CN107085755A - A kind of photovoltaic plant short term power Forecasting Methodology and system - Google Patents
A kind of photovoltaic plant short term power Forecasting Methodology and system Download PDFInfo
- Publication number
- CN107085755A CN107085755A CN201710339205.6A CN201710339205A CN107085755A CN 107085755 A CN107085755 A CN 107085755A CN 201710339205 A CN201710339205 A CN 201710339205A CN 107085755 A CN107085755 A CN 107085755A
- Authority
- CN
- China
- Prior art keywords
- sample
- sample data
- forecast
- data
- mrow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 28
- 238000012545 processing Methods 0.000 claims abstract description 80
- 238000012216 screening Methods 0.000 claims description 43
- 241001269238 Data Species 0.000 claims description 14
- 230000001174 ascending effect Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 abstract description 6
- 238000012423 maintenance Methods 0.000 description 9
- 230000009286 beneficial effect Effects 0.000 description 6
- 238000012360 testing method Methods 0.000 description 5
- 230000005855 radiation Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000001154 acute effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000005286 illumination Methods 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
- 238000007670 refining Methods 0.000 description 1
- 230000001172 regenerating effect Effects 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of photovoltaic plant short term power Forecasting Methodology and system, this method comprises the following steps:Read current time and its data for the previous period be used as forecast sample data;Denoising is carried out to forecast sample data and subtracts sample processing, then is normalized;It is modeled, obtains photovoltaic plant short term power forecast model;Prediction time data are obtained, then are normalized;Photovoltaic plant short term power forecast model will be inputted after normalized, obtain output vector, then carry out inverse normalized, obtain predicted value.This kind of system is further related to, the system includes:Obtain sample data module, sample data processing module, supporting vector machine model modeling module, acquisition prediction data module, short-term forecast module.By instant invention overcomes traditional distance model contingency it is strong, to noise-sensitive and the problem of limit to overall sample distribution sensitivity, improve the accuracy rate of photovoltaic plant short term power prediction.
Description
Technical field
The invention belongs to power system field of photovoltaic power generation, more particularly to a kind of photovoltaic plant short term power Forecasting Methodology and
System.
Background technology
Photovoltaic generation is as cleaning, regenerative resource, and its application prospect is boundless.Due to the output of photovoltaic power station power generation
The features such as power has the intermittence of acute variation, randomness, thus photovoltaic electric station grid connection the balance of power network will be produced it is huge
Big influence, while also bringing great challenge for power scheduling.
With large-scale photovoltaic power station access area power network, each photovoltaic plant service condition is different, is power scheduling
The unified each photovoltaic plant power output of prediction of department brings difficulty.Therefore, it is necessary to photovoltaic plant, with minimum condition, in advance
Its power output situation is surveyed, the prediction data of each photovoltaic plant power output is provided for power scheduling, power grid energy management is ensured
System formulates operation plan.Existing Forecasting Methodology requires higher to each photovoltaic plant, and operability is relatively low, it is impossible to meet
Actual motion requirement, and also because scheduling the plant maintenance such as forced partial outage or inverter and caused by noise sample, it is impossible to
Reflect the objective law of photovoltaic plant active power and meteorologic factor, at the same traditional distance model contingency it is strong, to noise-sensitive
And there is limitation to overall sample distribution sensitivity.
The content of the invention
The technical problems to be solved by the invention are:Existing Forecasting Methodology requires higher to each photovoltaic plant, operable
Property it is relatively low, it is impossible to meet actual motion requirement, and also making an uproar caused by dispatching the plant maintenance such as forced partial outage or inverter
Sound sample, it is impossible to reflect the objective law of photovoltaic plant active power and meteorologic factor, while traditional distance model contingency
By force, there is limitation to noise-sensitive and to overall sample distribution sensitivity so that can not show that accurate prediction photovoltaic plant is short
Phase power.
To solve technical problem above, the invention provides a kind of photovoltaic plant short term power Forecasting Methodology, this method
Comprise the following steps:
S1, read current time and its horizontal irradiation level of aeropause of photovoltaic plant for the previous period, numerical value day
Gas is forecast and the data of photovoltaic electric station grid connection point active power are used as forecast sample data;
Forecast sample data, using screening sample algorithm, are carried out denoising and subtract sample processing, then to denoising and subtracted at sample by S2
Forecast sample data after reason are normalized;
S3, using the forecast sample data after normalized as modeling sample, is modeled to supporting vector machine model,
Obtain photovoltaic plant short term power forecast model;
S4, obtains the horizontal irradiation level of aeropause, the prediction time of numerical weather forecast of the photovoltaic plant of prediction time
Prediction time data are normalized by data;
S5, photovoltaic plant short term power forecast model described in the prediction time data input after normalized is obtained
Output vector, and inverse normalized is carried out to output vector, obtain the predicted value of photovoltaic electric station grid connection point active power.
Beneficial effects of the present invention:Photovoltaic plant short term power Forecasting Methodology proposed by the present invention, at screening sample
Reason technology, screened out because scheduling the plant maintenance such as forced partial outage or inverter and caused by noise sample so that after screening
Sample data preferably reflects the objective law that photovoltaic plant is exerted oneself with meteorologic factor, while it is even to overcome traditional distance model
Right property is strong, there is the problem of limiting to noise-sensitive and to overall sample distribution sensitivity.In addition, SVMs input because
In son in addition to convenient value weather forecast, on the air that also added reflection photovoltaic plant location solar radiation variations physics law
The horizontal irradiation level in boundary, using its result of the test of method and practical application of the present invention, hence it is evident that improve photovoltaic plant short term power
The accuracy rate of prediction.
Further, in the S3 using the forecast sample data after normalized as modeling sample, it is specifically will
The horizontal irradiation level of aeropause, numerical weather forecast in forecast sample data after normalized are used as modeling input sample
This, while exporting sample using the photovoltaic electric station grid connection point active power in the forecast sample data after normalized as modeling
This, using input sample and modeling output sample is modeled while being modeled to supporting vector machine model, obtains photovoltaic plant short
Phase power prediction model.
Above-mentioned further beneficial effect:Using input modeling sample as the input of supporting vector machine model, it will model defeated
Go out sample as the output of supporting vector machine model, while being modeled to supporting vector machine model, obtain photovoltaic plant short-term
Power prediction model.It can thus make it that the foundation of model is accurate, substantially increase the precision of follow-up short-term forecast.
Further, the S2 includes:
Forecast sample data are carried out preliminary screening, the sample data of lack part data are directly deleted by S21;
Forecast sample data are equally divided into N number of power interval by S22 by photovoltaic plant installed capacity, wherein, N is basis
The precision of photovoltaic plant installed capacity and screening sample is determined;
S23, the forecast sample data after preliminary screening are respectively divided on N number of power interval, N number of subsample is obtained
Collection;
S24, denoising is carried out to each sample data that each subsample is concentrated;
S25, carries out subtracting sample processing to the sample data after denoising;
S26, the sample data for subtracting sample processing is normalized.
Above-mentioned further beneficial effect:Screen out because scheduling the plant maintenance such as forced partial outage or inverter and caused by make an uproar
Sound sample so that the sample data after screening preferably reflects the objective law that photovoltaic plant is exerted oneself with meteorologic factor, favorably
In prediction subsequently to photovoltaic plant short term power, while also greatling save the time, the cumbersome of calculating is reduced, is also improved
The precision of prediction.
Further, the S24 includes:
S241, chooses a sample data in a sub- sample set, calculates the sample data and its in this power interval
The between class distance of the inter- object distance of its sample data, the sample data and other power interval sample datas, and by this power area
Between inter- object distance and the between class distances of other power intervals be ranked up;
S242, according to the inter- object distance and between class distance after sequence, the sample data is calculated using k nearest neighbor distance model
K nearest neighbor distance between k nearest neighbor distance and class in class;
S243, judges whether k nearest neighbor distance is more than k nearest neighbor distance between class in class, if k nearest neighbor distance is more than K between class in class
Nearest neighbor distance, then delete the sample, while subtracting 1 by the sample number of the corresponding subsample collection of the sample data;If it is not, then denoising is whole
Only.
Further, the S25 includes:
S251, to the importance W of all sample datas after denoisingiIt is entered as 0, deleted sample number RiAssignment
For 1;
S252, calculates the importance W of all sample datasi, according to importance WiBy all sample data ascending sorts;
S253, selection importance WiMinimum sample data i and its nearest samples data j, by minimum sample data i
Delete, and sample number is subtracted 1;
Whether S254, judgement sample number meets sample reduction requirement, if meeting, and subtracts sample termination;Otherwise, continue executing with
S252。
Further, in the S242:K nearest neighbor distance between k nearest neighbor distance and class is calculated in the class, it is respectively in class
Euclidean distance dW(i)The Euclidean distance d between classB(i):
Wherein, di,jRepresent point xiWith point xjBetween Euclidean distance, Jw(i) it is preceding K farthest sample data i same power
The set of interval sample number, JB(i) it is the set of preceding K nearest sample data i other power interval sample numbers.
Further, sample data i importance W is calculated in the S252iFor:
Wherein, RiFor the deleted number of samples in sample data i periphery, λ be its span of coefficient of similarity for [0,
1], S (i) is sample data i same power interval similarity, dB(i)For Euclidean distance between class.
The invention further relates to a kind of photovoltaic plant short term power forecasting system, the system includes:Acquisition sample data module,
Sample data processing module, supporting vector machine model modeling module, acquisition prediction data module, short-term forecast module;
The acquisition sample data module, its be used to reading current time and its photovoltaic plant for the previous period it is big
The data of the horizontal irradiation level in the gas upper bound, numerical weather forecast and photovoltaic electric station grid connection point active power are used as forecast sample number
According to;
The sample data processing module, its be used for utilize screening sample algorithm, to forecast sample data carry out denoising and
Subtract sample processing, then denoising and the forecast sample data subtracted after sample processing are normalized;
The supporting vector machine model modeling module, it is used to regard the forecast sample data after normalized as modeling
Sample, is modeled to supporting vector machine model, obtains photovoltaic plant short term power forecast model;
It is described acquisition prediction data mould, its be used for the photovoltaic plant for obtaining prediction time the horizontal irradiation level of aeropause,
Prediction time data are normalized by the prediction time data of numerical weather forecast;
The short-term forecast module, it is used to the prediction data after normalized inputting the short-term work(of photovoltaic plant
Rate forecast model, obtains output vector, and inverse normalized is carried out to output vector, obtains photovoltaic electric station grid connection point wattful power
The predicted value of rate.
Beneficial effects of the present invention:Photovoltaic plant short term power Forecasting Methodology proposed by the present invention, at screening sample
Reason technology, screened out because scheduling the plant maintenance such as forced partial outage or inverter and caused by noise sample so that after screening
Sample data preferably reflects the objective law that photovoltaic plant is exerted oneself with meteorologic factor, while it is even to overcome traditional distance model
Right property is strong, there is the problem of limiting to noise-sensitive and to overall sample distribution sensitivity.In addition, SVMs input because
In son in addition to convenient value weather forecast, on the air that also added reflection photovoltaic plant location solar radiation variations physics law
The horizontal irradiation level in boundary, using its result of the test of method and practical application of the present invention, hence it is evident that improve photovoltaic plant short term power
The accuracy rate of prediction.
Further, the supporting vector machine model modeling module, it is specifically by the forecast sample number after normalized
The horizontal irradiation level of aeropause, numerical weather forecast in is as modeling input sample, while will be pre- after normalized
Photovoltaic electric station grid connection point active power in test sample notebook data exports sample as modeling, using modeling input sample and model defeated
Go out sample while being modeled to supporting vector machine model, obtain photovoltaic plant short term power forecast model.
Above-mentioned further beneficial effect:Using input modeling sample as the input of supporting vector machine model, it will model defeated
Go out sample as the output of supporting vector machine model, while being modeled to supporting vector machine model, obtain photovoltaic plant short-term
Power prediction model.It can thus make it that the foundation of model is accurate, substantially increase the precision of follow-up short-term forecast.
Further, the sample data processing module includes:At first sample data processing unit, the second sample data
Manage unit, the 3rd sample data processing unit, the 4th sample data processing unit, the 5th sample data processing unit, the 6th sample
Notebook data processing unit;
The first sample data processing unit, for carrying out preliminary screening to forecast sample data, by lack part number
According to sample data directly delete;
The second sample data processing unit, for forecast sample data to be equally divided into by photovoltaic plant installed capacity
N number of power interval, wherein, N is determined according to the precision of photovoltaic plant installed capacity and screening sample;
The 3rd sample data processing unit, for by the forecast sample data after preliminary screening, being respectively divided N number of
On power interval, N number of subsample collection is obtained;
The 4th sample data processing unit, for being gone to each sample data that each subsample is concentrated
Make an uproar processing;
The 5th sample data processing unit, for carrying out subtracting sample processing to the sample data after denoising;
The 6th sample data processing unit, for carrying out subtracting sample processing to the sample data after denoising.
Above-mentioned further beneficial effect:Screen out because scheduling the plant maintenance such as forced partial outage or inverter and caused by make an uproar
Sound sample, the sample data after screening preferably reflects the objective law that photovoltaic plant is exerted oneself with meteorologic factor, after being conducive to
The continuous prediction to photovoltaic plant short term power, while also greatling save the time, reduces the cumbersome of calculating, also improves prediction
Precision.
Brief description of the drawings
Fig. 1 is a kind of schematic diagram of photovoltaic plant short term power Forecasting Methodology of the present invention;
Fig. 2 is a kind of flow chart of photovoltaic plant short term power Forecasting Methodology of the present invention;
Fig. 3 is denoising of the invention and subtracts the flow chart that sample is handled;
A kind of schematic diagram of photovoltaic plant short term power forecasting system of Fig. 4 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 short term power Forecasting Methodology in the present embodiment 1, this method is included such as
Lower step:
S1, read current time and its horizontal irradiation level of aeropause of photovoltaic plant for the previous period, numerical value day
Gas is forecast and the data of photovoltaic electric station grid connection point active power are used as forecast sample data;
In the present embodiment 1 we be first start read current time and its photovoltaic plant for the previous period air
The data of the horizontal irradiation level in the upper bound, numerical weather forecast and photovoltaic electric station grid connection point active power, wherein numerical weather forecast
Data include:Solar shortwave radiation, scattering irradiance, direct projection irradiation level, high cloud amount, middle cloud amount, the historical data of low cloud cover,
Then all data these read again are used as forecast sample data.
Forecast sample data, using screening sample algorithm, are carried out denoising and subtract sample processing, then to denoising and subtracted at sample by S2
Forecast sample data after reason are normalized;
We are to use the screening sample algorithm based on power interval nearest neighbor distance in the present embodiment 1, to above-mentioned steps
All forecast sample data are read in S1, denoising is carried out to these forecast sample data and subtracts sample processing, sieve can be caused
Except because the scheduling plant maintenance such as forced partial outage or inverter and caused by noise sample, the sample after screening preferably reflects
The objective law that photovoltaic plant is exerted oneself with meteorologic factor, at the same overcome traditional distance model contingency it is strong, to noise-sensitive with
And there is the problem of limiting to overall sample distribution sensitivity.After we carry out denoising to forecast sample data and subtract sample processing,
New forecast sample data are obtained, then new forecast sample data are normalized again.
S3, using the forecast sample data after normalized as modeling sample, is modeled to supporting vector machine model,
Obtain photovoltaic plant short term power forecast model;
In the present embodiment 1 we be using the forecast sample data after being normalized in above-mentioned steps S2 as
Modeling sample, wherein preferably, we are that it is specifically will as modeling sample using the forecast sample data after normalized
The horizontal irradiation level of aeropause, the forecast sample data of numerical weather forecast after normalized in forecast sample data,
These forecast sample data are as modeling input sample, while by the photovoltaic plant in forecast sample data after normalized simultaneously
The forecast sample data of site active power, export sample using these forecast sample data as modeling, then again build these
Mould input sample and modeling output sample are modeled to supporting vector machine model simultaneously, obtain the prediction of photovoltaic plant short term power
Model.
S4, obtains the horizontal irradiation level of aeropause, the prediction time of numerical weather forecast of the photovoltaic plant of prediction time
Prediction time data are normalized by data;
We are the horizontal irradiation level of aeropause, the numerical value for the photovoltaic plant for first obtaining prediction time in the present embodiment 1
The prediction time data of the prediction time data of weather forecast, wherein numerical weather forecast include:Solar shortwave radiation, scattering spoke
Illumination, direct projection irradiation level, high cloud amount, the prediction time data of middle cloud amount and low cloud cover, are obtained after these prediction time data,
These obtained prediction time data are normalized again for we.So make it that the prediction photovoltaic plant subsequently obtained is short
Phase power is more accurate.
S5, photovoltaic plant short term power forecast model described in the prediction time data input after normalized is obtained
Output vector, and inverse normalized is carried out to output vector, obtain the predicted value of photovoltaic electric station grid connection point active power.
We are to above-mentioned by the prediction time data input after normalized in above-mentioned steps S3 in the present embodiment 1
In the photovoltaic plant short term power forecast model obtained in step S3, the output vector that can be exported, we are defeated by these again
Outgoing vector carries out inverse normalized, and finally obtaining us needs to obtain the prediction for the photovoltaic electric station grid connection point active power come
Value, that is, the photovoltaic plant short term power described in us.
Preferably, as shown in figure 3, we in the S2 also further to refining in the present embodiment 1, wherein described
S2 includes:
Forecast sample data are carried out preliminary screening, the sample data of lack part data are directly deleted by S21;
Forecast sample data are equally divided into N number of power interval by S22 by photovoltaic plant installed capacity, wherein, N is basis
The precision of photovoltaic plant installed capacity and screening sample is determined;
S23, the forecast sample data after preliminary screening are respectively divided on N number of power interval, N number of subsample is obtained
Collection;
S24, denoising is carried out to each sample data that each subsample is concentrated;
S25, carries out subtracting sample processing to the sample data after denoising;
S26, the sample data for subtracting sample processing is normalized.
In the present embodiment 1 preferably, we are that the forecast sample data obtained in above-mentioned steps S1 are carried out tentatively
Screening, wherein lack part data item is exactly caused the moment sample data can not directly delete correspondingly by preliminary screening
Remove, be so screen out because scheduling the plant maintenance such as forced partial outage or inverter and caused by noise sample, the sample after screening
Data preferably reflect the objective law that photovoltaic plant is exerted oneself with meteorologic factor, are conducive to follow-up to photovoltaic plant short term power
Prediction, while also greatling save the time, reduce the cumbersome of calculating, also improve the precision of prediction.
Wherein S22 also carries out subregion while preliminary screening is carried out to forecast sample data to forecast sample data, will
Forecast sample data are equally divided into N number of power interval by photovoltaic plant installed capacity, wherein, N is to be installed to hold according to photovoltaic plant
Amount determines that we are to choose photovoltaic plant installed capacity for 50Mw in the present embodiment 1, N values with the precision of screening sample
For 50;
After execution of step S22, the forecast sample data after above-mentioned S21 preliminary screenings are respectively divided N number of for we
On power interval, N number of subsample collection is obtained;Concentrated due to each subsample and include many sample datas, it would be desirable to
Denoising is carried out to each sample data that each subsample is concentrated;Then the sample data after denoising is entered again
Row subtracts sample processing;Finally the sample data for subtracting sample processing is normalized again for we.
Preferably, described in the present embodiment 1 in S24, we also further refine, including:
S241, chooses a sample data in a sub- sample set, calculates the sample data and its in this power interval
The between class distance of the inter- object distance of its sample data, the sample data and other power interval sample datas, and by this power area
Between inter- object distance and the between class distances of other power intervals be ranked up;
S242, according to the inter- object distance and between class distance after sequence, is calculated in the class of the sample using k nearest neighbor distance model
K nearest neighbor distance between k nearest neighbor distance and class;
S243, judges whether k nearest neighbor distance is more than k nearest neighbor distance between class in class, if k nearest neighbor distance is more than K between class in class
Nearest neighbor distance, then delete the sample, while subtracting 1 by the sample number of the corresponding subsample collection of the sample data;If it is not, then denoising is whole
Only.
We are first to choose a sub- sample set in N number of subsample collection in above-mentioned preferably S24, in a son wherein
A sample data is chosen in sample set, calculate the inter- object distance of other sample datas in the sample data and this power interval,
The between class distance of the sample data and other power interval sample datas, and by this power interval inter- object distance and other power areas
Between between class distance be ranked up.Such as:We choose the first sample data of the first subsample concentration, first subsample collection
Belong in the first power interval, first subsample collection includes:First sample data, the second sample data, the 3rd sample
Data ...;N number of power interval includes:First power interval, the second power interval, the 3rd power interval ...;We calculate
The inter- object distance of the sample data and other sample datas in this power interval is exactly by first sample data and the first power area
Between in other sample datas of the first subsample collection calculated, such as:First sample data and the second sample data, the 3rd sample
The calculating of notebook data, they are the inter- object distances for belonging to the sample data and other sample datas in this power interval, for me
The between class distance of the sample data that calculates and other power interval sample datas be exactly by first sample data and the second work(
The sample data of subsample collection during rate is interval is calculated.Inter- object distance and between class distance are being calculated, we incite somebody to action and will again
This power interval inter- object distance and other power interval between class distances are ranked up;
After sequence, we further according to this power interval inter- object distance after sequence and other power interval between class distances,
K nearest neighbor between k nearest neighbor distance and the class of other power intervals is calculated in the class of this power interval of the sample using k nearest neighbor distance model
Distance;
Calculated at us and obtain in class k nearest neighbor distance between k nearest neighbor distance and class, it would be desirable to judge k nearest neighbor distance in class
Whether k nearest neighbor distance between class is more than, if k nearest neighbor distance is more than between the class of other power intervals in the class of this power interval of the sample
K nearest neighbor distance, then delete the sample, while subtracting 1 by the sample number of the corresponding subsample collection of the sample data;If it is not, then denoising
Terminate.
Preferably, described in the present embodiment 1 in S25, we also further refine, including:
S251, to the importance W of all sample datas after denoisingiIt is entered as 0, deleted sample number RiAssignment
For 1;
S252, calculates the importance W of all sample datasi, according to importance WiBy all sample data ascending sorts;
S253, S253, selection importance WiMinimum sample data i and its nearest samples data j, by minimum sample
Data delete i, and subtract 1 by sample number;Wherein, Rj=Rj+Ri+1;
Whether S254, judgement sample number meets sample reduction requirement, if meeting, and subtracts sample termination;Otherwise, continue executing with
S252。
It should be noted that the sample data that we terminate in the present embodiment 1 is calculated, sample data therein terminates bar
Part is that remaining sample size is total sample
It is understood that being that the definition in the prior art for k nearest neighbor distance is:First give m dimension sample spaces RmAnd space
Point xi,
Wherein, di,jRepresent point xiWith point xjBetween Euclidean distance, J (i) be range points xiThe collection of nearest preceding K point
Close.
It is therefore preferred that described in the present embodiment 1 in S242:Calculate in the class k nearest neighbor between k nearest neighbor distance and class
Distance, it is respectively Euclidean distance d in classW(i)The Euclidean distance d between classB(i):
Wherein, di,jRepresent point xiWith point xjBetween Euclidean distance, Jw(i) it is preceding K farthest sample data i same power
The set of interval sample number, JB(i) it is the set of preceding K nearest sample data i other power interval sample numbers.
After above-mentioned step is carried out, in addition to:Wherein described step S252 calculates the importance W of all sample datasi,
Evaluated using following formula:
Wherein, RiFor the deleted number of samples in sample i periphery;λ is coefficient of similarity, the shadow for adjusting similarity
The degree of sound, span is [0,1], in the present embodiment, and λ values are the same power interval similarity that 0.5, S (i) is sample i,
dB(i)For Euclidean distance between class;
Wherein, S (i) is defined as sample i same power interval similarity:
Sample i same power interval similarity S (i) is that similarity is most between sample i and all same power interval samples
Big value;
Wherein, the similarity between sample i and sample j is defined as:
For sample K (Xi,Yj), O is the origin of coordinates, the similarity S between sample i and sample ji,jWith vectorWith to
AmountIncluded angle cosine represent:
What we also needed to parsing in the present embodiment 1 is that normalized is in the S26:
Wherein, xnFor data original value, xminFor the minimum value in data original value, xmaxMaximum in data original value
Value, XnFor the data after normalized.
Also include in the present embodiment 1:SVMs (SVM) model selection radial direction base (RBF) core is used as supporting vector
The kernel function of machine, RBF cores are:
K(Xi,Xj)=exp (- γ | | Xi-Xj||)2
The degree of accuracy that predicts the outcome of the inventive method is high, meets the requirement that operation of power networks is used;Error statistics explanation:
Predicting power of photovoltaic plant accuracy rate:
Wherein:PmkFor the actual power at K moment, PpkFor the pre- power scale at K moment, N is the total moment number of daily forecast, CapFor
Photovoltaic plant runs installed capacity.
Embodiment 2
The present embodiment 2 further relates to a kind of photovoltaic plant short term power forecasting system, and the system includes:Obtain sample data mould
Block, sample data processing module, supporting vector machine model modeling module, acquisition prediction data module, short-term forecast module;
The acquisition sample data module, its be used to reading current time and its photovoltaic plant for the previous period it is big
The data of the horizontal irradiation level in the gas upper bound, numerical weather forecast and photovoltaic electric station grid connection point active power are used as forecast sample number
According to;
The sample data processing module, its be used for utilize screening sample algorithm, to forecast sample data carry out denoising and
Subtract sample processing, then denoising and the forecast sample data subtracted after sample processing are normalized;
The supporting vector machine model modeling module, it is used to regard the forecast sample data after normalized as modeling
Sample, is modeled to supporting vector machine model, obtains photovoltaic plant short term power forecast model;
It is described acquisition prediction data mould, its be used for the photovoltaic plant for obtaining prediction time the horizontal irradiation level of aeropause,
Prediction time data are normalized by the prediction time data of numerical weather forecast;
The short-term forecast module, it is used to the prediction data after normalized inputting the short-term work(of photovoltaic plant
Rate forecast model, obtains output vector, and inverse normalized is carried out to output vector, obtains photovoltaic electric station grid connection point wattful power
The predicted value of rate.
Preferably, the supporting vector machine model modeling module described in the present embodiment 2, it is specifically for by normalized
The horizontal irradiation level of aeropause, numerical weather forecast in forecast sample data afterwards is as modeling input sample, while will return
The photovoltaic electric station grid connection point active power in forecast sample data after one change processing is defeated using modeling as modeling output sample
Enter sample and modeling output sample while being modeled to supporting vector machine model, obtain photovoltaic plant short term power prediction mould
Type.
Preferably, sample data processing module includes described in the present embodiment 2:First sample data processing unit,
Two sample data processing units, the 3rd sample data processing unit, the 4th sample data processing unit, the processing of the 5th sample data
Unit, the 6th sample data processing unit;
The first sample data processing unit, for carrying out preliminary screening to forecast sample data, by lack part number
According to sample data directly delete;
The second sample data processing unit, for forecast sample data to be equally divided into by photovoltaic plant installed capacity
N number of power interval, wherein, N is determined according to the precision of photovoltaic plant installed capacity and screening sample;
The 3rd sample data processing unit, for by the forecast sample data after preliminary screening, being respectively divided N number of
On power interval, N number of subsample collection is obtained;
The 4th sample data processing unit, for being gone to each sample data that each subsample is concentrated
Make an uproar processing;
The 5th sample data processing unit, for carrying out subtracting sample processing to the sample data after denoising;
The 6th sample data processing unit, for carrying out subtracting sample processing to the sample data after denoising.
In the present embodiment 2 preferably, we are the pre- test samples to being obtained in acquisition sample data module described above
Notebook data carries out preliminary screening, and lack part data item is exactly caused the moment can not sample correspondingly by wherein preliminary screening
Notebook data is directly deleted, be so screen out because scheduling the plant maintenance such as forced partial outage or inverter and caused by noise sample,
Sample data after screening preferably reflects the objective law that photovoltaic plant is exerted oneself with meteorologic factor, is conducive to follow-up to photovoltaic
The prediction of power station short term power, while also greatling save the time, reduces the cumbersome of calculating, also improves the precision of prediction.
Wherein described second sample data processing unit to forecast sample data while preliminary screening is carried out, also to pre-
Test sample notebook data carries out subregion, and forecast sample data are equally divided into N number of power interval by photovoltaic plant installed capacity, wherein, N
It is to be determined according to the precision of photovoltaic plant installed capacity and screening sample, we are to choose photovoltaic plant in the present embodiment 1
Installed capacity is 50Mw, and N values are 50;
After the second sample data processing unit has been performed, we are by first sample data processing unit described above
Forecast sample data after preliminary screening, are respectively divided on N number of power interval, obtain N number of subsample collection;Due to each increment
This concentration all includes many sample datas, it would be desirable to which each sample data that each subsample is concentrated is gone
Make an uproar processing;Then the sample data after denoising is carried out again subtracting sample processing;Finally we are again to subtracting the sample number that sample is handled
According to being normalized.
The system being related in the present embodiment 2 is, it is necessary to which explanation is that method described above can be used to be used
Content.
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 short term power Forecasting Methodology, it is characterised in that this method comprises the following steps:
S1, read current time and its photovoltaic plant for the previous period the horizontal irradiation level of aeropause, Numerical Weather it is pre-
Respond with and photovoltaic electric station grid connection point active power data as forecast sample data;
Forecast sample data, using screening sample algorithm, are carried out denoising and subtract sample processing, then to denoising and subtracted after sample processing by S2
Forecast sample data be normalized;
S3, using the forecast sample data after normalized as modeling sample, is modeled to supporting vector machine model, obtains
Photovoltaic plant short term power forecast model;
S4, obtains the horizontal irradiation level of aeropause, the prediction time data of numerical weather forecast of the photovoltaic plant of prediction time,
Prediction time data are normalized;
S5, photovoltaic plant short term power forecast model described in the prediction time data input after normalized is exported
Vector, and inverse normalized is carried out to output vector, obtain the predicted value of photovoltaic electric station grid connection point active power.
2. Forecasting Methodology according to claim 1, it is characterised in that by the forecast sample after normalized in the S3
Data as modeling sample, its be specifically by the horizontal irradiation level of aeropause in the forecast sample data after normalized,
Numerical weather forecast is as modeling input sample, while by the photovoltaic electric station grid connection in the forecast sample data after normalized
Point active power as modeling export sample, using model input sample and modeling output sample simultaneously to supporting vector machine model
It is modeled, obtains photovoltaic plant short term power forecast model.
3. Forecasting Methodology according to claim 1 or 2, it is characterised in that the S2 includes:
Forecast sample data are carried out preliminary screening, the sample data of lack part data are directly deleted by S21;
Forecast sample data are equally divided into N number of power interval by S22 by photovoltaic plant installed capacity, wherein, N is according to photovoltaic
Installed capacity of power station and the precision of screening sample are determined;
S23, the forecast sample data after preliminary screening are respectively divided on N number of power interval, obtain N number of subsample collection;
S24, denoising is carried out to each sample data that each subsample is concentrated;
S25, carries out subtracting sample processing to the sample data after denoising;
S26, the sample data for subtracting sample processing is normalized.
4. Forecasting Methodology according to claim 3, it is characterised in that the S24 includes:
S241, chooses a sample data in a sub- sample set, calculates the sample data and other samples in this power interval
The between class distance of the inter- object distance of notebook data, the sample data and other power interval sample datas, and by this power interval
The between class distance of inter- object distance and other power intervals is ranked up;
S242, according to the inter- object distance and between class distance after sequence, is calculated in the class of the sample data using k nearest neighbor distance model
K nearest neighbor distance between k nearest neighbor distance and class;
S243, judges whether k nearest neighbor distance is more than k nearest neighbor distance between class in class, if k nearest neighbor distance is more than k nearest neighbor between class in class
Distance, then delete the sample, while subtracting 1 by the sample number of the corresponding subsample collection of the sample data;If it is not, then denoising is terminated.
5. Forecasting Methodology according to claim 3, it is characterised in that the S25 includes:
S251, to the importance W of all sample datas after denoisingiIt is entered as 0, deleted sample number RiIt is entered as 1;
S252, calculates the importance W of all sample datasi, according to importance WiBy all sample data ascending sorts;
S253, selection importance WiMinimum sample data i and its nearest samples data j, minimum sample data i is deleted,
And sample number is subtracted 1;
Whether S254, judgement sample number meets sample reduction requirement, if meeting, and subtracts sample termination;Otherwise, S252 is continued executing with.
