CN107766990A - A kind of Forecasting Methodology of photovoltaic power station power generation power - Google Patents
A kind of Forecasting Methodology of photovoltaic power station power generation power Download PDFInfo
- Publication number
- CN107766990A CN107766990A CN201711103682.9A CN201711103682A CN107766990A CN 107766990 A CN107766990 A CN 107766990A CN 201711103682 A CN201711103682 A CN 201711103682A CN 107766990 A CN107766990 A CN 107766990A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- data
- max
- power
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention discloses a kind of Forecasting Methodology of photovoltaic power station power generation power, including, using photovoltaic plant history meteorological data, six kinds of Meteorological Characteristics are extracted in units of day, establish Meteorological Characteristics storehouse；The day characteristic in Meteorological Characteristics storehouse is clustered by KFCM algorithms, realizes that weather pattern is classified, and power data and meteorological data progress category label to each day；According to category label, a SVR submodel is established per the power data in one kind and meteorological data；The target day weather characteristics provided by numerical weather forecast, the weather pattern of target day, SVR submodels corresponding to selection are identified using SVM；ARIMA models are established using the Real-time Monitoring Data of target day, the real-time estimate of irradiation intensity and temperature is realized using rolling forecast mode；The selected SVR submodels of the predicted value of irradiation intensity and temperature input, obtain predicting power of photovoltaic plant result.The present invention improves the precision of prediction of photovoltaic power station power generation power.
Description
Technical field
The present invention relates to a kind of Forecasting Methodology of photovoltaic power station power generation power, belong to power system automation technology field.
Background technology
Photovoltaic generation at home and abroad obtains in recent years as a kind of main technology realization rate in Solar use
Rapidly development.Ended for the end of the year 2015, global photovoltaic adds up installed capacity and reaches 227GW.The accumulative installation of the photovoltaic generation of China
Capacity reaches 43.18GW, and wherein photovoltaic plant is 37.12GW, turns into the maximum country of global Photovoltaic generation installed capacity.Together
When, Chinese adding new capacity 15.13GW in 2015, and the whole world increase one of faster country.Photovoltaic plant it is extensive
Development brings more serious influence with application to the stabilization and the quality of power supply of power network, and photovoltaic consumption, which turns into, hinders photovoltaic industry
The major issue of development.On the one hand photovoltaic power prediction can provide important branch for the coordination control of power network with management and running
Support, photovoltaic digestion capability on the other hand can be improved, increase photovoltaic plant rate of return on investment.In time scale, photovoltaic power
Prediction can be divided into ultra-short term, short-term and medium-term and long-term.For the angle of operation of power networks, the cycle of prediction is shorter, to urgent feelings
The processing of condition and prevention state is more favourable.Therefore, the power prediction of the ultra-short term of photovoltaic plant is paid close attention to by height.
Traditional photovoltaic power prediction is divided into direct method and indirect method.Direct forecast model is to utilize irradiation intensity, temperature,
The related historical data such as humidity, wind speed establishes generated output regression model, and data are typically derived from numerical weather forecast
(NWP), modeling method includes statistical model, neutral net, SVMs etc..Indirect predictions are then divided into two stages, and first
Stage forecast solar irradiation intensity or other weather informations, second stage calculate generated output again.The Forecasting Methodology of irradiation intensity
Including the irradiation prediction based on cloud atlas, multiple regression, time series, Markov chain etc..In terms of direct prediction, Gaoyang etc. is (high
Sun, Zhang Biling, Mao Jingli, adaptive photovoltaic ultra-short term output forecast model [J] electric power network techniques of the Liu Yong based on machine learning,
2015,39(2):307-311.) propose the adaptive photovoltaic ultra-short term output forecast model based on SVMs (SVM)；
(De Giorgi MG, Congedo PM, the Malvoni M.Photovoltaic power such as M.G.De Giorgi
forecasting using statistical methods:impact of weather data[J].IET Science
Measurement&Technology,2014,8(3):The photovoltaic based on artificial neural network (ANN) 90-97.) is established to contribute
Forecast model；(Mellit A, Pavan AM, the Lughi V.Short-term forecasting of power such as A.Mellit
production in a large-scale photovoltaic plant[J].Solar Energy,2014,105:401-
413.) three kinds of different artificial neural networks (ANN) moulds of database development of solar irradiance, battery temperature and power output are used
Type, suitable for the photovoltaic power generation output forecasting of three kinds of typical weather (fine day, part cloudy day and cloudy day)；In terms of indirect predictions, irradiation
The prediction of intensity is the most key, (Arquez R, the Coimbra CFM.Intra-hour DNI forecasting such as R.Arquez
based on cloud tracking image analysis[J].Solar Energy,2013,91:327-336.) propose
Direct intensity of illumination Forecasting Methodology based on cloud layer tracking graphical analysis；H.D.Yang etc. (Yang HD, Kurtz B,
Nguyen D,et al.Solar irradiance forecasting using a ground-based sky imager
developed at UC San Diego[J].Solar Energy,2014,103:502-524.) propose ground cloud atlas point
The solar irradiation intensity prediction method of analysis；(Escrig H, Batlles FJ, the Alonso J.Cloud such as H.Escrig
detection,classification and motion estimation using geostationary satellite
imagery for cloud cover forecast[J].Energy,2013,55:853-859.) utilize geostationary satellite
Image carries out the detection of cloud layer, classification and mobile estimation to predict irradiation intensity.In recent years, hybrid prediction model has obtained rapidly
Development, its basic procedure can be divided into classification, recurrence, prediction three phases.H.T.Yang etc. (Yang HT, Huang CM,
Huang YC,et al.A Weather-Based Hybrid Method for 1-Day Ahead Hourly
Forecasting of PV Power Output[J].IEEE Transactions on Sustainable Energy,
2014,5(3):917-926.) propose the photovoltaic power generation output forecasting mixed model per hour based on weather pattern identification；
(Wang JD, Ran R, Song ZL, the et al.Short-Term Photovoltaic Power such as J.D.Wang
Generation Forecasting Based on Environmental Factors and GA-SVM[J].Journal
of Electrical Engineering&Technology,2017,12(1):64-71.) propose based on environmental factor and something lost
Pass SVM photovoltaic output Comprehensive Model.
There are two major defects in current photovoltaic power prediction algorithm, first, the precision of photovoltaic power prediction can not protect
Barrier, the power output of photovoltaic generating system exists with solar irradiation intensity closely to be contacted, but irradiation intensity can be hidden by cloud layer
The influence of gear, more obvious fluctuation is presented, and weather conditions are different, the severe degree of fluctuation also differs, fluctuation with
Machine has had a strong impact on the precision of photovoltaic power generation power prediction；Second, the time scale of prediction is longer, it is difficult to realizes real-time estimate.
The content of the invention
The technical problems to be solved by the invention are the defects of overcoming prior art, there is provided a kind of photovoltaic power station power generation power
Forecasting Methodology, employ SVR submodels based on weather pattern and carry out regression forecasting, eliminate other disturbing factors, predict
Precision gets a promotion.
To solve the above problems, the technical solution adopted in the present invention is as follows：A kind of prediction of photovoltaic power station power generation power
Method, comprise the following steps：
1) the history meteorological data and power data of photovoltaic plant are gathered, six kinds of Meteorological Characteristics are extracted in units of day, are built
Vertical Meteorological Characteristics storehouse；
2) the day Meteorological Characteristics data in Meteorological Characteristics storehouse are clustered by FCM Clustering Algorithm of Kernel, realized
Weather pattern is classified, and the power data to each day and meteorological data carry out category label on this basis；
3) according to the category label of step 2), a SVMs is established per the power data in one kind and meteorological data
Return submodel；
4) weather characteristics of the target day provided by numerical weather forecast data, mesh is identified using algorithm of support vector machine
Mark the weather pattern of day, Support vector regression submodel corresponding to selection target day；
5) autoregression integration moving average model is established using the Real-time Monitoring Data of target day, uses rolling forecast mode
Realize the real-time estimate of irradiation intensity and temperature；
6) by step 5) irradiation intensity and the predicted value input step 4 of temperature) in selection Support vector regression submodule
In type, final predicting power of photovoltaic plant result is obtained.
History meteorological data derives from the comprehensive monitoring system in photovoltaic plant in foregoing step 1), includes independent gas
As real time meteorological data is collected at station；The meteorological data includes real-time irradiation intensity and temperature；The power data derives from light
Overhead utility metering system.
Six kinds of foregoing Meteorological Characteristics are：[IRmax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax],
Wherein,
IRmax=max (IRi) it is maximum irradiation level, Tmax=max (Ti) it is the highest temperature, DIFFIRmax=max
(DIFFIRi) it is maximum fluctuation value, STDIRTo fluctuate standard deviation, MVIRTo fluctuate average, TDIRmaxTo fluctuate three order derivatives, IRi
And TiIt is i-th of real-time irradiation intensity and temperature data in day historical data,
DIFFIRi=IRi+1-IRiI=1,2 ..., s-1
S is the quantity of sampled point.
The idiographic flow of FCM kernel clustering is as follows in foregoing step 2)：
(2-1) prepares the data sample of cluster for the first time：xk=[IRmax, Tmax, DIFFIRmax, TDIR]kK=1,2,
3 ..., n,
N is sample number；
(2-2) initial clustering number C=2；
(2-3) random initializtion cluster centre νi；
(2-4) calculates degree of membership coefficient uik；
Wherein, K (xi,xj) it is gaussian kernel function,
δ is kernel function coefficient, and m takes 2；
(2-5) calculates new cluster centre：
Represent uikSquare；
(2-6) circulation step (2-3)~step (2-5), reaches end condition, then loop termination；
(2-7) calculates Cluster Validity coefficient V during C=2XB(C)；
(2-8) another C=C+1, circulation performs step (2-3)~step (2-7), until C=Cmax, obtain VXB(C+1), Cmax
To cluster number；
(2-9) judges optimum clustering number Copt, minimum VXB(C) C corresponding to is Copt；
(2-10) is with C=Copt, step (2-3)~step (2-6) is re-executed, obtains optimum cluster result, by sample xk
Carry out category label；
(2-11) secondary cluster：On the basis of first time clusters, by a few class data clustered respectively with xk'=
[MVIR, STDIR]k'K'=1,2,3 ..., n' are sample, and n' is sample number, then carries out second with FCM Clustering Algorithm of Kernel
Secondary cluster.
In foregoing step 3), the Support vector regression submodel, input as irradiation intensity IRiWith temperature Ti, output
For instantaneous power Pi。
In foregoing step 4), numerical weather forecast data source is in weather service supplier.
In foregoing step 4), the process that the weather pattern of target day is identified using algorithm of support vector machine is：Training branch
Vector machine model is held, the input of training data is the Meteorological Characteristics storehouse in step 1), includes six features
[IRmax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax], export as the weather pattern in step 2)；Identification process
In, by the feature [IR of target daymax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax]tarIt is input to multi-category support vector machines mould
Type, you can obtain the weather pattern of target day.
In foregoing step 5), the rolling forecast mode is the reality of several irradiation intensities and temperature using target day
When Monitoring Data ordered series of numbers establish two autoregressions integration moving average model time serieses respectively, predict the spoke of subsequent time respectively
According to intensity and temperature value；The actual measurement irradiation intensity and temperature data is recycled to correct the two autoregressions product respectively in subsequent cycle
Divide moving average model time series, predict again the irradiation intensity and temperature value of subsequent time respectively；Circulated, rolled with this
Prediction.
