefit
To lower aprons algorithm: in step S24, taking the range a of an end1n(t) the termination condition of ± Δ x as iteration;Take step
A in S251(t)≈1.In step S26, PF1Amplitude be a1(t), instantaneous frequency f1It (t) can be directly by s1n(t) it acquires, i.e.,
Are as follows:
S3. training sample is selected, the prediction model of crossover algorithm optimization RBF neural in length and breadth is established;
In the present embodiment, the prediction model for establishing crossover algorithm optimization RBF neural in length and breadth includes the following steps, process
Figure is as shown in Figure 2:
S31. it according to given training sample, determines the neuron number of neural network topology structure and each layer, and determines
Crossed longitudinally probability P v, population scale M, the maximum number of iterations T of crossover algorithm in length and breadthmaxgen;
S32. the particle to be optimized is encoded, in the solution space of coding, initial population X=[X is randomly generated1,
X2..., XM]T;
S33. fitness evaluation is carried out to particle each in initial population described in step S32 using following formula:
In formula, ptIndicate the reality output of neural network,Indicate the target output of neural network, N indicates training sample
Number;
S34. lateral cross operation is carried out according to the following formula, and the filial generation that lateral cross operation obtains is stored in matrix MShc, then count
Calculate matrix MShcIn all particles adaptive value, obtained adaptive value and the adaptive value of parent population X are compared, selected suitable
The better particle of response is stored in matrix D ShcIn, it is expressed as following formula:
MShc(i, d)=r1× X (i, d)+(1-r1) × X (j, d)+c1× (X (i, d)-X (j, d))
MShc(j, d)=r2× X (j, d)+(1-r2) × X (i, d)+c2× (X (j, d)-X (i, d))
I, j ∈ N (1, M), d ∈ N (1, D)
In formula, r1、r2It is the random number between [0,1];c1、c2It is the random number between [- 1,1];M is the model of population
It encloses;D is the dimension of variable;The d that X (i, d), X (j, d) respectively indicate parent particle X (i) and X (j) is tieed up;MShc(i, d), MShc
(j, d) respectively indicates X (i, d) and X (j, d) and ties up generation filial generation in d by lateral cross.Lateral cross probability of the invention takes
1, lateral cross is crossover operation of doing sums in two particles, and two particles are randomly generated with one-dimensional.
S35. crossed longitudinally operation is carried out according to the following formula, and the solution that crossed longitudinally operation obtains is stored in matrix MSvc, then calculate
Matrix MSvcIn all particles adaptive value, be compared with its parent population X, select the better particle of fitness be retained in square
Battle array DShcIn, it is expressed as following formula:
MSvc(i, d1)=rX (i, d1)+(1-r) X (i, d2)
I ∈ N (1, M), d1, d2∈ N (1, D), r ∈ [0,1]
In formula, MSvc(i, d1) be particle X (i) d1Peacekeeping d2Tie up the filial generation by generating after crossed longitudinally operation.
Crossed longitudinally is a kind of arithmetic crossover that all particles carry out between different dimensional, and bidimensional be random combine together.
S36. judge whether current iteration number k is greater than Tmaxgen, if so, terminate optimization, iteration ends, and by DSvcIn
One group of best solution of fitness is set as weight corresponding to neural network and threshold value;Otherwise, step S34 is gone to be changed again
Generation.
The present embodiment optimizes the weight and threshold value of radial base neural net input layer and hidden layer using crossover algorithm in length and breadth,
The defect that neural network easily falls into local optimum is compensated for, the generalization ability of neural network is improved.The training sample of the present embodiment
It is made of 600 history electricity price data, establishes crossover algorithm in length and breadth based on history electricity price data and optimize radial base neural net
Prediction model.
S4. crossover algorithm in length and breadth is all made of described in step S3 to PF components all in step S2 and optimizes RBF neural
Prediction model predicted to obtain the predicted value of PF component;
The present embodiment to all subsequences be all made of the prediction model of the optimization radial base neural net of crossover algorithm in length and breadth into
Row shifts to an earlier date 0.5h prediction.
