CN110263971A - Super short-period wind power combination forecasting method based on support vector machines - Google Patents
Super short-period wind power combination forecasting method based on support vector machines Download PDFInfo
- Publication number
- CN110263971A CN110263971A CN201910398404.3A CN201910398404A CN110263971A CN 110263971 A CN110263971 A CN 110263971A CN 201910398404 A CN201910398404 A CN 201910398404A CN 110263971 A CN110263971 A CN 110263971A
- Authority
- CN
- China
- Prior art keywords
- wind power
- data
- sequence
- value
- formula
- 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
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 27
- 238000013277 forecasting method Methods 0.000 title claims abstract description 16
- 239000002245 particle Substances 0.000 claims abstract description 46
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 17
- 238000005457 optimization Methods 0.000 claims abstract description 6
- 238000000034 method Methods 0.000 claims description 31
- 238000010606 normalization Methods 0.000 claims description 5
- 230000007547 defect Effects 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 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 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims 1
- 239000013589 supplement Substances 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013179 statistical model Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000000714 time series forecasting Methods 0.000 description 1
- 238000005303 weighing 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
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
Pretreated data are normalized firstly, carrying out linear interpolation replacement according to the data of time adjacent segments to wind power historical data to be processed for super short-period wind power combination forecasting method disclosed by the invention based on support vector machines；Secondly, processed wind power data are resolved into the intrinsic sequence of function and residual sequence using empirical mode decomposition；Then, are established by quantum particle swarm-supporting vector machine model and is trained optimization for the obtained eigenfunction sequence of decomposition and residual sequence, obtain the predicted value of each sequence；Finally, the predicted value of each sequence is superimposed to obtain final wind power prediction value, and carry out error assessment analysis.Prediction result of the invention directly predicted with support vector machines or without data characteristics decomposition result compared with all increase, while there is not the excessive situation of local error.Robustness is stronger compared with existing wind power prediction scheme, calculating speed faster, data requirements is few, prediction effect is more preferable.
Description
Technical field
The invention belongs to wind power prediction technical fields, are related to a kind of super short-period wind power based on support vector machines
Combination forecasting method.
Background technique
As the energy is petered out and appearance the problems such as environmental pollution is serious, the scale that wind-power electricity generation accesses power grid is more next
It is bigger.But wind energy has intermittent, randomness and the influence vulnerable to factors such as wind speed, wind direction, geographical location and meteorologies.When
When grid-connected scale reaches a certain level, it would be possible to a series of problem can be brought to electric system, as voltage disturbance, three-phase are uneven
Weighing apparatus, Optimized Operation even cause safety accident etc..Along with the influence of burst factor (typhoon, thunder and lightning weather etc.), so that wind-powered electricity generation
Power prediction precision is not able to satisfy always the demand of large-scale grid connection, limits making full use of and dissolving for renewable energy.Cause
This, carrying out accurate prediction to wind power prediction has very important theory significance and engineering application value.
For this purpose, scholars propose many wind power prediction algorithms one after another, it mainly include based on the pre- of statistical model
Method of determining and calculating, the prediction algorithm based on physical model and Combined model forecast algorithm.Prediction algorithm based on statistical model is common
Have grey method, Kalman filtering method, artificial neural network, support vector machines etc..Wherein, grey method can only be fitted
Linear discrete sequence, and often most of original series are all Continuous Nonlinears；Kalman filtering method is only applicable to noise clothes
From the signal of Gaussian Profile, so being not suitable for the wind power signal of noise non-gaussian distribution；Artificial neural network has very strong
Adaptive and Self-organization, generalization is stronger, can efficiently handle complexity problem, but exist easily fall into local optimum and
The problems such as convergence rate is slow；Support vector machines is current popular method, has global convergence and believes independent of experience
The advantages that breath, but it is slower to extensive, multidimensional the signal operation time.Prediction algorithm based on physical model refers mainly to numerical value
Weather forecast (NWP), it needs very accurately information such as roughness of ground surface, atmospheric density and wind speed, and input parameter is more and makes
Valence is high, and this method is rarely employed in the country.Combinational algorithm is broadly divided into after data processing and the combinational algorithm of intelligent algorithm and several
Two kinds of intelligent algorithm weighted array algorithm.Such as wavelet decomposition with and support vector machines combinational algorithm, wavelet decomposition is to the time
Span is big and Small Sample Database treatment effect is bad, while it is also relatively difficult to reasonably select wavelet basis；Several intelligent algorithm weightings
Although combination precision improves still no consideration data feature itself and human factor is relatively more, it is high-precision that it is not able to satisfy wind power
Spend the demand of prediction.
