CN107169683A - A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient - Google Patents
A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient Download PDFInfo
- Publication number
- CN107169683A CN107169683A CN201710526959.2A CN201710526959A CN107169683A CN 107169683 A CN107169683 A CN 107169683A CN 201710526959 A CN201710526959 A CN 201710526959A CN 107169683 A CN107169683 A CN 107169683A
- Authority
- CN
- China
- Prior art keywords
- weight coefficient
- prediction
- forecast model
- individual
- 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.)
- Pending
Links
- 238000013277 forecasting method Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000000034 method Methods 0.000 claims abstract description 15
- 238000004458 analytical method Methods 0.000 claims abstract description 11
- 230000005855 radiation Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000010219 correlation analysis Methods 0.000 claims description 6
- 238000000513 principal component analysis Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 abstract description 3
- 230000000052 comparative effect Effects 0.000 description 7
- 239000011159 matrix material Substances 0.000 description 6
- 238000011161 development Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000007704 transition 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
- G06Q10/06375—Prediction of business process outcome or impact based on a proposed change
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/067—Enterprise or organisation modelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Theoretical Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The step of a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient, this method, includes:A variety of Individual forecast models are set up using comprising the historical data nearest from prediction period, match value of each Individual forecast model to forecast sample point is obtained;Weight coefficient of each Individual forecast model in each forecast sample point is calculated by gray relative analysis method;BPNN network models are obtained to the match value of each forecast sample point and the training fitting of corresponding weight coefficient using each Individual forecast model;The single power prediction value of each Individual forecast model is worth to by the newest meteorological element prediction of prediction period, variable weight coefficient when being calculated using BPNN network models finally calculates the power prediction value of prediction period weighting;Above step is circulated, the power prediction value of prediction period is constantly updated.Compared with prior art, the weight coefficient real-time change of combination forecasting of the present invention, has the advantages that precision of prediction is high.
Description
Technical field
The present invention relates to generation of electricity by new energy and technical field of power systems, more particularly, to a kind of the grid-connected of variable weight coefficient
Photovoltaic plant short term power combination forecasting method.
Background technology
In recent years, solar energy development utilizes scale rapid expansion, and technological progress and industrial upgrading are accelerated, and cost significantly drops
It is low, it has also become the key areas of global energy transition." 12 " period, China's photovoltaic industry system constantly improve, technology is entered
Step is notable, and photovoltaic manufacture and application scale are at the forefront in the world.Solar energy thermal-power-generating technical research and equipment manufacturing obtain larger
Progress, is completed commercialization experimental power station, tentatively possesses large-scale development condition.By the end of the end of the year 2014, photovoltaic generation tires out
28,050,000 kilowatts of installed capacity is counted, " 12 " object of planning is fulfiled ahead of schedule, shows that China's photovoltaic market has contained huge hair
Open up potentiality.
" 13 " by be solar energy industry development critical period, basic task is industrial upgrading, reduces cost, expand
Using self market-oriented sustainable development independent of public subsidies being realized, as realizing the year two thousand twenty and the year two thousand thirty non-fossil energy
The important force that primary energy consumes the target of proportion 15% and 20% is accounted for respectively.By the end of the year 2016, China's photovoltaic generation is increased newly
34,540,000 kilowatts of installed capacity, adds up 77,420,000 kilowatts of installed capacity, and newly-increased and accumulative installed capacity is the whole world first.Its
In, photovoltaic plant adds up 67,100,000 kilowatts of installed capacity, accumulative 10,320,000 kilowatts of the installed capacity of distribution.Annual generated energy 662
Hundred million kilowatt hours, account for the 1% of the annual gross generation of China.Because exerting oneself for photovoltaic plant is easily influenceed by weather element, tool
The features such as having uncertain and unstability, will can be to the peace of power system with the continuous improvement of grid-connected permeability
Full stable operation produces certain influence, so accurate in the urgent need to being carried out to power output of the photovoltaic plant within following a period of time
Really prediction.And for Individual forecast model common at present, the difficulty for improving its precision of prediction is larger, and each Individual forecast
The degree of accuracy of model and applicability are different, and prediction exists uncertain.Probability is a kind of probabilistic mode of expression, and
Combined prediction (or DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM) is then an effective way for obtaining probability forecast.
In the research for the meteorologic factor exerted oneself at present for analyzing influence photovoltaic, the meteorology exerted oneself due to influence photovoltaic plant
Factor is complicated and changeable, and actual Changes in weather situation is also much more complex, and is difficult to pair to enter to photovoltaic related meteorologic factor of exerting oneself
Row analysis comprehensively, at present common photovoltaic forecast model be mostly to consider a small amount of meteorologic factor, such as with solar irradiance with
Temperature etc. is input variable, and seldom consider to introduce the associated environmental impacts such as air articulation index, sunshine time, the temperature difference because
Element.Above-mentioned meteorologic factor is all taken into account when modeling, can equally increase the complexity of forecast model, moreover each gas
As being also the presence of multicollinearity relation between factor, it is unfavorable for the foundation of forecast model, so in when of setting up forecast model institute
The weather environment key element of selection and how to reduce raising of the multicollinearity relation between meteorologic factor for model prediction accuracy
Seem and be even more important.
When setting up model in addition to above-mentioned considered meteorologic factor condition, the selection of model algorithm is for prediction essence
No less important for degree, is before based on Individual forecast model, and for Individual forecast mostly for predicting power of photovoltaic plant
Model, the difficulty for improving its precision of prediction is larger, and the degree of accuracy of each Individual forecast model and applicability are different, and prediction is deposited
In uncertainty.
Combined prediction common at present is usually one kind to the progress that predicts the outcome obtained by a variety of Individual forecast models
Combination, but different predicting the outcome for single model is different certainly, this is accomplished by each Individual forecast model
Output result determines a weights, and the forecast model of equal weight is identical due to the weight of Individual forecast model prediction result, and this is just
Relatively large deviation can occur due to predicting the outcome for some Individual forecast model, predicting the outcome for influence combination forecasting is chosen
The main purpose of appropriate weight is to eliminate Individual forecast method relatively large deviation that may be present, the accuracy of raising prediction.
For these reasons, traditional photovoltaic power Forecasting Methodology is difficult to meet photovoltaic plant and power system practical, letter
The requirement of list and predictive ability.
The content of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide a kind of more accurately variable weight
The grid-connected photovoltaic power station short term power combination forecasting method of weight coefficient.
The purpose of the present invention can be achieved through the following technical solutions:
The step of a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient, this method, includes:
S1, using comprising the historical data nearest from prediction period a variety of Individual forecast models are set up, obtain each single
Match value of the forecast model to forecast sample point;
S2, weight coefficient of each Individual forecast model in each forecast sample point is calculated by gray relative analysis method;
The fitting of S3, each Individual forecast model respectively obtained using step S1 and step S2 to each forecast sample point
Value and the training fitting of corresponding weight coefficient obtain BPNN network models;
S4, the single power prediction for being worth to by the prediction of prediction period newest meteorological element each Individual forecast model
Value, the BPNN network models then obtained using step S3 calculate when variable weight of each Individual forecast model in prediction period
Coefficient, using obtained single power prediction value and when variable weight coefficient calculate the power prediction value of prediction period;
S5, circulation step S1~S4, constantly update the power prediction value of prediction period.
The step S1 is specifically included:
S11, acquisition photovoltaic plant the history meteorological element data nearest from prediction period and history photovoltaic power output number
According to as sample data, the sample data includes training sample and forecast sample;
S12, the training sample data obtained using step S11, fitting is trained by minimizing training error, is set up
A variety of Individual forecast models, obtain match value of each Individual forecast model to forecast sample.
The step S2 is specifically included:Each Individual forecast model is calculated in forecast sample point by gray relative analysis method
Grey incidence coefficient between the match value and observation at place, regard obtained grey incidence coefficient as correspondence Individual forecast model
Weight coefficient.
The step S3 is specifically included:Plan of each Individual forecast model that step S1 is obtained to each forecast sample point
Conjunction value is as input, and each corresponding weight coefficient of Individual forecast model that step S2 is obtained is as output, and training is obtained
BPNN network models.
The step S4 is specifically included:
The newest meteorological element predicted value of S41, acquisition prediction period;
S42, when going out using the obtained newest meteorological element predictor calculations of step S41 each Individual forecast model to prediction
The single power prediction value of section;
In S43, the BPNN network models for obtaining the single power prediction value input step S3 that step S42 is obtained, obtain
When variable weight coefficient of each Individual forecast model in prediction period;
S44, for each prediction period, the single power prediction value that each Individual forecast model is obtained is multiplied by corresponding
When variable weight coefficient after again all superposition, obtain the power prediction value of the prediction period.
A variety of Individual forecast models include:Supporting vector machine model, grey correlation analysis combination supporting vector machine mould
Type and principal component analysis combination supporting vector machine model.
The meteorological element includes:Total solar radiation on hour horizontal plane, outside hour ground when horizontal plane global radiation, sunshine
Number, hourly average temperature, the hour highest temperature, the hour lowest temperature, relative humidity, station pressure and wind speed.
Compared with prior art, the present invention has advantages below:
1st, the meteorologic factors of a variety of influence photovoltaic plant power outputs are taken into account, respectively according to PCA and
Gray relative analysis method carries out dimension-reduction treatment to meteorologic factor, reduces the meteorologic factor number considered in Individual forecast model,
The complexity of traditional Individual forecast model is simplified while considering meteorologic factor.
2nd, due to the uncertainty of climatic factor, photovoltaic power prediction is inevitably present uncertainty, and probability is one
The probabilistic mode of expression is planted, and combined prediction or DATA PROCESSING IN ENSEMBLE PREDICTION SYSTEM are then an effective way for obtaining probability forecast, this hair
The thought of the combined prediction of bright proposition, not only conforms with the reality of Meteorological Science and photovoltaic generation, is also beneficial to power system peak regulation
And Electric Power Network Planning.
3rd, grey incidence coefficient then is passed through into BPNN network moulds as the weights of various Individual forecast models in combined prediction
Variable weight coefficient combination forecasting when type is set up, realizes different Individual forecast models different power at different predicted time nodes
The change of weight coefficient, can make up the big shortcoming of Individual forecast model predictive error, reduce forecast model and extreme error occur
Probability, improve the precision of prediction of model.
4th, the algorithm structure of the combination forecasting is simple, after the completion of BPNN network models are set up, it is only necessary to input single
The variable weight coefficient when predicted value of forecast model is with regard to that can obtain, can efficiently meet the prediction requirement to photovoltaic plant power output.
Brief description of the drawings
Fig. 1 is the grid-connected photovoltaic power station short term power combination forecasting method flow chart of variable weight coefficient of the present invention;
Fig. 2 is the inventive method logical schematic;
Fig. 3 is that model of the present invention sets up process schematic;
Fig. 4 is the curve map of power observation and the power prediction value of 3 kinds of Individual forecast models;
Fig. 5 is the residual error histogram of error that predicts the outcome of SVM models;
Fig. 6 is the residual error histogram of error that predicts the outcome of GRA-SVM models;
Fig. 7 is the residual error histogram of error that predicts the outcome of PCA-SVM models;
Fig. 8 is the power prediction value of variable weight coefficient combination forecasting of the present invention and the curve map of observation;
Fig. 9 is the residual error histogram of error that predicts the outcome of variable weight coefficient combination forecasting of the present invention.
Embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.The present embodiment is with technical solution of the present invention
Premised on implemented, give detailed embodiment and specific operating process, but protection scope of the present invention is not limited to
Following embodiments.
This method is proposed a kind of on the basis of the Individual forecast model and equal weight forecast model of completion is had built up
Variable weight coefficient combination forecasting realizes prediction to photovoltaic plant power output, can be to the weights of a variety of Individual forecast models
Coefficient is adjusted in real time.The method of proposition obtains variable weight coefficient combination die by grey correlation analysis combination BP neural network
Type, realize with supporting vector machine model (SVM), grey correlation analysis combination supporting vector machine model (GRA-SVM) and it is main into
Analysis combination supporting vector machine model (PCA-SVM) three kinds of Individual forecast models change in different predicted time point weight coefficients
Power prediction.SVM, GRA-SVM and PCA-SVM tri- is set up according to history meteorological element data and photovoltaic power observation first
Individual forecast model is planted, the weight coefficient of three kinds of Individual forecast models is drawn using grey correlation analysis, then basis is obtained
BP neural network model is set up in the match value training that weight coefficient combines three kinds of Individual forecast models;According to newest meteorological element
Predicted value obtains the single power prediction value of three kinds of Individual forecast models, calculates single pre- by the BP neural network model of foundation
The when variable weight coefficient of model is surveyed, single power prediction value when superposition is multiplied by after variable weight coefficient obtains final power prediction
Value.
As shown in Figures 1 to 3, a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient, this method
The step of include:
S1, using comprising the historical data nearest from prediction period a variety of Individual forecast models are set up, obtain each single
Forecast model specifically includes following steps to the match value of forecast sample point:
S11, data acquisition and pretreatment:
Obtain the photovoltaic plant history meteorological element data nearest from prediction period and history photovoltaic power output data are made
For sample data, meteorological element include total solar radiation on hour horizontal plane, the outer horizontal plane global radiation in hour ground (astronomical to radiate),
Sunshine time, hourly average temperature, the hour highest temperature, the hour lowest temperature, relative humidity, station pressure and wind speed, according to
Acquired meteorological element data calculate a hour temperature difference, articulation index etc., and the data to acquisition are standardized, sample
Notebook data includes training sample and forecast sample;
S12, the training sample data obtained using step S11, fitting is trained by minimizing training error, is set up
3 kinds of Individual forecast models, obtain match value of each Individual forecast model to forecast sample:
S121, set up SVM models:
Support vector regression is a kind of intelligence machine learning algorithm, is applied to solve nonlinear regression problem, this
Statistical learning algorithm is to be based on structural risk minimization for principle, can realize preferably fitting in the case of less sample
, at present, in grid-connected photovoltaic power station power output forecasting problem, there is power station history power output data deficiencies and meteorology in effect
The problems such as data is not comprehensive, chooses supporting vector machine model very suitable for photovoltaic power generation output forecasting as Individual forecast model;
S122, set up GRA-SVM models:According to grey correlation theory, calculate each meteorologic factor of history meteorological data with
Between photovoltaic power output in the size of the degree of association, the present embodiment, choose the larger meteorologic factor of preceding 5 associations angle value and be used as branch
Hold the mode input of vector machine;
S123, set up PCA-SVM models:According to Theory of Principal Components Analysis, in the present embodiment, to 8 original meteorological variables
Dimension-reduction treatment is carried out, synthesis draws 3 principal component factor amounts as the input of supporting vector machine model;
S2, match value and observation of each Individual forecast model at forecast sample point are calculated by gray relative analysis method
Grey incidence coefficient between value, regard obtained grey incidence coefficient as the weight coefficient for corresponding to Individual forecast model;If v-th
Forecast model match value p at w-th of forecast sample pointv(w) the grey incidence coefficient ξ between observationv(w), in the present embodiment
In, match value of the combination forecasting in w-th of period of forecast sample collection is:
P (w)=ξ1(w)p1(w)+ξ2(w)p2(w)+ξ3(w)p3(w);
S3, each Individual forecast model for obtaining step S1, as input, are incited somebody to action the match value of each forecast sample point
Each corresponding weight coefficient of Individual forecast model that step S2 is obtained obtains BPNN network models as output, training;
S4, the power prediction value for going out by the newest meteorological element predictor calculation of prediction period prediction period:
The newest meteorological element predicted value of S41, acquisition prediction period;
S42, when going out using the obtained newest meteorological element predictor calculations of step S41 each Individual forecast model to prediction
The single power prediction value p of sectionv;
In S43, the BPNN network models for obtaining the single power prediction value input step S3 that step S42 is obtained, obtain
When variable weight coefficient λ of each Individual forecast model in prediction periodv;
S44, for each prediction period, the single power prediction value that each Individual forecast model is obtained is multiplied by corresponding
When variable weight coefficient after again all superposition set up variable weight coefficient combination forecasting, obtain the power prediction of the prediction period
Value:
P=λ1p1+λ2p2+λ3p3;
S5, circulation step S1~S4, constantly update the power prediction value of prediction period.
The specific calculation procedure of grey incidence coefficient is further analyzed as follows in above-mentioned steps S2:
S21, reference sequences and comparative sequences selection:
X is set first as gray system factor set, there is xi∈ X are factor of system.xiK-th of data point of sequence is expressed as xi
(k), wherein k=1,2,3 ..., n, n are key element sample data number.The reference sequences for defining factor of system are X0={ x0(1),
x0(2),x0(3),…,x0(n)};
The sequence as comparative sequences of other index compositions being included in index system table, are designated as Xi:
Xi={ xi(1),xi(2),xi(3),…,xi(n)}
Wherein, i=1,2,3 ..., m, m are the number of comparative sequences;
It is reference sequences to choose photovoltaic plant power output, and meteorologic factor is comparative sequences, carries out meteorologic factor and photovoltaic
The calculating of grey incidence coefficient between station output;
S22, nondimensionalization processing:
In order to eliminate influence, it is necessary to right resulting on grey correlation analysis due to dimension disunity between each sequence
Each Meteorological series data of input are handled, and reject unusual and incomplete data, and remainder data is normalized place
Reason, normalization formula is as follows:
In formula, xmax、xminMaximum and minimum value respectively in specified sequence;
S23, calculating grey incidence coefficient:
Each index for calculating meteorologic factor comparative sequences using following formula corresponds to the pass of each index of power output reference sequences
Contact number:
In formula, ρ is resolution ratio, is changed between 0~1, and it is 0.5 typically to take ρ, and the uncertain of resolution ratio can cause to close
The unstable of number result of calculation is contacted, resolution ratio ρ takes different values to be likely to result in each comparative sequences and reference sequences grey
The tagmeme of the degree of association changes;
S24, grey relational grade are calculated:
Due to incidence coefficient ξi(k) number is more, is not easy to compare, therefore by ordered series of numbers XiWith reference sequences X0The pass of each point
Contact number, which is added, averages:
By riThe ordered series of numbers lined up successively is relating sequence, is that can determine that each ordered series of numbers that compares to reference sequence according to rank order
The importance of influence degree.
From formula (3), grey relational grade riUsual span is 0~1, and its value shows comparative sequences closer to 1
Influence to reference sequences is bigger.
Table 1 is the grey between the meteorological variables and photovoltaic plant power output calculated using grey correlation theory
The value of the degree of association, has been ranked up to the grey relational grade between meteorological variables and power:R (total radiation)>R (articulation index)
>R (wind speed)>R (sunshine time)>R (hourly average temperature)>R (the hour temperature difference)>R (station pressure)>R (relative humidity).This reality
Apply example selection and input of larger preceding 5 meteorological variables of the station output degree of association as supporting vector machine model, construction
GRA-SVM models are predicted.
Grey relational grade between the meteorologic factor of table 1 and photovoltaic plant power output
Specifically included in above-mentioned steps S123 the step of principal component analysis:
S1231, original meteorological variables data are standardized:
If original variable there are s, x is used respectively1,x2,x3,…,xsRepresent, original variable is standardized, standard
Changing formula is:
In formula, x* ljRepresent result of j-th of variable after standardization at l-th of sample, xljIt is j-th of variable at l-th
Data value at sample,sjThe average and variance of respectively j-th variable;
S1232, the simple correlation coefficient matrix R for calculating variable:
R=YTY(N-1)
Wherein, Y is the matrix after original variable is standardized, and N is the number of samples of variable;
S1233, the characteristic value δ for calculating correlation matrix R1≥δ2≥δ3≥…≥δs>=0 and corresponding unit character to
Measure μ1,μ2,μ3,…,μs。
S1234, the selected principal component number of determination:
The variance contribution ratio of q-th of factor is:
The accumulative variance contribution ratio of the preceding t factor is:
The factor adds up the explanation degree of the original variable information of t factor pair before variance contribution ratio reaction, usually, selected
The quantity of the principal component factor depend on the factor and add up variance contribution ratio, when accumulative variance contribution ratio is more than 75%~85%, say
Bright extracted factor number can react the most information of original variable.
Table 2 show all principal component factors of 30 days of the different meteorologic factors calculated according to correlation matrix
Characteristic value and variance contribution value, it is seen that the accumulative variance contribution ratio of preceding 3 main genes be 81.89%, so 3 before selective extraction
The individual principal component factor is enough the most information instead of original meteorological variables;
The factor of table 2 adds up variance contribution table
By above-mentioned steps, calculateJust each principal component is obtained, table 3 show preceding 3 principal component scores systems
Matrix number, the order of magnitude of score coefficient shows percentage contribution size of the variable to corresponding principal component.Wherein, P represents day
Average gas pressure, Kt represents articulation index, and H represents total solar radiation on horizontal plane, and Td represents daily temperature range, and T represents that day puts down
Equal temperature, S represents sunshine time, and RR represents relative humidity, and V represents wind speed.Preceding t characteristic value therein and corresponding feature to
Amount is the initial solution of factorial analysis.Then the principal component factor values calculated are trained plan as SVM input
Close, construction PCA-SVM models are predicted.
The composition score coefficient matrix of table 3
Fig. 4 represents the curve of predict the outcome curve and the actual observed value of 3 kinds of Individual forecast models, it is seen then that Individual forecast
The prediction curve of model is substantially consistent with photovoltaic plant actual observed value curve, but model is in some predicted time points and sight
Measured value difference is larger, and such as GRA-SVM models predicting the outcome at 9 days 12 May differs greatly with observation.
Fig. 5~7 represent the residual error histogram of error that predicts the outcome of 3 kinds of Individual forecast models respectively.As shown in figure 5, SVM moulds
The main integrated distribution of residual error of type is between both sides -0.2~-0.1 and 0.1~0.3, in -0.1 to 0.1 interval almost distribution-free;
As shown in fig. 6, the residual error error distribution of GRA-SVM models is larger, and still it is distributed at more than positive direction 0.5, explanation
Model prediction result there is a possibility that extreme residual error error occur, and this also causes the ME errors of model larger;As shown in fig. 7,
The residual distribution scope of PCA-SVM models is although relatively small, but also mainly concentrate on both sides -0.3~-0.1 and 0.1~
0.3 is interval, and the left and right of histogram distribution is more symmetrical, and this make it that the ME calculation errors of its model are smaller.From 3 Individual forecasts
On the whole, the deviation that predicts the outcome of PCA-SVM models is relatively minimal for ME errors and residual error the histogram of error distribution of model,
Meteorological variables are handled using PCA, the input quantity of SVM network models is reduced, simplified model complexity, SVM models are pre-
Survey result error to take second place, GRA-SVM model prediction result error relative maximums, this is probably to screen due in GRA-SVM passing through
Afterwards, only select to export photovoltaic 5 larger meteorological variables of influence as input, made the precision of prediction of GRA-SVM models
Reduction.
Fig. 8 show the variable weight coefficient combination forecasting of this method proposition and the curve of observation, can be with from figure
Find out, the curve that variable weight coefficient combination forecasting predicts the outcome is close with observation curve.