S5. it is superimposed the predicted value of PF component all in step S4, obtains actual prediction result.
Embodiment two
The present embodiment is that first embodiment is predicting the application in different model Research on electricity price prediction: the present embodiment is first to original
Electricity price data carry out the decomposition of local mean value (LMD), using crossover algorithm in length and breadth (CSO) optimization radial base neural net (RBF)
Model predicts all PF components, by the prediction result of all PF components
Superposition obtains electricity price actual prediction value.By prediction model LMD-CSO-RBF and the LMD-CSO-BP mould of the present embodiment
Type, CSO-RBF model and RBF model do error comparison and prediction time consuming analysis, and error comparison and prediction are time-consuming such as table 1
Shown, the predicted value of LMD-CSO-RBF model and the comparing result of actually measured value are as shown in Figure 3.
The different model Research on electricity price prediction error comparisons of table 1
As it can be seen from table 1 CSO-RBF model is predicted compared to traditional RBF, precision of prediction is improved, and at four kinds
In model, highest precision of prediction is LMD-CSO-RBF model, and compared to LMD-CSO-BP model, prediction is time-consuming
It greatly shortens.In addition, from figure 3, it can be seen that electricity price and actually measured electricity price phase that LMD-CSO-RBF model prediction obtains
Closely, it is seen that LMD-CSO-RBF model precision of prediction with higher.
Claims (10)
1. the Short-term electricity price forecasting method of RBF neural is decomposed and optimized based on local mean value, which is characterized in that including following
Step:
S1. history electricity price data are obtained and history electricity price data are pre-processed;
S2. it is decomposed using local mean value and history electricity price data described in step S1 is decomposed into several PF components;
S3. training sample is selected, the prediction model of crossover algorithm optimization RBF neural in length and breadth is established;
S4. the pre- of crossover algorithm optimization RBF neural in length and breadth is all made of described in step S3 to PF components all in step S2
Model is surveyed to be predicted to obtain the predicted value of PF component;
S5. it is superimposed the predicted value of PF component all in step S4, obtains actual prediction result.
2. the Short-term electricity price forecasting method according to claim 1 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, the history electricity price data include continuous 1 week~4 weeks electricity price data in step S1, temporal resolution is
0.25h~1h, one day includes 24~96 data points.
3. the Short-term electricity price forecasting method according to claim 1 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, in step S1, the history electricity price data include continuous 2 weeks electricity price data, temporal resolution 0.5h, one
It includes 48 data points.
4. the Short-term electricity price forecasting method according to claim 1 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, local mean value described in step S2 decompose the following steps are included:
S21. using history electricity price data as original signal x (t), all Local Extremum n in original signal x (t)i(i=1,2 ...,
M), M indicates extreme value points, if any two adjacent extreme point is niAnd ni+1, then have:
mi=(ni+ni+1)/2
ai=| ni+ni+1|/2
In formula, miFor two neighboring extreme point niAnd ni+1Average value, aiFor envelope estimated value;
S22. by miAnd aiDiscrete point is wired to broken line with straight line respectively, using moving average method to miAnd aiIt is handled, is obtained
To local mean value function m11(t) and local envelope function a11(t);
S23. by local mean value function m11(t) it separates, obtains from original signal x (t):
h11(t)=x (t)-m11(t)
And to h11(t) it is demodulated, is obtained:
s11(t)=h11(t)/a11(t)
If S24. s11It (t) is pure FM signal, then s11(t) local envelope function meets a12(t)=1;If a12(t) ≠ 1, then
Iteration is to s as the following formula1n(t) become pure FM Function, i.e. s1n(t) envelope estimation function a1(n+1)(t)=1;Obtain following public affairs
Formula:
h11(t)=x (t)-m11(t)
h12(t)=s11(t)-m12(t)
