CN109636054A - Solar energy power generating amount prediction technique based on classification and error combination prediction - Google Patents
Solar energy power generating amount prediction technique based on classification and error combination prediction Download PDFInfo
- Publication number
- CN109636054A CN109636054A CN201811570233.XA CN201811570233A CN109636054A CN 109636054 A CN109636054 A CN 109636054A CN 201811570233 A CN201811570233 A CN 201811570233A CN 109636054 A CN109636054 A CN 109636054A
- Authority
- CN
- China
- Prior art keywords
- prediction
- error
- neural network
- data
- mpso
- 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 39
- 238000013528 artificial neural network Methods 0.000 claims abstract description 74
- 239000011159 matrix material Substances 0.000 claims abstract description 73
- 238000010248 power generation Methods 0.000 claims description 32
- 230000001186 cumulative effect Effects 0.000 claims description 12
- 230000001932 seasonal effect Effects 0.000 claims description 6
- 230000005611 electricity Effects 0.000 description 4
- 238000007619 statistical method Methods 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 241001269238 Data Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 1
- 239000002803 fossil fuel Substances 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention discloses the solar energy power generating amount prediction technique of a kind of classification and error combination prediction, including S1, prediction data determined it according to the time, its weather pattern is determined using KNN algorithm according to the meteorological data of forecast date and history meteorological data;Corresponding combination forecasting is used after S2, classification, is predicted to obtain respective prediction output respectively using MPSO-BP neural network and gray model GM (1,1);S3, the error matrix obtained using sample training data, acquire the weight matrix of respective combined prediction;S4, two parts prediction output valve is combined according to weight matrix, finally obtains solar energy power generating amount.The weight of each output point is obtained further according to error after this method classification, the prediction result obtained from is relatively reliable, realizes the reliable prediction to solar energy power generating amount.
Description
Technical field
The invention belongs to technical field of solar utilization technique, and in particular to a kind of sun based on classification and error combination prediction
It can method for forecasting photovoltaic power generation quantity.
Background technique
Nowadays, the conventional fossil fuel energy is increasingly depleted, while will also result in very big harm to environment in use process.And
Renewable energy is the inexhaustible energy, and for the sustainable development of human society, countries in the world are one after another mesh
Light has invested renewable energy, and solar power generation is then the main Land use systems of renewable energy, is the main of smart grid
Component part.The common-denominator target that smart grid is made great efforts is the utilization rate for greatly improving environmentally friendly renewable energy, and micro-capacitance sensor
Technology is to realize the key technology of the target, but uncontrollable characteristic for having of renewable energy power generation gives our micro-capacitance sensor again
Energy management brings bigger difficulty, causes serious influence and threat to micro-capacitance sensor economy, safe and stable operation, because
This finds suitable method promotion micro-capacitance sensor reliability and validity is very important.
Currently in some aspects progress obtained of microgrid energy management highly significant, but to realize efficiently
Energy management, need accurately to predict network load and renewable energy power generation.The prediction side of existing solar power generation amount
Method, predominantly statistical method and Artificial Neural Network, statistical method are the utilizations by for statistical analysis to historical data
Probability theory find out in it rule and for predicting;And individually Artificial Neural Network using sample data as input,
Prediction model is established, to predict the following generated energy;Both the above method is for regular and periodically stronger data
Information can reach higher precision of prediction, but solar energy has the features such as randomness, fluctuation, with both methods, prediction
Effect is just very unsatisfactory, is unable to satisfy the needs of existing energy management, greatly limit microgrid energy management efficiency and
Reliability.
Therefore, it is particularly significant to find a kind of method that can carry out reliable prediction to solar energy power generating.
Summary of the invention
For above-mentioned deficiency in the prior art, the solar energy provided by the invention based on classification and error combination prediction
Volt generated energy prediction technique solves in existing method for forecasting photovoltaic power generation quantity, and prediction effect is undesirable, it is existing to be unable to satisfy
Energy management limits the problem of efficiency and reliability of microgrid energy management.
In order to achieve the above object of the invention, the technical solution adopted by the present invention are as follows: predicted based on classification and error combination
Solar energy power generating amount prediction technique, comprising the following steps:
S1, the meteorological data that solar energy power generating amount prediction day is obtained from weather station;
S2, it determines mid-season weather pattern belonging to the meteorological data, the meteorological data for predicting day is input to the weather
Under type in corresponding trained MPSO-BP neural network, the first solar power generation amount output time series are obtained
The meteorological data for predicting day is input to corresponding trained gray model GM (1,1) under the weather pattern simultaneously
In, obtain the second solar power generation amount output time series
S3, by the first solar power generation amount output time seriesWith corresponding trained MPSO-BP neural network
In combined prediction matrix WBPIt is multiplied, obtains the first solar power generation premeasuring YBP;
Simultaneously by the second solar power generation amount output time seriesWith corresponding trained gray model GM (1,1)
In combined prediction matrix WGMIt is multiplied, obtains the second solar power generation premeasuring YGM;
S4, by the first solar power generation premeasuring YBPWith the second solar power generation premeasuring YGMIt is added, obtains solar energy
Lie prostrate the prediction generated energy Y of generated energy prediction dayP。
Further, the meteorological data in the step S1 includes highest temperature value, lowest temperature value, every three hours
Difference, difference, the relative humidity of minimum temperature and previous Daily minimum temperature of temperature value, maximum temperature and previous max. daily temperature
With the uitraviolet intensity of numeralization.