Fig. 9 show the residual error histogram of error that predicts the outcome of the variable weight coefficient combination forecasting of this method proposition,
The residual error histogram of error left avertence of built-up pattern, and residual error error is concentrated mainly between -0.1~0.1, but the prediction of model
As a result it is small probability event larger residual error error occur.
Above result of the test shows:This method can grid-connected photovoltaic power station more accurate than Individual forecast model realization
The prediction exerted oneself in short term.
Claims (7)
1. a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient, it is characterised in that the step of this method
Suddenly include:
S1, using comprising the historical data nearest from prediction period a variety of Individual forecast models are set up, obtain each Individual forecast
Match value of the model to forecast sample point;
S2, weight coefficient of each Individual forecast model in each forecast sample point is calculated by gray relative analysis method;
S3, each Individual forecast model respectively obtained using step S1 and step S2 to the match value of each forecast sample point and
Corresponding weight coefficient training fitting obtains BPNN network models;
S4, the single power prediction value for being worth to by the prediction of prediction period newest meteorological element each Individual forecast model, so
The BPNN network models obtained afterwards using step S3 calculate when variable weight coefficient of each Individual forecast model in prediction period,
Using obtained single power prediction value and when variable weight coefficient calculate the power prediction value of prediction period;
S5, circulation step S1~S4, constantly update the power prediction value of prediction period.
2. a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient according to claim 1, its
It is characterised by, the step S1 is specifically included:
S11, acquisition photovoltaic plant the history meteorological element data nearest from prediction period and history photovoltaic power output data are made
For sample data, the sample data includes training sample and forecast sample;
S12, the training sample data obtained using step S11, are trained fitting by minimizing training error, set up a variety of
Individual forecast model, obtains match value of each Individual forecast model to forecast sample.
3. a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient according to claim 1, its
It is characterised by, the step S2 is specifically included:Each Individual forecast model is calculated in forecast sample by gray relative analysis method
Match value at point and the grey incidence coefficient between observation, regard obtained grey incidence coefficient as correspondence Individual forecast model
Weight coefficient.
4. a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient according to claim 1, its
It is characterised by, the step S3 is specifically included:Each Individual forecast model that step S1 is obtained is to each forecast sample point
Match value is as input, and each corresponding weight coefficient of Individual forecast model that step S2 is obtained is as output, and training is obtained
BPNN network models.
5. a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient according to claim 2, its
It is characterised by, the step S4 is specifically included:
The newest meteorological element predicted value of S41, acquisition prediction period;
S42, using the obtained newest meteorological element predictor calculations of step S41 go out each Individual forecast model to prediction period
Single power prediction value;
In S43, the BPNN network models for obtaining the single power prediction value input step S3 that step S42 is obtained, each is obtained
When variable weight coefficient of the Individual forecast model in prediction period;
S44, for each prediction period, the single power prediction value that each Individual forecast model is obtained is multiplied by corresponding time-varying
All superpositions again, obtain the power prediction value of the prediction period after weight coefficient.
6. a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient according to claim 1, its
It is characterised by, a variety of Individual forecast models include:Supporting vector machine model, grey correlation analysis combination supporting vector machine mould
Type and principal component analysis combination supporting vector machine model.
7. a kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient according to claim 5, its
It is characterised by, the meteorological element includes:Total solar radiation on hour horizontal plane, outside hour ground when horizontal plane global radiation, sunshine
Number, hourly average temperature, the hour highest temperature, the hour lowest temperature, relative humidity, station pressure and wind speed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710526959.2A CN107169683A (en) | 2017-06-30 | 2017-06-30 | A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710526959.2A CN107169683A (en) | 2017-06-30 | 2017-06-30 | A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107169683A true CN107169683A (en) | 2017-09-15 |
Family
ID=59826970
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710526959.2A Pending CN107169683A (en) | 2017-06-30 | 2017-06-30 | A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107169683A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107991721A (en) * | 2017-11-21 | 2018-05-04 | 上海电力学院 | It is a kind of based on astronomical and meteorological envirment factor by when scattering ratio Forecasting Methodology |
CN108537379A (en) * | 2018-04-04 | 2018-09-14 | 北京科东电力控制***有限责任公司 | Adaptive variable weight combination load forecasting method and device |
CN110737876A (en) * | 2019-09-26 | 2020-01-31 | 国家电网公司华北分部 | Regional power grid photovoltaic power prediction optimization method and device |
CN111260154A (en) * | 2020-02-17 | 2020-06-09 | 河海大学 | Short-term solar radiation prediction method and device based on CNN-LSTM |
CN111310964A (en) * | 2018-12-12 | 2020-06-19 | 华北电力大学扬中智能电气研究中心 | Load prediction method and device |
CN111367349A (en) * | 2018-12-26 | 2020-07-03 | 株洲中车时代电气股份有限公司 | Photovoltaic MPPT control method and system based on prediction model |
CN111798055A (en) * | 2020-07-06 | 2020-10-20 | 国网山东省电力公司电力科学研究院 | Variable weight combined photovoltaic output prediction method based on grey correlation degree |
CN111931981A (en) * | 2020-07-06 | 2020-11-13 | 安徽天尚清洁能源科技有限公司 | Photovoltaic power generation ultra-short-term prediction method based on machine learning multi-model combination |
CN112418502A (en) * | 2020-11-13 | 2021-02-26 | 国网冀北电力有限公司计量中心 | Photovoltaic power generation prediction method based on combined model |
CN112686472A (en) * | 2021-01-22 | 2021-04-20 | 国网河南省电力公司许昌供电公司 | Power prediction method for distributed photovoltaic equivalent power station |
CN113657677A (en) * | 2021-08-20 | 2021-11-16 | 济南大学 | Transformer oil temperature prediction method and system based on variable weight combined model |
CN115882455A (en) * | 2023-02-20 | 2023-03-31 | 国网山东省电力公司滨州供电公司 | Distributed photovoltaic power generation prediction method, system and terminal |
WO2023009070A3 (en) * | 2021-07-27 | 2023-05-25 | Envision Digital International Pte. Ltd. | Method and apparatus for forecasting optical power, computer device and storage medium |
CN116187685A (en) * | 2023-01-16 | 2023-05-30 | 南通电力设计院有限公司 | Calculation method for maximum capacity of power grid admittance photovoltaic based on peak shaving constraint |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663513A (en) * | 2012-03-13 | 2012-09-12 | 华北电力大学 | Combination forecast modeling method of wind farm power by using gray correlation analysis |
CN103023065A (en) * | 2012-11-20 | 2013-04-03 | 广东工业大学 | Wind power short-term power prediction method based on relative error entropy evaluation method |
CN104915736A (en) * | 2015-06-29 | 2015-09-16 | 东北电力大学 | Method for improving accuracy of wind power combined prediction based on improved entropy weight method |
CN104978611A (en) * | 2015-07-06 | 2015-10-14 | 东南大学 | Neural network photovoltaic power generation output prediction method based on grey correlation analysis |
CN106529724A (en) * | 2016-11-14 | 2017-03-22 | 吉林大学 | Wind power prediction method based on grey-combined weight |
-
2017
- 2017-06-30 CN CN201710526959.2A patent/CN107169683A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663513A (en) * | 2012-03-13 | 2012-09-12 | 华北电力大学 | Combination forecast modeling method of wind farm power by using gray correlation analysis |
CN103023065A (en) * | 2012-11-20 | 2013-04-03 | 广东工业大学 | Wind power short-term power prediction method based on relative error entropy evaluation method |
CN104915736A (en) * | 2015-06-29 | 2015-09-16 | 东北电力大学 | Method for improving accuracy of wind power combined prediction based on improved entropy weight method |
CN104978611A (en) * | 2015-07-06 | 2015-10-14 | 东南大学 | Neural network photovoltaic power generation output prediction method based on grey correlation analysis |
CN106529724A (en) * | 2016-11-14 | 2017-03-22 | 吉林大学 | Wind power prediction method based on grey-combined weight |
Non-Patent Citations (2)
Title |
---|
李芬 等: "基于PCA-BPNN的并网光伏电站发电量", 《可再生能源》 * |
陈嵩: "组合预测技术及其在功率预测中的应用", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107991721A (en) * | 2017-11-21 | 2018-05-04 | 上海电力学院 | It is a kind of based on astronomical and meteorological envirment factor by when scattering ratio Forecasting Methodology |
CN107991721B (en) * | 2017-11-21 | 2020-05-08 | 上海电力学院 | Time-by-time scattering ratio prediction method based on astronomical and meteorological environment factors |
CN108537379A (en) * | 2018-04-04 | 2018-09-14 | 北京科东电力控制***有限责任公司 | Adaptive variable weight combination load forecasting method and device |
CN108537379B (en) * | 2018-04-04 | 2021-11-16 | 北京科东电力控制***有限责任公司 | Self-adaptive variable weight combined load prediction method and device |
CN111310964B (en) * | 2018-12-12 | 2024-04-26 | 长春理工大学 | Load prediction method and device |
CN111310964A (en) * | 2018-12-12 | 2020-06-19 | 华北电力大学扬中智能电气研究中心 | Load prediction method and device |