h1n(t)=s1(n-1)(t)-m1n(t)
In formula,
s11(t)=h11(t)/a11(t)
s12(t)=h12(t)/a12(t)
s1n(t)=h1n(t)-a1n(t)
Stopping criterion for iteration are as follows:
S25. all envelope estimation functions generated in an iterative process are multiplied to obtain envelope signal a1(t),
S26. by envelope signal a1(t) and pure FM signal s1n(t) it is multiplied and obtains first PF component PF in original signal1
(t), PF1(t)=a1(t)s1n(t);
S27. by PF1(t) it is separated from original signal x (t), obtains a new signal u1(t), u1(t)=x (t)-PF1
(t);
S28. by u1(t) step S21~S27 is repeated as original signal, recycled k times, until uk(t) it is for a monotonic function
Only:
u1(t)=x (t)-PF1(t)
u2(t)=u1(t)-PF2(t)
uk(t)=uk-1(t)-PFk(t)
S29. original signal x (t) is by k PF component and uk(t) it reconstructs, it may be assumed that
5. the Short-term electricity price forecasting method according to claim 4 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, taking the range a of an end in step S241n(t) the termination condition of ± Δ x as iteration;It takes in step S25
a1(t)≈1。
6. the Short-term electricity price forecasting method according to claim 4 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, in step S26, PF1Amplitude be a1(t), instantaneous frequency f1It (t) can be directly by s1n(t) it acquires, i.e., are as follows:
7. the Short-term electricity price forecasting method according to claim 1 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, in step S3, establish the prediction model of crossover algorithm optimization RBF neural in length and breadth the following steps are included:
S31. it according to given training sample, determines the neuron number of neural network topology structure and each layer, and determines in length and breadth
Crossed longitudinally probability P v, population scale M, the maximum number of iterations T of crossover algorithmmaxgen;
S32. the particle to be optimized is encoded, in the solution space of coding, initial population X=[X is randomly generated1,X2,...,
XM]T;
S33. fitness evaluation is carried out to particle each in initial population described in step S32 using following formula:
In formula, ptIndicate the reality output of neural network,Indicate the target output of neural network, N indicates number of training;
S34. lateral cross operation is carried out according to the following formula, and the filial generation that lateral cross operation obtains is stored in matrix MShc, then calculate square
Battle array MShcIn all particles adaptive value, obtained adaptive value and the adaptive value of parent population X are compared, fitness is selected
Better particle is stored in matrix D ShcIn, it is expressed as following formula:
MShc(i, d)=r1×X(i,d)+(1-r1)×X(j,d)+c1×(X(i,d)-X(j,d))
MShc(j, d)=r2×X(j,d)+(1-r2)×X(i,d)+c2×(X(j,d)-X(i,d))
I, j ∈ N (1, M), d ∈ N (1, D)
In formula, r1、r2It is the random number between [0,1];c1、c2It is the random number between [- 1,1];M is the range of population;D is
The dimension of variable;The d that X (i, d), X (j, d) respectively indicate parent particle X (i) and X (j) is tieed up;MShc(i,d)、MShc(j,d)
It respectively indicates X (i, d) and X (j, d) and generation filial generation is tieed up in d by lateral cross;
S35. crossed longitudinally operation is carried out according to the following formula, and the solution that crossed longitudinally operation obtains is stored in matrix MSvc, then calculating matrix
MSvcIn all particles adaptive value, be compared with its parent population X, select the better particle of fitness be retained in matrix
DShcIn, it is expressed as following formula:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
I ∈ N (1, M), d1,d2∈ N (1, D), r ∈ [0,1]
In formula, MSvc(i,d1) be particle X (i) d1Peacekeeping d2Tie up the filial generation by generating after crossed longitudinally operation;
S36. judge whether current iteration number k is greater than Tmaxgen, if so, terminate optimization, iteration ends, and by DSvcMiddle adaptation
It spends one group of best solution and is set as weight corresponding to neural network and threshold value;Otherwise, it goes to step S34 and carries out iteration again.
8. the Short-term electricity price forecasting method according to claim 7 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, the training sample is made of 400~800 history electricity price data in step S31.