With immediate prior art ratio, the invention has the advantages that：
1st, the history meteorological data and go out force data that the present invention is monitored using photovoltaic plant weather station, close to the actual shape in scene
Condition；
2nd, present invention employs two layers of Clustering Model, different features is respectively adopted and is clustered, classification more refinement with
Rationally, scientific basis is provided for the selection of submodel；
3rd, present invention employs the SVR submodels based on weather pattern to carry out regression forecasting, eliminates other disturbing factors,
Precision of prediction gets a promotion；
4th, present invention employs ARIMA time series models, provided by photovoltaic plant weather station monitoring system real-time
Data, realize the real-time estimate of 5 minute time intervals.
Brief description of the drawings
Fig. 1 is the flow chart of the inventive method；
Fig. 2 is KFCM modeling procedure figures；
Fig. 3 is V in embodimentXBIndex changing trend diagram；Fig. 3 (a) is index change after first layer cluster；Fig. 3 (b) is the
A+B indexs change after two strata classes；Fig. 3 (c) is C+D indexs change after second layer cluster；
Fig. 4 is irradiation intensity ARIMA time series forecasting results in embodiment；Fig. 4 (a) is the prediction result on May 19,
Fig. 4 (b) is the prediction result on June 7, and Fig. 4 (c) is the prediction result on July 21, and Fig. 4 (d) is the August prediction result of 22 days；
Fig. 5 is temperature ARIMA time series forecasting results in embodiment；Fig. 5 (a) be May 19 prediction result, Fig. 5
(b) it is the prediction result on June 7, Fig. 5 (c) is the prediction result on July 21, and Fig. 5 (d) is the August prediction result of 22 days；
Fig. 6 is power SVR model prediction results in embodiment；Fig. 6 (a) is the prediction result on May 19, and Fig. 6 (b) is 6
The prediction result on the moon 7, Fig. 6 (c) are the prediction result on July 21, and Fig. 6 (d) is the August prediction result of 22 days.
Embodiment
The invention will be further described below.Following examples are only used for the technical side for clearly illustrating the present invention
Case, and can not be limited the scope of the invention with this.
The invention provides a kind of Forecasting Methodology of photovoltaic power station power generation power, as shown in figure 1, comprising the following steps：
Step 1, using photovoltaic plant history meteorological data, six kinds of Meteorological Characteristics are extracted in units of day, established meteorological special
Levy storehouse.
Classification regression modeling has been demonstrated there is more preferable precision for relative block mold, it is advantageous that excluding other
Interference of the X factor to regression model in weather conditions, realizes becoming more meticulous for data model.History meteorological data source
Comprehensive monitoring system in photovoltaic plant, wherein containing the meteorological data collected by independent weather station, make in the present invention
The real-time irradiation intensity and temperature data and photovoltaic plant metering system that mainly photovoltaic weather station is gathered are provided
Power data.Irradiation intensity and temperature are that photovoltaic is contributed to influence two maximum key elements.Meanwhile the fluctuation of irradiation intensity
And influence the principal element of precision of forecasting model.The meaning of weather pattern classification is the approximate weather pattern of fluctuation situation
One kind is classified as, so as to establish independent recurrence submodel with exclusive PCR, improves precision of prediction.Based on this, the present invention selects
Six kinds of Meteorological Characteristics are：[IRmax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax], it is defined as follows：
(1) maximum irradiation level IRmax=max (IRi)；
(2) highest temperature Tmax=max (Ti)；
The maximum instantaneous power value that the two index reflections photovoltaic module can be sent, wherein, IRiAnd TiIt is a day history number
I-th of irradiation intensity and temperature data in；
(3) maximum fluctuation value DIFFIRmax=max (DIFFIRi)；
For discrete data of the sample rate in the case of constant, generally use first-order difference DIFFIRTo substitute first derivative；
DIFFIRi=IRi+1-IRi(i=1,2 ..., s-1) (1)
S is the quantity of sampled point；
(4) standard deviation STD is fluctuatedIR：DIFFIRiStandard deviation, reflect irradiation level fluctuation intensity；
(5) average MV is fluctuatedIR：DIFFIRiAverage, reflection irradiation level fluctuation overall amplitude；
(6) three order derivative TD are fluctuatedIRmax：Quick change of three order derivatives to weather condition is more more sensitive than other exponent numbers.
Step 2, the day Meteorological Characteristics data in Meteorological Characteristics storehouse are entered by FCM kernel clustering (KFCM) algorithm
Row cluster, realize that weather pattern is classified, the power data to each day and meteorological data carry out category label, gas on this basis
Image data refers to that the irradiation intensity that photovoltaic weather station gathered and temperature, power data refer to the output work of photovoltaic plant
Rate.
In view of in practical application, category of model is thinner, its disturbing factor is fewer；But excessive classification number can serious shadow
Computational efficiency is rung, is unfavorable for real-time.Therefore, using layering KFCM cluster modes in the present invention, for different features not
Clustered on same level.Specifically, exactly first once clustered according to maximum value tag, then by first cluster result
Carry out secondary cluster with residue character respectively again.As shown in Fig. 2 by KFCM algorithms by the day Meteorological Characteristics in Meteorological Characteristics storehouse
Data are clustered, and realize that weather pattern is classified, the power data of each day and meteorological data are all carried out into class on this basis
Do not mark, two layers of cluster of KFCM clusters is as follows：
(1) according to maximum value tag [IRmax, Tmax, DIFFIRmax, TDIRmax] perform cluster for the first time；
(2) initial clustering result will be again with remaining feature [STDIR, MVIR] perform secondary cluster.
Two layers of cluster idiographic flow of KFCM is as follows：
(1) the data sample x of cluster for the first time is preparedk=[IRmax, Tmax, DIFFIRmax, TDIR]kK=1,2,3 ..., n,
N is sample number；
(2) initial clustering number C=2；
(3) random initializtion cluster centre νi；
(4) degree of membership coefficient u is calculatedik；
K is gaussian kernel function, and δ is kernel function coefficient, and m typically takes 2：
(5) new cluster centre is calculated：
Because m takes 2,Represent uikSquare.
(6) circulation step (3)~step (5), judges end condition, generally there are two kinds：Cycle-index and cluster centre
Deviation, if circulation reaches certain number, or the front and rear cluster centre calculated twice is less than the threshold value of setting apart from its difference, then
Loop termination；
(7) Cluster Validity coefficient V during C=2 is calculatedXB(C)；
(8) another C=C+1, circulation performs step (3)~step (7), until C=Cmax, obtain VXB(C+1), CmaxFor cluster
Number；
(9) optimum clustering number C is judgedopt, minimum VXB(C) C corresponding to is Copt；
(10) with C=Copt, step (3)~step (6) is re-executed, obtains optimum cluster result, by sample xkCarry out class
Do not mark；
(11) secondary cluster：On the basis of first time clusters, by a few class data clustered respectively with xk'=[MVIR,
STDIR]k'K'=1,2,3 ..., n' are sample, and n' is sample number, then carry out second with KFCM algorithms and cluster.
Step 3, according to category label, a support vector regression is established per the power data in one kind and meteorological data
SVR submodels.
According to the result clustered in step 2, if historical data is divided into Ganlei, SVR is could set up per one kind
Model.Specifically, a SVR submodel is established per the power data in one kind and meteorological data, its category label derives from
The result of two layers of cluster of KFCM, some SVR submodels trained, is inputted as irradiation intensity IRiWith temperature Ti, export to be instantaneous
Power Pi, for the model established labeled as SVR I, SVR II ... SVR N, N is the final classification number of cluster.
SVR is the mode that SVM is used for regression modeling, and its core concept is to build Optimal Separating Hyperplane, and ensures from super
The distance between the nearest sample of plan range and hyperplane maximum.SVM is for two classification problems：(xi,yi), i=1,2 ...,
L, xi∈Rn, yi∈ { -1 ,+1 }, inerrably it is divided into two classifications completely by optimal separating hyper plane wx+b=0.Therefore, construct
Optimal hyperlane problem can be converted into optimization problem：
Its constraints is：
yi((w·xi)+b)≥1-ξi, i=1,2 ..., l (7)
W is optimal hyperlane normal vector, and b is threshold value, and M is punishment parameter, ξiFor slack variable, l is the number of point, and l is individual
Point is divided into two classes, (xi,yi) represent plane coordinates on a point.
Method of Lagrange multipliers can be utilized to solve above mentioned problem.If expanding to nonlinear problem, mapping can be utilized
Sample in lower dimensional space is mapped as in higher dimensional space by φ (x), and now object function is：
αiFor Lagrange multiplier；Selection of kernel function gaussian kernel function.
Step 4, the weather characteristics of the target day provided by numerical weather forecast data, target is identified using SVM algorithm
The weather pattern of day, SVR submodels corresponding to selection.
The premise of SVR submodels selection is the weather pattern for needing to judge target day, is carried by numerical weather forecast data
The weather characteristics of the target day of confession, the weather pattern of target day, SVR submodels, target corresponding to selection are identified using SVM algorithm
The weather characteristics of day are consistent with the computational methods of six features in step 1.Numerical weather forecast data source is in weather service
Supplier.
SVM modelings in the present invention are also to be established by characteristic, and the input of training data is the meteorology in step 1
Feature database, include six feature [IRmax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax]；Export as the classification mark in step 2
Label.In identification process, by the feature [IR of target daymax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax]tarThe more classification SVM of input
Model, you can obtain the weather pattern of target day.
Step 5, autoregression integration moving average model (ARIMA) is established using the Real-time Monitoring Data of target day, is used
Rolling forecast mode realizes the real-time estimate of irradiation intensity and temperature.
Hybrid prediction model provided by the invention, it is that real-time temperature and irradiation intensity are carried out by ARIMA time serieses
Prediction, SVR submodels are recycled to realize that instantaneous power returns.ARIMA models are established using the Real-time Monitoring Data of target day, are made
The real-time estimate of irradiation intensity and temperature is realized with rolling forecast mode, ARIMA models are expressed as ARIMA (p, q, d), and p is certainly
Regression order, q are moving average exponent number, and d is difference processing exponent number.Its basic step is as follows：
(1) difference processing, original time series [Xt] handle to obtain stationary time series data [XA through d order differencest].This
Time series includes irradiation intensity sequence [X in inventiont-E] and temperature sequence [Xt-T]。
(2) Model Identification determines rank with parameter, calculates the auto-correlation and deviation―related function of data sample, principium identification model class
Type (AR, MA, ARMA).In general, the expression formula for the ARIMA models established is as follows：
Wherein, aiFor autoregressive coefficient, bjFor moving average coefficient.et-jIt is white noise sequence, is independent error.Using red
Pond information criterion (AIC) information criterion defines p and q value.
(3) parameter Estimation, a is carried out using Correlation MomentiAnd bjParameter Estimation, finally give ARIMA (p, q, d) model.
(4) data prediction, ARIMA model realization Single-step Predictions are passed through.
Further, rolling forecast mode is the Real-time Monitoring Data for several irradiation intensities and temperature for utilizing target day
Ordered series of numbers (IRtAnd Tt, t=1,2 ..., g, g is sequence samples number) and two ARIMA time serieses are established respectively, under predicting respectively
The irradiation intensity and temperature value at one moment (are designated as IR 'n+1With T 'n+1)；The actual measurement irradiation intensity and temperature are recycled in subsequent cycle
Degrees of data IRn+1And Tn+1Two ARIMA time serieses are corrected respectively, predict again the irradiation intensity and temperature value of subsequent time respectively
(it is designated asWith).Circulated with this, carry out rolling forecast.
The SVR submodels selected in the predicted value input step four of step 6, irradiation intensity and temperature, obtain final light
Overhead utility power prediction result.
Embodiment
Using effective sample between Suzhou City of Jiangsu Province Wujiang area photovoltaic plant in April, 2016 to 2 months 2017 in the present invention
This 31397, photovoltaic output hybrid prediction model is established, selects four typical meteorological states：Fine day (July 21), cloudy day (5
Months 19 days), sleet (June 7), cloudy (August 22 days) as test sample (586) with test modeling effect, remainder data
Sample is as training sample, totally 30811.
First, the foundation in photovoltaic history Meteorological Characteristics storehouse and KFCM weather patterns rationally cluster
Modeling data collection will extract daily Meteorological Characteristics, effective feature set bag in of the invention according to the method for step 1
Containing 261 samples, represent 261 days.
According to the KFCM modeling and optimization flows in step 2, Layered Clustering Model is established, the cluster numbers of two-layer model are all
For C=2~10.VXBVariation tendency is as shown in Figure 3.As can be seen that after each strata class, VXB(2) numerical value is minimum.Therefore, press
According to described in step 2, it is believed that the optimum clustering number of two layers of cluster is all 2, and all samples are finally classified into 4 classes, are designated as A classes,
B classes, C classes and D classes.Classification results are as shown in table 1.
The KFCM cluster results of table 1
2nd, SVM modelings identify with target day weather category
By cluster, 261 samples are divided into 4 classes.It is special to establish SVM weather patterns identification model and testing its precision
Four class samples in collection are used for establishing more classification SVM models, select 70% to be used as training sample, residue 30% in four class samples
As test sample, optimized parameter is obtained by cross-validation method.Training data is 261 feature samples, is exported as classification mark
Label.The result of weather pattern test is as shown in table 2.
The weather pattern recognition result of the SVM models of table 2
In 78 test samples altogether, there are 4 to reach 94.87% by misclassification, the accuracy of classification.Also,
In four class samples, only there is misclassification in B classes sample, illustrates the weather that the precision of the identification of SVM classifier is higher, is established
Type identification model accurately the weather pattern of identification prediction day and can select suitable SVR to return submodel.
3rd, ARIMA time series modelings and SVM regression forecastings
As it was previously stated, according to KFCM result, the irradiation intensity of 261 days, temperature and power data sample also will be by altogether
A, B, C, the classes of D tetra- are classified as, and gives class label respectively.One SVR of Sample Establishing in same class is returned into submodel, obtained
Four submodels SUB-A, SUB-B, SUB-C, SUB-D, input as irradiation intensity IRi, temperature Ti, export as the wink at corresponding moment
When power Pi.It is same to obtain the optimized parameter of SVR models using by cross-validation method.
Then, it is predicted the identification of day weather pattern.The weather information of prediction day is transfused to the more disaggregated models of SVM, obtains
To the category attribute of prediction day.The prediction day selected in the present invention for：(July 21), cloudy (May 19), sleet (June 7
Day), cloudy (August 22 days), obtain category attribute after inputting SVM models respectively and be：D, B, C, B.The submodel of corresponding selection is
SUB-D, SUB-B, SUB-C, SUB-B.
Real-time estimate is mixed to realize, according to the description of step 5, establishes IRi,TiARIMA time series models go forward side by side
Row rolling forecast.According to time series modeling and the needs of engineer applied, 20 IR a few days ago are predicted in selectioni,Ti(i=1-20) number
According to establishing two ARIMA models respectively, the data that on-line monitoring system of the data source in prediction day provides, the sampling interval is
5min.By taking July 21 as an example, separately begin first Monitoring Data occur from 6: 15.Collect 20 field monitoring IRi,TiNumber
According to, 7: 55 timesharing start ARIMA modeling and prediction work, respectively obtain ARIMAIRAnd ARIMAT.Then, predict next
The IR ' of sampled point (after 5min, that is, 8 points)i+1,T′i+1.The predicted value will be inputted in submodel SUB-D, predicted
Instantaneous power P.Again, 8 when, new actual monitoring data IRi+1,Ti+1It can be obtained from monitoring system, for ARIMAIR
And ARIMATCarry out real time correction.Predict the IR of (5 minutes) in next step again at 8 pointsi+2,Ti+2And P.Such circulating rolling, realizes all day
The real-time estimate of instantaneous power.Final four prediction days ARIMA model IRi,TiPrediction result such as Fig. 4, shown in Fig. 5, SVR
Model regression result is as shown in Figure 6.Using average absolute percent error εMAPEWith root-mean-square error εRMSERefer to as precision of prediction
Mark, IRi,TiAnd P precision of predictions are shown in Table 3.
The precision of prediction of the hybrid prediction model of table 3
Described above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, without departing from the technical principles of the invention, some improvement and deformation can also be made, these are improved and deformation
Also it should be regarded as protection scope of the present invention.
Claims (8)
1. a kind of Forecasting Methodology of photovoltaic power station power generation power, it is characterised in that comprise the following steps：
1) the history meteorological data and power data of photovoltaic plant are gathered, six kinds of Meteorological Characteristics are extracted in units of day, establish gas
As feature database；
2) the day Meteorological Characteristics data in Meteorological Characteristics storehouse are clustered by FCM Clustering Algorithm of Kernel, realizes weather
Classification of type, the power data to each day and meteorological data carry out category label on this basis；
3) according to the category label of step 2), a Support vector regression is established per the power data in one kind and meteorological data
Submodel；
4) weather characteristics of the target day provided by numerical weather forecast data, target day is identified using algorithm of support vector machine
Weather pattern, Support vector regression submodel corresponding to selection target day；
5) autoregression integration moving average model is established using the Real-time Monitoring Data of target day, is realized using rolling forecast mode
The real-time estimate of irradiation intensity and temperature；
6) by step 5) irradiation intensity and the predicted value input step 4 of temperature) in selection Support vector regression submodel in,
Obtain final predicting power of photovoltaic plant result.
A kind of 2. Forecasting Methodology of photovoltaic power station power generation power according to claim 1, it is characterised in that the step 1)
Middle history meteorological data derives from the comprehensive monitoring system in photovoltaic plant, and real-time weather number is collected comprising independent weather station
According to；The meteorological data includes real-time irradiation intensity and temperature；The power data derives from photovoltaic plant metering system.
A kind of 3. Forecasting Methodology of photovoltaic power station power generation power according to claim 2, it is characterised in that six kinds of gas
As being characterized as：[IRmax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax],
Wherein,
IRmax=max (IRi) it is maximum irradiation level, Tmax=max (Ti) it is the highest temperature, DIFFIRmax=max (DIFFIRi) for most
Great fluctuation process value, STDIRTo fluctuate standard deviation, MVIRTo fluctuate average, TDIRmaxTo fluctuate three order derivatives, IRiAnd TiIt is a day history number
I-th of the real-time irradiation intensity and temperature data in,
DIFFIRi=IRi+1-IRiI=1,2 ..., s-1
S is the quantity of sampled point.
A kind of 4. Forecasting Methodology of photovoltaic power station power generation power according to claim 3, it is characterised in that the step 2)
The idiographic flow of middle FCM kernel clustering is as follows：
(2-1) prepares the data sample of cluster for the first time：xk=[IRmax, Tmax, DIFFIRmax, TDIR]kK=1,2,3 ..., n,
N is sample number；
(2-2) initial clustering number C=2；
(2-3) random initializtion cluster centre νi；
(2-4) calculates degree of membership coefficient uik；
<mrow>
<msub>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>k</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>/</mo>
<mo>(</mo>
<mrow>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mn>2</mn>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mrow>
<mn>1</mn>
<mo>/</mo>
<mrow>
<mo>(</mo>
<mi>m</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</msup>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>c</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>/</mo>
<mo>(</mo>
<mrow>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>v</mi>
<mi>j</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mn>2</mn>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mrow>
<mn>1</mn>
<mo>/</mo>
<mrow>
<mo>(</mo>
<mi>m</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, K (xi,xj) it is gaussian kernel function,
δ is kernel function coefficient, and m takes 2；
(2-5) calculates new cluster centre：
<mrow>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>k</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mi>K</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>k</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mi>K</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
Represent uikSquare；
(2-6) circulation step (2-3)~step (2-5), reaches end condition, then loop termination；
(2-7) calculates Cluster Validity coefficient V during C=2XB(C)；
<mrow>
<msub>
<mi>V</mi>
<mrow>
<mi>X</mi>
<mi>B</mi>
</mrow>
</msub>
<mrow>
<mo>(</mo>
<mi>C</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>C</mi>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msubsup>
<mi>u</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mi>m</mi>
</msubsup>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>v</mi>
<mi>i</mi>
</msub>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
</mrow>
<mrow>
<mi>n</mi>
<mrow>
<mo>(</mo>
<munder>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
<mrow>
<mi>j</mi>
<mo>&NotEqual;</mo>
<mi>k</mi>
</mrow>
</munder>
<mo>|</mo>
<mo>|</mo>
<msub>
<mi>v</mi>
<mi>j</mi>
</msub>
<mo>-</mo>
<msub>
<mi>v</mi>
<mi>k</mi>
</msub>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
(2-8) another C=C+1, circulation performs step (2-3)~step (2-7), until C=Cmax, obtain VXB(C+1), CmaxIt is poly-
Class number；
(2-9) judges optimum clustering number Copt, minimum VXB(C) C corresponding to is Copt；
(2-10) is with C=Copt, step (2-3)~step (2-6) is re-executed, obtains optimum cluster result, by sample xkCarry out
Category label；
(2-11) secondary cluster：On the basis of first time clusters, by a few class data clustered respectively with xk'=[MVIR,
STDIR]k'K'=1,2,3 ..., n' are sample, and n' is sample number, then carry out second with FCM Clustering Algorithm of Kernel and gather
Class.
A kind of 5. Forecasting Methodology of photovoltaic power station power generation power according to claim 1, it is characterised in that the step 3)
In, the Support vector regression submodel, input as irradiation intensity IRiWith temperature Ti, export as instantaneous power Pi。
A kind of 6. Forecasting Methodology of photovoltaic power station power generation power according to claim 1, it is characterised in that the step 4)
In, numerical weather forecast data source is in weather service supplier.
A kind of 7. Forecasting Methodology of photovoltaic power station power generation power according to claim 6, it is characterised in that the step 4)
In, the process that the weather pattern of target day is identified using algorithm of support vector machine is：Training Support Vector Machines model, training data
Input be step 1) in Meteorological Characteristics storehouse, include six feature [IRmax, Tmax, DIFFIRmax, STDIR, MVIR, TDIRmax],
Export as the weather pattern in step 2)；In identification process, by the feature [IR of target daymax, Tmax, DIFFIRmax, STDIR,
MVIR, TDIRmax]tarIt is input to multi-category support vector machines model, you can obtain the weather pattern of target day.
A kind of 8. Forecasting Methodology of photovoltaic power station power generation power according to claim 1, it is characterised in that the step 5)
In, the rolling forecast mode is built respectively using several irradiation intensities of target day and the Real-time Monitoring Data ordered series of numbers of temperature
Vertical two autoregressions integration moving average model time series, the irradiation intensity and temperature value of subsequent time are predicted respectively；It is next
The actual measurement irradiation intensity and temperature data is recycled to correct the two autoregressions integration moving average model time respectively in circulation
Sequence, the irradiation intensity and temperature value of subsequent time are predicted again respectively；Circulated with this, carry out rolling forecast.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711103682.9A CN107766990B (en) | 2017-11-10 | 2017-11-10 | Method for predicting power generation power of photovoltaic power station |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711103682.9A CN107766990B (en) | 2017-11-10 | 2017-11-10 | Method for predicting power generation power of photovoltaic power station |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107766990A true CN107766990A (en) | 2018-03-06 |
| CN107766990B CN107766990B (en) | 2021-06-08 |
Family
ID=61273364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711103682.9A Active CN107766990B (en) | 2017-11-10 | 2017-11-10 | Method for predicting power generation power of photovoltaic power station |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107766990B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108985513A (en) * | 2018-07-17 | 2018-12-11 | 合肥工业大学 | Industrial enterprise's Electric Load Prediction System |
| CN109102102A (en) * | 2018-06-25 | 2018-12-28 | 南京邮电大学 | Based on the photovoltaic of multivariate phase space reconstruction and SVR power output short term prediction method |
| CN109388863A (en) * | 2018-09-21 | 2019-02-26 | 国网天津市电力公司 | A kind of distributed photovoltaic power output power forecasting method based on ARIMA model |
| CN109685257A (en) * | 2018-12-13 | 2019-04-26 | 国网青海省电力公司 | A kind of photovoltaic power generation power prediction method based on Support vector regression |
| CN109871996A (en) * | 2019-02-14 | 2019-06-11 | 浙江中控太阳能技术有限公司 | A kind of tower photo-thermal power station mirror field cloud monitoring System and method for based on photovoltaic panel |
| CN109919353A (en) * | 2019-01-11 | 2019-06-21 | 杭州电子科技大学 | A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence |
| CN110009385A (en) * | 2019-01-11 | 2019-07-12 | 国网浙江杭州市临安区供电有限公司 | A kind of photovoltaic power generation user group division methods based on multifactor mapping |
| CN110059441A (en) * | 2019-04-29 | 2019-07-26 | 广东电网有限责任公司 | A kind of photovoltaic plant modeling method and photovoltaic plant model output calibration method |
| CN110289605A (en) * | 2019-06-27 | 2019-09-27 | 昆明理工大学 | A kind of novel mixed compensation route intelligence direction protection method based on instantaneous power set of curves |
| CN110929953A (en) * | 2019-12-04 | 2020-03-27 | 国网山东省电力公司电力科学研究院 | Photovoltaic power station ultra-short term output prediction method based on cluster analysis |
| CN111079980A (en) * | 2019-11-22 | 2020-04-28 | 天合云能源互联网技术(杭州)有限公司 | Optical power prediction method based on self-adaptive classification strategy and hybrid optimization SVR |
| CN111738327A (en) * | 2020-06-18 | 2020-10-02 | 河海大学常州校区 | Ultra-short-term irradiation prediction method based on typical cloud shielding irradiation difference |
| CN111967660A (en) * | 2020-08-05 | 2020-11-20 | 河海大学 | Ultra-short-term photovoltaic prediction residual error correction method based on SVR |
| CN112633572A (en) * | 2020-12-21 | 2021-04-09 | 国网甘肃省电力公司电力科学研究院 | Day classification dual-model photovoltaic power generation combined prediction method based on prediction meteorological data |
| CN112987802A (en) * | 2021-02-22 | 2021-06-18 | 山东理工大学 | Photovoltaic power generation method and device, electronic equipment and storage medium |
| CN113033910A (en) * | 2021-04-09 | 2021-06-25 | 电子科技大学中山学院 | Photovoltaic power generation power prediction method, storage medium and terminal equipment |
| CN113240153A (en) * | 2021-04-09 | 2021-08-10 | 水发兴业能源(珠海)有限公司 | Photovoltaic power generation data prediction method and device, computing equipment and storage medium |
| CN115632396A (en) * | 2022-11-02 | 2023-01-20 | 广州瑞鑫智能制造有限公司 | Photovoltaic energy storage system for air compression station |
| CN116109007A (en) * | 2023-03-23 | 2023-05-12 | 阿里巴巴达摩院(杭州)科技有限公司 | Power generation power determination method, server and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106649919A (en) * | 2016-09-13 | 2017-05-10 | 上海电气电站环保工程有限公司 | Method and system for predicting carbon content of fly ash in coal-fired power plant boiler |
-
2017
- 2017-11-10 CN CN201711103682.9A patent/CN107766990B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106649919A (en) * | 2016-09-13 | 2017-05-10 | 上海电气电站环保工程有限公司 | Method and system for predicting carbon content of fly ash in coal-fired power plant boiler |
Non-Patent Citations (2)
| Title |
|---|
| 彭璐等: ""基于模糊C- 均值聚类的支持向量机"", 《工业控制计算机》 * |
| 赫卫国等: ""基于ARIMA和SVR的光伏电站超短期功率预测"", 《广东电力》 * |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109102102A (en) * | 2018-06-25 | 2018-12-28 | 南京邮电大学 | Based on the photovoltaic of multivariate phase space reconstruction and SVR power output short term prediction method |
| CN108985513A (en) * | 2018-07-17 | 2018-12-11 | 合肥工业大学 | Industrial enterprise's Electric Load Prediction System |
| CN109388863A (en) * | 2018-09-21 | 2019-02-26 | 国网天津市电力公司 | A kind of distributed photovoltaic power output power forecasting method based on ARIMA model |
| CN109388863B (en) * | 2018-09-21 | 2023-05-26 | 国网天津市电力公司 | ARIMA model-based distributed photovoltaic output power prediction method |
| CN109685257A (en) * | 2018-12-13 | 2019-04-26 | 国网青海省电力公司 | A kind of photovoltaic power generation power prediction method based on Support vector regression |
| CN109919353B (en) * | 2019-01-11 | 2022-07-05 | 杭州电子科技大学 | Distributed photovoltaic prediction method of ARIMA model based on spatial correlation |
| CN110009385A (en) * | 2019-01-11 | 2019-07-12 | 国网浙江杭州市临安区供电有限公司 | A kind of photovoltaic power generation user group division methods based on multifactor mapping |