CN111367349A (en) * | 2018-12-26 | 2020-07-03 | 株洲中车时代电气股份有限公司 | Photovoltaic MPPT control method and system based on prediction model |
CN110737876B (en) * | 2019-09-26 | 2023-08-08 | 国家电网公司华北分部 | Regional power grid photovoltaic power prediction optimization method and device |
CN110737876A (en) * | 2019-09-26 | 2020-01-31 | 国家电网公司华北分部 | Regional power grid photovoltaic power prediction optimization method and device |
CN111260154A (en) * | 2020-02-17 | 2020-06-09 | 河海大学 | Short-term solar radiation prediction method and device based on CNN-LSTM |
CN111798055A (en) * | 2020-07-06 | 2020-10-20 | 国网山东省电力公司电力科学研究院 | Variable weight combined photovoltaic output prediction method based on grey correlation degree |
CN111931981A (en) * | 2020-07-06 | 2020-11-13 | 安徽天尚清洁能源科技有限公司 | Photovoltaic power generation ultra-short-term prediction method based on machine learning multi-model combination |
CN112418502A (en) * | 2020-11-13 | 2021-02-26 | 国网冀北电力有限公司计量中心 | Photovoltaic power generation prediction method based on combined model |
CN112418502B (en) * | 2020-11-13 | 2024-04-30 | 国网冀北电力有限公司计量中心 | Photovoltaic power generation prediction method based on combined model |
CN112686472A (en) * | 2021-01-22 | 2021-04-20 | 国网河南省电力公司许昌供电公司 | Power prediction method for distributed photovoltaic equivalent power station |
CN112686472B (en) * | 2021-01-22 | 2022-09-20 | 国网河南省电力公司许昌供电公司 | Power prediction method for distributed photovoltaic equivalent power station |
WO2023009070A3 (en) * | 2021-07-27 | 2023-05-25 | Envision Digital International Pte. Ltd. | Method and apparatus for forecasting optical power, computer device and storage medium |
CN113657677B (en) * | 2021-08-20 | 2024-02-27 | 济南大学 | Transformer oil temperature prediction method and system based on variable weight combination model |
CN113657677A (en) * | 2021-08-20 | 2021-11-16 | 济南大学 | Transformer oil temperature prediction method and system based on variable weight combined model |
CN116187685A (en) * | 2023-01-16 | 2023-05-30 | 南通电力设计院有限公司 | Calculation method for maximum capacity of power grid admittance photovoltaic based on peak shaving constraint |
CN116187685B (en) * | 2023-01-16 | 2023-11-24 | 南通电力设计院有限公司 | Calculation method for maximum capacity of power grid admittance photovoltaic based on peak shaving constraint |
CN115882455A (en) * | 2023-02-20 | 2023-03-31 | 国网山东省电力公司滨州供电公司 | Distributed photovoltaic power generation prediction method, system and terminal |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107169683A (en) | A kind of grid-connected photovoltaic power station short term power combination forecasting method of variable weight coefficient | |
Chang et al. | An improved neural network-based approach for short-term wind speed and power forecast | |
CN107766990B (en) | Method for predicting power generation power of photovoltaic power station | |
Liu et al. | Random forest solar power forecast based on classification optimization | |
Wang et al. | Bi-directional long short-term memory method based on attention mechanism and rolling update for short-term load forecasting | |
Du et al. | Multi-step ahead forecasting in electrical power system using a hybrid forecasting system | |
Tahmasebifar et al. | A new hybrid model for point and probabilistic forecasting of wind power | |
CN110414788A (en) | A kind of power quality prediction technique based on similar day and improvement LSTM | |
CN105373857A (en) | Photovoltaic power station irradiance prediction method | |
CN103117546A (en) | Ultrashort-term slide prediction method for wind power | |
CN105069521A (en) | Photovoltaic power plant output power prediction method based on weighted FCM clustering algorithm | |
CN111695724B (en) | Wind speed prediction method based on hybrid neural network model | |
CN114792156A (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
CN111915092A (en) | Ultra-short-term wind power prediction method based on long-time and short-time memory neural network | |
Banik et al. | Wind power generation probabilistic modeling using ensemble learning techniques | |
CN113822418A (en) | Wind power plant power prediction method, system, device and storage medium | |
CN113537582B (en) | Photovoltaic power ultra-short-term prediction method based on short-wave radiation correction | |
Park et al. | Multi-layer RNN-based short-term photovoltaic power forecasting using IoT dataset | |
CN115860177A (en) | Photovoltaic power generation power prediction method based on combined machine learning model and application thereof | |
CN109583645A (en) | A kind of public building short-term load forecasting method | |
CN105160441A (en) | Real-time power load forecasting method based on integrated network of incremental transfinite vector regression machine | |
CN113610328A (en) | Power generation load prediction method | |
Çevik et al. | Day ahead wind power forecasting using complex valued neural network | |
Amer et al. | Solar power prediction based on Artificial Neural Network guided by feature selection for Large-scale Solar Photovoltaic Plant | |
Tao et al. | On comparing six optimization algorithms for network-based wind speed forecasting |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170915 |