Further, the meteorological data affiliated season in the step S2 includes spring, summer, fall and winter;Each
Season includes three kinds of fine day, cloudy day and rainy day weather patterns;
There are a corresponding trained MPSO-BP neural network and training under each mid-season each weather pattern
Good gray model GM (1,1).
Further, the gray model GM (1,1) are as follows:
In formula, y(1)(ki+1j) indicate to predict the cumulative sequential value of the i+1 output point of j-th obtained of sample;
y(1)(kij) indicate j-th of sample i-th of output point cumulative sequential value;
M is output point number in sample;
U, a model parameter to be asked;
U and a is determined according to the data being input in gray model.
Further, the method for corresponding gray model GM (1,1) has under one weather pattern of training in the step S2
Body are as follows:
A1, the history meteorological data that three kinds of weather patterns in Various Seasonal are obtained from weather station;
A2, the history meteorological data under same mid-season same weather pattern is divided into one kind;
A3, according to the date data in similar history meteorological data, obtain in the corresponding date practical photovoltaic hair
Electricity historical data, and be input in gray model GM (1,1) as training sample;
A4, the data according to the preceding j × m+i point of training sample, calculate the cumulative sequences y of gray model(1), and establish
Its corresponding calculating matrix;
A5, according to calculating matrix, the u and a in gray model are calculated by least square method, and carry it into cumulative sequence
y(1)In, obtain y(1)(ki+1j), complete the training of gray model GM (1,1).
Further, in the step S2, the method tool of corresponding MPSO-BP neural network under one weather pattern of training
Body are as follows:
B1, the history meteorological data that three kinds of weather patterns in Various Seasonal are obtained from weather station;
B2, the history meteorological data under same mid-season same weather pattern is divided into one kind;
B3, according to the date data in similar history meteorological data, obtain in the corresponding date practical photovoltaic hair
Electricity historical data;
B4, training data is inputted using history meteorological data as sample, by practical solar energy power generating amount historical data
In every half an hour solar energy power generating amount as sample export training data;
B5, the maximum frequency of training of MPSO-BP neural network is set as EPmaxConvergence error ε is predicted with expectationE;
B6, sample input training data is input in MPSO-BP neural network, is to export sample output training data
Target is trained, until frequency of training reaches the maximum frequency of training EP of settingmaxOr prediction error value ε when trainingpIt is less than
The expectation prediction error value ε of settingEWhen, complete the training of MPSO-BP neural network.
Further, the combined prediction matrix W in the trained MPSO-BP neural network of determination in the step S3BP
With the combined prediction matrix W in the trained gray model GM (1,1) of determinationGMMethod specifically:
C1, when completing the training of MPSO-BP neural network, determine i-th of output of j-th sample input training data
The error of point are as follows:
In formula,I-th of output point for j-th of sample input training data in MPSO-BP neural network is defeated
Prediction solar energy power generating amount out;
kijI-th of output point that j-th of sample to be input in MPSO-BP neural network inputs training data is corresponding
History meteorological data;
yreal(kij) it is the corresponding practical solar energy power generating of i-th of output point that j-th of sample inputs training data
Measure historical data;
Meanwhile when completing the training of gray model GM (1,1), determine that i-th of j-th of sample input training data is defeated
The error put out are as follows:
In formula,I-th of output point for j-th of sample input training data in gray model GM (1,1) is defeated
Prediction solar energy power generating amount out;
kijI-th of output point that training data is inputted to be input to j-th of sample of gray model GM (1,1) corresponding is gone through
History meteorological data;
C2, the error that the corresponding output point of training data is inputted according to each sample, obtain in MPSO-BP neural network accidentally
Poor matrix are as follows:
In formula, Et BPFor error matrix in neural network;
M is the number of output point;
Meanwhile the error of the corresponding output point of training data is inputted according to each sample, it obtains in gray model GM (1,1)
Error matrix are as follows:
C3, it sums to the error of output point each in error matrix, obtains the prediction error moments in MPSO-BP neural network
Battle array are as follows:
In formula, a is a-th of training error sample;
Meanwhile to each summing with the error of the corresponding output point of neural network output in error matrix, grey is obtained
Prediction error matrix in model GM (1,1) are as follows:
C4, according to the prediction error moments in the prediction error matrix and gray model GM (1,1) in MPSO-BP neural network
Battle array determines the weight W of the prediction error matrix of MPSO-BP neural network respectivelyBP kWith the prediction error of gray model GM (1,1)
The weight W of matrixGM k:
Wherein, the weight of the prediction error matrix of MPSO-BP neural network are as follows:
The weight of the prediction error matrix of gray model GM (1,1) are as follows:
C5, according to MPSO-BP neural network prediction error matrix weight and gray model GM (1,1) prediction error
The weight of matrix determines the combined prediction matrix W in trained MPSO-BP neural network respectivelyBPWith trained grey mould
Combined prediction matrix W in type GM (1,1)GM;
Wherein, the combined prediction matrix W in trained MPSO-BP neural networkBPAre as follows:
Combined prediction matrix W in trained gray model GM (1,1)GMAre as follows:
The invention has the benefit that
1, the present invention is using KNN to the advanced row classification processing of data, it is ensured that has higher standard in different weather datas
True property, there is higher accuracy in the case where sample is few.
2, the calculating ginseng present invention employs the error amount based on each output point as the combined prediction weight of respective point
Number has higher precision, more fitting actual conditions compared to the method evidence of the Combining weights of traditional all output points of unification.
3, the present invention uses the highest temperature, and the lowest temperature is often separated by the temperature data of three hours, the highest with proxima luce (prox. luc)
The difference of temperature, the difference with the low temperature of proxima luce (prox. luc), the input parameter for the prediction that the uitraviolet intensity of numeralization is used as improve
The accuracy of the generated energy of prediction.
Detailed description of the invention
Fig. 1 is the solar energy power generating amount prediction technique flow chart in the present invention based on classification and error combination prediction.
Specific embodiment
A specific embodiment of the invention is described below, in order to facilitate understanding by those skilled in the art this hair
It is bright, it should be apparent that the present invention is not limited to the ranges of specific embodiment, for those skilled in the art,
As long as various change is in the spirit and scope of the present invention that the attached claims limit and determine, these variations are aobvious and easy
See, all are using the innovation and creation of present inventive concept in the column of protection.
As shown in Figure 1, the solar energy power generating amount prediction technique based on classification and error combination prediction, including following step
It is rapid:
S1, the meteorological data that solar energy power generating amount prediction day is obtained from weather station;
Meteorological data therein include highest temperature value, lowest temperature value, the temperature value every three hours, maximum temperature with
Difference, relative humidity and the ultraviolet light of numeralization of the difference of previous max. daily temperature, minimum temperature and previous Daily minimum temperature
Intensity.
S2, it determines mid-season weather pattern belonging to the meteorological data, the meteorological data for predicting day is input to the weather
Under type in corresponding trained MPSO-BP neural network, the first solar power generation amount output time series are obtained
The meteorological data for predicting day is input to corresponding trained gray model GM (1,1) under the weather pattern simultaneously
In, obtain the second solar power generation amount output time series
Meteorological data affiliated season in above-mentioned steps S2 includes spring, summer, fall and winter;Each season includes
Three kinds of fine day, cloudy day and rainy day weather patterns, fine day, cloudy day and rainy day as three kinds of typical weather patterns, for different seasons
The meteorological data of the lower atypia weather (such as cloudy, shower weather pattern) of section, can be used K- close on algorithm (KNN) calculate and
The distance of the corresponding meteorological data of typical weather type is classified as realizing atypia apart from nearest typical weather type
Weather typing is into typical weather.It may be implemented accurately to classify under less sample using KNN algorithm.
There are a corresponding trained MPSO-BP neural network and training under each mid-season each weather pattern
Good gray model GM (1,1).