9. the Short-term electricity price forecasting method according to claim 1 decomposed based on local mean value and optimize RBF neural,
It is characterized in that, carrying out 0.2h~0.75h prediction in advance to all PF components in step S4.
10. the Electric Price Forecasting side according to claim 9 decomposed based on local mean value and optimize RBF neural
Method, which is characterized in that in step S4,0.5h prediction in advance is carried out to all PF components.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810876676.5A CN109242532B (en) | 2018-08-03 | 2018-08-03 | Short-term electricity price prediction method based on local mean decomposition and RBF neural network optimization |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810876676.5A CN109242532B (en) | 2018-08-03 | 2018-08-03 | Short-term electricity price prediction method based on local mean decomposition and RBF neural network optimization |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109242532A true CN109242532A (en) | 2019-01-18 |
CN109242532B CN109242532B (en) | 2022-04-26 |
Family
ID=65070170
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810876676.5A Active CN109242532B (en) | 2018-08-03 | 2018-08-03 | Short-term electricity price prediction method based on local mean decomposition and RBF neural network optimization |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109242532B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110658905A (en) * | 2019-09-23 | 2020-01-07 | 珠海格力电器股份有限公司 | Early warning method, early warning system and early warning device for equipment running state |
CN110866631A (en) * | 2019-10-09 | 2020-03-06 | 重庆大学 | Method for predicting atmospheric pollution condition based on integrated gate recursion unit neural network GRU |
CN111311335A (en) * | 2020-03-13 | 2020-06-19 | 广东红海湾发电有限公司 | Electric power market frequency modulation auxiliary service quotation strategy support system and method |
CN111414977A (en) * | 2020-03-09 | 2020-07-14 | 西南交通大学 | Weighted integration temperature sensitive point combination selection method for machine tool spindle thermal error modeling |
CN111751133A (en) * | 2020-06-08 | 2020-10-09 | 南京航空航天大学 | Intelligent fault diagnosis method of deep convolutional neural network model based on non-local mean embedding |
CN112100913A (en) * | 2020-09-08 | 2020-12-18 | 中国电力科学研究院有限公司 | Data-driven clear electricity price boundary simulation method and system and storage medium |
WO2021077977A1 (en) * | 2019-10-23 | 2021-04-29 | 华能大理风力发电有限公司 | Method for analyzing transaction data in wind power bidding market, device, apparatus, and medium |
CN117217843A (en) * | 2023-10-18 | 2023-12-12 | 河北建投电力科技服务有限公司 | Power selling electricity price data processing method based on cloud computing |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102542167A (en) * | 2011-12-31 | 2012-07-04 | 东北电力大学 | Wind-speed time series forecasting method for wind power station |
CN107480829A (en) * | 2017-08-25 | 2017-12-15 | 广东工业大学 | A kind of Short-term electricity price forecasting method, apparatus and system |
-
2018
- 2018-08-03 CN CN201810876676.5A patent/CN109242532B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102542167A (en) * | 2011-12-31 | 2012-07-04 | 东北电力大学 | Wind-speed time series forecasting method for wind power station |
CN107480829A (en) * | 2017-08-25 | 2017-12-15 | 广东工业大学 | A kind of Short-term electricity price forecasting method, apparatus and system |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110658905A (en) * | 2019-09-23 | 2020-01-07 | 珠海格力电器股份有限公司 | Early warning method, early warning system and early warning device for equipment running state |
CN110658905B (en) * | 2019-09-23 | 2023-08-04 | 珠海格力电器股份有限公司 | Early warning method, early warning system and early warning device for equipment operation state |
CN110866631A (en) * | 2019-10-09 | 2020-03-06 | 重庆大学 | Method for predicting atmospheric pollution condition based on integrated gate recursion unit neural network GRU |
WO2021077977A1 (en) * | 2019-10-23 | 2021-04-29 | 华能大理风力发电有限公司 | Method for analyzing transaction data in wind power bidding market, device, apparatus, and medium |
