CN107609774A - A kind of photovoltaic power Forecasting Methodology based on mind evolutionary Optimization of Wavelet neutral net - Google Patents
A kind of photovoltaic power Forecasting Methodology based on mind evolutionary Optimization of Wavelet neutral net Download PDFInfo
- Publication number
- CN107609774A CN107609774A CN201710811772.7A CN201710811772A CN107609774A CN 107609774 A CN107609774 A CN 107609774A CN 201710811772 A CN201710811772 A CN 201710811772A CN 107609774 A CN107609774 A CN 107609774A
- Authority
- CN
- China
- Prior art keywords
- wavelet
- data
- sub
- group
- prediction
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000005457 optimization Methods 0.000 title claims abstract description 25
- 230000007935 neutral effect Effects 0.000 title claims abstract description 10
- 238000013528 artificial neural network Methods 0.000 claims abstract description 23
- 238000005096 rolling process Methods 0.000 claims abstract description 8
- 238000013519 translation Methods 0.000 claims abstract description 3
- 230000006870 function Effects 0.000 claims description 32
- 230000001537 neural effect Effects 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 9
- 238000010248 power generation Methods 0.000 claims description 6
- 235000009413 Ratibida columnifera Nutrition 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 230000035800 maturation Effects 0.000 claims description 3
- FNMKZDDKPDBYJM-UHFFFAOYSA-N 3-(1,3-benzodioxol-5-yl)-7-(3-methylbut-2-enoxy)chromen-4-one Chemical compound C1=C2OCOC2=CC(C2=COC=3C(C2=O)=CC=C(C=3)OCC=C(C)C)=C1 FNMKZDDKPDBYJM-UHFFFAOYSA-N 0.000 claims description 2
- 230000004807 localization Effects 0.000 claims description 2
- 238000002715 modification method Methods 0.000 claims description 2
- 238000003062 neural network model Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 239000013589 supplement Substances 0.000 claims description 2
- 241000510442 Ratibida peduncularis Species 0.000 claims 2
- 230000007812 deficiency Effects 0.000 abstract description 2
- 244000286916 Ratibida columnifera Species 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- 238000013277 forecasting method Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000003915 air pollution Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000009194 climbing Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
Classifications
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Photovoltaic Devices (AREA)
Abstract
The invention discloses a kind of photovoltaic power Forecasting Methodology based on mind evolutionary Optimization of Wavelet neutral net.Photovoltaic historical data and air quality data are obtained from photovoltaic power forecasting system database module；The forecast model of wavelet neural network is established to historical data, by introducing two new parameters of yardstick and translation, improves the Function approximation capabilities of forecast model；And the weights and threshold value of BP neural network are optimized using mind evolutionary, effectively solve the deficiency of photovoltaic power ultra-short term precision of prediction present in existing forecast model, be easily trapped into local minimum problem.It is improved simultaneously in modified weight link using momentum arithmetic method, so as to lift convergence rate；Rolling forecast is carried out using the data predicted as input data, so as to realize the multi-step prediction across time interval.For the Forecasting Methodology compared with conventional photovoltaic Forecasting Methodology, precision of prediction is higher, and convergence rate is faster.
Description
Technical field
The present invention relates to electric power system control and electric powder prediction, in particular it relates to which a kind of be based on mind evolutionary
The photovoltaic power Forecasting Methodology of Optimization of Wavelet neutral net.
Background technology
As what renewable and clean energy resource generated electricity greatly develops, photovoltaic generation is as renewable energy power generation the most ripe
One of mode, worldwide utilized extensively.It is larger by inside even from weather yet with photovoltaic output situation
And change is very fast, photovoltaic, which is contributed, has mutability and intermittence, and which increase the difficulty of grid-connected traffic control.Photovoltaic is contributed
Ultra-short term prediction refer to lie prostrate grid-connected power prediction from forecast moment to following interior focusing in 4 hours, every 15 minutes future positions,
Every 15 minutes rolling forecasts once, and will meet root-mean-square error≤10% of prediction.Therefore, Accurate Prediction photovoltaic is contributed to carrying
Have great importance for high security of system and stability and whole dispatching of power netwoks.
