CN107256437A - A kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology and system - Google Patents
A kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology and system Download PDFInfo
- Publication number
- CN107256437A CN107256437A CN201710340400.0A CN201710340400A CN107256437A CN 107256437 A CN107256437 A CN 107256437A CN 201710340400 A CN201710340400 A CN 201710340400A CN 107256437 A CN107256437 A CN 107256437A
- Authority
- CN
- China
- Prior art keywords
- data
- prediction
- irradiation level
- variance
- photovoltaic plant
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 32
- 238000004364 calculation method Methods 0.000 claims abstract description 8
- 230000001186 cumulative effect Effects 0.000 claims description 28
- 238000005286 illumination Methods 0.000 claims description 7
- 241001269238 Data Species 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract 1
- 230000009286 beneficial effect Effects 0.000 description 6
- 238000005259 measurement Methods 0.000 description 5
- 238000012937 correction Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000005096 rolling process Methods 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—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)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Photovoltaic Devices (AREA)
Abstract
The present invention relates to a kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology and system, this method includes：According to the support vector regression model after training, obtain predicting irradiation level；Similarity is calculated, while historical values data of weather forecast is ranked up according to Similarity value；The variance of historical values weather forecast prediction irradiation level and actual irradiation level is calculated, while being weighted to variance accumulative；Weight shared by the prediction irradiation level of weight and support vector regression model shared by evaluation weather forecast prediction irradiation level；Calculating obtains prediction time tpPhotovoltaic plant ultra-short term predicts irradiation level, and the invention further relates to a kind of forecasting system, the system includes：Train support vector regression model module, similarity calculation module, variance computing module, weight computation module, ultra-short term irradiation level prediction module.Precision of prediction can significantly be improved in prediction by the present invention, while computational efficiency and performance reach the requirement of prediction, the need for fully meeting photovoltaic generation Real-Time Scheduling.
Description
Technical field
The invention belongs to photovoltaic prediction field, more particularly to a kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology and it is
System.
Background technology
Solar energy power generating has become countries in the world competitively as the most technology of Practical significance in Solar use
Research and the focus of application.But the intrinsic height of photovoltaic generation relies on weather condition, randomness and fluctuation are big, and prediction is difficult
The characteristics of, again limit the large-scale application of photovoltaic generation.
The power output of photovoltaic generation depends greatly on the solar radiation quantity that photovoltaic panel can be received, and holds
It is vulnerable to the influence of weather conditions.The photovoltaic panel inclined-plane irradiation level surveyed installed in the environment monitor of photovoltaic plant, it is impossible to
Consider the fluctuation and randomness of irradiation level, its precision of prediction is relatively low, violent or predicted time yardstick is changed in weather conditions
It is worse compared with long-term prediction effect.Following several hours spokes are especially predicted based on inclined-plane irradiance measurement history value in the prior art
During illumination, fail the following several hours Changes in weather factors of reaction, so as to cause photovoltaic plant ultra-short term irradiation level forecasting inaccuracy
Really.
The content of the invention
The technical problems to be solved by the invention are：It is existing following several small based on the prediction of inclined-plane irradiance measurement history value
When irradiation level when, fail the following several hours Changes in weather factors of reaction, so as to cause photovoltaic plant ultra-short term irradiation level to be predicted
It is inaccurate.
, should the invention provides a kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology to solve technical problem above
Method comprises the following steps：
S1, utilizes the current time t of reading0And its actual irradiance data of the photovoltaic plant of certain time period is instructed before
Practice support vector regression model, the support vector regression model after being trained, while when utilizing the model after training to prediction
Carve tpIrradiation level be predicted, obtain the first prediction irradiation level；
S2, reads prediction time tpAnd tpFront and rear tp-Ts、tp+TsIt is first data of the numerical weather forecast at moment, current
Moment t0And its second data of the numerical weather forecast of certain time period, and calculate first data and each moment before
Second data similarity, while the second data of the numerical weather forecast are ranked up according to Similarity value,
Obtain the 3rd data of numerical weather forecast；
S3, reads actual irradiation level and the first prediction of the photovoltaic plant corresponding with each moment in the 3rd data
Irradiation level, and the first variance of the second prediction irradiation level and actual irradiation level of the first data is calculated respectively and described first pre-
The second variance of irradiation level and actual irradiation level is surveyed, while using forgetting factor respectively to the first variance and the second party
Difference is weighted accumulative；
S4, according to the first weighted cumulative variance and the second weighted cumulative variance drawn after weighted cumulative, calculates described the
The second weight shared by the first weight and the first prediction irradiation level shared by two prediction irradiation level；
S5, according to the described first prediction irradiation level, the second prediction irradiation level, the first weight and the second weight, is calculated
Obtain prediction time tpPhotovoltaic plant ultra-short term prediction irradiation level.
Beneficial effects of the present invention：In the prediction of ultra-short term irradiation level, existing algorithm is overcome using the method for the present invention
Only for actual measurement irradiation degree series be predicted, to irradiation level variation tendency hold ability the problem of, while in the present invention with
Based on the prediction irradiation level of support vector regression (SVR) model, the trend data provided with reference to numerical weather forecast, invention
Screening sample and irradiation level correction algorithm based on similarity and forgetting factor, can significantly be carried using the algorithm in prediction
High precision of prediction, while the computational efficiency and performance of the algorithm have reached the requirement of 15 minutes rolling forecasts of photovoltaic plant, completely
The need for meeting photovoltaic generation Real-Time Scheduling.