| CN110009385B (en) * | 2019-01-11 | 2023-12-12 | 国网浙江杭州市临安区供电有限公司 | Multi-factor mapping-based photovoltaic power generation user group division method |
| CN109919353A (en) * | 2019-01-11 | 2019-06-21 | 杭州电子科技大学 | A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence |
| CN109871996A (en) * | 2019-02-14 | 2019-06-11 | 浙江中控太阳能技术有限公司 | A kind of tower photo-thermal power station mirror field cloud monitoring System and method for based on photovoltaic panel |
| CN110059441A (en) * | 2019-04-29 | 2019-07-26 | 广东电网有限责任公司 | A kind of photovoltaic plant modeling method and photovoltaic plant model output calibration method |
| CN110059441B (en) * | 2019-04-29 | 2022-07-19 | 广东电网有限责任公司 | Photovoltaic power station modeling method and photovoltaic power station model output correction method |
| CN110289605A (en) * | 2019-06-27 | 2019-09-27 | 昆明理工大学 | A kind of novel mixed compensation route intelligence direction protection method based on instantaneous power set of curves |
| CN111079980B (en) * | 2019-11-22 | 2021-06-29 | 天合云能源互联网技术(杭州)有限公司 | Optical power prediction method based on self-adaptive classification strategy and hybrid optimization SVR |
| CN111079980A (en) * | 2019-11-22 | 2020-04-28 | 天合云能源互联网技术(杭州)有限公司 | Optical power prediction method based on self-adaptive classification strategy and hybrid optimization SVR |
| CN110929953A (en) * | 2019-12-04 | 2020-03-27 | 国网山东省电力公司电力科学研究院 | Photovoltaic power station ultra-short term output prediction method based on cluster analysis |
| CN111738327A (en) * | 2020-06-18 | 2020-10-02 | 河海大学常州校区 | Ultra-short-term irradiation prediction method based on typical cloud shielding irradiation difference |
| CN111738327B (en) * | 2020-06-18 | 2022-09-13 | 河海大学常州校区 | Ultrashort-term irradiation prediction method based on typical cloud shielding irradiation difference |
| CN111967660B (en) * | 2020-08-05 | 2022-09-23 | 河海大学 | Ultra-short-term photovoltaic prediction residual error correction method based on SVR |
| CN111967660A (en) * | 2020-08-05 | 2020-11-20 | 河海大学 | Ultra-short-term photovoltaic prediction residual error correction method based on SVR |
| CN112633572A (en) * | 2020-12-21 | 2021-04-09 | 国网甘肃省电力公司电力科学研究院 | Day classification dual-model photovoltaic power generation combined prediction method based on prediction meteorological data |
| CN112987802A (en) * | 2021-02-22 | 2021-06-18 | 山东理工大学 | Photovoltaic power generation method and device, electronic equipment and storage medium |
| CN112987802B (en) * | 2021-02-22 | 2022-08-09 | 山东理工大学 | Photovoltaic power generation method and device, electronic equipment and storage medium |
| CN113240153A (en) * | 2021-04-09 | 2021-08-10 | 水发兴业能源(珠海)有限公司 | Photovoltaic power generation data prediction method and device, computing equipment and storage medium |
| CN113033910A (en) * | 2021-04-09 | 2021-06-25 | 电子科技大学中山学院 | Photovoltaic power generation power prediction method, storage medium and terminal equipment |
| CN115632396A (en) * | 2022-11-02 | 2023-01-20 | 广州瑞鑫智能制造有限公司 | Photovoltaic energy storage system for air compression station |
| CN116109007A (en) * | 2023-03-23 | 2023-05-12 | 阿里巴巴达摩院(杭州)科技有限公司 | Power generation power determination method, server and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107766990B (en) | 2021-06-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107766990A (en) | A kind of Forecasting Methodology of photovoltaic power station power generation power | |
| Du et al. | Multi-step ahead forecasting in electrical power system using a hybrid forecasting system | |
| Qing et al. | Hourly day-ahead solar irradiance prediction using weather forecasts by LSTM | |
| Xuan et al. | A multi-energy load prediction model based on deep multi-task learning and ensemble approach for regional integrated energy systems | |
| Raza et al. | An ensemble framework for day-ahead forecast of PV output power in smart grids | |
| De Giorgi et al. | Comparison of strategies for multi-step ahead photovoltaic power forecasting models based on hybrid group method of data handling networks and least square support vector machine | |
| Chen et al. | Online 24-h solar power forecasting based on weather type classification using artificial neural network | |
| Wu et al. | A novel hybrid model for short-term forecasting in PV power generation | |
| Yang et al. | Day-ahead forecasting of photovoltaic output power with similar cloud space fusion based on incomplete historical data mining | |
| Melzi et al. | Hourly solar irradiance forecasting based on machine learning models | |
| CN103390199A (en) | Photovoltaic power generation capacity/power prediction device | |
| Wang et al. | A comprehensive review for wind, solar, and electrical load forecasting methods | |
| CN112215428B (en) | Photovoltaic power generation power prediction method and system based on error correction and fuzzy logic | |
| WO2024051524A1 (en) | Joint prediction method and apparatus for hydraulic, wind and photovoltaic generation power | |
| Xiao et al. | A novel method based on time series ensemble model for hourly photovoltaic power prediction | |
| Eseye et al. | Short-term wind power forecasting using a double-stage hierarchical hybrid GA-ANN approach | |
| CN114169445A (en) | Day-ahead photovoltaic power prediction method, device and system based on CAE and GAN hybrid network | |
| CN112927097A (en) | Photovoltaic power generation short-term prediction method based on GRA-ABC-Elman model | |
| Xiaolan et al. | One-month ahead prediction of wind speed and output power based on EMD and LSSVM | |
| Kim et al. | Solar radiation forecasting based on the hybrid CNN-CatBoost model | |
| Yu et al. | Comparison of short-term solar irradiance forecasting methods when weather conditions are complicated | |
| Munsif et al. | CT-NET: A Novel Convolutional Transformer-Based Network for Short-Term Solar Energy Forecasting Using Climatic Information. | |
| Zhong et al. | PV power short-term forecasting model based on the data gathered from monitoring network | |
| CN116502074A (en) | Model fusion-based photovoltaic power generation power prediction method and system | |
| Abdellatif et al. | Forecast of Solar Photovoltaic Power Output Based on Polycrystalline Panel-based Employing Various Ensemble Machine Learning Methods |
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 |