The widely used photovoltaic power generation output forecasting method of recent domestic can substantially be divided into two kinds：One kind is direct method
(statistic law) is mainly according to mathematical statistics prediction theory and method；Another kind is that indirect method (Physical) is to be based on photovoltaic generation
System physical electricity generating principle is predicted.What these methods had can not embody the nonlinear characteristic of photovoltaic output；Some needs Pang
Big database costs dearly as support and precision has much room for improvement；It is and more prominent with the seriousness of air quality problems
Go out, should be predicted air pollution index AQI as the factor for needing to consider.
For in theory, by choosing suitable network structure and node in hidden layer, BP neural network can be with any
Precision approaches given function or signal.But in actual applications, it is difficult to which the determination network structure of science, learning training are optimal
During flexible strategy, its BP algorithm there is convergence rate it is slow, be easily trapped into the inherent shortcomings such as local minimum, it is short-term negative so as to have impact on
The reliability and accuracy of lotus forecast.
The content of the invention
In order to overcome the existing technical problem of the above, the present invention provides a kind of photovoltaic output ultra-short term prediction method, and adopts
The weights and threshold value of BP neural network are optimized with mind evolutionary, effectively solves and exists in existing forecast model
Photovoltaic power ultra-short term precision of prediction deficiency, be easily trapped into local minimum problem.But evolution algorithm existing defects are restrained
Slow-footed defect.Wavelet analysis is combined with BP neural network after optimization, introduces two new parameters of yardstick and translation,
Make wavelet network that there is the higher free degree, so as to have more effective Function approximation capabilities, while adopted in modified weight link
It is improved with momentum arithmetic method, so as to lift convergence rate so that the model prediction accuracy is higher, and convergence rate is more
It hurry up.
Technical scheme is as follows：
A kind of photovoltaic power generation output forecasting method of the neural wavelet network based on mind evolutionary optimization, including following step
Suddenly：
(1) acquisition and pretreatment of data：Obtained from the data processing module of photovoltaic power forecasting system specify when
Between in the range of photovoltaic go out force data and air quality data, and thus obtain training set and forecast set, and be normalized.
(2) generation of initial weight and threshold value
(2.1) initial population, winning sub- population and interim sub- population are randomly generated.
(2.2) sub- population operation similartaxis：In the range of sub-group, the individual process for turning into victor and competing is called convergent.
One sub-group is during convergent, if no longer producing new victor, claims the sub-group ripe.When sub-group is ripe
When, the convergent process of the sub-group terminates, and goes to step 2.3, while supplements new population.
(2.3) sub- population operation dissimilation：Each sub-group carries out global competition, if the score of an interim sub-group is higher than certain
Individual ripe winning sub-group, the then interim sub-group that the winning sub-group is won substitute.Individual in former winning sub-group
It is released；If the score of a ripe interim sub-group is less than the score of any one winning sub-group, the interim subgroup
Body is gone out of use, and individual therein is released.The individual being released re-started in global scope search for and formed it is new interim
Colony.In whole solution space, each sub-group constantly detects point new in solution space to be competed as victor.
(2.4) when meeting iteration stopping condition, mind evolutionary terminates optimization process.Now, optimum individual is found
Output, according to coding rule, is parsed to the optimum individual searched out, so as to obtain the weights of corresponding neural wavelet network
And threshold value
(3) prediction wavelet neural network is created to be predicted
(3.1) threshold value and value information provided according to above-mentioned optimum individual creates neural wavelet network and it carried out just
Beginningization, the inputoutput data by the use of the training set data after normalization as neural wavelet network are excellent using additional momentum
Change further training wavelet neural network until convergence.
(3.2) the forecast set data after normalized are inputted to the above-mentioned wavelet neural network trained, to photovoltaic
Output is predicted, and is exported progress renormalization output to network and obtained final prediction result.
(3.3) it is the data obtained after prediction are corresponding as one of the prediction input value at next 15min time points, modification
Other forecast set input values, the neural network model trained using this carries out rolling forecast, until having predicted that following 4 is small
When interior every 15 minutes data point prediction.
(4) prediction data that this photovoltaic power generation output forecasting system is completed is transferred into power network dispatching system to use.