Further, in the S1, if the current time t read0And its actual spoke of certain time period photovoltaic plant before
In illumination data, there is loss or invalid data, then use the actual irradiance data with losing or invalid data is adjacent
Support vector regression model is trained as alternate data, and using alternate data.
Above-mentioned further beneficial effect：Exist lose or invalid data, using with lose or invalid data phase
Adjacent actual irradiance data can so cause data close, be not in the phenomenon of fracture as alternate data, while
It ensure that the raising of the precision of prediction in subsequent step.
Further, the S1 includes：
S11, reads current time t0And its actual irradiance data of the photovoltaic plant of certain time period before；
S12, is divided into continuous multigroup training sample data, and utilize every group of training sample by the actual irradiance data
Data train support vector regression model；
S13, using the support vector regression model after training to prediction time tpIrradiation level be predicted, obtain first
Predict irradiation level.
Above-mentioned further beneficial effect：It is divided into continuous multigroup, such purpose successively using by the data read
Having levels property of data can be caused, the error produced in data is greatly reduced, the follow-up middle data precision read can be caused
Greatly improve.
Further, the data read are divided into continuous multigroup training sample data successively in the S12, and utilized
Every group of training sample data train support vector regression model, wherein m is individual continuous in every group of training sample data
Actual irradiance data is trains the input of support vector regression model, and it is described to train support vector regression model to be output as
The m continuous actual follow-up w of irradiance datastepActual irradiance data, wherein, m be phase space number, wstepFor
wstepWalk prediction steps (wstep=1...Nts), NtsFor prediction steps,Predicted time length is Tfp, ultra-short term spoke
The time scale of illumination prediction is Ts。
Further, this method also includes between S2 and S3：Second data are ranked up according to Similarity value,
The data at each moment and prediction time t in first data in the 3rd data are calculated using Sigmoid functionspData
Between forgetting factor.
Above-mentioned further beneficial effect：Calculated after being ranked up to similarity using Sigmoid functions, after can causing
Continuous calculating is also surveyed precision and gradually stepped up, and is not in that data the phenomenon for omitting missing occur.
Further, in the S2, including first data and the current time t are calculated0And its before certain for the moment
Between any historical juncture t in sectionhAnd front and rear th-Ts、th+TsThe similarity of the data at moment.
Above-mentioned further beneficial effect：Choose in any historical juncture t in the periodhAnd front and rear th-Ts、th+Ts
The data at moment participate in the calculating of similarity and forgetting factor, are not to use all data, choose data with being spaced such that,
The error that can be reduced between data, greatly improves follow-up precision of prediction.
Further, also include in the S2：Read prediction time tpFront and rear tp-Ts、tp+TsThe Numerical Weather at moment is pre-
Report；
The historical juncture thAnd its front and rear th-Ts、th+TsSecond data at moment respectively with prediction time tpAnd prediction
Moment tpFront and rear tp-Ts、tp+TsFirst data at moment are corresponding one by one, and according to first in the numerical weather forecast
Data carry out variance calculating with the second data, while the obtained variance is weighted accumulative.
Further, in the S4, the first weight according to shared by below equation calculates the second prediction irradiation level is：
Wherein, se2For the first weighted cumulative variance, se1For the second weighted cumulative variance.
Further, in the S5, prediction time t is calculated according to below equationpPhotovoltaic plant ultra-short term prediction irradiation
Spend and be：
ModGhi=weights × Gtire+(1.0-weights)×GhiSVR
Wherein, weights is the first weight, GtireFor the second prediction irradiation level, GhiSVRFor the first prediction irradiation level.
The invention further relates to a kind of photovoltaic plant ultra-short term irradiation level forecasting system, the system includes：Train supporting vector
Regression model module, similarity calculation module, variance computing module, weight computation module, ultra-short term irradiation level prediction module；
The training support vector regression model module, it is used to utilize the current time t read0And its before certain for the moment
Between section photovoltaic plant actual irradiance data training support vector regression model, the support vector regression mould after being trained
Type, while using the model after training to prediction time tpIrradiation level be predicted, obtain the first prediction irradiation level；
The similarity calculation module, it is used to read prediction time tpNumerical weather forecast the first data, current
Moment t0And its second data of the numerical weather forecast of certain time period, and calculate first data and each moment before
Second data similarity, while the second data of the numerical weather forecast are ranked up according to Similarity value,
Obtain the 3rd data of numerical weather forecast；
The variance computing module, it is used to read the photovoltaic plant corresponding with each moment in the 3rd data
Actual irradiation level and the first prediction irradiation level, and calculate respectively the first data the second prediction irradiation level and actual irradiation level the
The second variance of one variance and the first prediction irradiation level and actual irradiation level, while using forgetting factor respectively to described
First variance and the second variance are weighted accumulative；
The weight computation module, it is used to be weighted according to the first weighted cumulative variance drawn after weighted cumulative and second
Accumulative variance, calculates the second power shared by the first weight and the first prediction irradiation level shared by the second prediction irradiation level
Weight；
The ultra-short term irradiation level prediction module, it is used for according to the described first prediction irradiation level, the second prediction spoke
Illumination, the first weight and the second weight, calculating obtain prediction time tpPhotovoltaic plant ultra-short term prediction irradiation level.
Beneficial effects of the present invention：In the prediction of ultra-short term irradiation level, existing algorithm is overcome using the method for the present invention
Only for actual measurement irradiation degree series be predicted, to irradiation level variation tendency hold ability the problem of, while in the present invention with
Based on the prediction irradiation level of support vector regression (SVR) model, the trend data provided with reference to numerical weather forecast, invention
Screening sample and irradiation level correction algorithm based on similarity and forgetting factor, can significantly be carried using the algorithm in prediction
High precision of prediction, while the computational efficiency and performance of the algorithm have reached the requirement of 15 minutes rolling forecasts of photovoltaic plant, completely
The need for meeting photovoltaic generation Real-Time Scheduling.
Brief description of the drawings
Fig. 1 is a kind of flow chart of photovoltaic plant ultra-short term irradiation level Forecasting Methodology of the present invention；
Fig. 2 is a kind of schematic diagram of photovoltaic plant ultra-short term irradiation level Forecasting Methodology of the present invention；
Fig. 3 is a kind of schematic diagram of photovoltaic plant ultra-short term irradiation level forecasting system of the present invention.
Embodiment
The principle and feature of the present invention are described below in conjunction with accompanying drawing, the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the present invention.
Embodiment 1
As depicted in figs. 1 and 2, a kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology in the embodiment of the present invention, the party
Method comprises the following steps：
S1, utilizes the current time t of reading0And its actual irradiance data of the photovoltaic plant of certain time period is instructed before
Practice support vector regression model, the support vector regression model after being trained, while when utilizing the model after training to prediction
Carve tpIrradiation level be predicted, obtain the first prediction irradiation level；
It is first to read current time t in the present embodiment 10And its actual irradiation of the photovoltaic plant of certain time period before
Then these data read are trained support vector regression model, the support vector regression mould after being trained by degrees of data
Type, such as：Read current time 8:00, and 8:The actual irradiance data of the photovoltaic plant of 3 days before 00, by current time
8:00 and the actual irradiance data of the photovoltaic plant of the period of 3 days before, support vector regression (SVR) model is trained,
Use newly-established support vector regression model and the first data current time 8:00 and the photovoltaic electric of certain time period before
Actual irradiance data of standing is as input, to prediction time tpIrradiation level be predicted, such as：It is 8 to prediction time:15 points
The irradiation level of clock is predicted, and obtains the first prediction irradiation level.
S2, reads prediction time tpAnd prediction time tpFront and rear tp-Ts、tp+TsThe of the numerical weather forecast at moment
One data, current time t0And its second data of the numerical weather forecast of certain time period, and calculate first number before
According to the similarity with second data at each moment, while by the second data of the numerical weather forecast according to similarity
Value is ranked up, and obtains the 3rd data of numerical weather forecast；The period conduct of nearest 31 days is used in this embodiment
Current time t0Certain time period before；
It is first to read prediction time t in the present embodiment 1pAnd prediction time tpFront and rear tp-Ts、tp+TsThe number at moment
It is worth the first data of weather forecast, such as：It is 8 for prediction time:The data of 30 points of selected numerical weather forecasts when
Carve as 8:15 points, 8:30 points and 8：45 points, also read current time t0And its numerical weather forecast of certain time period before
The second data, such as：Read current time 8:00, and the same day 8:The numerical weather forecast of 31 day this period before 00
Data, read after the data of these numerical weather forecasts, and calculate described the second of first data and each moment
The similarity of data, such as：It is 8 to prediction time:30 points, choose 8:15 points, 8:30 points and 8:45 points of Numerical Weather is pre-
The data of report and current time 8:00, and 8:The data of the numerical weather forecast of 31 day this period are carried out a pair before 00
The calculating of many data similarities, for example：8:15 points the first data correspondence the previous day 9:15 points of the second data, 8:30 points
First data correspondence the previous day 9:30 points of the second data, 8：45 points the first data correspondence the previous day 9：45 points of the second number
According to calculating 8 respectively:15 points, 8:30 points and 8 points of data of correspondence first of 45 minutes and each gas of the previous day the second value data weather forecast
As the variance of key element, the variance for obtaining each meteorological element is subjected to standard deviation normalized and weighting is handled, then by after processing
8：15 points, 8:30 points and 8：45 points of data are added up, and obtain prediction time 8:30 point of first data and the previous day 9：30
Divide the similarity of the second data.Obtain after similarity, then the second data of numerical weather forecast are arranged according to Similarity value
Sequence, obtains the 3rd data of numerical weather forecast.In addition, being chosen in this implementation in history day and 2 hours before and after prediction time
Within numerical weather forecast participate in the calculating of similarity and forgetting factor, other numerical weather forecasts will not participate in similarity and
The calculating of forgetting factor.
S3, reads actual irradiation level and the first prediction of the photovoltaic plant corresponding with each moment in the 3rd data
Irradiation level, and the first variance of the second prediction irradiation level and actual irradiation level of the first data is calculated respectively and described first pre-
The second variance of irradiation level and actual irradiation level is surveyed, while using forgetting factor respectively to the first variance and the second party
Difference is weighted accumulative；
It is first to read to that is to say that each moment is corresponding in the 3rd data with the second data after sequence in the present embodiment 1
Photovoltaic plant actual irradiation level and the prediction irradiation level of support vector regression model, after having read, then calculate respectively
The prediction irradiation level of one data is irradiated with the variance of actual irradiation level and the prediction irradiation level of support vector regression model with actual
The variance of degree, then using the prediction irradiation level using forgetting factor respectively to first data after calculating obtains variance
It is weighted with the variance of the variance of actual irradiation level and the prediction irradiation level of support vector regression model and actual irradiation level tired
Meter.
In addition, except calculating irradiation level in the present embodiment 1, in addition to：Environment temperature, wind speed, humidity, total amount of cloud, height
The variance of the meteorological elements such as cloud amount, middle cloud amount, low cloud cover, air pressure and wind direction, and the variance of 3 identical meteorological elements is added up.
S4, according to the first weighted cumulative variance and the second weighted cumulative variance drawn after weighted cumulative, calculates described the
The second weight shared by the first weight and the first prediction irradiation level shared by two prediction irradiation level；
Be weighted in the present embodiment 1 in above-mentioned steps S3 it is accumulative after obtain the prediction irradiation level of first data with
The weighted cumulative of the weighted cumulative variance of actual irradiation level and the prediction irradiation level and actual irradiation level of support vector regression model
Variance, weight and support vector regression model according to shared by the prediction irradiation level of obtained weighted cumulative variance the first data of calculating
Prediction irradiation level shared by weight.
S5, according to the described first prediction irradiation level, the second prediction irradiation level, the first weight and the second weight, is calculated
Obtain prediction time tpPhotovoltaic plant ultra-short term prediction irradiation level.
Above-mentioned steps S1 calculates the prediction irradiation level of obtained first data, supporting vector into S4 in the present embodiment
Weight shared by the prediction irradiation level of forecast of regression model irradiation level and first data and support vector regression model it is pre-
Weight shared by irradiation level is surveyed, then calculates and obtains prediction time tpPhotovoltaic plant ultra-short term predicts irradiation level.
It should be noted that it is T to predicted time length that this method, which is, in the embodiment of the present invention 1fp, time scale is (in advance
Survey step-length) it is TsUltra-short term irradiation level prediction, be divided into NtsStepIt is predicted, wherein to wstepWalk (wstep=
1...Nts) Forecasting Methodology is exactly steps of the above-mentioned described S1 into S5.
By above-mentioned method in the prediction of ultra-short term irradiation level, in order to overcome existing algorithm only for actual measurement irradiation level sequence
Row are predicted, and the shortcoming of ability is held to irradiation level variation tendency, are irradiated with the prediction of support vector regression (SVR) model
Based on degree, the trend data provided with reference to numerical weather forecast has invented the screening sample based on similarity and forgetting factor
With irradiation level correction algorithm.Practical application shows that the algorithm can significantly improve prediction in the prediction after advanced 1 hour
Precision, while the computational efficiency and performance of the algorithm have reached the requirement of 15 minutes rolling forecasts of photovoltaic plant, fully meets light
The need for lying prostrate generating Real-Time Scheduling.
Preferably, current time t in the S10And its actual irradiance data of the photovoltaic plant of certain time period before
Including：If in the actual irradiance data of photovoltaic plant, there is loss or invalid data, then use with losing or illegally counting
According to adjacent actual irradiance data as alternate data, alternate data is trained into support vector regression model.
There will be loss or invalid data in the present embodiment 1, use the reality with losing or invalid data is adjacent
Irradiance data can so cause data close, be not in the phenomenon of fracture, while after also ensure that as alternate data
The raising of precision of prediction in continuous step.
Preferably, above-mentioned steps S1 includes in the present embodiment 1：
S11, reads current time t0And its actual irradiance data of the photovoltaic plant of certain time period before；
S12, is divided into multigroup training sample data by actual irradiance data, and every group of training sample input is continuous for m
Actual irradiance data, training sample is output as the m continuous actual follow-up w of irradiance datastepIndividual actual irradiation level
Data, wherein, m is phase space number, can be set, and is T to predicted time lengthfp, time scale (prediction step) is TsIt is super
Short time irradiation degree is predicted, is divided into NtsStepIt is predicted, wstepFor wstepWalk (wstep=1...Nts) prediction step
Suddenly；
S13, using the support vector regression model after training to prediction time tpIrradiation level be predicted, obtain first
Predict irradiation level.
Used in the present embodiment 1 and the data read are divided into continuous multigroup successively, such purpose can cause
Having levels property of data, greatly reduces the error produced in data, and the follow-up middle data precision read can be caused to greatly improve.
Preferably, further relate to that the data read are divided into continuous multigroup successively in step s 12 in the present embodiment 1,
And support vector regression model is trained using grouped data, while training support vector regression model to be output as described m even
Continue the actual follow-up w of irradiance datastepIndividual actual irradiance data, wherein, m is phase space number, can be set, the present embodiment
In, value is 4.
Preferably, this method also includes between S2 and S3 in the present embodiment 1：By second data according to similarity
Value is ranked up, and the data at each moment and prediction time t in the 3rd data are calculated using Sigmoid functionspBetween something lost
Forget the factor.