Summary of the invention
The object of the present invention is to provide a kind of the super short-period wind power combination forecasting method based on support vector machines, solution
Wind power prediction model precision is not high under the conditions of the prior art, it is difficult to meet the problem of large-scale grid connection requires.
The technical scheme adopted by the invention is that the super short-period wind power combination forecasting method based on support vector machines,
Specific operation process includes the following steps:
Step 1, data prediction
Linear interpolation replacement is carried out according to the data of time adjacent segments to wind power historical data to be processed；
Step 2, pretreated data are normalized；
Step 3, the intrinsic sequence of function will be resolved into through the processed wind power data of step 2 using empirical mode decomposition
And residual sequence；
Step 4, quantum particle swarm-support vector machines is established respectively to the obtained eigenfunction sequence of decomposition and residual sequence
Model is trained optimization, obtains the predicted value of each sequence；
Step 5, the predicted value of each sequence is superimposed to obtain final wind power prediction value.
Other features of the invention also reside in,
It is long for the data interval time of defect in step 1 in process of data preprocessing, then use same time
Section, similar meteorological condition and adjacent several days data replace；For wrong data according to the data weighted average of front/rear 5min
It is supplemented.
Preferably, the process of data normalization is as follows in step 2:
It is normalized using formula (1):
Wherein, xmaxAnd xminThe maxima and minima in initial data is respectively indicated, normalization range is [0,1].
Preferably, detailed process is as follows for step 3:
Step 3.1, all maximum points in original wind power sequence X (t) are found out, then with cubic spline function pair
Maximum point is handled, and the envelope up and down that curve forms original wind power sequence X (t) is finally connected into；
Step 3.2, it averages to the envelope up and down of original wind power sequence X (t), as shown in formula (7):
In formula, e+(t) indicate that coenvelope line, e- (t) indicate lower envelope line, m1(t) envelope mean value up and down is indicated；
Average value and original wind power sequence are made the difference, wind power data h is obtained1(t), if h1(t) meet eigen mode
Function definition, i.e. for function in entire time range, the number of Local Extremum and zero crossing must equal or most difference one
It is a, then it is denoted as first intrinsic mode functions, otherwise by h1(t) K iteration screening is repeated, until meeting intrinsic mode functions
Until definition requires, i.e. h1k(t) become first IMF, remember C1(t)=h1k(t)；
Step 3.3, by C in X (t)1(t) it separates, obtains a difference signal r1(t), difference signal r1(t)
As new original signal, step 3.1- step 3.2 is repeated, next intrinsic mode functions will be obtained, be denoted as C2(t), repeatedly
N times are carried out, n intrinsic mode functions have just been obtained, as shown in formula (8):
In formula, { r1(t)、r2(t)…rn(t) } difference signal, { C are indicated1(t)、C2(t)...Cn(t) } eigen mode letter is indicated
Number.
Work as rn(t) or Cn(t) decomposable process terminates when meeting the following conditions:
Wherein SdValue is between 0.2~0.3；
In conclusion nonstationary time series is broken down into n IMF and residual error, as shown in formula (10):
Preferably, detailed process is as follows for step 4:
Step 4.1, in random initializtion population particle Position And Velocity, the position expression parameter (C, σ) of particle
Current value；
Step 4.2, the fitness value for calculating all particles updates the optimal position of individual of each particle according to its fitness
It sets and the global optimum position of entire population；
Step 4.3, mbest is calculated according to formula (11)；
Step 4.4, allow f from iter/iter in formula (12)maxIt is incremented to itermax/itermax, calculate each particle
Random point ppij；
Step 4.5, it according to the position of formula (13) more new particle, show that most recent parameters combine, works as XijMinimum i.e. individual is most
Algorithm terminates when excellent；Otherwise, return step 4.2；
Step 4.6, SVM is assigned to according to the optimized parameter that quantum particle swarm optimizing obtains；
PPij=f × Pij+(1-f)×Pgj (12)
Xij=PPij±a×|mbestj-Xij|×ln(1/u) (13)
Wherein, f, u are the random numbers on section [0,1]；Mbest is the average value of population individual extreme value pbest；PPij
For PijAnd PgjBetween random point；A is the particle shrinkage expansion coefficient of quantum particle swarm, when i-th ter times iteration, a=b+c ×
(itermax-iter)/itermax, indicate that a is decremented to b from b+c with the increase of the number of iterations, b and c value depends on the circumstances full
Sufficient a is linear from 1 to 0.5 to be reduced, itermaxIt is the maximum times of iteration；
When finding optimized parameter using quantum particle swarm, the position of each particle represents support vector machines parameter (C, σ)
Current value, to each particle be supported vector machine training, until by formula (13) export optimized parameter, then use quantum
Population-decomposition of supporting vector machine model prediction steps 3 obtains each subsequence.
Preferably, final wind power prediction value step 5 obtained is by using root-mean-square error RMSE, average exhausted
To error MAE, relative coefficient r, largest prediction error δmax, relative error RE do error assessment index, carry out error assessment point
Analysis, as shown in formula (14)-formula (18):
δmax=max (| PMi-PPi|) (17)
In formula, PMiFor the actual power at i moment,For the average value of all sample actual powers, PPiIt is pre- for the i moment
Power scale；For the average value of all prediction powers；Cap is the booting total capacity of wind power plant；N is all number of samples.
The invention has the advantages that the super short-period wind power combination forecasting method based on support vector machines, is being supported
On the basis of vector machine (SVM) prediction scheme, easily there is excessive office to improve it in non-gaussian distribution time series forecasting
The case where portion's error, reference quantum particle swarm (QPSO) to optimize SVM parameter (C, σ), are searching for early period using QPSO,
Particle has biggish solution space, by a period of time search close to globally optimal solution, then in the search later period to region of search
It is limited, finer search quickening convergence rate is carried out in the neighborhood centered on globally optimal solution and avoids part most
It is excellent.Wind power time series is non-stationary series simultaneously, puts down original wind power data to make data tranquilization introduce EMD
Steadyization processing, then predicts super short-period wind power with QPSO-SVM model.Method prediction result of the invention and support to
There is not the excessive situation of local error compared to improving 5% or so in the direct prediction result of amount machine.With existing wind power
Prediction scheme compared to robustness is stronger, calculating speed faster, data requirements is few, more mutually agrees with engineering practice, prediction effect is more
It is good.
Detailed description of the invention
Fig. 1 is the flow chart of the super short-period wind power combination forecasting method of the invention based on support vector machines；
Fig. 2 is the parameter optimization signal of the super short-period wind power combination forecasting method of the invention based on support vector machines
Figure；
Fig. 3 is the EMD exploded view of wind speed and power in the embodiment of the present invention；
Fig. 4 is the power prediction value of subsequence in the embodiment of the present invention and the comparison diagram of actual value；
Fig. 5 is the prediction result and existing method prediction result comparison diagram of the embodiment of the present invention.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
Super short-period wind power combination forecasting method based on support vector machines of the invention, as shown in Figure 1, concrete operations
Process includes the following steps:
Step 1, data prediction
Linear interpolation replacement is carried out according to the data of time adjacent segments to wind power historical data to be processed；
It is long for the data interval time of defect in step 1 in process of data preprocessing, then use same time
Section, similar meteorological condition and adjacent several days data replace；For wrong data according to the data weighted average of front/rear 5min
It is supplemented；
Step 2, pretreated data are normalized；
The process of data normalization is as follows in step 2:
It is normalized using formula (1):
Wherein, xmaxAnd xminThe maxima and minima in initial data is respectively indicated, normalized range is [0 1]；
Step 3, eigen mode will be resolved into through the processed wind power data of step 2 using empirical mode decomposition (EMD)
Function (IMF) sequence and residual error (r) sequence；
Detailed process is as follows for step 3:
Step 3.1, all maximum points in original wind power sequence X (t) are found out, then with cubic spline function pair
Maximum point is handled, and the envelope up and down that curve forms original wind power sequence X (t) is finally connected into；
Step 3.2, it averages to the envelope up and down of original wind power sequence X (t), as shown in formula (7):
In formula, e+(t) coenvelope line, e are indicated-(t) lower envelope line, m are indicated1(t) envelope mean value up and down is indicated；
Mean value and original wind power sequence are made the difference, wind power data h is obtained1(t), if h1(t) meet eigen mode letter
Number (IMF) definition, i.e. function is in entire time range, the necessary equal or most differences of the number of Local Extremum and zero crossing
One, then it is denoted as first intrinsic mode functions, otherwise by h1(t) K iteration screening is repeated, until meeting intrinsic mode functions
Definition require until, i.e. h1k(t) become first IMF, remember C1(t)=h1k(t)；
Step 3.3, by C in X (t)1(t) it separates, obtains a difference signal r1(t), difference signal r1(t)
As new original signal, step 3.1- step 3.2 is repeated, next intrinsic mode functions will be obtained, be denoted as C2(t), repeatedly
N times are carried out, n intrinsic mode functions have just been obtained, as shown in formula (8):
In formula, { r1(t)、r2(t)…rn(t) } difference signal, { C are indicated1(t)、C2(t)...Cn(t) } eigen mode letter is indicated
Number；
Work as rn(t) or Cn(t) decomposable process terminates when meeting the following conditions:
Wherein SdValue is between 0.2~0.3；
In conclusion nonstationary time series is broken down into n IMF and residual error, as shown in formula (10):
Step 4, quantum particle swarm-support vector machines is established respectively to the obtained eigenfunction sequence of decomposition and residual sequence
Model is trained optimization, obtains the predicted value of each sequence；
Detailed process is as follows for step 4, as shown in Figure 2:
Step 4.1, in random initializtion population particle Position And Velocity, the position expression parameter (C, σ) of particle
Current value；
Step 4.2, the fitness value for calculating all particles updates the optimal position of individual of each particle according to its fitness
It sets and the global optimum position of entire population；
Step 4.3, mbest is calculated according to formula (11)；
Step 4.4, allow f from iter/iter in formula (12)maxIt is incremented to itermax/itermax, calculate each particle
Random point ppij；
Step 4.5, it according to the position of formula (13) more new particle, show that most recent parameters combine, works as XijMinimum i.e. individual is most
Algorithm terminates when excellent；Otherwise, return step 4.2；
Step 4.6, the optimized parameter (C, σ) obtained according to quantum particle swarm optimizing is assigned to SVM；
PPij=f × Pij+(1-f)×Pgj (12)
Xij=PPij±a×|mbestj-Xij|×ln(1/u) (13)
Wherein, f, u are the random numbers on section [0,1]；Mbest is the average value of population individual extreme value pbest；PPij
For PijAnd PgjBetween random point；A is the particle shrinkage expansion coefficient of quantum particle swarm, when i-th ter times iteration, a=b+c ×
(itermax-iter)/itermax, a is indicated as the increase of the number of iterations is decremented to b from b+c, b and c value depends on the circumstances,
Meet value a linear reduction, iter from 1 to 0.5maxIt is the maximum times of iteration；
When finding optimized parameter using quantum particle swarm, the position of each particle represents support vector machines parameter (C, σ)
Current value, to each particle be supported vector machine training, until by formula (13) export optimized parameter, therefore, by (C, σ)
Quantum particle swarm-supporting vector machine model can be obtained by bringing traditional support vector machine into, finally according to model built prediction steps 3
Decompose obtained each subsequence.
Step 5, the predicted value of each sequence is superimposed to obtain final wind power prediction value；
By obtained final wind power prediction value by using root-mean-square error RMSE, mean absolute error MAE, phase
Close property coefficient r, largest prediction error δmax, relative error RE do error assessment index, carry out error assessment analysis, such as formula
(14) shown in-formula (18):
δmax=max (| PMi-PPi|) (17)
In formula, PMiFor the actual power at i moment,For the average value of all sample actual powers, PPiIt is pre- for the i moment
Power scale；For the average value of all prediction powers；Cap is the booting total capacity of wind power plant；N is all number of samples.
Embodiment
Step 1, the measured data of the every 5min of certain wind power plant 1# blower year September is chosen as test object, to history number
According to the data incomplete being likely to occur, linear interpolation replacement is carried out according to the data of time adjacent segments, if when the data break of defect
Between it is long, can be replaced with same time period, similar meteorological condition, adjacent several days data；It can basis for wrong data
Front/rear 5min data weighted average is supplemented；
Step 2, the wind power data of input are normalized；
Step, 3, EMD decomposition is carried out respectively to wind speed and power, as shown in figure 3, by can after training to normalization data
Wind speed and Power Decomposition are obtained 6 IMF sequences and 1 residual sequence；
Step 4, the subsequence obtained to decomposition establishes QPSO-SVM model respectively and obtains the predicted value of each sequence；
Analyzed by Multi simulation running, parameter setting is as follows: penalty coefficient and kernel functional parameter search range be C=[0,
150], σ=[0.1,10]；It is 20 that maximum number of iterations, which takes 100, particle populations number m,.Inertia weight coefficient w=[0.4,0.9],
C1=C2=2；The predicted value of each subsequence as shown in figure 4, as shown in Figure 4 IMF1~IMF5 prediction result with actual value very
Close, IMF6 prediction result is bigger than normal, main cause be it is very close with IMF5 frequency, caused by being EMD algorithm self-defect；
Step 5, the predicted value of each subsequence is superimposed to obtain final wind power prediction value；
Error assessment analysis is carried out, the results are shown in Table 1, from table 1 and Fig. 5 is it is found that with cross validation and grid search is passed through
Support vector machine method precision of prediction after method optimization probably improves 4% or so, and essence can be improved with model prediction of the present invention
Degree 5% or so.Although simultaneously it can be seen that the root-mean-square error after EMD is decomposed without the root-mean-square error that EMD is decomposed than mentioning
High by 0.78%, mean absolute error improves 0.11%, but the former worst error 177kW smaller than the latter, illustrates that EMD drops
The low non-stationary influence to prediction model of data, precision of prediction increase really.
1 error assessment analytical table of table
Method of the invention is first handled original wind power data tranquilization with empirical mode decomposition, then dosage
Seed subgroup searches out SVM optimal parameter and establishes QPSO-SVM model and goes to predict each subsequence, and finally superposition obtains most
Whole predicted value.By embodiment, the result shows that, present invention precision on the basis of SVM is individually predicted improves 5% or so, reaches
Effectively improve the purpose of wind power prediction precision.Effectively prevent going out for the excessive situation of local prediction error in prediction result
It is existing.Meet State Grid Corporation of China's standard.If can apply to electric system wind power prediction, wind-electricity integration can be effectively reduced to being
The phenomenon that system impact, there is very important theory significance and engineering application value.
Claims (6)
1. the super short-period wind power combination forecasting method based on support vector machines, which is characterized in that specific operation process includes
Following steps:
Step 1, data prediction
Linear interpolation replacement is carried out according to the data of time adjacent segments to wind power historical data to be processed；
Step 2, pretreated data are normalized；
Step 3, the intrinsic sequence of function and residual will be resolved into through the processed wind power data of step 2 using empirical mode decomposition
Difference sequence；
Step 4, quantum particle swarm-supporting vector machine model is established respectively to the obtained eigenfunction sequence of decomposition and residual sequence
It is trained optimization, obtains the predicted value of each sequence；
Step 5, the predicted value of each sequence is superimposed to obtain final wind power prediction value.
2. the super short-period wind power combination forecasting method based on support vector machines as described in claim 1, which is characterized in that
It is long for the data interval time of defect in the step 1 in process of data preprocessing, then use same time period, class
Replace like meteorological condition and adjacent several days data；Wrong data is carried out according to the data weighted average of front/rear 5min
Supplement.
3. the super short-period wind power combination forecasting method based on support vector machines as described in claim 1, which is characterized in that
The process of data normalization is as follows in the step 2:
It is normalized using formula (1):
Wherein, x indicates initial data, xmaxAnd xminRespectively indicate the maxima and minima in initial data, normalized model
It encloses for [0 1].
4. the super short-period wind power combination forecasting method based on support vector machines as described in claim 1, which is characterized in that
Detailed process is as follows for the step 3:
Step 3.1, all maximum points in original wind power sequence X (t) are found out, then with cubic spline function to very big
Value point is handled, and the envelope up and down that curve forms original wind power sequence X (t) is finally connected into；
Step 3.2, it averages to the envelope up and down of original wind power sequence X (t), as shown in formula (7):
In formula, e+(t) indicate that coenvelope line, e- (t) indicate lower envelope line, m1(t) envelope mean value up and down is indicated；
Average value and original wind power sequence are made the difference, wind power data h is obtained1(t), if h1(t) meet intrinsic mode functions
Definition, i.e. function in entire time range, the number of Local Extremum and zero crossing must equal or most differences one, then
It is denoted as first intrinsic mode functions, otherwise by h1(t) K iteration screening, the definition until meeting intrinsic mode functions is repeated
Until it is required that, i.e. h1k(t) become first IMF, remember C1(t)=h1k(t)；
Step 3.3, by C in X (t)1(t) it separates, obtains a difference signal r1(t), difference signal r1(t) conduct
New original signal repeats step 3.1- step 3.2, will obtain next intrinsic mode functions, be denoted as C2(t), n is repeated
It is secondary, n intrinsic mode functions have just been obtained, as shown in formula (8):
In formula, { r1(t)、r2(t)…rn(t) } difference signal, { C are indicated1(t)、C2(t)...Cn(t) } intrinsic mode functions are indicated；
Work as rn(t) or Cn(t) decomposable process terminates when meeting the following conditions:
Wherein SdValue is between 0.2~0.3；
In conclusion nonstationary time series is broken down into n IMF and residual error, as shown in formula (10):
5. the super short-period wind power combination forecasting method based on support vector machines as claimed in claim 4, which is characterized in that
Detailed process is as follows for the step 4:
Step 4.1, in random initializtion population particle Position And Velocity, the position expression parameter (C, σ) of particle it is current
Value；
Step 4.2, the fitness value for calculating all particles, according to its fitness, update each particle personal best particle and
The global optimum position of entire population；
Step 4.3, mbest is calculated according to formula (11)；
Step 4.4, allow f from iter/iter in formula (12)maxIt is incremented to itermax/itermax, it is random to calculate each particle
Point ppij；
Step 4.5, it according to the position of formula (13) more new particle, show that most recent parameters combine, works as XijWhen minimum is that individual is optimal
Algorithm terminates；Otherwise, return step 4.2；
Step 4.6, SVM is assigned to according to the optimized parameter that quantum particle swarm optimizing obtains；
PPij=f × Pij+(1-f)×Pgj (12)
Xij=PPij±a×|mbestj-Xij|×ln(1/u) (13)
Wherein, f, u are the random numbers on section [0,1]；Mbest is the average value of population individual extreme value pbest；PPijFor Pij
And PgjBetween random point；A is the particle shrinkage expansion coefficient of quantum particle swarm, when i-th ter times iteration, a=b+c ×
(itermax-iter)/itermax, a is indicated as the increase of the number of iterations is decremented to b from b+c, b and c value depends on the circumstances,
Meet a linear reduction from 1 to 0.5, itermax is the maximum times of iteration；
When finding optimized parameter using quantum particle swarm, the position of each particle represents working as support vector machines parameter (C, σ)
Preceding value is supported vector machine training to each particle, until exporting optimized parameter by formula (13), then uses quanta particle
The decomposition of group-supporting vector machine model prediction steps 3 obtains each subsequence.