CN111414977A (en) * | 2020-03-09 | 2020-07-14 | 西南交通大学 | Weighted integration temperature sensitive point combination selection method for machine tool spindle thermal error modeling |
CN111414977B (en) * | 2020-03-09 | 2023-06-20 | 西南交通大学 | Weighted integration temperature sensitive point combination selection method for machine tool spindle thermal error modeling |
CN111311335A (en) * | 2020-03-13 | 2020-06-19 | 广东红海湾发电有限公司 | Electric power market frequency modulation auxiliary service quotation strategy support system and method |
CN111311335B (en) * | 2020-03-13 | 2023-08-18 | 广东红海湾发电有限公司 | Power market frequency modulation auxiliary service quotation strategy support system and method |
CN111751133A (en) * | 2020-06-08 | 2020-10-09 | 南京航空航天大学 | Intelligent fault diagnosis method of deep convolutional neural network model based on non-local mean embedding |
CN112100913A (en) * | 2020-09-08 | 2020-12-18 | 中国电力科学研究院有限公司 | Data-driven clear electricity price boundary simulation method and system and storage medium |
CN117217843A (en) * | 2023-10-18 | 2023-12-12 | 河北建投电力科技服务有限公司 | Power selling electricity price data processing method based on cloud computing |
CN117217843B (en) * | 2023-10-18 | 2024-04-19 | 河北建投电力科技服务有限公司 | Power selling electricity price data processing method based on cloud computing |
Also Published As
Publication number | Publication date |
---|---|
CN109242532B (en) | 2022-04-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109242532A (en) | The Short-term electricity price forecasting method of RBF neural is decomposed and optimized based on local mean value | |
Deng et al. | An improved differential evolution algorithm and its application in optimization problem | |
Yuan et al. | Multi-objective optimal power flow based on improved strength Pareto evolutionary algorithm | |
Chen et al. | Wind speed forecasting using nonlinear-learning ensemble of deep learning time series prediction and extremal optimization | |
CN104636801B (en) | A kind of prediction transmission line of electricity audible noise method based on Optimized BP Neural Network | |
Lu et al. | Environmental/economic dispatch problem of power system by using an enhanced multi-objective differential evolution algorithm | |
Cai et al. | A hybrid CPSO–SQP method for economic dispatch considering the valve-point effects | |
Datta et al. | A binary-real-coded differential evolution for unit commitment problem | |
Gan et al. | A hybrid algorithm to optimize RBF network architecture and parameters for nonlinear time series prediction | |
Younes et al. | Multi-objective economic emission dispatch solution using hybrid FFA (firefly algorithm) and considering wind power penetration | |
Sefrioui et al. | A hierarchical genetic algorithm using multiple models for optimization | |
Gupta et al. | Comparison of Heuristic techniques: A case of TSP | |
Ran et al. | Short-term load forecasting based on CEEMDAN and Transformer | |
CN108647824A (en) | A kind of electricity price ultra-short term prediction method | |
Sun et al. | Solving the economic dispatch problem with a modified quantum-behaved particle swarm optimization method | |
Han et al. | Information-utilization-method-assisted multimodal multiobjective optimization and application to credit card fraud detection | |
CN105305490A (en) | Active distribution network planning method considering optimal economical efficiency of power quality | |
CN109242021A (en) | A kind of classification prediction technique based on multistage mixed model | |
CN110222908A (en) | A kind of short-term electricity demand forecasting method based on data-driven | |
Chang | Short-term load forecasting using radial basis function neural network | |
CN112819225A (en) | Carbon market price prediction method based on BP neural network and ARIMA model | |
Zhang et al. | A swarm based double Q-learning for optimal PV array reconfiguration with a coordinated control of hydrogen energy storage system | |
Sang et al. | Encoding carbon emission flow in energy management: A compact constraint learning approach | |
Liu et al. | DPSO based octagonal steiner tree algorithm for VLSI routing | |
Wang et al. | TATCN: time series prediction model based on time attention mechanism and TCN |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |