CN109919353A - A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence - Google Patents
A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence Download PDFInfo
- Publication number
- CN109919353A CN109919353A CN201910028384.0A CN201910028384A CN109919353A CN 109919353 A CN109919353 A CN 109919353A CN 201910028384 A CN201910028384 A CN 201910028384A CN 109919353 A CN109919353 A CN 109919353A
- Authority
- CN
- China
- Prior art keywords
- photovoltaic
- weather
- power station
- data
- power
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Photovoltaic Devices (AREA)
Abstract
The distributed photovoltaic prediction technique of the invention discloses a kind of ARIMA model based on spatial coherence, is related to power domain.Currently, distributed photovoltaic precision of prediction is low.The present invention is based on the photovoltaics of different weather type to go out force data, calculates power station to be predicted and other power station history go out the rank correlation coefficient of force data;Choosing N number of power station that rank correlation coefficient is greater than the set value and/or correlation is best is that correlation slave station is included in the prediction model that different weather type is established in the ARIMA model based on spatial coherence;In conjunction with the daily forecast weather pattern that meteorological department provides, the prediction model of corresponding weather is selected to carry out power output prediction, if dividing weather pattern to carry out ARIMA modeling using power station data to be predicted without correlation power station is matched to;It is matched according to the forecast information that weather forecast department provides, utilizes corresponding model realization photovoltaic ultra-short term power prediction.The technical program forecasting accuracy is high, and has adaptivity in reply Changes in weather.
Description
Technical field
The present invention relates to power domain more particularly to a kind of distributed photovoltaics of the ARIMA model based on spatial coherence
Prediction technique.
Background technique
Centralized photovoltaic plant is compared, there are the following problems for distributed photovoltaic power prediction: 1) component takes up little area,
It is distributed more dispersed, and causes operating status irregular since O&M is not perfect;2) due to the influence of meteorologic factor, photovoltaic
Randomness and fluctuation is presented in power output;3) single photovoltaic user lacks the meteorological data of actual measurement, and precision of prediction is not high.Therefore
Distributed photovoltaic predicts that difficulty is higher, and the method for prediction cannot be copied mechanically and applied indiscriminately fully according to centralized photovoltaic plant prediction technique.
Distributed photovoltaic prediction technique is broadly divided into direct forecast methods and indirect prediction method at present, and indirect prediction method passes through
Weather data is predicted indirectly, and then predicts photovoltaic power output, mainly there is the methods of numerical weather forecast and ground cloud atlas.
Indirect prediction method is to predict on the basis of relevant weather data photovoltaic, but numerical weather forecast precision prescribed is higher,
It is domestic at present to have high-precision weather monitoring point less, and direct forecast methods are sent out using practical power output data statistic analysis
Now certain rule, including time series method, multiple linear regression method, Grey Theory Forecast method etc., its advantage is that not needing gas
The extra datas such as image data can achieve the purpose that prediction merely with the practical force data out of photovoltaic, but its precision of prediction is low, meter
It calculates complicated.
Summary of the invention
The technical problem to be solved in the present invention and the technical assignment of proposition are to be improved and changed to prior art
Into, a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence is provided, with reach Accurate Prediction distribution
The purpose of the power output of formula photovoltaic.For this purpose, the present invention takes following technical scheme.
A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence, comprising the following steps:
1) the practical force data out of distributed photovoltaic is obtained, the history of interception photovoltaic power output effective time goes out force data;
2) go out force data to the reality of each user to carry out marking change;
3) weather pattern is divided according to daily output average level;
4) photovoltaic based on different weather type goes out force data, calculates power station to be predicted and other power station history power output number
According to rank correlation coefficient;
5) choosing N number of power station that rank correlation coefficient is greater than the set value and/or correlation is best is that correlation slave station is included in base
The prediction model of different weather type is established in the ARIMA model of spatial coherence;
6) the daily forecast weather pattern for combining meteorological department to provide, the prediction model of the corresponding weather of selection contribute pre-
It surveys, if not considering correlation without correlation power station is matched to and dividing weather pattern to carry out using power station data to be predicted
ARIMA modeling;
7) forecast information that need to predict that photovoltaic power output area weather forecasting department provides is obtained;And according to weather forecast portion
The forecast information that door provides is matched, and corresponding model realization photovoltaic ultra-short term power prediction is utilized.
The technical program based on the large-scale area photovoltaic grouping method having proposed, further screening divide in group power station with
Power station to be predicted has the photovoltaic plant of spatial coherence, divides weather pattern to establish ARIMA model, department mentions with weather forecast
The forecast information of confession is matched, and corresponding model realization photovoltaic ultra-short term power prediction is utilized.Introducing correlation power station data
On the basis of, history is gone out into force data and is classified to weather, with reference to the weather forecast that meteorological department provides, chooses corresponding weather
Model carries out power forecast, has adaptivity in reply Changes in weather.
Preferably, in step 4), the correlation between two power stations is measured using Spearman rank correlation coefficient,
Formula is as follows:
The history for extracting all same dimensions in power station in the same area goes out force data, the data based on different weather classification
Sample calculates separately its relative coefficient for going out force data with power station history to be predicted, screens the highest N number of electricity of related coefficient
It stands or related coefficient is greater than correlation slave station of the power station of a certain given threshold as power station to be predicted.
Preferably, in step 5), photovoltaic plant of the correlation greater than 0.8 is extracted as correlation photovoltaic slave station.
Preferably, power station X power output prediction model to be predicted are as follows:
Wherein εtIt is the random disturbances at current time, coefficient gamma=(α0,α(l,x),β(l,x),α(l,i),β(l,i)), 1≤i≤
N,0≤l≤Ls;N is the quantity of the photovoltaic plant with correlation, and L is parameter training sample length, and p, q are the mould of ARIMA
Type order,
Coefficient gamma is calculated by least squares estimate:
LSE=| | Px-Xγ||2
It is that the reality of power station X to be predicted goes out force data, X=[E, Xx,X1,X2,…
XN] be N number of correlation power station and power station to be predicted composition coefficient matrix, wherein [1,1 ..., 1] E=TIt is the unit column of L × 1
Vector,
1≤i≤N;
Preferably, in step 3), the photovoltaic power output with the per day power output characterization whole day of photovoltaic is horizontal, calculation formula are as follows:
N be can generating dutation section length, PiFor the power generating value of i-th of moment photovoltaic;
Work as PaverageWhen >=0.4, judge weather pattern for fine day;As 0.3≤PaverageWhen < 0.4, weather pattern is judged
It is more days;As 0.2≤PaverageWhen < 0.3, judge weather pattern for the cloudy day;As 0≤PaverageWhen < 0.2, weather class is judged
Type is the rainy day.
Preferably, prediction model uses ARIMA (p, d, q) model, and prediction model is the group of d order difference and ARMA (p, q)
It closes;In data prediction, stationarity detection need to be carried out first to sequence, nonstationary time series can pass through the difference of limited times
Point form stationary time series is modeled again;Come using AIC (Akaike information criterion) information criterion
It carries out determining rank, chooses p when AIC value minimum, for q as model order, AIC calculation formula is as follows
Model parameter can be calculated by least squares estimate, i.e., to objective function:
Minimization obtainsAs least-squares estimation parameters obtained.
Finally examine residual sequence { εtWhite noise, if residual sequence is white noise sequence, i.e. model parameter is complete
The useful information of abstraction sequence is valid model, if non-white noise sequence, also remaining useful information is to be extracted in sequence, is needed
Again model of fit.
Preferably, before carrying out distributed photovoltaic prediction, large-scale area photovoltaic first is carried out to distributed photovoltaic and divides group,
Further screening divides in group power station the photovoltaic plant with power station to be predicted with spatial coherence, and weather pattern is divided to establish
ARIMA model, the forecast information provided with weather forecast department are matched, and the short-term function of corresponding model realization photovoltaic is utilized
Rate prediction.
Preferably, distributed photovoltaic carry out large-scale area photovoltaic divide group the following steps are included:
A) with each hour data for a sample, data is carried out after marking change processing, 5:00-20:00 point 15 is intercepted
A hour history goes out force data, goes out force data to photovoltaic history by section and carries out weather pattern match cognization, each single electricity
It stands with the weather pattern index at 15 moment the power output historical data indicated, weather pattern is divided into: it is fine day, fine with occasional clouds, it is more
Cloud, the nether world are cloudy, the cloudy day, shower, thunder shower, light rain, heavy rain, heavy or torrential rain;
B) clustering successively is carried out to the data of each integral point, with generating the distributed photovoltaic under five kinds of weather patterns
Location map is managed, is clustered again by K-means, the geographical location of force data is gone out to the photovoltaic under every a kind of weather pattern
Subregion;
C) under each weather pattern, the size of photovoltaic power output is subordinated to a section, i.e. photovoltaic power output has phase each other
Like property;The geographical location piecemeal of each photovoltaic plant is positioned by look-up table, if photovoltaic plant all subordinates under five kinds of weather patterns
In a geographical location block, only middle from same geographic region suitable power station need to select to refer to power station as preliminary, i.e., without
By under what weather pattern, the power output of photovoltaic plant is all followed with reference to power station, referred to as harmonious power station;If different weather patterns
Under, photovoltaic plant is subordinated to different geographic regions, that is, is subordinated to different reference power stations, is known as power station of being discord, then needing
In addition it is elected to be with reference to power station;
D) calculating is discord the accounting rate in power station and main switching station quantity in classification K value setting range, according to accounting rate,
Determine the optimal K value of clustering;
E) clustering is carried out by K value to whole distributed photovoltaic geographical locations again, obtained a result as with space
The compartmentalization that the wide-area distribution type photovoltaic of correlation draws group is shown, obtains photovoltaic power generation user group division result.
Preferably, when the matching of weather pattern, according to the accounting value of practical daily generation and day rated generation amount reference value
K presses section to weather typing, distributed area and corresponding five class weather patterns, as 1≤k < 0.8, it is believed that weather pattern is fine
It, it is fine with occasional clouds;As 0.8≤k < 0.6, it is believed that weather pattern is cloudy, the nether world is cloudy;As 0.6≤k < 0.4, it is believed that day
Gas type is the cloudy day;As 0.4≤k < 0.2, it is believed that weather pattern is shower, thunder shower, light rain;As 0.2≤k < 0, recognize
It is heavy rain, heavy or torrential rain for weather pattern;
When discrete sampling, day actual power generation be 15 minutes sampled values of day interval, then have:
Wherein i is photovoltaic user serial number, and j is photovoltaic day acquisition points, P (i, j) be i-th photovoltaic user's j moment from
Dissipate sampled value, PmaxIt is discrete by daily generation herein due to lacking rated generation amount data for the reference value of day rated generation amount
The history maximum value of sampled value is estimated;When at interval of the data of acquisition in 15 minutes, there is within one day 4 × 24=96 point, i.e. j=1,
2,…,96。
Preferably, cluster analysis result is fluctuated dependent on initial cluster center and cluster number, has certain shakiness
It is qualitative;Select suitable initial cluster center position most important to analysis cluster result;Assuming that data set to be clustered is X
={ xi|xi∈RP, i=1,2 ..., n }, K initial position center is C1,C2,…,CK, use W1,W2,…,WKIndicate K class institute
The sample set for including, all sample sets are W;
Define 1 sample xi,xjBetween Euclidean distance:
Define 2 sample xiTo the average value of all sample distances:
Define the average distance of 3 data set samples
Define 4 data point xiDensity d ensity (xi)
Density (x)=p ∈ C | d (x, p)≤cmean*θ}
θ is density radius coefficient;
First according to the density for defining each sample data in 3-4 calculating data set, the maximum sample of density is foundInitial center C as first class1, the sample in density radius is deleted, i.e.,
W=W-W1
Mentioned above principle is repeated, finds the maximum sample of density in W againAs the C classification it is initial in
The heart, and enable
And so on until find out K initial cluster center C1,C2,…,CK;
Use the Euclidean distance between vector as the foundation of classification, calculation formula is as follows:
Wherein, dijFor the Euclidean distance between i-th of standard vector and j-th of standard vector, n is each standard vector
Dimension.
The utility model has the advantages that
The technical program screening divides in group power station the photovoltaic plant with power station to be predicted with spatial coherence, divides weather
Type establishes ARIMA model, and the forecast information provided with weather forecast department is matched, and utilizes corresponding model realization photovoltaic
Ultra-short term power prediction.On the basis of introducing correlation power station data, history is gone out into force data and is classified to weather, referred to
The weather forecast that meteorological department provides chooses corresponding synoptic model and carries out power forecast, has in reply Changes in weather adaptive
Ying Xing.
The technical program divides user's photovoltaic under every class weather pattern according to region using clustering method
Block, analyzes the power output consistency of different zones, and optimization user's photovoltaic divides group position, is conducive to select most representational
It is least to utilize with reference to power station as the installation location of weather monitoring point or the weather station reference position of purchase meteorological data
Meteorological data predicts the power of each user's photovoltaic in group, to optimize meteorological station location selection both to ensure that photovoltaic is pre-
Precision is surveyed, and reduces the cost for introducing meteorological data, improves economy;Even if being incited somebody to action in addition, not taking increase weather station
The meteorological consistency in large-scale distributed power station divides group, can by the distributed power station of meteorological consistency set up divide group when
Between space photovoltaic predict.Effectively improve the accurate of prediction.
Detailed description of the invention
Fig. 1 is flow chart of the invention.
Fig. 2 is ARIMA modeling procedure of the invention.
Fig. 3 is that photovoltaic power generation user group of the invention divides flow chart.
Specific embodiment
Technical solution of the present invention is described in further detail below in conjunction with Figure of description.
The technical program is primarily based on the large-scale area photovoltaic grouping method having proposed, and further screening divides group power station
In with power station to be predicted have spatial coherence photovoltaic plant, divide weather pattern to establish ARIMA model, with weather forecast portion
The forecast information that door provides is matched, and corresponding 15 minutes ultra-short term power predictions of model realization photovoltaic are utilized.
As shown in Figure 1, a kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence, including it is following
Step:
1) the practical force data out of distributed photovoltaic is obtained, the history of interception photovoltaic power output effective time goes out force data;
2) go out force data to the reality of each user to carry out marking change;
3) weather pattern is divided according to daily output average level;
4) photovoltaic based on different weather type goes out force data, calculates power station to be predicted and other power station history power output number
According to rank correlation coefficient;
5) choosing N number of power station that rank correlation coefficient is greater than the set value and/or correlation is best is that correlation slave station is included in base
The prediction model of different weather type is established in the ARIMA model of spatial coherence;
6) the daily forecast weather pattern for combining meteorological department to provide, the prediction model of the corresponding weather of selection contribute pre-
It surveys, if not considering correlation without correlation power station is matched to and dividing weather pattern to carry out using power station data to be predicted
ARIMA modeling;
7) forecast information that need to predict that photovoltaic power output area weather forecasting department provides is obtained;And according to weather forecast portion
The forecast information that door provides is matched, and corresponding model realization photovoltaic ultra-short term power prediction is utilized.
The technical program based on the large-scale area photovoltaic grouping method having proposed, further screening divide in group power station with
Power station to be predicted has the photovoltaic plant of spatial coherence, divides weather pattern to establish ARIMA model, department mentions with weather forecast
The forecast information of confession is matched, and corresponding model realization photovoltaic ultra-short term power prediction is utilized.Introducing correlation power station data
On the basis of, history is gone out into force data and is classified to weather, with reference to the weather forecast that meteorological department provides, chooses corresponding weather
Model carries out power forecast, has adaptivity in reply Changes in weather.
Explanation is further explained with regard to the following point in distributed photovoltaic prediction below:
One, weather pattern divides
Per day contribute of photovoltaic belongs to the scope of average value, refers to the average value that photovoltaic power output generates electricity during it can generate electricity.
It is symmetrical that practical photovoltaic statistical data shows that ideal photovoltaic power curve should obey, therefore the per day power output of application photovoltaic is more
The photovoltaic power output that whole day can be characterized is horizontal, is formulated are as follows:
Wherein n be can generating dutation section length, PiFor the power generating value of i-th of moment photovoltaic, to wide-area distribution type photovoltaic
User should be first to P to remove difference caused by different installed capacity equidimensionalsiMark change processing.The per day power output of photovoltaic
Size directly reflect the same day weather conditions, size can be used as divide weather pattern foundation, if daily output is horizontal
It is worth larger, then shows fine, sunshine amplitude is higher, and it is corresponding opposite, if daily output level is lower, show that it is vaporous
Condition is poor, and sunshine amplitude is lower.Its per day force threshold setting such as table 1 out for corresponding to weather pattern.
1 weather pattern of table and corresponding model parameter
Two, volt power output spatial coherence matching
Spatial coherence refers under same dimension that the degree of correlation between different photovoltaic plant power outputs track, it has measured two
The related intimate degree of a Variable Factors, measured using Spearman rank correlation coefficient in the technical program two power stations it
Between correlation, formula is as follows:
The history for extracting all same dimensions in power station in the same area goes out force data, the data based on different weather classification
Sample calculates separately its relative coefficient for going out force data with power station history to be predicted, screens the highest N number of electricity of related coefficient
It stands or related coefficient is greater than correlation slave station of the power station of a certain given threshold as power station to be predicted.
Three, ARIMA model modeling
ARMA model is the model of currently used fitting stationary time series, be system to history from
The memory of body state and the noise into system, i.e. sequence p history observation and preceding q before the value of t moment is are a random
The multiple linear function of interference, is simply denoted as ARMA (p, q):
xt=α0+α1xt-1+α2xt-2+…+αpxt-p+εt-β1εt-1-β2εt-2-…βqεt-q
Wherein error term εtIt is the random disturbances of t moment, is the white noise sequence that mean value is zero.And ARIMA (p, d, q)
Model is the combination of d order difference and ARMA (p, q).In data prediction, stationarity detection need to be carried out first to sequence, it is non-
Stationary time series can form stationary time series by the difference of limited times and be modeled again.Use AIC (Akaike
Information criterion) information criterion determine rank, choose p when AIC value minimum, q as model order,
AIC calculation formula is as follows
Model parameter can be calculated by least squares estimate, i.e., to objective function:
Minimization obtainsAs least-squares estimation parameters obtained.
Finally examine residual sequence { εtWhite noise, if residual sequence is white noise sequence, i.e. model parameter is complete
The useful information of abstraction sequence is valid model, if non-white noise sequence, also remaining useful information is to be extracted in sequence, is needed
Again model of fit.
Four, the ARIMA model based on spatial coherence
The technical program introduces, and there is the photovoltaic plant of spatial coherence to change to the ARIMA model in power station to be predicted
Into, using multiple photovoltaic plants with correlation it is practical go out force data establish the ARIMA model in power station to be predicted, to improve
The precision of prediction of model.
Assuming that power station X to be predicted has N number of photovoltaic plant with correlation, p, q are the model order of ARIMA, power output
Prediction model is defined as follows:
Wherein εtIt is the random disturbances at current time, coefficient gamma=(α0,α(l,x),β(l,x),α(l,i),β(l,i)), 1≤i≤
N,0≤l≤Ls;It can be calculated by least squares estimate:
LSE=| | Px-Xγ||2
Setting parameter training sample length is L,It is the practical power output of power station X to be predicted
Data, X=[E, Xx,X1,X2,…XN] be N number of correlation power station and power station to be predicted composition coefficient matrix, wherein E=[1,
1,…,1]TIt is the unit column vector of L × 1,
1≤i≤N
For simplified model calculating, phase can be controlled by changing the threshold value of relative coefficient when matching correlation power station
Closing property slave station quantity avoids X matrix dimension is excessively high model is caused to calculate complexity.
Fig. 2 illustrates the modeling procedure of ARIMA model, and ARIMA model is the core of the technical program, in photovoltaic history
The power prediction that data analysis modeling realizes photovoltaic is carried out on the basis of time series of contributing.ARIMA (p, d, q) model is d scale
Divide the combination with ARMA (p, q).In data prediction, stationarity detection, nonstationary time series need to be carried out first to sequence
Stationary time series can be formed by the difference of limited times to be modeled again.
As shown in figure 3, photovoltaic power generation user group includes the step of division
A) with each hour data for a sample, data is carried out after marking change processing, 5:00-20:00 point 15 is intercepted
A hour history goes out force data, goes out force data to photovoltaic history by section and carries out weather pattern match cognization, each single electricity
It stands with the weather pattern index at 15 moment the power output historical data indicated, weather pattern is divided into: it is fine day, fine with occasional clouds, it is more
Cloud, the nether world are cloudy, the cloudy day, shower, thunder shower, light rain, heavy rain, heavy or torrential rain;
B) clustering successively is carried out to the data of each integral point, with generating the distributed photovoltaic under five kinds of weather patterns
Location map is managed, is clustered again by K-means, the geographical location of force data is gone out to the photovoltaic under every a kind of weather pattern
Subregion;
C) under each weather pattern, the size of photovoltaic power output is subordinated to a section, i.e. photovoltaic power output has phase each other
Like property;The geographical location piecemeal of each photovoltaic plant is positioned by look-up table, if photovoltaic plant all subordinates under five kinds of weather patterns
In a geographical location block, only middle from same geographic region suitable power station need to select to refer to power station as preliminary, i.e., without
By under what weather pattern, the power output of photovoltaic plant is all followed with reference to power station, referred to as harmonious power station;If different weather patterns
Under, photovoltaic plant is subordinated to different geographic regions, that is, is subordinated to different reference power stations, is known as power station of being discord, then needing
In addition it is elected to be with reference to power station;
D) calculating is discord the accounting rate in power station and main switching station quantity in classification K value setting range, according to accounting rate,
Determine the optimal K value of clustering;
E) clustering is carried out by K value to whole distributed photovoltaic geographical locations again, obtained a result as with space
The compartmentalization that the wide-area distribution type photovoltaic of correlation draws group is shown, obtains photovoltaic power generation user group division result.
The technical program is that meteorological site at least disposes or propose that more photovoltaic users are based on the function of " space-time association "
Rate prediction provides foundation;The influence that meteorology contributes to photovoltaic is divided into two class of macroclimate and miniclimate first: macroclimate is mainly
Sunshine or five class weather patterns influence, and account for the ratio of nominal output by the practical power output of photovoltaic to divide, thus by historical data
Time segments division is five class weather pattern sample clusters;Miniclimate is considered that photovoltaic Installation Elevation, temperature, humidity and surrounding are geographical
The broad sense miniclimate such as environment influences, and to five class weather pattern sample cluster of history, carries out the clustering of space correlation, is used
Family photovoltaic region divides;Unsocial user's photovoltaic point quantity and subregion meteorology consistency are optimal to determine in comprehensive piecemeal
Region segment partition scheme is that user's photovoltaic divides group tactful.The technical program is using clustering method under every class weather pattern
User's photovoltaic carries out piecemeal according to region, analyzes the power output consistency of different zones, and optimization user's photovoltaic divides group position, has
Conducive to select it is most representational with reference to power station as the installation location of weather monitoring point or buy meteorological data weather station
Reference position, to predict the power of each user's photovoltaic in group using least meteorological data, to both optimize meteorology
Station location is selected to ensure photovoltaic precision of prediction, and reduces the cost for introducing meteorological data, improves economy;In addition, i.e.
Just increase weather station is not taken, divides the meteorological consistency in large-scale distributed power station to group, can pass through point of meteorological consistency
Cloth power station, which is set up, divides the time and space photovoltaic of group to predict.
Photovoltaic power generation user group, which divides, mainly uses K-means clustering method, according to the distance or similarity of data itself
Several groups are divided into, the principle of division is that sample minimizes and group distance maximization in group.The specific step of the algorithm
It is rapid as follows:
1) data scrubbing carries out quality analysis, including the analysis of shortage of data value, data outliers processing to initial data
Deng.
2) data prediction is normalized data and standardizes, for eliminating the difference between dimension.
3) cluster feature is extracted, and most effective cluster feature is extracted from data, is converted to feature vector.
4) it clusters, for feature vector, selects optimal cluster number and distance function, execute cluster or grouping.
5) Cluster Assessment refers to and evaluates the suitable evaluation function of the effect selection of cluster.
Explanation is further explained with regard to the following point in the division of photovoltaic power generation user group below:
One, photovoltaic power output process is associated with principle with meteorological:
1) macroclimate influences
Since photovoltaic power output influence factor is numerous, precision of prediction is closely related with state of weather.Photovoltaic power generation quantity influences
The correlation maximum of solar irradiance and photovoltaic power output in the factor, even linearly related, it is major influence factors, referred to as
Macroclimate influences.
2) influence of geographic properties miniclimate
Under macroclimate the same terms, the microclimatic factors such as landform where temperature, humidity, photovoltaic also play photovoltaic power output
The effect of can not ignore is close under identical macroclimate, belonging to the power output process characteristic of the photovoltaic of a certain region, due to geographical special
Property miniclimate influence, and there is deviation in the photovoltaic process characteristic of different geographical, becomes miniclimate influence.
Two, the determination of state of weather:
Since photovoltaic plant lacks history meteorological data, the feature of weather pattern is extracted with the historical data that photovoltaic is contributed
Vector is the key that cluster.Since intensity of illumination directly affects photovoltaic power output, and since cloud amount is less when fine day, intensity of illumination
In maximum value, the accounting of photovoltaic power output practical daily generation and day rated generation amount reference value can reflect to a certain extent
Value is maximum, with the decrease of intensity of illumination, photovoltaic contribute practical daily generation and day rated generation amount reference value accounting value by
Step is reduced.Therefore the ratio for proposing day actual power generation and day rated generation amount is weather pattern index K, when discrete sampling, day
Actual power generation is 15 minutes sampled values of day interval, then has:
Wherein i is photovoltaic user serial number, and j is photovoltaic day acquisition points, P (i, j) be i-th photovoltaic user's j moment from
Dissipate sampled value, PmaxIt is discrete by daily generation herein due to lacking rated generation amount data for the reference value of day rated generation amount
The history maximum value of sampled value is estimated.Assuming that having within one day 4 × 24=96 point, i.e. j=at interval of the data of acquisition in 15 minutes
1,2,…,96。
According to the accounting value of practical daily generation and day rated generation amount reference value by section to weather typing, distributed area
Between and corresponding five class weather patterns such as table 1:
1 weather pattern index of table is compareed with weather pattern
Three, initial center position selects:
Cluster analysis result is fluctuated dependent on initial cluster center and cluster number, has certain unstability.Choosing
It is most important to analysis cluster result to select suitable initial cluster center position.Assuming that data set to be clustered is X={ xi|xi
∈RP, i=1,2 ..., n }, K initial position center is C1,C2,…,CK, use W1,W2,…,WKIndicate that K class is included
Sample set, all sample sets are W.
Define 1 sample xi,xjBetween Euclidean distance:
Define 2 sample xiTo the average value of all sample distances:
Define the average distance of 3 data set samples
Define 4 data point xiDensity d ensity (xi)
Density (x)=p ∈ C | d (x, p)≤cmean*θ } θ be density radius coefficient
First according to the density for defining each sample data in 3-4 calculating data set, the maximum sample of density is foundInitial center C as first class1, the sample in density radius is deleted, i.e.,
W=W-W1
Mentioned above principle is repeated, finds the maximum sample of density in W againAs the C classification it is initial in
The heart, and enable
And so on until find out K initial cluster center C1,C2,…,CK。
Four, foundation is clustered
So-called cluster is according to the condition for referring to data classification.Use herein Euclidean distance between vector as classification according to
According to calculation formula is as follows:
Wherein, dijFor the Euclidean distance between i-th of standard vector and j-th of standard vector, n is each standard vector
Dimension.
Five, region division process
The history of 92 days 6-8 months in 2017 of the city Ti Mou distributed photovoltaic user goes out force data, 9 points of interception morning to evening
The historical data of 1 hour 8 hour of upper 5 point interval is as data source.
By taking 11:00 in morning as an example, force data is gone out to photovoltaic history by section to weather pattern index first and carries out weather class
Type match cognization;Then it is clustered by K-means, the geographical location for going out force data to the photovoltaic under every a kind of weather pattern is drawn
Molecular domains;Cluster class number range is set as K=[3,10].Illustrating geographical location clustering block quantity is 6 pieces, each
The location distribution of photovoltaic user after being clustered under weather pattern;For convenience of differentiation to each subarea number 1-6.
Under each weather pattern, the size of photovoltaic power output is subordinated to a geographic area, i.e. the photovoltaic power output in the region
There is similitude each other.The geographic location area of each user's photovoltaic is positioned respectively, if user's photovoltaic under five kinds of weather patterns
It is all subordinated to a determining geographic location area, need to only select suitable power station as preliminary ginseng from same geographic area
Power station is examined, i.e., whatsoever under weather pattern, the power output of user's photovoltaic all follows this identical with reference to power station, we are known as
Harmonious power station.If under different weather patterns, user's photovoltaic is subordinated to different geographic areas, that is, it is subordinated to different references
Power station, we are known as power station of being discord;For power station of being discord, then different weather type, it is difficult to follow same reference electricity
It stands.It calculates the photovoltaic slave station of each region overlay and calculates quantity accounting of the power station in whole power stations of being discord, such as table 2.
The harmonious power station of 2 11:00 period of table and power station accounting of being discord
With geographical location piecemeal class number increase, power station accounting of being discord not necessarily with cluster class number become larger and become larger (or
Become smaller);According to the analysis of this county domain case, it has been suggested that its fluctuation, such as the power station quantity of being discord of K=8, just than K=7
Want small.
As geographical location piecemeal class number increases, region area reduces, and the variance of weather pattern index is gradually reduced, area
Meteorology tends to influence that is unified, therefore considering mima type microrelief, microclimate in domain, and the subregion area of division should not be too large.Therefore it is comprehensive
Consideration is discord power station accounting and meteorological consistency, and in 15 minutes data samples, user's photovoltaic that cluster numbers are 8 divides group
For optimal dividing, the method that the data sample selection at other moment represents station is same as described above, can obtain the data of different moments
Under sample, the optimal dividing class number optimized has K=7, K=8, K=9.
The optimal classification of user photovoltaic group
The calculated result of different classifications quantity K based on user's photovoltaic historical data is shown in Table 3;As can be seen that in most optimal sorting
When class K=7, two kinds of errors are all minimum values, illustrate this area distribution preferably go out representative power station history go out force data with
The fluctuation situation in region is closest.
3 two kinds of application condition results of table
A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence shown in figure 1 above -3 is this
The specific embodiment of invention has embodied substantive distinguishing features of the present invention and progress, needs can be used according to actual, at this
Under the enlightenment of invention, the equivalent modifications of shape, structure etc., the column in the protection scope of this programme are carried out to it.
Claims (10)
1. a kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence, it is characterised in that including following step
It is rapid:
1) the practical force data out of distributed photovoltaic is obtained, the history of interception photovoltaic power output effective time goes out force data;
2) go out force data to the reality of each user to carry out marking change;
3) weather pattern is divided according to daily output average level;
4) photovoltaic based on different weather type goes out force data, calculates power station to be predicted and other power station history go out the order of force data
Related coefficient;
5) choosing N number of power station that rank correlation coefficient is greater than the set value and/or correlation is best is that correlation slave station is included in based on sky
Between correlation ARIMA model in establish the prediction model of different weather type;
6) the daily forecast weather pattern for combining meteorological department to provide selects the prediction model of corresponding weather to carry out power output prediction, if
Without correlation power station is matched to, then correlation is not considered and divides weather pattern to carry out ARIMA using power station data to be predicted
Modeling;
7) forecast information that need to predict that photovoltaic power output area weather forecasting department provides is obtained;And it is provided according to weather forecast department
Forecast information matched, utilize corresponding model realization photovoltaic ultra-short term power prediction.
2. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 1,
It is characterized in that: in step 4), the correlation between two power stations is measured using Spearman rank correlation coefficient, formula is such as
Under:
It extracts the history of all same dimensions in power station in the same area and goes out force data, based on the data sample of different weather classification,
Calculate separately its relative coefficient for going out force data with power station history to be predicted, the highest N number of power station of screening related coefficient or phase
Relationship number is greater than correlation slave station of the power station of a certain given threshold as power station to be predicted.
3. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 1,
It is characterized in that: in step 5), extracting photovoltaic plant of the correlation greater than 0.8 as correlation photovoltaic slave station.
4. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 1,
It is characterized in that: power station X power output prediction model to be predicted are as follows:
Wherein εtIt is the random disturbances at current time, coefficient gamma=(α0,α(l,x),β(l,x),α(l,i),β(l,i)), 1≤i≤N, 0≤l
≤Ls;N is the quantity of the photovoltaic plant with correlation, and L is parameter training sample length, and p, q are the model order of ARIMA,
Coefficient gamma is calculated by least squares estimate:
LSE=| | Px-Xγ||2
It is that the reality of power station X to be predicted goes out force data, X=[E, Xx,X1,X2,…XN] it is N number of
The coefficient matrix in correlation power station and power station to be predicted composition, wherein [1,1 ..., 1] E=TIt is the unit column vector of L × 1,
1≤i≤N;
5. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 1,
Be characterized in that: in step 3), the photovoltaic power output with the per day power output characterization whole day of photovoltaic is horizontal, calculation formula are as follows:
N be can generating dutation section length, PiFor the power generating value of i-th of moment photovoltaic;
Work as PaverageWhen >=0.4, judge weather pattern for fine day;As 0.3≤PaverageWhen < 0.4, judge that weather pattern is more
It;As 0.2≤PaverageWhen < 0.3, judge weather pattern for the cloudy day;As 0≤PaverageWhen < 0.2, judge weather pattern for rain
It.
6. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 1,
Be characterized in that: prediction model uses ARIMA (p, d, q) model, and prediction model is the combination of d order difference and ARMA (p, q);In number
When Data preprocess, stationarity detection need to be carried out first to sequence, nonstationary time series can form flat by the difference of limited times
Steady time series is modeled again;It carries out determining rank using AIC (Akaike information criterion) information criterion,
P when AIC value minimum is chosen, for q as model order, AIC calculation formula is as follows
Model parameter can be calculated by least squares estimate, i.e., to objective function:
Minimization obtainsAs least-squares estimation parameters obtained.
Finally examine residual sequence { εtWhite noise, if residual sequence be white noise sequence, i.e., model parameter completely extract sequence
The useful information of column is valid model, if non-white noise sequence, also remaining useful information is to be extracted in sequence, need to be intended again
Molding type.
7. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 1,
It is characterized in that: before carrying out distributed photovoltaic prediction, large-scale area photovoltaic first being carried out to distributed photovoltaic and divides group, further
Screening divides in group power station the photovoltaic plant with power station to be predicted with spatial coherence, and weather pattern is divided to establish ARIMA model, with
The forecast information that weather forecast department provides is matched, and is predicted using the short term power of corresponding model realization photovoltaic.
8. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 7,
Be characterized in that: distributed photovoltaic carry out large-scale area photovoltaic divide group the following steps are included:
A) with each hour data for a sample, data are carried out after marking change processing, interception 5:00-20:00 point is 15 small
When history go out force data, force data is gone out to photovoltaic history by section and carries out weather pattern match cognization, each single power station is used
The weather pattern index at 15 moment is divided into the power output historical data that indicates, weather pattern: it is fine day, fine with occasional clouds, it is cloudy, negative
Between it is cloudy, cloudy day, shower, thunder shower, light rain, heavy rain, heavy or torrential rain;
B) clustering successively is carried out to the data of each integral point, generates the distributed photovoltaic geographical location under five kinds of weather patterns
Distribution map is clustered again by K-means, and the geographical location subregion of force data is gone out to the photovoltaic under every a kind of weather pattern;
C) under each weather pattern, the size of photovoltaic power output is subordinated to a section, i.e. photovoltaic power output has similitude each other;
The geographical location piecemeal of each photovoltaic plant is positioned by look-up table, if photovoltaic plant is all subordinated to one under five kinds of weather patterns
Whatsoever geographical location block only middle from same geographic region need to select suitable power station as power station is tentatively referred to, i.e.,
Under weather pattern, the power output of photovoltaic plant is all followed with reference to power station, referred to as harmonious power station;If under different weather patterns, photovoltaic
Power station is subordinated to different geographic regions, that is, is subordinated to different reference power stations, is known as power station of being discord, then needs in addition to be elected to be ginseng
Examine power station;
D) the accounting rate that be discord power station and main switching station quantity are calculated in classification K value setting range determines poly- according to accounting rate
The optimal K value of alanysis;
E) clustering is carried out by K value to whole distributed photovoltaic geographical locations again, obtained a result as with space correlation
Property wide-area distribution type photovoltaic draw group compartmentalization show, obtain photovoltaic power generation user group division result.
9. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 8,
It is characterized in that:
When the matching of weather pattern, ratio k is accounted for by section to day according to practical daily generation and day rated generation amount reference value
Gas classification, distributed area and corresponding five class weather patterns, as 1≤k < 0.8, it is believed that weather pattern is fine day, fine with occasional clouds;When
When 0.8≤k < 0.6, it is believed that weather pattern is cloudy, the nether world is cloudy;As 0.6≤k < 0.4, it is believed that weather pattern is the cloudy day;When
When 0.4≤k < 0.2, it is believed that weather pattern is shower, thunder shower, light rain;As 0.2≤k < 0, it is believed that weather pattern be heavy rain,
Heavy or torrential rain;
When discrete sampling, day actual power generation be 15 minutes sampled values of day interval, then have:
Wherein i is photovoltaic user serial number, and j is photovoltaic day acquisition points, and P (i, j) is discrete the adopting at i-th of photovoltaic user's j moment
Sample value, PmaxFor the reference value of day rated generation amount, herein due to lacking rated generation amount data, by daily generation discrete sampling
The history maximum value of value is estimated;When at interval of the data of acquisition in 15 minutes, there is within one day 4 × 24=96 point, i.e. j=1,2 ...,
96。
10. a kind of distributed photovoltaic prediction technique of ARIMA model based on spatial coherence according to claim 8,
It is characterized by: cluster analysis result is fluctuated dependent on initial cluster center and cluster number, there is certain unstability;
Select suitable initial cluster center position most important to analysis cluster result;Assuming that data set to be clustered is X={ xi|xi
∈RP, i=1,2 ..., n }, K initial position center is C1,C2,…,CK, use W1,W2,…,WKIndicate the sample that K class is included
This set, all sample sets are W;
Define 1 sample xi,xjBetween Euclidean distance:
Define 2 sample xiTo the average value of all sample distances:
Define the average distance of 3 data set samples
Define 4 data point xiDensity d ensity (xi)
Density (x)=p ∈ C | d (x, p)≤cmean* θ }
θ is density radius coefficient;
First according to the density for defining each sample data in 3-4 calculating data set, the maximum sample of density is foundAs
The initial center C of first class1, the sample in density radius is deleted, i.e.,
W=W-W1
Mentioned above principle is repeated, finds the maximum sample of density in W againAs the initial center of the C classification, and
It enables
And so on until find out K initial cluster center C1,C2,…,CK;
Use the Euclidean distance between vector as the foundation of classification, calculation formula is as follows:
Wherein, dijFor the Euclidean distance between i-th of standard vector and j-th of standard vector, n is the dimension of each standard vector.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910028384.0A CN109919353B (en) | 2019-01-11 | 2019-01-11 | Distributed photovoltaic prediction method of ARIMA model based on spatial correlation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910028384.0A CN109919353B (en) | 2019-01-11 | 2019-01-11 | Distributed photovoltaic prediction method of ARIMA model based on spatial correlation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109919353A true CN109919353A (en) | 2019-06-21 |
CN109919353B CN109919353B (en) | 2022-07-05 |
Family
ID=66960251
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910028384.0A Active CN109919353B (en) | 2019-01-11 | 2019-01-11 | Distributed photovoltaic prediction method of ARIMA model based on spatial correlation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109919353B (en) |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110298603A (en) * | 2019-07-08 | 2019-10-01 | 华北电力大学(保定) | Distributed photovoltaic power system capacity estimation method |
CN110865928A (en) * | 2019-11-26 | 2020-03-06 | 上海新炬网络技术有限公司 | Method for realizing capacity prediction based on ARIMA prediction model and gray prediction model |
CN111563611A (en) * | 2020-04-13 | 2020-08-21 | 北京工业大学 | Cloud data center renewable energy space-time prediction method for orientation graph convolutional network |
CN111784030A (en) * | 2020-06-12 | 2020-10-16 | 国网冀北电力有限公司电力科学研究院 | Distributed photovoltaic power prediction method and device based on spatial correlation |
CN112286993A (en) * | 2020-10-30 | 2021-01-29 | 合肥阳光智维科技有限公司 | Method and device for detecting abnormal power generation group string in photovoltaic power station |
CN112446554A (en) * | 2020-12-18 | 2021-03-05 | 阳光电源股份有限公司 | Power prediction model establishing method, power prediction method and device |
CN112651537A (en) * | 2019-10-10 | 2021-04-13 | 国网河北省电力有限公司 | Photovoltaic power generation ultra-short term power prediction method and system |
CN112668806A (en) * | 2021-01-17 | 2021-04-16 | 中国南方电网有限责任公司 | Photovoltaic power ultra-short-term prediction method based on improved random forest |
CN112884360A (en) * | 2021-03-18 | 2021-06-01 | 国家电网有限公司 | Distributed photovoltaic power station comprehensive effect evaluation method, system, equipment and medium |
CN113392533A (en) * | 2021-06-25 | 2021-09-14 | 湖南联智科技股份有限公司 | GNSS positioning data post-processing method based on ARIMA model and X-11 decomposition |
CN113704558A (en) * | 2021-08-25 | 2021-11-26 | 成都信息工程大学 | Automatic meteorological station automatic clustering method and system based on differential sequence |
CN113919606A (en) * | 2021-12-14 | 2022-01-11 | 山东建筑大学 | Intelligent site selection method and system for distributed photovoltaic power station |
CN114386334A (en) * | 2022-01-19 | 2022-04-22 | 浙江大学 | Runoff rolling forecasting method based on distributed hydrological runoff simulation substitution model |
CN114648251A (en) * | 2022-04-19 | 2022-06-21 | 湖南大学 | Method and system for building space-time correlation model of distributed photovoltaic power station |
CN114725982A (en) * | 2022-01-18 | 2022-07-08 | 中国农业大学 | Distributed photovoltaic cluster refined division and modeling method |
CN114757440A (en) * | 2022-05-10 | 2022-07-15 | 云南电网有限责任公司电力科学研究院 | Distributed wind power prediction method, model training method, device and medium |
CN114862023A (en) * | 2022-05-11 | 2022-08-05 | 济南大学 | Distributed photovoltaic power prediction method and system based on four-dimensional point-by-point meteorological forecast |
CN116596162A (en) * | 2023-07-05 | 2023-08-15 | 浙江正泰新能源开发有限公司 | Photovoltaic power generation amount measuring and calculating method, device, equipment and storage medium |
CN116667330A (en) * | 2023-05-30 | 2023-08-29 | 安徽固太新能源有限公司 | Photovoltaic output prediction method and device |
CN117749089A (en) * | 2024-02-19 | 2024-03-22 | 北京智芯微电子科技有限公司 | Photovoltaic power station abnormality identification method, device, equipment and medium |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105005937A (en) * | 2015-04-28 | 2015-10-28 | 国家电网公司 | Photovoltaic power station output sequential simulation method based on clearness indexes |
US20160026740A1 (en) * | 2007-02-12 | 2016-01-28 | Locus Energy, Inc. | Weather and satellite model for estimating solar irradiance |
CN105488592A (en) * | 2015-12-02 | 2016-04-13 | 国家电网公司 | Method for predicting generated energy of photovoltaic power station |
EP3161527A1 (en) * | 2014-06-30 | 2017-05-03 | Siemens Aktiengesellschaft | Solar power forecasting using mixture of probabilistic principal component analyzers |
CN107437149A (en) * | 2017-08-07 | 2017-12-05 | 华北电力大学(保定) | The determination method and system that a kind of photovoltaic plant is contributed |
CN107563565A (en) * | 2017-09-14 | 2018-01-09 | 广西大学 | A kind of short-term photovoltaic for considering Meteorology Factor Change decomposes Forecasting Methodology |
CN107766990A (en) * | 2017-11-10 | 2018-03-06 | 河海大学 | A kind of Forecasting Methodology of photovoltaic power station power generation power |
CN108053048A (en) * | 2017-11-06 | 2018-05-18 | 中国电力科学研究院有限公司 | A kind of gradual photovoltaic plant ultra-short term power forecasting method of single step and system |
CN108764548A (en) * | 2018-05-18 | 2018-11-06 | 杭州电子科技大学 | The online short term prediction method of photovoltaic generation dynamically associated based on sky brightness information |
CN108808730A (en) * | 2018-06-12 | 2018-11-13 | 国网山东省电力公司聊城供电公司 | Consider the distribution network system reserve capacity for load variation in power computational methods and system of photovoltaic time space distribution |
US20180358812A1 (en) * | 2017-05-05 | 2018-12-13 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and Methods for Coordinating Distributed Energy Storage |
-
2019
- 2019-01-11 CN CN201910028384.0A patent/CN109919353B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160026740A1 (en) * | 2007-02-12 | 2016-01-28 | Locus Energy, Inc. | Weather and satellite model for estimating solar irradiance |
EP3161527A1 (en) * | 2014-06-30 | 2017-05-03 | Siemens Aktiengesellschaft | Solar power forecasting using mixture of probabilistic principal component analyzers |
CN105005937A (en) * | 2015-04-28 | 2015-10-28 | 国家电网公司 | Photovoltaic power station output sequential simulation method based on clearness indexes |
CN105488592A (en) * | 2015-12-02 | 2016-04-13 | 国家电网公司 | Method for predicting generated energy of photovoltaic power station |
US20180358812A1 (en) * | 2017-05-05 | 2018-12-13 | The Board Of Trustees Of The Leland Stanford Junior University | Systems and Methods for Coordinating Distributed Energy Storage |
CN107437149A (en) * | 2017-08-07 | 2017-12-05 | 华北电力大学(保定) | The determination method and system that a kind of photovoltaic plant is contributed |
CN107563565A (en) * | 2017-09-14 | 2018-01-09 | 广西大学 | A kind of short-term photovoltaic for considering Meteorology Factor Change decomposes Forecasting Methodology |
CN108053048A (en) * | 2017-11-06 | 2018-05-18 | 中国电力科学研究院有限公司 | A kind of gradual photovoltaic plant ultra-short term power forecasting method of single step and system |
CN107766990A (en) * | 2017-11-10 | 2018-03-06 | 河海大学 | A kind of Forecasting Methodology of photovoltaic power station power generation power |
CN108764548A (en) * | 2018-05-18 | 2018-11-06 | 杭州电子科技大学 | The online short term prediction method of photovoltaic generation dynamically associated based on sky brightness information |
CN108808730A (en) * | 2018-06-12 | 2018-11-13 | 国网山东省电力公司聊城供电公司 | Consider the distribution network system reserve capacity for load variation in power computational methods and system of photovoltaic time space distribution |
Non-Patent Citations (4)
Title |
---|
ALEXANDER PHINIKARIDES ET AL.: "ARIMA modeling of the performance of different photovoltaic technologies", 《IEEE》 * |
XW´EGNON GHISLAIN AGOUA ET AL.: "Short-Term Spatio-Temporal Forecasting of Photovoltaic Power Production", 《IEEE TRANSACTIONS ON SUSTAINABLE ENERGY》 * |
于若英 等: "考虑天气和空间相关性的光伏电站输出功率修复方法", 《电网技术》 * |
*** 等: "基于空间相关性的分布式光伏超短期预测技术研究", 《陕西电力》 * |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110298603A (en) * | 2019-07-08 | 2019-10-01 | 华北电力大学(保定) | Distributed photovoltaic power system capacity estimation method |
CN110298603B (en) * | 2019-07-08 | 2021-12-14 | 华北电力大学(保定) | Distributed photovoltaic system capacity estimation method |
CN112651537A (en) * | 2019-10-10 | 2021-04-13 | 国网河北省电力有限公司 | Photovoltaic power generation ultra-short term power prediction method and system |
CN110865928A (en) * | 2019-11-26 | 2020-03-06 | 上海新炬网络技术有限公司 | Method for realizing capacity prediction based on ARIMA prediction model and gray prediction model |