6. the super short-period wind power combination forecasting method based on support vector machines as described in claim 1, which is characterized in that
The final wind power prediction value that step 5 is obtained is by using root-mean-square error RMSE, mean absolute error MAE, correlation
Property coefficient r, largest prediction error δmax, relative error RE do error assessment index, error assessment analysis is carried out, such as formula (14)-
Shown in formula (18):
δmax=max (| PMi-PPi|) (17)
In formula, PMiFor the actual power at i moment,For the average value of all sample actual powers, PPiFor the pre- measurement of power at i moment
Rate；For the average value of all prediction powers；Cap is the booting total capacity of wind power plant；N is all number of samples.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910398404.3A CN110263971A (en) | 2019-05-14 | 2019-05-14 | Super short-period wind power combination forecasting method based on support vector machines |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910398404.3A CN110263971A (en) | 2019-05-14 | 2019-05-14 | Super short-period wind power combination forecasting method based on support vector machines |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110263971A true CN110263971A (en) | 2019-09-20 |
Family
ID=67914682
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910398404.3A Pending CN110263971A (en) | 2019-05-14 | 2019-05-14 | Super short-period wind power combination forecasting method based on support vector machines |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110263971A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111079980A (en) * | 2019-11-22 | 2020-04-28 | 天合云能源互联网技术(杭州)有限公司 | Optical power prediction method based on self-adaptive classification strategy and hybrid optimization SVR |
| CN112036672A (en) * | 2020-11-06 | 2020-12-04 | 中国电力科学研究院有限公司 | New energy power generation ultra-short term power prediction method and system based on iterative correction |
| CN112528560A (en) * | 2020-12-04 | 2021-03-19 | 贵州电网有限责任公司 | Fan output simulation method for grid-connected detection of wind driven generator |
| CN112668078A (en) * | 2020-12-24 | 2021-04-16 | 青岛理工大学 | Method for identifying damage of rusted reinforced concrete beam after fire disaster |
| CN113361115A (en) * | 2021-06-11 | 2021-09-07 | 仲恺农业工程学院 | Method for predicting dissolved oxygen change of industrial aquaculture water of tilapia |
| CN115496627A (en) * | 2022-11-22 | 2022-12-20 | 国网天津市电力公司电力科学研究院 | Method and system for evaluating response potential of adjustable resource |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103400052A (en) * | 2013-08-22 | 2013-11-20 | 武汉大学 | Combined method for predicting short-term wind speed in wind power plant |
| CN106778846A (en) * | 2016-12-02 | 2017-05-31 | 华北电力大学 | A kind of method for forecasting based on SVMs |
| US20170329314A1 (en) * | 2014-11-26 | 2017-11-16 | Shenyang Institute Of Automation, Chinese Academy Of Sciences | Modbus tcp communication behaviour anomaly detection method based on ocsvm dual-outline model |
| CN107507097A (en) * | 2017-07-03 | 2017-12-22 | 上海电力学院 | A kind of short-term wind power prediction method |
| CN108052968A (en) * | 2017-12-08 | 2018-05-18 | 哈尔滨工程大学 | A kind of perception intrusion detection method of QSFLA-SVM |
| CN111697621A (en) * | 2020-06-23 | 2020-09-22 | 石家庄铁道大学 | Short-term wind power prediction method based on EWT-PDBN combination |
-
2019
- 2019-05-14 CN CN201910398404.3A patent/CN110263971A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103400052A (en) * | 2013-08-22 | 2013-11-20 | 武汉大学 | Combined method for predicting short-term wind speed in wind power plant |
| US20170329314A1 (en) * | 2014-11-26 | 2017-11-16 | Shenyang Institute Of Automation, Chinese Academy Of Sciences | Modbus tcp communication behaviour anomaly detection method based on ocsvm dual-outline model |
| CN106778846A (en) * | 2016-12-02 | 2017-05-31 | 华北电力大学 | A kind of method for forecasting based on SVMs |
| CN107507097A (en) * | 2017-07-03 | 2017-12-22 | 上海电力学院 | A kind of short-term wind power prediction method |
| CN108052968A (en) * | 2017-12-08 | 2018-05-18 | 哈尔滨工程大学 | A kind of perception intrusion detection method of QSFLA-SVM |