After being ranked up in the present embodiment to phase knowledge and magnanimity can so that precision is also surveyed in follow-up calculating gradually steps up, and
It is not in the phenomenon that data omit missing.Wherein use Sigmoid functions as forgetting function, calculate forgetting factor W,
Sigmoid functional forms are as follows：
Wherein, ε is gradient, and its value is relevant with Effective Numerical weather forecast historical record bar number, the present embodiment value
For the sequence number that 0.00807949, i is the 3rd data sorting.
Preferably, described in the present embodiment 1 in S2, including first data and the current time t are calculated0And its
Any historical juncture t in certain time period beforehAnd front and rear th-Ts、th+TsThe similarity of the data at moment.
The calculating of calculating and the forgetting factor preferably for the similarity is in the current time t0And
In second data of the numerical weather forecast of its certain time period previous, choose in any historical juncture t in the periodhAnd
Front and rear th-Ts、th+TsThe data at moment participate in the calculating of similarity and forgetting factor.
In the embodiment of the present invention 1 chosen in any historical juncture t in the periodhAnd front and rear th-Ts、th+TsWhen
The data at quarter participate in the calculating of similarity and forgetting factor, are not to use all data, choose data with being spaced such that, can
To reduce the error between data, follow-up precision of prediction is greatly improved, such as：At above-mentioned described current time 8:Before 00
31 days this periods, selection is that this is the calculating that data in the period participate in similarity and forgetting factor, such as in advance
It is 8 to survey the moment：30 points, then what is selected is 8:15 points, 8:30 points and 8:45 point of first data, calculates the previous day 9:30 points this
The similarity calculating method of moment and prediction time is：Choose the previous day 9:15 points, 9:30 and 9:45 second data, are counted respectively
Calculate the first data 8:15 points and the second data day before yesterday 9:15 points, the first data 8:30 points and the second data day before yesterday 9:30 and first
Data 8:45 points and the second data day before yesterday 9:The variance of 45 each meteorological elements of fractional value weather forecast, by each meteorological element of calculating
Variance handled by standard deviation normalized and weighting, then by each meteorological element data accumulation after processing, obtain first
The data prediction moment 8:30 points and second data the previous day 9:30 points of Similarity value.
By the second data, the similarity of the second data and the first data prediction moment is calculated respectively according to the method described above, and
According to Similarity value, the second data are ranked up, the 3rd data are obtained.By the 3rd data of sequence, successively using Sigmoid
Function calculates forgetting factor W, Sigmoid functional form as follows as forgetting function：
Wherein, ε is gradient, and its value is relevant with Effective Numerical weather forecast historical record bar number, the present embodiment value
For the sequence number that 0.00807949, i is the 3rd data sorting.
Preferably, also include in the S2：Read prediction time tpFront and rear tp-Ts、tp+TsThe Numerical Weather at moment is pre-
Report；
The historical juncture thAnd its front and rear th-Ts、th+TsSecond data at moment respectively with prediction time tpAnd prediction
Moment tpFront and rear tp-Ts、tp+TsFirst data at moment are corresponding one by one, calculated according to each meteorologic factor of numerical weather forecast
The variance of each meteorologic factor, and variance is weighted accumulative.
What is chosen in the present embodiment 1 is to read prediction time tpFront and rear tp-Ts、tp+TsThe numerical weather forecast at moment,
Such as：Prediction time is 8:The data of 30 numerical weather forecast, and 8:15、8:The data difference of 45 numerical weather forecast
With the day before yesterday 9:15,9:30,9:The data at 45 these three time points are corresponding one by one, and computational methods are entered according to above-mentioned calculation formula
Row is calculated and can then obtained.
Preferably, in the S4, the first weight according to shared by below equation calculates the second prediction irradiation level is：
Wherein, se2For the first weighted cumulative variance, se1For the second weighted cumulative variance.
Weight shared by calculating the prediction irradiation level of the first data in the S4 in the present embodiment is：
Wherein, se2For the weighted cumulative variance of the prediction irradiation level and actual irradiation level of support vector regression model, se1For
The weighted cumulative variance of the prediction irradiation level and actual irradiation level of numerical weather forecast, weights is prediction time tpNumerical value day
Weight shared by gas forecast prediction irradiation level.
According to the different factors of influence irradiation level in the present embodiment 1, each meteorological element variance after adding up carries out comprehensive
Weighting processing is closed, wherein, the weight of irradiation level is 0.8, and the weight of temperature is 0.04, and the weight of wind speed is the 0.01, power of humidity
Weight is that the 0.01, weight of high cloud amount is that the 0.02, weight of middle cloud amount is that the 0.02, weight of low freight volume is 0.02, solar zenith angle
The weight of cosine is that the 0.04, weight of photovoltaic panel solar incident angle cosine is 0.04.
Preferably, in the S5, prediction time t is calculated according to below equationpPhotovoltaic plant ultra-short term prediction irradiation level
For：
ModGhi=weights × Gtire+(1.0-weights)×GhiSVR
Wherein, ModGhi is prediction time tpPhotovoltaic plant ultra-short term irradiation level, weights is the first weight, GtireFor
Second prediction irradiation level, GhiSVRFor the first prediction irradiation level.
, wherein it is desired to which parsing is that weights is that the first weight namely refers to prediction time t in the present embodiment 1pNumerical value
Weight, Gti shared by the prediction irradiation level of weather forecastreNamely refer to prediction time t for the second prediction irradiation levelpNumerical Weather is pre-
The prediction irradiation level of report, GhiSVRNamely refer to prediction time t for the first prediction irradiation levelpSupport vector regression model prediction spoke
Illumination.
Embodiment 2
As shown in Figure 3, a kind of photovoltaic plant ultra-short term irradiation level forecasting system, the system are further related in the present embodiment 2
Including：Train support vector regression model module, similarity calculation module, variance computing module, weight computation module, ultra-short term
Irradiation level prediction module；
The training support vector regression model module, it is used to utilize the current time t read0And its before certain for the moment
Between section photovoltaic plant actual irradiance data training support vector regression model, the support vector regression mould after being trained
Type, while using the model after training to prediction time tpIrradiation level be predicted, obtain the first prediction irradiation level；
The similarity calculation module, it is used to read prediction time tpNumerical weather forecast the first data, current
Moment t0And its second data of the numerical weather forecast of certain time period, and calculate first data and each moment before
Second data similarity, while the second data of the numerical weather forecast are ranked up according to Similarity value,
Obtain the 3rd data of numerical weather forecast；
The variance computing module, it is used to read the photovoltaic plant corresponding with each moment in the 3rd data
Actual irradiation level and the first prediction irradiation level, and calculate respectively the first data the second prediction irradiation level and actual irradiation level the
The second variance of one variance and the first prediction irradiation level and actual irradiation level, while using forgetting factor respectively to described
First variance and the second variance are weighted accumulative；
The weight computation module, it is used to be weighted according to the first weighted cumulative variance drawn after weighted cumulative and second
Accumulative variance, calculates the second power shared by the first weight and the first prediction irradiation level shared by the second prediction irradiation level
Weight；
The ultra-short term irradiation level prediction module, it is used for according to the described first prediction irradiation level, the second prediction spoke
Illumination, the first weight and the second weight, calculating obtain prediction time tpPhotovoltaic plant ultra-short term prediction irradiation level.
The content of all methods being mentioned in the present embodiment 2 in the system for quoting example 1 above.
In this manual, identical embodiment or example are necessarily directed to the schematic representation of above-mentioned term.
Moreover, specific features, structure, material or the feature of description can be in any one or more embodiments or example with suitable
Mode is combined.In addition, in the case of not conflicting, those skilled in the art can be by the difference described in this specification
The feature of embodiment or example and non-be the same as Example or example is combined and combined.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included in the scope of the protection.
Claims (10)
1. a kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology, it is characterised in that this method comprises the following steps：
S1, utilizes the current time t of reading0And its actual irradiance data training of the photovoltaic plant of certain time period before
Vector regression model is held, the support vector regression model after being trained, while using the model after training to prediction time tp
Irradiation level be predicted, obtain the first prediction irradiation level；
S2, reads prediction time tpAnd tpFront and rear tp-Ts、tp+TsFirst data of the numerical weather forecast at moment, current time
t0And its second data of the numerical weather forecast of certain time period, and calculate the institute of first data and each moment before
The similarity of the second data is stated, while the second data of the numerical weather forecast are ranked up according to Similarity value, is obtained
3rd data of numerical weather forecast；
S3, reads actual irradiation level and the first prediction irradiation of the photovoltaic plant corresponding with each moment in the 3rd data
Degree, and the second prediction irradiation level of the first data and the first variance of actual irradiation level and the first prediction spoke are calculated respectively
Illumination and the second variance of actual irradiation level, while being entered respectively to the first variance and the second variance using forgetting factor
Row weighted cumulative；
S4, according to the first weighted cumulative variance and the second weighted cumulative variance drawn after weighted cumulative, calculates described second pre-
Survey the second weight shared by the first weight and the first prediction irradiation level shared by irradiation level；
S5, according to the described first prediction irradiation level, the second prediction irradiation level, the first weight and the second weight, calculating is obtained
Prediction time tpPhotovoltaic plant ultra-short term prediction irradiation level.
2. Forecasting Methodology according to claim 1, it is characterised in that in the S1, if the current time t read0And its it
In the preceding actual irradiance data of certain time period photovoltaic plant, there is loss or invalid data, then use and lose or non-
The adjacent actual irradiance data of method data trains support vector regression model as alternate data, and using alternate data.
3. Forecasting Methodology according to claim 2, it is characterised in that the S1 includes：
S11, reads current time t0And its actual irradiance data of the photovoltaic plant of certain time period before；
S12, is divided into continuous multigroup training sample data, and utilize every group of training sample data by the actual irradiance data
Train support vector regression model；
S13, using the support vector regression model after training to prediction time tpIrradiation level be predicted, obtain the first prediction
Irradiation level.
4. Forecasting Methodology according to claim 3, it is characterised in that be divided into the data read successively in the S12
Continuous multigroup training sample data, and train support vector regression model, wherein institute using every group of training sample data
State m continuous actual irradiance datas in every group of training sample data and be the input of training support vector regression model, and instruct
Practice support vector regression model and be output as the m continuous actual follow-up w of irradiance datastepThe actual irradiation number of degrees
According to, wherein, m is phase space number, wstepFor wstepWalk prediction steps (wstep=1...Nts), NtsFor prediction steps,TfpFor predicted time length, TsThe time scale predicted for ultra-short term irradiation level.