S3, by the first solar power generation amount output time seriesIn corresponding trained MPSO-BP neural network
Combined prediction matrix WBPIt is multiplied, obtains the first solar power generation premeasuring YBP;
Simultaneously by the second solar power generation amount output time seriesWith corresponding trained gray model GM (1,1)
In combined prediction matrix WGMIt is multiplied, obtains the second solar power generation premeasuring YGM;
Wherein, the corresponding solar power generation premeasuring Y of MPSO-BP neural networkBPAre as follows:
The corresponding solar power generation premeasuring Y of gray model GM (1,1)GMAre as follows:
S4, by the first solar power generation premeasuring YBPWith the second solar power generation premeasuring YGMIt is added, obtains solar energy
Lie prostrate the prediction generated energy Y of generated energy prediction dayP。
Wherein, the prediction generated energy Y of solar energy power generating amount prediction dayPAre as follows:
Gray model GM (1,1) in above-mentioned steps S2 are as follows:
In formula, y(1)(ki+1j) indicate to predict the cumulative sequential value of the i+1 output point of j-th obtained of sample;
y(1)(kij) indicate j-th of sample i-th of output point cumulative sequential value;
M is output point number in sample;
U, a are model parameter to be asked;
The undetermined parameter that the data being input in gray model according to u and a determine;
The method of corresponding gray model GM (1,1) under a mid-season weather pattern is trained in the step S2
Specifically:
A1, the history meteorological data that three kinds of weather patterns in Various Seasonal are obtained from weather station;
A2, the history meteorological data under same mid-season same weather pattern is divided into one kind;
A3, according to the date data in similar history meteorological data, obtain in the corresponding date practical photovoltaic hair
Electricity historical data, and be input in gray model GM (1,1) as training sample;
A4, the data according to the preceding j × m+i point of training sample, calculate the cumulative sequences y of gray model(1), and establish
Its corresponding calculating matrix;
A5, according to calculating matrix, the u and a in gray model are calculated by least square method, and carry it into cumulative sequence
y(1)In, obtain y(1)(ki+1j), complete the training of gray model GM (1,1).
In above-mentioned steps A5, y is obtained(1)(ki+1j) after, gray model is are as follows:
Enable y(1)(k11)=y (k11), the predicted value of each historical data point of training sample can be obtained.
In above-mentioned steps S2, the side of corresponding MPSO-BP neural network under one mid-season weather pattern of training
Method specifically:
B1, the history meteorological data that three kinds of weather patterns in Various Seasonal are obtained from weather station;
B2, the history meteorological data under same mid-season same weather pattern is divided into one kind;
B3, according to the date data in similar history meteorological data, obtain in the corresponding date practical photovoltaic hair
Electricity historical data;
B4, training data is inputted using history meteorological data as sample, by practical solar energy power generating amount historical data
In every half an hour solar energy power generating amount as sample export training data;
B5, the maximum frequency of training of MPSO-BP neural network is set as EPmaxConvergence error ε is predicted with expectationE;
B6, sample input training data is input in MPSO-BP neural network, is to export sample output training data
Target is trained, until frequency of training reaches the maximum frequency of training EP of settingmaxOr prediction error value ε when trainingpIt is less than
The expectation prediction error value ε of settingEWhen, complete the training of MPSO-BP neural network;
In above-mentioned steps B6, the convergence error of setting MPSO-BP neural network is εBP=0.5 × εE,
Every one group of sample of input inputs training data, and MPSO-BP neural network exports corresponding data, has a prediction to receive
Hold back error value epsilonp, andWherein,For root at this time
According to the prediction generated energy that gray model GM (1,1) and MPSO-BP neural network ensemble are predicted, YrealIt is instructed to be inputted with sample
Practice the corresponding practical solar energy power generating amount historical data of data, works as εp≤εEWhen, then the MPSO-BP neural network prediction mould
Type convergence, even if MPSO-BP neural network does not meet convergence error at this time, which does not continue to learn yet
It practises;Conversely, the neural network, which continues input sample, inputs training data, until frequency of training reaches the maximum frequency of training of setting
EPmaxOr prediction error value ε when trainingPLess than the expectation prediction error value ε of settingE, complete the instruction of MPSO-BP neural network
Practice.
The combined prediction matrix W in the trained MPSO-BP neural network of determination in above-mentioned steps S3BPIt is trained with determining
Combined prediction matrix W in good gray model GM (1,1)GMMethod specifically:
C1, when completing the training of MPSO-BP neural network, determine i-th of output of j-th sample input training data
The error of point are as follows:
In formula,I-th of output point for j-th of sample input training data in MPSO-BP neural network is defeated
Prediction solar energy power generating amount out;
kijI-th of output point that j-th of sample to be input in MPSO-BP neural network inputs training data is corresponding
History meteorological data;
yreal(kij) it is the corresponding practical solar energy power generating of i-th of output point that j-th of sample inputs training data
Measure historical data;
Meanwhile when completing the training of gray model GM (1,1), determine that i-th of j-th of sample input training data is defeated
The error put out are as follows:
In formula,I-th of output point for j-th of sample input training data in gray model GM (1,1) is defeated
Prediction solar energy power generating amount out;
kijI-th of output point that training data is inputted to be input to j-th of sample of gray model GM (1,1) corresponding is gone through
History meteorological data;
C2, the error that the corresponding output point of training data is inputted according to each sample, obtain in MPSO-BP neural network accidentally
Poor matrix are as follows:
In formula, Et BPFor error matrix in neural network;
M is the number of output point;
Meanwhile the error of the corresponding output point of training data is inputted according to each sample, it obtains in gray model GM (1,1)
Error matrix are as follows:
C3, it sums to the error of output point each in error matrix, obtains the prediction error moments in MPSO-BP neural network
Battle array are as follows:
In formula, a is a-th of training error sample;
Meanwhile to each summing with the error of the corresponding output point of neural network output in error matrix, grey is obtained
Prediction error matrix in model GM (1,1) are as follows:
C4, according to the prediction error moments in the prediction error matrix and gray model GM (1,1) in MPSO-BP neural network
Battle array determines the weight W of the prediction error matrix of MPSO-BP neural network respectivelyBP kWith the prediction error of gray model GM (1,1)
The weight W of matrixGM k:
Wherein, the weight of the prediction error matrix of MPSO-BP neural network are as follows:
The weight of the prediction error matrix of gray model GM (1,1) are as follows:
C5, according to MPSO-BP neural network prediction error matrix weight and gray model GM (1,1) prediction error
The weight of matrix determines the combined prediction matrix W in trained MPSO-BP neural network respectivelyBPWith trained grey mould
Combined prediction matrix W in type GM (1,1)GM;
Wherein, the combined prediction matrix W in trained MPSO-BP neural networkBPAre as follows:
Combined prediction matrix W in trained gray model GM (1,1)GMAre as follows:
The invention has the benefit that
1, the present invention is using KNN to the advanced row classification processing of data, it is ensured that has higher standard in different weather datas
True property, there is higher accuracy in the case where sample is few.