Further, the photovoltaic in the range of the specified time described in step (1) is contributed and air quality index data are
Refer in chronological sequence tactic one group of data of a certain number of constant duration：
Input=[input (t- (n+r) Δs t), input (t- (n+r-1) Δs t), input (t- (n+r-2) Δs
t),...,input(t-Δt)]
Wherein input=AQId,AQId-1,pd-1,pn-1,pn-2,pn-3, AQI in formuladFor same day air quality index, AQId-1
For the previous day air quality index, pd-1For the previous day synchronization photovoltaic output situation, pn-1, pn-2, pn-3Prediction is represented respectively
Point first three data point photovoltaic goes out force data；The then training set input and output of forecasting system are respectively：
Input：
Output：
The input and output of forecast set are respectively
Input：Pn_prediction=[AQId(t),AQId-1(t),pd-1(t),pn-1(t),pn-2(t),pn-3(t)]
Output：Tn_prediction=pn(d)
All data are as all converted to the number between [0,1] by the normalized, and the present invention uses minimum most
Small method normalized, functional form are：
xk=(xk-xmean)/(xmax-xmin)
Wherein xmeanFor the average of data sequence, xmaxAnd xminFor data sequence maximin, after normalized,
The difference of the order of magnitude between each dimension data is eliminated, avoids causing network because inputoutput data order of magnitude difference is larger
Predict that error is larger.In step 3.2, equally to be obtained truly using renormalization for predicting that obtained photovoltaic goes out force data
Value.
Further, different population is encoded in step 2 and takes mind evolutionary operation to include：
1 encodes and randomly generates initial population, because wavelet neural network structure has determined as 6-10-1, so power
Value/threshold value total number has determined as 81, for each weight threshold, between [0,1] be uniformly distributed produce it is multigroup with
Machine number, as initial weights and threshold value colony.
2 individual scoring function settings, the inverse of the mean square error of present invention selection training set is as each individual and son kind
The scoring function of group, function expression are：
Y in formulaiRepresent the network output valve of i-th of training sample, tiIndicate target output, p is number of training.
3 pairs of each populations are iterated convergent, alienation, selection operation generation new population, and scoring function in use 2 respectively
Assessed
4 optimization end conditions judge, judge whether Neural Network Optimization end condition meets, if optimization end condition is expired
Foot, then mind evolutionary optimization process terminates, and exports optimum individual, and is decoded according to coding rule, and it is small to produce nerve
The initial weight and threshold information of wave network；Otherwise, return and re-optimization is carried out to each population, optimal conditions are evolution iteration
Number reaches the upper limit or continuous some band mind evolutionaries optimize wooden plaque function value and do not changed.
The wavelet neural network built in step 3 is according to usable condition of the present invention, using 6-10-1 structures.Consider
Mexican Hat functions are the second dervative of Gauss functions, and it has good localization in time domain and frequency, used herein
Wavelet basis function Mexican Hat morther wavelet basic functions, its mathematical formulae are：
The additional guide vanes that neural wavelet network uses are optimized in step 3.1, because traditional neural wavelet network uses
The weights and wavelet basis function parameter of gradient modification method corrective networks, easily cause convergence rate slow.Increase momentum can be used
The method of item improves e-learning efficiency, and the weights and parameters revision formula for increasing momentum term are：
ak(i+1)=ak(i)+Δak(i+1)+k·(ak(i)-ak(i-1))
bk(i+1)=bk(i)+Δbk(i+1)+k·(bk(i)-bk(i-1))
In formulaFor wavelet neural network weights, ajFor the contraction-expansion factor of wavelet basis function, bjFor the flat of wavelet basis function
The factor is moved, k is momentum term learning rate.
Rolling forecast refers to input the predicted value of the future position as the input value of next future position in step 3.3,
Rolling forecast, the predicted value until having predicted all future positions in following 4 hours are carried out, prediction uses to be trained in step 3.2
Good neural wavelet network, being brought directly to forecast set input can show that forecast set exports.