5. according to any described Forecasting Methodology of Claims 1-4, it is characterised in that this method also includes between S2 and S3：
Second data are ranked up according to Similarity value, each moment in the 3rd data is calculated using Sigmoid functions
Data and prediction time t in first datapData between forgetting factor.
6. Forecasting Methodology according to claim 5, it is characterised in that in the S2, including calculate first data with
The current time t0And its any historical juncture t in certain time period beforehAnd front and rear th-Ts、th+TsThe phase of the data at moment
Like degree.
7. Forecasting Methodology according to claim 6, it is characterised in that also include in the S2：Read prediction time tpIt is front and rear
Tp-Ts、tp+TsThe numerical weather forecast at moment；
The historical juncture thAnd its front and rear th-Ts、th+TsSecond data at moment respectively with prediction time tpAnd prediction time
tpFront and rear tp-Ts、tp+TsFirst data at moment are corresponding one by one, and according to the first data in the numerical weather forecast
Variance calculating is carried out with the second data, while the obtained variance is weighted accumulative.
8. Forecasting Methodology according to claim 1, it is characterised in that in the S4, calculates second pre- according to below equation
Surveying the first weight shared by irradiation level is：
<mrow>
<mi>w</mi>
<mi>e</mi>
<mi>i</mi>
<mi>g</mi>
<mi>h</mi>
<mi>t</mi>
<mi>s</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>&Sigma;se</mi>
<mn>2</mn>
</msub>
</mrow>
<mrow>
<msub>
<mi>&Sigma;se</mi>
<mn>1</mn>
</msub>
<mo>+</mo>
<msub>
<mi>&Sigma;se</mi>
<mn>2</mn>
</msub>
</mrow>
</mfrac>
<mo>;</mo>
</mrow>
Wherein, se2For the first weighted cumulative variance, se1For the second weighted cumulative variance.
9. Forecasting Methodology according to claim 1, it is characterised in that in S5, prediction time t is calculated according to below equationp's
Photovoltaic plant ultra-short term predicts that irradiation level is：
ModGhi=weights × Gtire+(1.0-weights)×GhiSVR
Wherein, ModGhi is prediction time tpPhotovoltaic plant ultra-short term irradiation level, weights is the first weight, GtireFor second
Predict irradiation level, GhiSVRFor the first prediction irradiation level.
10. the prediction system of any described photovoltaic plant ultra-short term irradiation level Forecasting Methodology in a kind of utilization claim 1 to 9
System, it is characterised in that the system includes：Support vector regression model module, similarity calculation module, variance is trained to calculate mould
Block, weight computation module, ultra-short term irradiation level prediction module；
The training support vector regression model module, it is used to utilize the current time t read0And its certain time period before
Photovoltaic plant actual irradiance data training support vector regression model, the support vector regression model after being trained,
Simultaneously using the model after training to prediction time tpIrradiation level be predicted, obtain the first prediction irradiation level；
The similarity calculation module, it is used to read prediction time tpAnd tpFront and rear tp-Ts、tp+TsNumerical weather forecast
The first data, current time t0And its second data of the numerical weather forecast of certain time period, and calculate described before
The similarity of one data and second data at each moment, while by the second data of the numerical weather forecast according to phase
It is ranked up like angle value, obtains the 3rd data of numerical weather forecast；
The variance computing module, it is used for the reality for reading the photovoltaic plant corresponding with each moment in the 3rd data
Irradiation level and the first prediction irradiation level, and the first party of the second prediction irradiation level and actual irradiation level of the first data is calculated respectively
The second variance of poor and described first prediction irradiation level and actual irradiation level, while using forgetting factor respectively to described first
Variance and the second variance are weighted accumulative；
The weight computation module, it is used for according to the first weighted cumulative variance drawn after weighted cumulative and the second weighted cumulative
Variance, calculates the second weight shared by the first weight and the first prediction irradiation level shared by the second prediction irradiation level；
The ultra-short term irradiation level prediction module, its be used for according to described first prediction irradiation level, it is described second prediction irradiation level,
First weight and the second weight, calculating obtain prediction time tpPhotovoltaic plant ultra-short term prediction irradiation level.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710340400.0A CN107256437B (en) | 2017-05-15 | 2017-05-15 | Photovoltaic power station ultra-short-term irradiance prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710340400.0A CN107256437B (en) | 2017-05-15 | 2017-05-15 | Photovoltaic power station ultra-short-term irradiance prediction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107256437A true CN107256437A (en) | 2017-10-17 |
| CN107256437B CN107256437B (en) | 2020-11-27 |
Family
ID=60027611
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710340400.0A Active CN107256437B (en) | 2017-05-15 | 2017-05-15 | Photovoltaic power station ultra-short-term irradiance prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107256437B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108776843A (en) * | 2018-04-12 | 2018-11-09 | 天津大学 | Photovoltaic ultra-short term power forecasting method based on Boltzmann machine |