2, the calculating ginseng present invention employs the error amount based on each output point as the combined prediction weight of respective point
Number has higher precision, more fitting actual conditions compared to the method evidence of the Combining weights of traditional all output points of unification.
3, the present invention uses the highest temperature, and the lowest temperature is often separated by the temperature data of three hours, the highest with proxima luce (prox. luc)
The difference of temperature, the difference with the low temperature of proxima luce (prox. luc), the input parameter for the prediction that the uitraviolet intensity of numeralization is used as improve
The accuracy of the generated energy of prediction.
Claims (7)
1. the solar energy power generating amount prediction technique based on classification and error combination prediction, which is characterized in that including following step
It is rapid:
S1, the meteorological data that solar energy power generating amount prediction day is obtained from weather station;
S2, it determines mid-season weather pattern belonging to the meteorological data, the meteorological data for predicting day is input to the weather pattern
Under in corresponding trained MPSO-BP neural network, obtain the first solar power generation amount output time series
The meteorological data for predicting day is input under the weather pattern in corresponding trained gray model GM (1,1) simultaneously,
Obtain the second solar power generation amount output time series
S3, by the first solar power generation amount output time seriesWith the group in corresponding trained MPSO-BP neural network
Close prediction matrix WBPIt is multiplied, obtains the first solar power generation premeasuring YBP;
Simultaneously by the second solar power generation amount output time seriesIn corresponding trained gray model GM (1,1)
Combined prediction matrix WGMIt is multiplied, obtains the second solar power generation premeasuring YGM;
S4, by the first solar power generation premeasuring YBPWith the second solar power generation premeasuring YGMIt is added, obtains photovoltaic hair
The prediction generated energy Y of power quantity predicting dayP。
2. the solar energy power generating amount prediction technique according to claim 1 based on classification and error combination prediction,
Be characterized in that, the meteorological data in the step S1 include highest temperature value, lowest temperature value, every three hours temperature values,
Difference, relative humidity and the numerical value of maximum temperature and the difference of previous max. daily temperature, minimum temperature and previous Daily minimum temperature
The uitraviolet intensity of change.
3. the solar energy power generating amount prediction technique according to claim 1 based on classification and error combination prediction,
It is characterized in that, the meteorological data affiliated season in the step S2 includes spring, summer, fall and winter;Each season wraps
Include three kinds of fine day, cloudy day and rainy day weather patterns;
There is corresponding trained MPSO-BP neural network and trained under each mid-season each weather pattern
Gray model GM (1,1).
4. the solar energy power generating amount prediction technique according to claim 3 based on classification and error combination prediction,
It is characterized in that, the gray model GM (1,1) are as follows:
In formula, y(1)(ki+1 j) indicate to predict the cumulative sequential value of the i+1 output point of j-th obtained of sample;
y(1)(kij) indicate j-th of sample i-th of output point cumulative sequential value;
M is output point number in sample;
U, a model parameter to be asked;
U and a is determined according to the data being input in gray model.