The advantage of the invention is that：
First, colony is divided into winning sub-group and interim sub-group, the convergent and operation dissimilation defined on this basis,
Detected and developed respectively, in system operation, both functions are mutually coordinated and keep certain independence, are easy to distinguish
Improve efficiency.Meanwhile mind evolutionary can remember the evolution information of iteration population each time and individual, these information can be with
Convergent and alienation is instructed to be carried out towards favourable direction.MEA ability of searching optimum is extremely strong, can be concurrently big with higher efficiency
Scale searches for globally optimal solution, reduces the dependence to initial data accuracy, avoids system from occurring being absorbed in local minimum and show
As so that precision of prediction improves, and stability is strengthened.
2nd, neural wavelet network is built as initial value using the weight threshold after optimization, and entered by additional guide vanes
Row optimization so that forecast model is provided with more preferable Function approximation capabilities, and convergence rate is accelerated, while improves precision of prediction.
Brief description of the drawings
Fig. 1 is the photovoltaic forecasting system pie graph described in embodiment
Fig. 2 is mind evolutionary system construction drawing in embodiment
Fig. 3 is mind evolutionary flow chart in embodiment
Fig. 4 is the photovoltaic power forecast model figure based on neural wavelet network in embodiment
Fig. 5 is the overview flow chart of Forecasting Methodology in embodiment
Fig. 6 is using present example photovoltaic power generation output forecasting result figure
Embodiment
The present invention is further described with example below in conjunction with the accompanying drawings.
Fig. 1 is the overview flow chart of Forecasting Methodology in embodiment, is the foundation of prediction data first, is predicted in the present invention
In model, predict the data source taken for SCADA/EMS system acquisitions to history photovoltaic go out force data and measurement module and survey
The history air quality AQI indexes measured.The initial data input that generation prediction needs after data processing module arrangement is defeated
Forecasting system is sent into after going out data format.Data after predicted resume module will feed back to data processing module and be rolled after processing
The dynamic next round that enters is predicted, and prediction result is submitted into control centre and is scheduled for personnel's reference.
Further：Fig. 2, Fig. 3 are respectively mind evolutionary system construction drawing and algorithm flow chart.Mind evolutionary
Basic ideas in solution space at random generation certain scale individual, according to score (correspond to genetic algorithm in adaptation
Functional value is spent, characterizes individual to the adaptability of environment) search out several winning individuals and temporary individuals of highest scoring.Point
Not centered on these winning individuals and temporary individual, some new individuals are produced around each individual, it is some so as to obtain
Individual winning sub-group and interim sub-group.Operation similartaxis is performed inside each sub-group, until the subgroup body maturation, and with this
Score of the score of optimum individual (i.e. center) as the sub-group in sub-group.After the body maturation of subgroup, by each sub-group
Score is puted up on global advertisement plate, and operation dissimilation is performed between sub-group, is completed between winning sub-group and interim sub-group
Replace, the process that individual discharges in discarded, sub-group, so as to calculate global optimum's individual and its score.
Fig. 4 is the photovoltaic power forecast model figure based on neural wavelet network, the wavelet neural network knot that the present invention uses
Structure is 6-10-1 structures, and neural network topology structure is as shown in the figure.Inputoutput data is as schemed, wherein AQIdFor same day air
Performance figure, AQId-1For the previous day air quality index, pd-1For the previous day synchronization photovoltaic output situation, pn-1, pn-2,
pn-3Represent that future position first three data point photovoltaic goes out force data respectively.Output is that the photovoltaic of prediction data point is contributed.Hidden layer
The transmission function of node selects wavelet basis function Mexican Hat morther wavelet basic functions, and its mathematical formulae is：
Fig. 5 is the overview flow chart of Forecasting Methodology.Its mentality of designing is mainly using mind evolutionary to wavelet neural
The initial weight and threshold value of network optimize.First, according to the topological structure of wavelet neural network, using weights and threshold value as
Solution, space encoder is mapped to by solution space, each corresponding individual of coding.Then calculated by scoring function using mind-evolution
Method, by continuous iteration, optimum individual is exported, and in this, as initial weight and threshold value, carry out building neural wavelet network simultaneously
It is trained, until convergence.Finally photovoltaic is contributed with the network model trained and is predicted.
Principle to illustrate the invention, simulating, verifying analysis is carried out with a photovoltaic measured data.The data taken are certain
Photovoltaic plant 15min time interval measured datas.