| CN111697621A (en) * | 2020-06-23 | 2020-09-22 | 石家庄铁道大学 | Short-term wind power prediction method based on EWT-PDBN combination |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111079980A (en) * | 2019-11-22 | 2020-04-28 | 天合云能源互联网技术(杭州)有限公司 | Optical power prediction method based on self-adaptive classification strategy and hybrid optimization SVR |
| CN111079980B (en) * | 2019-11-22 | 2021-06-29 | 天合云能源互联网技术(杭州)有限公司 | Optical power prediction method based on self-adaptive classification strategy and hybrid optimization SVR |
| CN112036672A (en) * | 2020-11-06 | 2020-12-04 | 中国电力科学研究院有限公司 | New energy power generation ultra-short term power prediction method and system based on iterative correction |
| CN112528560A (en) * | 2020-12-04 | 2021-03-19 | 贵州电网有限责任公司 | Fan output simulation method for grid-connected detection of wind driven generator |
| CN112668078A (en) * | 2020-12-24 | 2021-04-16 | 青岛理工大学 | Method for identifying damage of rusted reinforced concrete beam after fire disaster |
| CN112668078B (en) * | 2020-12-24 | 2022-05-17 | 青岛理工大学 | Method for identifying damage of rusted reinforced concrete beam after fire disaster |
| CN113361115A (en) * | 2021-06-11 | 2021-09-07 | 仲恺农业工程学院 | Method for predicting dissolved oxygen change of industrial aquaculture water of tilapia |
| CN115496627A (en) * | 2022-11-22 | 2022-12-20 | 国网天津市电力公司电力科学研究院 | Method and system for evaluating response potential of adjustable resource |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110263971A (en) | Super short-period wind power combination forecasting method based on support vector machines | |
| CN107766990B (en) | Method for predicting power generation power of photovoltaic power station | |
| CN106447098B (en) | Photovoltaic ultra-short-term power prediction method and device | |
| CN112529282A (en) | Wind power plant cluster short-term power prediction method based on space-time graph convolutional neural network | |
| CN105868853B (en) | Method for predicting short-term wind power combination probability | |
| CN102184453A (en) | Wind power combination predicting method based on fuzzy neural network and support vector machine | |
| CN109086928A (en) | Photovoltaic plant realtime power prediction technique based on SAGA-FCM-LSSVM model | |
| CN109711609A (en) | Photovoltaic plant output power predicting method based on wavelet transformation and extreme learning machine | |
| CN109523084A (en) | A kind of ultrashort-term wind power prediction method based on pivot analysis and machine learning | |
| CN112418346B (en) | Numerical weather forecast total radiation system error classification calculation method | |
| CN108428017A (en) | Wind power interval prediction method based on core extreme learning machine quantile estimate | |
| Ye et al. | Combined approach for short-term wind power forecasting based on wave division and Seq2Seq model using deep learning | |
| CN108717579A (en) | A kind of short-term wind-electricity power interval prediction method | |
| CN111242355A (en) | Photovoltaic probability prediction method and system based on Bayesian neural network | |
| CN107895202A (en) | Short-term wind speed forecasting method based on various visual angles wind speed mode excavation | |
| CN110866633A (en) | Micro-grid ultra-short term load prediction method based on SVR support vector regression | |
| CN114970952B (en) | Photovoltaic output short-term prediction method and system considering environmental factors | |
| Pan et al. | Probabilistic short-term wind power forecasting using sparse Bayesian learning and NWP | |
| CN109858667A (en) | It is a kind of based on thunder and lightning weather to the short term clustering method of loading effects | |
| Chen et al. | Graph neural network-based wind farm cluster speed prediction | |
| CN116167508B (en) | Short-term photovoltaic output rapid prediction method and system based on meteorological factor decomposition | |
| CN116128211A (en) | Wind-light-water combined short-term optimization scheduling method based on wind-light uncertainty prediction scene | |
| Chong et al. | Short-term PV generation forecasting based on weather type clustering and improved GPR model | |
| Han et al. | Analysis of electric heating load characteristics in south hebei power grid | |
| Liu et al. | A novel photovoltaic power output forecasting method based on weather type clustering and wavelet support vector machines regression |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190920 |
|
| RJ01 | Rejection of invention patent application after publication |