5. the solar energy power generating amount prediction technique according to claim 4 based on classification and error combination prediction,
It is characterized in that, the method for corresponding gray model GM (1,1) under a weather pattern is trained in the step S2 specifically:
A1, the history meteorological data that three kinds of weather patterns in Various Seasonal are obtained from weather station;
A2, the history meteorological data under same mid-season same weather pattern is divided into one kind;
A3, according to the date data in similar history meteorological data, obtain the practical solar energy power generating amount in the corresponding date
Historical data, and be input in gray model GM (1,1) as training sample;
A4, the data according to the preceding j × m+i point of training sample, calculate the cumulative sequences y of gray model(1), and it is right to establish its
The calculating matrix answered;
A5, according to calculating matrix, the u and a in gray model are calculated by least square method, and carry it into cumulative sequences y(1)
In, obtain y(1)(ki+1 j), complete the training of gray model GM (1,1).
6. the solar energy power generating amount prediction technique according to claim 3 based on classification and error combination prediction,
It is characterized in that, in the step S2, the method for corresponding MPSO-BP neural network under one weather pattern of training specifically:
B1, the history meteorological data that three kinds of weather patterns in Various Seasonal are obtained from weather station;
B2, the history meteorological data under same mid-season same weather pattern is divided into one kind;
B3, according to the date data in similar history meteorological data, obtain the practical solar energy power generating amount in the corresponding date
Historical data;
B4, training data is inputted using history meteorological data as sample, it will be every in practical solar energy power generating amount historical data
Training data is exported as sample every half an hour solar energy power generating amount;
B5, the maximum frequency of training of MPSO-BP neural network is set as EPmaxConvergence error ε is predicted with expectationE;
B6, sample input training data is input in MPSO-BP neural network, to export sample output training data as target
It is trained, until frequency of training reaches the maximum frequency of training EP of settingmaxOr prediction error value ε when trainingpLess than setting
Expectation prediction error value εEWhen, complete the training of MPSO-BP neural network.
7. the solar energy power generating amount prediction technique according to claim 6 based on classification and error combination prediction,
It is characterized in that, the combined prediction matrix W in the trained MPSO-BP neural network of the determination in the step S3BPIt is instructed with determining
The combined prediction matrix W in gray model GM (1,1) perfectedCMMethod specifically:
C1, when completing the training of MPSO-BP neural network, determine i-th of output point of j-th sample input training data
Error are as follows:
In formula,I-th of output point output of training data is inputted for j-th of sample in MPSO-BP neural network
Predict solar energy power generating amount;
kijJ-th of sample to be input in MPSO-BP neural network inputs the corresponding history of i-th of output point of training data
Meteorological data;
yreal(kij) gone through for the corresponding practical solar energy power generating amount of i-th of output point of j-th of sample input training data
History data;
Meanwhile when completing the training of gray model GM (1,1), i-th of output point of j-th of sample input training data is determined
Error are as follows:
In formula,I-th of output point output of training data is inputted for j-th of sample in gray model GM (1,1)
Predict solar energy power generating amount;
kijTo be input to the corresponding history gas of i-th of output point that j-th of sample of gray model GM (1,1) inputs training data
Image data;
C2, the error that the corresponding output point of training data is inputted according to each sample, obtain error moments in MPSO-BP neural network
Battle array are as follows:
In formula, Et BPFor error matrix in neural network;
M is the number of output point;
Meanwhile the error of the corresponding output point of training data is inputted according to each sample, obtain error in gray model GM (1,1)
Matrix are as follows:
C3, it sums to the error of output point each in error matrix, obtains the prediction error matrix in MPSO-BP neural network
Are as follows:
In formula, a is a-th of training error sample;
Meanwhile to each summing with the error of the corresponding output point of neural network output in error matrix, gray model is obtained
Prediction error matrix in GM (1,1) are as follows:
C4, according to the prediction error matrix in the prediction error matrix and gray model GM (1,1) in MPSO-BP neural network,
The weight W of the prediction error matrix of MPSO-BP neural network is determined respectivelyBP kWith the prediction error matrix of gray model GM (1,1)
Weight WGM k:
Wherein, the weight of the prediction error matrix of MPSO-BP neural network are as follows:
The weight of the prediction error matrix of gray model GM (1,1) are as follows:
C5, according to MPSO-BP neural network prediction error matrix weight and gray model GM (1,1) prediction error matrix
Weight, determine the combined prediction matrix W in trained MPSO-BP neural network respectivelyBPWith trained gray model GM
(1,1) the combined prediction matrix W inGM;
Wherein, the combined prediction matrix W in trained MPSO-BP neural networkBPAre as follows:
Combined prediction matrix W in trained gray model GM (1,1)GMAre as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811570233.XA CN109636054A (en) | 2018-12-21 | 2018-12-21 | Solar energy power generating amount prediction technique based on classification and error combination prediction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811570233.XA CN109636054A (en) | 2018-12-21 | 2018-12-21 | Solar energy power generating amount prediction technique based on classification and error combination prediction |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109636054A true CN109636054A (en) | 2019-04-16 |
Family
ID=66076313
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811570233.XA Pending CN109636054A (en) | 2018-12-21 | 2018-12-21 | Solar energy power generating amount prediction technique based on classification and error combination prediction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109636054A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110598896A (en) * | 2019-07-26 | 2019-12-20 | 陕西省水利电力勘测设计研究院 | Photovoltaic power prediction method based on prediction error correction |
CN111178450A (en) * | 2019-12-31 | 2020-05-19 | 上海三一重机股份有限公司 | Method and device for evaluating state of welding seam |
CN111798055A (en) * | 2020-07-06 | 2020-10-20 | 国网山东省电力公司电力科学研究院 | Variable weight combined photovoltaic output prediction method based on grey correlation degree |
CN111814826A (en) * | 2020-06-08 | 2020-10-23 | 武汉理工大学 | Rapid detection and rating method for residual energy of retired power battery |
CN111931981A (en) * | 2020-07-06 | 2020-11-13 | 安徽天尚清洁能源科技有限公司 | Photovoltaic power generation ultra-short-term prediction method based on machine learning multi-model combination |
CN114048896A (en) * | 2021-10-27 | 2022-02-15 | 国核自仪系统工程有限公司 | Method, system, equipment and medium for predicting photovoltaic power generation data |
JP2022550619A (en) * | 2019-11-14 | 2022-12-02 | エンヴィジョン デジタル インターナショナル ピーティーイー.エルティーディー. | Methods for Processing Solar Radiation Predictions, Methods for Training Stacked Generalized Models, and Apparatuses Therefor |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103390199A (en) * | 2013-07-18 | 2013-11-13 | 国家电网公司 | Photovoltaic power generation capacity/power prediction device |
CN104021427A (en) * | 2014-06-10 | 2014-09-03 | 上海电力学院 | Method for predicting daily generating capacity of grid-connected photovoltaic power station based on factor analysis |
CN104463349A (en) * | 2014-11-11 | 2015-03-25 | 河海大学 | Photovoltaic generated power prediction method based on multi-period comprehensive similar days |
CN104820877A (en) * | 2015-05-21 | 2015-08-05 | 河海大学 | Photovoltaic system generation power prediction method based on cloud adaptive PSO-SNN |
CN104978611A (en) * | 2015-07-06 | 2015-10-14 | 东南大学 | Neural network photovoltaic power generation output prediction method based on grey correlation analysis |
CN104992248A (en) * | 2015-07-07 | 2015-10-21 | 中山大学 | Microgrid photovoltaic power station generating capacity combined forecasting method |
CN105160423A (en) * | 2015-09-14 | 2015-12-16 | 河海大学常州校区 | Photovoltaic power generation prediction method based on Markov residual error correction |
CN107358318A (en) * | 2017-06-29 | 2017-11-17 | 上海电力学院 | Based on GM(1,1)The urban power consumption Forecasting Methodology of model and Grey Markov chain predicting model |
CN107563573A (en) * | 2017-09-29 | 2018-01-09 | 南京航空航天大学 | A kind of Forecasting Methodology of the solar power generation amount based on adaptive learning mixed model |
-
2018
- 2018-12-21 CN CN201811570233.XA patent/CN109636054A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103390199A (en) * | 2013-07-18 | 2013-11-13 | 国家电网公司 | Photovoltaic power generation capacity/power prediction device |
CN104021427A (en) * | 2014-06-10 | 2014-09-03 | 上海电力学院 | Method for predicting daily generating capacity of grid-connected photovoltaic power station based on factor analysis |
CN104463349A (en) * | 2014-11-11 | 2015-03-25 | 河海大学 | Photovoltaic generated power prediction method based on multi-period comprehensive similar days |
CN104820877A (en) * | 2015-05-21 | 2015-08-05 | 河海大学 | Photovoltaic system generation power prediction method based on cloud adaptive PSO-SNN |
CN104978611A (en) * | 2015-07-06 | 2015-10-14 | 东南大学 | Neural network photovoltaic power generation output prediction method based on grey correlation analysis |
CN104992248A (en) * | 2015-07-07 | 2015-10-21 | 中山大学 | Microgrid photovoltaic power station generating capacity combined forecasting method |
CN105160423A (en) * | 2015-09-14 | 2015-12-16 | 河海大学常州校区 | Photovoltaic power generation prediction method based on Markov residual error correction |
CN107358318A (en) * | 2017-06-29 | 2017-11-17 | 上海电力学院 | Based on GM(1,1)The urban power consumption Forecasting Methodology of model and Grey Markov chain predicting model |