1st, input data：Input data form is as shown in table 1
Table 1
Optimize threshold value result through mind evolutionary
Neural wavelet network is built by the use of optimal solution obtained above as initial weight and threshold value
Neural wavelet network parameter is arranged to：
Input node number：6；
Output node number：1；
Implicit node number：10；
Weights and threshold learning speed：0.01；
Scale parameter learning rate：0.001；
Network iterations：1000；
Input data is (part) as follows
By forecast set input model, prediction result is drawn, by result renormalization, obtains predicted value, prediction result such as table 2
It is shown
Table 2
Normalize root-mean-square error:
Prediction effect figure is as shown in fig. 6, with reference to table 2 and Fig. 6, it can be seen that this simulation model simulation result realizes list substantially
Data point tolerance is no more than 10% precision of prediction, one day overall average relative error 5% or so, and meets to carry in claims
The root-mean-square error for the prediction arrived≤10% requires that prediction effect is preferable.When photovoltaic contributes climbing decline and fluctuation, do not go out
Local optimum and the situation of over-fitting are now absorbed in, tracking prediction photovoltaic output can be changed well, forecast model tool of the present invention
There is good practical value.
Finally it should be noted that：The preferred embodiments of the present invention are the foregoing is only, are not intended to limit the invention,
Although the present invention is described in detail with reference to the foregoing embodiments, for those skilled in the art, it still may be used
To be modified to the technical scheme described in foregoing embodiments, or equivalent substitution is carried out to which part technical characteristic.
All any modification, equivalent substitution and improvements within present disclosure and principle, made etc., it should be included in the present invention's
Within protection domain.
Claims (6)
- A kind of 1. photovoltaic power Forecasting Methodology based on mind evolutionary Optimization of Wavelet neutral net, it is characterised in that including Following steps：(1) acquisition and pretreatment of data：The time model specified is obtained from the data processing module of photovoltaic power forecasting system Enclose interior photovoltaic and go out force data and air quality data, and thus obtain training set and forecast set, and be normalized；(2) initial weight and threshold value are generated；(3) prediction wavelet neural network is created to be predicted；(4) prediction data that this photovoltaic power generation output forecasting system is completed is transferred into power network dispatching system to use.
- A kind of 2. photovoltaic power prediction side based on mind evolutionary Optimization of Wavelet neutral net according to claim 1 Method, it is characterised in that step (2) comprises the following steps：(2.1) initial population, winning sub- population and interim sub- population are randomly generated；(2.2) sub- population operation similartaxis：In the range of sub-group, the individual process for turning into victor and competing is called convergent；One Sub-group is during convergent, if no longer producing new victor, claims the sub-group ripe；When sub-group maturation, The convergent process of the sub-group terminates, and goes to step 2.3, while supplements new population；(2.3) sub- population operation dissimilation：Each sub-group carries out global competition, if the score of an interim sub-group be higher than some into Ripe winning sub-group, the then interim sub-group that the winning sub-group is won substitute；Individual in former winning sub-group is released Put；If the score of a ripe interim sub-group is less than the score of any one winning sub-group, the interim sub-group quilt Discarded, individual therein is released；The individual being released re-starts in global scope searches for and is formed new interim colony； In whole solution space, each sub-group constantly detects point new in solution space to be competed as victor；(2.4) when meeting iteration stopping condition, mind evolutionary terminates optimization process；Now, optimum individual output is found, According to coding rule, the optimum individual searched out is parsed, so as to obtain the weights and threshold of corresponding neural wavelet network Value.