CN110865928B (en) * | 2019-11-26 | 2024-04-26 | 上海新炬网络技术有限公司 | Method for realizing capacity prediction based on ARIMA prediction model and gray prediction model |
CN111563611B (en) * | 2020-04-13 | 2023-11-24 | 北京工业大学 | Cloud data center renewable energy space-time prediction method for graph rolling network |
CN111563611A (en) * | 2020-04-13 | 2020-08-21 | 北京工业大学 | Cloud data center renewable energy space-time prediction method for orientation graph convolutional network |
CN111784030A (en) * | 2020-06-12 | 2020-10-16 | 国网冀北电力有限公司电力科学研究院 | Distributed photovoltaic power prediction method and device based on spatial correlation |
CN111784030B (en) * | 2020-06-12 | 2023-10-24 | 国网冀北电力有限公司电力科学研究院 | Distributed photovoltaic power prediction method and device based on spatial correlation |
CN112286993A (en) * | 2020-10-30 | 2021-01-29 | 合肥阳光智维科技有限公司 | Method and device for detecting abnormal power generation group string in photovoltaic power station |
CN112446554A (en) * | 2020-12-18 | 2021-03-05 | 阳光电源股份有限公司 | Power prediction model establishing method, power prediction method and device |
CN112446554B (en) * | 2020-12-18 | 2024-05-14 | 阳光电源股份有限公司 | Power prediction model establishment method, power prediction method and device |
CN112668806B (en) * | 2021-01-17 | 2022-09-06 | 中国南方电网有限责任公司 | Photovoltaic power ultra-short-term prediction method based on improved random forest |
CN112668806A (en) * | 2021-01-17 | 2021-04-16 | 中国南方电网有限责任公司 | Photovoltaic power ultra-short-term prediction method based on improved random forest |
CN112884360A (en) * | 2021-03-18 | 2021-06-01 | 国家电网有限公司 | Distributed photovoltaic power station comprehensive effect evaluation method, system, equipment and medium |
CN113392533B (en) * | 2021-06-25 | 2022-05-27 | 湖南联智科技股份有限公司 | GNSS positioning data post-processing method based on ARIMA model and X-11 decomposition |
CN113392533A (en) * | 2021-06-25 | 2021-09-14 | 湖南联智科技股份有限公司 | GNSS positioning data post-processing method based on ARIMA model and X-11 decomposition |
CN113704558A (en) * | 2021-08-25 | 2021-11-26 | 成都信息工程大学 | Automatic meteorological station automatic clustering method and system based on differential sequence |
CN113919606B (en) * | 2021-12-14 | 2022-03-15 | 山东建筑大学 | Intelligent site selection method and system for distributed photovoltaic power station |
CN113919606A (en) * | 2021-12-14 | 2022-01-11 | 山东建筑大学 | Intelligent site selection method and system for distributed photovoltaic power station |
CN114725982A (en) * | 2022-01-18 | 2022-07-08 | 中国农业大学 | Distributed photovoltaic cluster refined division and modeling method |
CN114386334A (en) * | 2022-01-19 | 2022-04-22 | 浙江大学 | Runoff rolling forecasting method based on distributed hydrological runoff simulation substitution model |
CN114648251A (en) * | 2022-04-19 | 2022-06-21 | 湖南大学 | Method and system for building space-time correlation model of distributed photovoltaic power station |
CN114757440A (en) * | 2022-05-10 | 2022-07-15 | 云南电网有限责任公司电力科学研究院 | Distributed wind power prediction method, model training method, device and medium |
CN114757440B (en) * | 2022-05-10 | 2024-07-16 | 云南电网有限责任公司电力科学研究院 | Distributed wind power prediction method, model training method, device and medium |
CN114862023A (en) * | 2022-05-11 | 2022-08-05 | 济南大学 | Distributed photovoltaic power prediction method and system based on four-dimensional point-by-point meteorological forecast |
CN116667330A (en) * | 2023-05-30 | 2023-08-29 | 安徽固太新能源有限公司 | Photovoltaic output prediction method and device |
CN116596162B (en) * | 2023-07-05 | 2023-10-20 | 浙江正泰新能源开发有限公司 | Photovoltaic power generation amount measuring and calculating method, device, equipment and storage medium |
CN116596162A (en) * | 2023-07-05 | 2023-08-15 | 浙江正泰新能源开发有限公司 | Photovoltaic power generation amount measuring and calculating method, device, equipment and storage medium |
CN117749089A (en) * | 2024-02-19 | 2024-03-22 | 北京智芯微电子科技有限公司 | Photovoltaic power station abnormality identification method, device, equipment and medium |
CN117749089B (en) * | 2024-02-19 | 2024-05-10 | 北京智芯微电子科技有限公司 | Photovoltaic power station abnormality identification method, device, equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN109919353B (en) | 2022-07-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109919353A (en) | A kind of distributed photovoltaic prediction technique of the ARIMA model based on spatial coherence | |
CN110008982A (en) | A kind of weather monitoring point selection method based on photovoltaic power generation power output cluster | |
CN113379109B (en) | Runoff forecasting method based on prediction model self-adaption | |
CN107766990B (en) | Method for predicting power generation power of photovoltaic power station | |
CN106446494B (en) | Honourable power forecasting method based on wavelet packet-neural network | |
CN108985965A (en) | A kind of photovoltaic power interval prediction method of combination neural network and parameter Estimation | |
CN106651200A (en) | Electrical load management method and system for industrial enterprise aggregate user | |
CN109934395B (en) | Multi-hydropower-region short-term power load prediction method based on time-sharing and regional meteorological data | |
CN105373857A (en) | Photovoltaic power station irradiance prediction method | |
CN106505631B (en) | Intelligent wind power wind power prediction system | |
CN109558975A (en) | A kind of integrated approach of a variety of prediction results of electric load probability density | |
CN102184337A (en) | Dynamic combination analysis method of new energy generating capacity influenced by meteorological information | |
Liu et al. | Solar forecasting by K-Nearest Neighbors method with weather classification and physical model | |
CN110009385A (en) | A kind of photovoltaic power generation user group division methods based on multifactor mapping | |
Song et al. | An indoor temperature prediction framework based on hierarchical attention gated recurrent unit model for energy efficient buildings | |
CN110163444A (en) | A kind of water demand prediction method based on GASA-SVR | |
Zhang et al. | Multi-quantile recurrent neural network for feeder-level probabilistic energy disaggregation considering roof-top solar energy | |
CN114483485B (en) | Method for improving wind speed prediction of Nudging wind farm observation data | |
Martí et al. | Wind power prediction in complex terrain: from the synoptic scale to the local scale | |
CN109858667A (en) | It is a kind of based on thunder and lightning weather to the short term clustering method of loading effects | |
CN109934396A (en) | A kind of regional network for the load fining prediction technique based on two stages reduction | |
Zhang et al. | Hybrid model for renewable energy and load forecasting based on data mining and EWT | |
CN113570414A (en) | Electricity price prediction method for optimizing deep neural network based on improved Adam algorithm | |
CN116454870A (en) | Regional distributed photovoltaic short-term power prediction method | |
Salcedo-Sanz et al. | Wind power field reconstruction from a reduced set of representative measuring points |
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 |