6. Forecasting Methodology according to claim 4, it is characterised in that in the S242:Calculate k nearest neighbor distance in the class
The k nearest neighbor distance between class, it is respectively to calculate Euclidean distance d in classW(i)The Euclidean distance d between classB(i):
<mrow>
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>d</mi>
<mi>w</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>k</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<msub>
<mi>d</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
</msub>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msub>
<mi>J</mi>
<mi>w</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>d</mi>
<mi>B</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>k</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<msub>
<mi>d</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
</msub>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msub>
<mi>J</mi>
<mi>B</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
<mo>;</mo>
</mrow>
Wherein, di,jRepresent point xiWith point xjBetween Euclidean distance, Jw(i) it is preceding K farthest sample data i same power intervals
The set of sample number, JB(i) it is the set of preceding K nearest sample data i other power interval sample numbers.
7. the Forecasting Methodology according to claim 5 or 6, it is characterised in that the important of sample data i is calculated in the S252
Spend WiFor:
<mrow>
<msub>
<mi>W</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>-</mo>
<mi>&lambda;</mi>
<mi>S</mi>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
</mrow>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>d</mi>
<mi>B</mi>
</msub>
<mo>(</mo>
<mi>i</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mfrac>
<mo>;</mo>
</mrow>
Wherein, RiFor the deleted number of samples in sample data i periphery, λ is that its span of coefficient of similarity is [0,1], S
(i) the same power interval similarity for being sample data i, dB(i)For Euclidean distance between class.
8. a kind of forecasting system of any described Forecasting Methodology in utilization claim 1 to 7, it is characterised in that the system bag
Include:Obtain sample data module, sample data processing module, supporting vector machine model modeling module, obtain prediction data module,
Short-term forecast module;
The acquisition sample data module, its be used to reading current time and its photovoltaic plant for the previous period air
The data of the horizontal irradiation level in boundary, numerical weather forecast and photovoltaic electric station grid connection point active power are used as forecast sample data;
The sample data processing module, it is used to utilize screening sample algorithm, carries out denoising to forecast sample data and subtract sample
Processing, then denoising and the forecast sample data subtracted after sample processing are normalized;
The supporting vector machine model modeling module, it is used to regard the forecast sample data after normalized as modeling sample
This, is modeled to supporting vector machine model, obtains photovoltaic plant short term power forecast model;
The acquisition prediction data mould, it is used for the horizontal irradiation level of aeropause, the numerical value for the photovoltaic plant for obtaining prediction time
Prediction time data are normalized by the prediction time data of weather forecast;
The short-term forecast module, it is used for the short-term work(of photovoltaic plant described in the prediction time data input after normalized
Rate forecast model, obtains output vector, and inverse normalized is carried out to output vector, obtains photovoltaic electric station grid connection point wattful power
The predicted value of rate.
9. forecasting system according to claim 8, it is characterised in that the supporting vector machine model modeling module, it has
Body is that the horizontal irradiation level of aeropause in the forecast sample data after normalized, numerical weather forecast is defeated as modeling
Enter sample, while being exported the photovoltaic electric station grid connection point active power in the forecast sample data after normalized as modeling
Sample, using input sample and modeling output sample is modeled while being modeled to supporting vector machine model, obtains photovoltaic plant
Short term power forecast model.
10. forecasting system according to claim 8, it is characterised in that the sample data processing module includes:First sample
Notebook data processing unit, the second sample data processing unit, the 3rd sample data processing unit, the processing of the 4th sample data are single
Member, the 5th sample data processing unit, the 6th sample data processing unit;
The first sample data processing unit, for carrying out preliminary screening to forecast sample data, by lack part data
Sample data is directly deleted;
The second sample data processing unit, it is N number of for forecast sample data to be equally divided into by photovoltaic plant installed capacity
Power interval;
The 3rd sample data processing unit, for by the forecast sample data after preliminary screening, N number of power to be respectively divided
On interval, N number of subsample collection is obtained;
The 4th sample data processing unit, for being carried out to each sample data that each subsample is concentrated at denoising
Reason;
The 5th sample data processing unit, for carrying out subtracting sample processing to the sample data after denoising;
The 6th sample data processing unit, for carrying out subtracting sample processing to the sample data after denoising.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710339205.6A CN107085755A (en) | 2017-05-15 | 2017-05-15 | A kind of photovoltaic plant short term power Forecasting Methodology and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710339205.6A CN107085755A (en) | 2017-05-15 | 2017-05-15 | A kind of photovoltaic plant short term power Forecasting Methodology and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107085755A true CN107085755A (en) | 2017-08-22 |
Family
ID=59607322
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710339205.6A Pending CN107085755A (en) | 2017-05-15 | 2017-05-15 | A kind of photovoltaic plant short term power Forecasting Methodology and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107085755A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107506867A (en) * | 2017-08-31 | 2017-12-22 | 国网冀北电力有限公司廊坊供电公司 | A kind of photovoltaic power generation output forecasting method based on dynamic modeling |