CN107563573A (en) * | 2017-09-29 | 2018-01-09 | 南京航空航天大学 | A kind of Forecasting Methodology of the solar power generation amount based on adaptive learning mixed model |
Non-Patent Citations (3)
Title |
---|
刘静宜: "基于组合模型的光伏电站发电功率短期预测方法研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
孙佳等: "基于改进灰色模型与BP神经网络模型组合的风力发电量预测研究", 《水电能源科学》 * |
师彪等: "改进粒子群—BP神经网络模型的短期电力负荷预测", 《计算机应用》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110598896A (en) * | 2019-07-26 | 2019-12-20 | 陕西省水利电力勘测设计研究院 | Photovoltaic power prediction method based on prediction error correction |
JP2022550619A (en) * | 2019-11-14 | 2022-12-02 | エンヴィジョン デジタル インターナショナル ピーティーイー.エルティーディー. | Methods for Processing Solar Radiation Predictions, Methods for Training Stacked Generalized Models, and Apparatuses Therefor |
JP7369868B2 (en) | 2019-11-14 | 2023-10-26 | エンヴィジョン デジタル インターナショナル ピーティーイー.エルティーディー. | Methods for processing solar radiation prediction, methods for training stacked generalized models, and apparatus thereof |
CN111178450A (en) * | 2019-12-31 | 2020-05-19 | 上海三一重机股份有限公司 | Method and device for evaluating state of welding seam |
CN111178450B (en) * | 2019-12-31 | 2023-07-14 | 上海三一重机股份有限公司 | Weld joint state evaluation method and device |
CN111814826A (en) * | 2020-06-08 | 2020-10-23 | 武汉理工大学 | Rapid detection and rating method for residual energy of retired power battery |
CN111814826B (en) * | 2020-06-08 | 2022-06-03 | 武汉理工大学 | Rapid detection and rating method for residual energy of retired power battery |
CN111798055A (en) * | 2020-07-06 | 2020-10-20 | 国网山东省电力公司电力科学研究院 | Variable weight combined photovoltaic output prediction method based on grey correlation degree |
CN111931981A (en) * | 2020-07-06 | 2020-11-13 | 安徽天尚清洁能源科技有限公司 | Photovoltaic power generation ultra-short-term prediction method based on machine learning multi-model combination |
CN114048896A (en) * | 2021-10-27 | 2022-02-15 | 国核自仪系统工程有限公司 | Method, system, equipment and medium for predicting photovoltaic power generation data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109636054A (en) | Solar energy power generating amount prediction technique based on classification and error combination prediction | |
CN104573879B (en) | Photovoltaic plant based on optimal similar day collection goes out force prediction method | |
Tang et al. | Entropy method combined with extreme learning machine method for the short-term photovoltaic power generation forecasting | |
CN109858673A (en) | A kind of photovoltaic generating system power forecasting method | |
CN102930358B (en) | A kind of neural net prediction method of photovoltaic power station power generation power | |
CN103049798B (en) | A kind of short-term power generation power Forecasting Methodology being applied to photovoltaic generating system | |
CN105184678A (en) | Method for constructing photovoltaic power station generation capacity short-term prediction model based on multiple neural network combinational algorithms | |
CN106529814A (en) | Distributed photovoltaic ultra-short-term forecasting method based on Adaboost clustering and Markov chain | |
Oudjana et al. | Short term photovoltaic power generation forecasting using neural network | |
CN110729764B (en) | Optimal scheduling method for photovoltaic power generation system | |
CN105631558A (en) | BP neural network photovoltaic power generation system power prediction method based on similar day | |
CN106295899B (en) | Wind power probability density Forecasting Methodology based on genetic algorithm Yu supporting vector quantile estimate | |
CN106251001A (en) | A kind of based on the photovoltaic power Forecasting Methodology improving fuzzy clustering algorithm | |
CN109002915A (en) | Photovoltaic plant short term power prediction technique based on Kmeans-GRA-Elman model | |
CN109934395B (en) | Multi-hydropower-region short-term power load prediction method based on time-sharing and regional meteorological data | |
CN110175421B (en) | Novel water-light complementary multi-target optimization operation method | |
CN107256436B (en) | Predictive matching and consumption control method for heat storage electric boiler and clean energy | |
CN104050517A (en) | Photovoltaic power generation forecasting method based on GRNN | |
CN109376950A (en) | A kind of polynary Load Forecasting based on BP neural network | |
CN104376371B (en) | A kind of distribution based on topology is layered load forecasting method | |
CN102129511A (en) | System for forecasting short-term wind speed of wind power station based on MATLAB | |
CN103400204A (en) | Forecasting method for solar photovoltaic electricity generation amount based on SVM (support vector machine) - Markov combination method | |
CN108075471B (en) | Multi-objective constraint optimization power grid scheduling strategy based on stochastic power output prediction | |
CN112994115A (en) | New energy capacity configuration method based on WGAN scene simulation and time sequence production simulation | |
CN110866633A (en) | Micro-grid ultra-short term load prediction method based on SVR support vector regression |
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: 20190416 |
|
RJ01 | Rejection of invention patent application after publication |