- A kind of 3. photovoltaic power prediction side based on mind evolutionary Optimization of Wavelet neutral net according to claim 1 Method, it is characterised in that step (3) includes：(3.1) threshold value and value information provided according to above-mentioned optimum individual creates neural wavelet network and it is carried out initially Change, the inputoutput data by the use of the training set data after normalization as neural wavelet network, optimized using additional momentum Further wavelet neural network is trained until convergence；(3.2) the forecast set data after normalized are inputted to the above-mentioned wavelet neural network trained, photovoltaic is contributed It is predicted, and network is exported and carries out renormalization output, obtains final prediction result；(3.3) using the data obtained after prediction as one of the prediction input value at next 15min time points, amendment it is corresponding other Forecast set input value, the neural network model trained using this carries out rolling forecast, until having predicted in following 4 hours The data of every 15 minutes future positions, so as to realize the multi-step prediction across time interval；Wherein, the inputoutput data refers in chronological sequence tactic one group of number of a certain number of constant duration According to：Input=[input (t- (n+r) Δs t), input (t- (n+r-1) Δs t), input (t- (n+r-2) Δ t) ..., Input (t- Δs t)] wherein input=AQId,AQId-1,pd-1,pn-1,pn-2,pn-3, AQI in formuladFor same day air quality index, AQId-1For the previous day air quality index, pd-1For the previous day synchronization photovoltaic output situation, pn-1, pn-2, pn-3Represent respectively Future position first three data point photovoltaic goes out force data；The then training set input and output of forecasting system are respectively：Input：Output：The input and output of forecast set are respectivelyInput：Pn_prediction=[AQId(t),AQId-1(t),pd-1(t),pn-1(t),pn-2(t),pn-3(t)]Output：Tn_prediction=pn(d)All data are as all converted to the number between [0,1] by the normalized, and the present invention is using minimum minimum method Normalized, functional form are：xk=(xk-xmean)/(xmax-xmin)Wherein xmeanFor the average of data sequence, xmaxAnd xminFor data sequence maximin, after normalized, eliminate The difference of the order of magnitude between each dimension data, avoids causing neural network forecast because inputoutput data order of magnitude difference is larger Error is larger；In step 3.2, equally actual value is obtained using renormalization for predicting that obtained photovoltaic goes out force data.
- A kind of 4. photovoltaic power prediction side based on mind evolutionary Optimization of Wavelet neutral net according to claim 2 Method, it is characterised in that encoded in the step 2 to different population and take mind evolutionary operation to include：1) encode and randomly generate initial population, because wavelet neural network structure has determined as 6-10-1, so weights/threshold Value total number has determined as 81, and for each weight threshold, multigroup random number is produced to be uniformly distributed between [0,1], As initial weights and threshold value colony；2) individual scoring function setting, the inverse of the mean square error of present invention selection training set is as each individual and sub- population Scoring function, function expression are：Y in formulaiRepresent the network output valve of i-th of training sample, tiIndicate target output, p is number of training；3) convergent, alienation, selection operation generation new population are iterated respectively to each population, and scoring function enters in use 2 Row is assessed4) optimize end condition to judge, judge whether Neural Network Optimization end condition meets, if optimization end condition meets, Mind evolutionary optimization process terminates, and exports optimum individual, and is decoded according to coding rule, produces neural wavelet network Initial weight and threshold information；Otherwise, return and re-optimization is carried out to each population, optimal conditions are that evolution iterations reaches Optimize wooden plaque function value to the upper limit or continuous some band mind evolutionaries not change.
- A kind of 5. photovoltaic power prediction side based on mind evolutionary Optimization of Wavelet neutral net according to claim 3 Method, it is characterised in that the wavelet neural network built in the step 3 is according to usable condition of the present invention, using 6-10-1 structures； In view of the second dervative that Mexican Hat functions are Gauss functions, it has good localization in time domain and frequency, this Literary grace wavelet basis function Mexican Hat morther wavelet basic functions, its mathematical formulae are：The additional guide vanes that neural wavelet network uses are optimized in the step 3.1, because traditional neural wavelet network uses The weights and wavelet basis function parameter of gradient modification method corrective networks, easily cause convergence rate slow；Increase momentum can be used The method of item improves e-learning efficiency, and the weights and parameters revision formula for increasing momentum term are：ak(i+1)=ak(i)+Δak(i+1)+k·(ak(i)-ak(i-1))bk(i+1)=bk(i)+Δbk(i+1)+k·(bk(i)-bk(i-1))In formulaFor wavelet neural network weights, ajFor the contraction-expansion factor of wavelet basis function, bjFor wavelet basis function translation because Son, k are momentum term learning rate.