| CN109884896A (en) * | 2019-03-12 | 2019-06-14 | 河海大学常州校区 | A kind of photovoltaic tracking system optimization tracking based on similar day irradiation prediction |
| CN110956270A (en) * | 2019-10-18 | 2020-04-03 | 广州丰石科技有限公司 | XGBOOST algorithm-based solar irradiance prediction method |
| CN116167531A (en) * | 2023-04-26 | 2023-05-26 | 天津福天科技有限公司 | Photovoltaic power generation prediction method based on digital twin |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103390116A (en) * | 2013-08-07 | 2013-11-13 | 华北电力大学（保定） | Method for predicting electricity generation power of photovoltaic power station in step-by-step way |
| CN103559561A (en) * | 2013-11-13 | 2014-02-05 | 上海电气集团股份有限公司 | Super-short-term prediction method of photovoltaic power station irradiance |
| CN105373857A (en) * | 2015-11-27 | 2016-03-02 | 许昌许继软件技术有限公司 | Photovoltaic power station irradiance prediction method |
| US20160306906A1 (en) * | 2007-02-12 | 2016-10-20 | Locus Energy, Inc. | Solar irradiance modeling augmented with atmospheric water vapor data |
| CN106485093A (en) * | 2016-11-10 | 2017-03-08 | 哈尔滨工程大学 | Based on the solar irradiance time series synthetic method for improving Markov chain |
-
2017
- 2017-05-15 CN CN201710340400.0A patent/CN107256437B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160306906A1 (en) * | 2007-02-12 | 2016-10-20 | Locus Energy, Inc. | Solar irradiance modeling augmented with atmospheric water vapor data |
| CN103390116A (en) * | 2013-08-07 | 2013-11-13 | 华北电力大学（保定） | Method for predicting electricity generation power of photovoltaic power station in step-by-step way |
| CN103559561A (en) * | 2013-11-13 | 2014-02-05 | 上海电气集团股份有限公司 | Super-short-term prediction method of photovoltaic power station irradiance |
| CN105373857A (en) * | 2015-11-27 | 2016-03-02 | 许昌许继软件技术有限公司 | Photovoltaic power station irradiance prediction method |
| CN106485093A (en) * | 2016-11-10 | 2017-03-08 | 哈尔滨工程大学 | Based on the solar irradiance time series synthetic method for improving Markov chain |
Non-Patent Citations (2)
| Title |
|---|
| 代倩等: "基于天气类型聚类识别的光伏系统短期", 《中国电机工程学报》 * |
| 龚莺飞等: "光伏功率预测技术", 《电力系统自动化》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108776843A (en) * | 2018-04-12 | 2018-11-09 | 天津大学 | Photovoltaic ultra-short term power forecasting method based on Boltzmann machine |
| CN109884896A (en) * | 2019-03-12 | 2019-06-14 | 河海大学常州校区 | A kind of photovoltaic tracking system optimization tracking based on similar day irradiation prediction |
| CN110956270A (en) * | 2019-10-18 | 2020-04-03 | 广州丰石科技有限公司 | XGBOOST algorithm-based solar irradiance prediction method |
| CN116167531A (en) * | 2023-04-26 | 2023-05-26 | 天津福天科技有限公司 | Photovoltaic power generation prediction method based on digital twin |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107256437B (en) | 2020-11-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Random forest solar power forecast based on classification optimization | |
| Durrani et al. | Photovoltaic yield prediction using an irradiance forecast model based on multiple neural networks | |
| CN103020487B (en) | A kind of photovoltaic plant irradiance predicted value modification method | |
| Monteiro et al. | Short‐Term Forecasting Models for Photovoltaic Plants: Analytical versus Soft‐Computing Techniques | |
| CN105046374B (en) | A kind of power interval prediction technique based on core extreme learning machine model | |
| Tao et al. | Forecasting power output for grid-connected photovoltaic power system without using solar radiation measurement | |
| Paulescu et al. | Structured, physically inspired (gray box) models versus black box modeling for forecasting the output power of photovoltaic plants | |
| CN103390202B (en) | A kind of photovoltaic plant output power predicting method chosen based on set of metadata of similar data | |
| Schulz et al. | Post-processing numerical weather prediction ensembles for probabilistic solar irradiance forecasting | |
| Gao et al. | Predicting day-ahead solar irradiance through gated recurrent unit using weather forecasting data | |
| CN103106544B (en) | A kind of photovoltaic generation prognoses system based on T-S Fuzzy neutral net | |
| Yang et al. | Classification and summarization of solar irradiance and power forecasting methods: A thorough review | |
| CN107256437A (en) | A kind of photovoltaic plant ultra-short term irradiation level Forecasting Methodology and system | |
| CN112200377A (en) | Photovoltaic medium-long term power generation capacity forecasting method and device based on SARIMAX model | |
| CN104978611A (en) | Neural network photovoltaic power generation output prediction method based on grey correlation analysis | |
| CN101741085A (en) | Method for forecasting short-term wind-electricity power | |
| CN103955749A (en) | Sun direct radiation value prediction method and system | |
| CN106485093B (en) | Based on the solar irradiance time series synthetic method for improving Markov chain | |
| Perveen et al. | An ANFIS‐based model for solar energy forecasting and its smart grid application | |
| CN103473621A (en) | Wind power station short-term power prediction method | |
| CN103761578A (en) | Solar irradiation predicting method based on multiple linear regression | |
| CN106446440A (en) | Short-term photovoltaic generation power prediction method based on online sequential extreme learning machine | |
| CN105846425A (en) | Economic dispatching method based on general wind power forecasting error model | |
| Xie et al. | A backpropagation neural network improved by a genetic algorithm for predicting the mean radiant temperature around buildings within the long-term period of the near future | |
| Overen et al. | Solar energy resources and photovoltaic power potential of an underutilised region: a case of Alice, South Africa |
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 |