CN110533265A (en) * | 2019-09-20 | 2019-12-03 | 云南电网有限责任公司电力科学研究院 | A kind of power distribution network lightning stroke method for early warning and power distribution network are struck by lightning prior-warning device |
CN112052996A (en) * | 2020-09-03 | 2020-12-08 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
CN112580862A (en) * | 2020-12-08 | 2021-03-30 | 国家电网有限公司 | Method and device for predicting short-term real-time power generation power of distributed photovoltaic system |
CN114006369A (en) * | 2021-10-28 | 2022-02-01 | 华北电力大学 | Regional wind and light station power joint prediction method and device, electronic equipment and storage medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100082172A1 (en) * | 2008-09-25 | 2010-04-01 | Korea Electric Power Corporation | Load forecasting analysis system for calculating customer baseline load |
CN103927598A (en) * | 2014-04-22 | 2014-07-16 | 国家电网公司 | Photovoltaic generation power prediction method based on self-learning radial basis function |
CN103559561B (en) * | 2013-11-13 | 2016-03-09 | 上海电气集团股份有限公司 | A kind of ultra-short term prediction method of photovoltaic plant irradiance |
-
2017
- 2017-05-15 CN CN201710339205.6A patent/CN107085755A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100082172A1 (en) * | 2008-09-25 | 2010-04-01 | Korea Electric Power Corporation | Load forecasting analysis system for calculating customer baseline load |
CN103559561B (en) * | 2013-11-13 | 2016-03-09 | 上海电气集团股份有限公司 | A kind of ultra-short term prediction method of photovoltaic plant irradiance |
CN103927598A (en) * | 2014-04-22 | 2014-07-16 | 国家电网公司 | Photovoltaic generation power prediction method based on self-learning radial basis function |
Non-Patent Citations (5)
Title |
---|
HAO QUAN ET AL: "Short-Term Load and Wind Power Forecasting Using Neural Network-Based Prediction Intervals", 《IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 》 * |
ZIMING ZHU ET AL: "Short term forecast of wind power generation based on SVM with pattern matching", 《2016 IEEE INTERNATIONAL ENERGY CONFERENCE (ENERGYCON)》 * |
侯佑华 等: "大规模风电入网的有功功率波动特性分析及发电计划仿真", 《电网技术》 * |
杨锡运 等: "基于粒子群优化的核极限学习机模型的风电功率区间预测方法", 《中国电机工程学报》 * |
陈圣兵 等: "基于近邻距离的大规模样本集去噪与减样", 《计算机工程》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107506867A (en) * | 2017-08-31 | 2017-12-22 | 国网冀北电力有限公司廊坊供电公司 | A kind of photovoltaic power generation output forecasting method based on dynamic modeling |
CN107506867B (en) * | 2017-08-31 | 2021-09-21 | 国网冀北电力有限公司廊坊供电公司 | Photovoltaic output prediction method based on dynamic modeling |
CN110533265A (en) * | 2019-09-20 | 2019-12-03 | 云南电网有限责任公司电力科学研究院 | A kind of power distribution network lightning stroke method for early warning and power distribution network are struck by lightning prior-warning device |
CN112052996A (en) * | 2020-09-03 | 2020-12-08 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
CN112052996B (en) * | 2020-09-03 | 2022-02-25 | 哈尔滨工业大学 | Small hydropower station cluster power prediction method based on convolutional neural network technology |
CN112580862A (en) * | 2020-12-08 | 2021-03-30 | 国家电网有限公司 | Method and device for predicting short-term real-time power generation power of distributed photovoltaic system |
CN114006369A (en) * | 2021-10-28 | 2022-02-01 | 华北电力大学 | Regional wind and light station power joint prediction method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107085755A (en) | A kind of photovoltaic plant short term power Forecasting Methodology and system | |
CN101769788B (en) | Method for forecasting optical output power and electric energy production of photovoltaic power station | |
CN104978608B (en) | A kind of wind electric powder prediction device and prediction technique | |
CN102810861B (en) | Generating capacity prediction method and system for photovoltaic generating system | |
CN103729685A (en) | Method for photovoltaic power station group region power prediction on basis of improved RBF neural network | |
CN113496311A (en) | Photovoltaic power station generated power prediction method and system | |
CN104182804B (en) | A kind of electricity-generating method a few days ago for predicting that uncertain small power station of exerting oneself is coordinated with large medium-size station | |
CN105069521A (en) | Photovoltaic power plant output power prediction method based on weighted FCM clustering algorithm | |
CN105701562B (en) | Training method, applicable method for predicting generated power and respective system | |
CN106451541A (en) | Island type microgrid energy control method and control system | |
CN105226648A (en) | A kind of distributed power source distribution network planning method based on large data | |
CN107528344A (en) | A kind of light storage integrated generating device is incorporated into the power networks control method and system | |
CN111009895A (en) | Microgrid optimal scheduling method, system and equipment | |
CN110400056A (en) | Cascade hydropower based on honourable space-time characterisation Optimization Scheduling and device a few days ago | |
CN106159998A (en) | Regenerative resource micro-capacitance sensor Optimal Scheduling | |
CN114611842B (en) | Whole-county roof distributed photovoltaic power prediction method | |
Yang et al. | Sub-minute probabilistic solar forecasting for real-time stochastic simulations | |
Rahim et al. | A survey and comparison of leading-edge uncertainty handling methods for power grid modernization | |
CN116470543A (en) | Operation control method, device, equipment and medium of virtual power plant | |
Ni et al. | A review of line loss analysis of the low-voltage distribution system | |
CN106485603A (en) | A kind of short-term wind speed forecasting method being applied to wind-power electricity generation | |
CN113255982A (en) | Medium-long term optimized scheduling method for wind-light-water complementary system | |
CN109546647B (en) | Safety and stability evaluation method for power system containing wind, light and water storage | |
CN117013531A (en) | Capacity domain assessment method of power distribution network | |
CN105528517B (en) | Based on neural network, wavelet decomposition predicting power of photovoltaic plant method, system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170822 |
|
RJ01 | Rejection of invention patent application after publication |