- A kind of 6. photovoltaic power prediction side based on mind evolutionary Optimization of Wavelet neutral net according to claim 3 Method, it is characterised in that rolling forecast refers to using the predicted value of the future position as the defeated of next future position in the step 3.3 Enter value input, carry out rolling forecast, the predicted value until having predicted all future positions in following 4 hours, prediction uses step The neural wavelet network trained in 3.2, being brought directly to forecast set input can show that forecast set exports.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710811772.7A CN107609774B (en) | 2017-09-11 | 2017-09-11 | Photovoltaic power prediction method for optimizing wavelet neural network based on thought evolution algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710811772.7A CN107609774B (en) | 2017-09-11 | 2017-09-11 | Photovoltaic power prediction method for optimizing wavelet neural network based on thought evolution algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107609774A true CN107609774A (en) | 2018-01-19 |
| CN107609774B CN107609774B (en) | 2023-12-08 |
Family
ID=61063013
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710811772.7A Active CN107609774B (en) | 2017-09-11 | 2017-09-11 | Photovoltaic power prediction method for optimizing wavelet neural network based on thought evolution algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107609774B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165792A (en) * | 2018-09-14 | 2019-01-08 | 国网天津市电力公司 | A kind of photovoltaic short-term output power prediction technique based on SOA-WNN |
| CN109255728A (en) * | 2018-09-27 | 2019-01-22 | 上海电力学院 | The photovoltaic generation power neural network prediction method of chaos phase space optimal reconfiguration |
| CN109818775A (en) * | 2018-12-14 | 2019-05-28 | 南昌大学 | Short-term network method for predicting based on adaptive differential evolution algorithm Optimization of Wavelet neural network |
| CN110927584A (en) * | 2019-12-09 | 2020-03-27 | 天津市捷威动力工业有限公司 | Neural network-based battery life extension prediction method |
| CN111292124A (en) * | 2020-01-18 | 2020-06-16 | 河北工程大学 | Water demand prediction method based on optimized combined neural network |
| CN114154583A (en) * | 2021-12-08 | 2022-03-08 | 深圳博沃智慧科技有限公司 | Water quality prediction method of wavelet analysis coupling LSTM neural network |
| CN114707743A (en) * | 2022-04-15 | 2022-07-05 | 西安邮电大学 | Air quality prediction method and system based on adaptive gate control circulation neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105139264A (en) * | 2015-06-06 | 2015-12-09 | 安徽工程大学 | Photovoltaic generation capacity prediction method based on particle swarm algorithm wavelet neural network |
| CN105375507A (en) * | 2015-07-10 | 2016-03-02 | 华北电力大学（保定） | Power two-stage interactive optimization scheduling system of virtual power plant in haze environment |
| CN105631517A (en) * | 2015-12-17 | 2016-06-01 | 河海大学 | Photovoltaic power generation power short term prediction method based on mind evolution Elman neural network |
| WO2016210102A1 (en) * | 2015-06-23 | 2016-12-29 | Qatar Foundation For Education, Science And Community Development | Method of forecasting for solar-based power systems |
-
2017
- 2017-09-11 CN CN201710811772.7A patent/CN107609774B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105139264A (en) * | 2015-06-06 | 2015-12-09 | 安徽工程大学 | Photovoltaic generation capacity prediction method based on particle swarm algorithm wavelet neural network |
| WO2016210102A1 (en) * | 2015-06-23 | 2016-12-29 | Qatar Foundation For Education, Science And Community Development | Method of forecasting for solar-based power systems |
| CN105375507A (en) * | 2015-07-10 | 2016-03-02 | 华北电力大学（保定） | Power two-stage interactive optimization scheduling system of virtual power plant in haze environment |
| CN105631517A (en) * | 2015-12-17 | 2016-06-01 | 河海大学 | Photovoltaic power generation power short term prediction method based on mind evolution Elman neural network |
Non-Patent Citations (2)
| Title |
|---|
| 刘爱国等: "基于遗传算法小波神经网络的光伏微网发电预测", 《电测与仪表》 * |
| 杨文海等: "计及雾霾影响的短期光伏出力预测和负荷预测", 《中国电力》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165792A (en) * | 2018-09-14 | 2019-01-08 | 国网天津市电力公司 | A kind of photovoltaic short-term output power prediction technique based on SOA-WNN |
| CN109255728A (en) * | 2018-09-27 | 2019-01-22 | 上海电力学院 | The photovoltaic generation power neural network prediction method of chaos phase space optimal reconfiguration |
| CN109255728B (en) * | 2018-09-27 | 2021-09-07 | 上海电力学院 | Photovoltaic power generation power neural network prediction method based on chaotic phase space optimization reconstruction |
| CN109818775A (en) * | 2018-12-14 | 2019-05-28 | 南昌大学 | Short-term network method for predicting based on adaptive differential evolution algorithm Optimization of Wavelet neural network |
| CN109818775B (en) * | 2018-12-14 | 2021-09-28 | 南昌大学 | Short-term network flow prediction method |
| CN110927584A (en) * | 2019-12-09 | 2020-03-27 | 天津市捷威动力工业有限公司 | Neural network-based battery life extension prediction method |
| CN110927584B (en) * | 2019-12-09 | 2022-05-10 | 天津市捷威动力工业有限公司 | Neural network-based battery life extension prediction method |
| CN111292124A (en) * | 2020-01-18 | 2020-06-16 | 河北工程大学 | Water demand prediction method based on optimized combined neural network |
| CN114154583A (en) * | 2021-12-08 | 2022-03-08 | 深圳博沃智慧科技有限公司 | Water quality prediction method of wavelet analysis coupling LSTM neural network |
| CN114154583B (en) * | 2021-12-08 | 2024-05-24 | 深圳博沃智慧科技有限公司 | Water quality prediction method of wavelet analysis coupled LSTM neural network |
| CN114707743A (en) * | 2022-04-15 | 2022-07-05 | 西安邮电大学 | Air quality prediction method and system based on adaptive gate control circulation neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107609774B (en) | 2023-12-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107609774A (en) | A kind of photovoltaic power Forecasting Methodology based on mind evolutionary Optimization of Wavelet neutral net | |
| CN106532778B (en) | Method for calculating maximum access capacity of distributed photovoltaic grid connection | |
| CN102129511B (en) | System for forecasting short-term wind speed of wind power station based on MATLAB | |
| CN110474339B (en) | Power grid reactive power control method based on deep power generation load prediction | |
| CN110516840A (en) | Short term prediction method based on the wind light generation power output for improving random forest method | |
| CN104077632B (en) | A kind of wind electric field power prediction method based on deep neural network | |
| CN109978284B (en) | Photovoltaic power generation power time-sharing prediction method based on hybrid neural network model | |
| CN103324980A (en) | Wind power station wind speed prediction method | |
| CN102102626A (en) | Method for forecasting short-term power in wind power station | |
| CN104732300B (en) | A kind of neutral net wind power short term prediction method theoretical based on Fuzzy divide | |
| CN114565259B (en) | Distribution network line loss calculation method for improving sparrow search optimization BP neural network | |
| CN105426956A (en) | Ultra-short-period photovoltaic prediction method | |
| CN105631517A (en) | Photovoltaic power generation power short term prediction method based on mind evolution Elman neural network | |
| CN105389634A (en) | Combined short-term wind power prediction system and method | |
| CN110866640A (en) | Power load prediction method based on deep neural network | |
| CN109242136A (en) | A kind of micro-capacitance sensor wind power Chaos-Genetic-BP neural network prediction technique | |
| CN105787592A (en) | Wind turbine generator set ultra-short period wind power prediction method based on improved RBF network | |
| CN113555881A (en) | Wind power plant frequency modulation fan sequencing method | |
| CN101719193B (en) | Method for forecasting service life of brake | |
| CN107358006A (en) | A kind of Lorenz disturbance wind speed forecasting methods based on principal component analysis | |
| CN113095477A (en) | Wind power prediction method based on DE-BP neural network | |
| CN101710062B (en) | Forecast method of friction coefficient of brake boot | |
| CN112686472A (en) | Power prediction method for distributed photovoltaic equivalent power station | |
| CN116415708B (en) | Power grid robust planning method considering confidence level | |
| Cao et al. | Solar energy forecasting in short term based on the ASO-BPNN model |
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 |