CN109711609B - Photovoltaic power station output power prediction method based on wavelet transformation and extreme learning machine - Google Patents
Photovoltaic power station output power prediction method based on wavelet transformation and extreme learning machine Download PDFInfo
- Publication number
- CN109711609B CN109711609B CN201811539356.7A CN201811539356A CN109711609B CN 109711609 B CN109711609 B CN 109711609B CN 201811539356 A CN201811539356 A CN 201811539356A CN 109711609 B CN109711609 B CN 109711609B
- Authority
- CN
- China
- Prior art keywords
- photovoltaic power
- power station
- output power
- data
- historical
- 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.)
- Active
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention relates to a photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine. Firstly, extracting a prediction data set of a photovoltaic power station from a historical power generation and meteorological environment parameter monitoring data set of the photovoltaic power station; secondly, preprocessing a photovoltaic power station prediction data set; secondly, extracting features from historical power data of the photovoltaic power station by adopting a PCA algorithm, and performing secondary classification by utilizing a K-means algorithm to obtain a smooth type and a fluctuating type; and finally, acquiring meteorological characteristic parameters of the day to be predicted through NWP to generate a test set, judging the type of the test set according to Euclidean distance, and traversing to find an optimal training set. The smooth type directly utilizes the extreme learning machine network to predict the output power of the photovoltaic power station. The fluctuation type data needs to be subjected to feature extraction and prediction one by one through a WT algorithm and reconstructed into a predicted value. The photovoltaic power station output power prediction method based on the extreme learning machine can effectively improve the accuracy of photovoltaic power station output power prediction.
Description
Technical Field
The invention relates to a photovoltaic power station output power prediction technology, in particular to a photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine.
Background
Because the factors influencing the photovoltaic power generation are unstable, the output power time sequence change of the photovoltaic power station is unsmooth, and the fluctuation, intermittence and randomness are high, when a large number of photovoltaic power generation stations are incorporated into the existing national household appliance network system, the balance state of the whole power generation and power utilization system can be greatly influenced, meanwhile, great challenges are brought to the safe and stable operation and the electric energy production and utilization of the whole power system, the output power prediction of the photovoltaic power generation stations is deeply researched, and the academic research and practical application values are high. If the output power of the photovoltaic power station can be accurately predicted, the power grid electric power can be timely scheduled in advance, so that the influence on the balance state of a power generation and power utilization system after a large number of photovoltaic power stations are merged into a power grid system can be effectively reduced, and meanwhile, the influence on the safe and stable operation of the power grid system by the photovoltaic power stations can be effectively reduced or even avoided. Certain guarantee is provided for the photovoltaic power station to be capable of being incorporated into the existing power grid system in a large scale, and the method has important significance for promoting and developing large-scale photovoltaic power station grid connection. Through the prediction of the power output of the photovoltaic power station, the optimization of a power grid system can be further improved, the power generation and the power consumption of the power grid system are more controllable and adjustable, the operation cost of the power system is reduced, and each power dispatching is reasonable. The influence of grid connection of a photovoltaic power station on the whole power system is reduced, the utilization rate of solar energy resources is improved, and the win-win situation of economic benefit and social benefit is realized.
In recent years, various power prediction methods and techniques have been proposed in succession. The existing mainstream photovoltaic output power prediction method can be wholly divided into two categories: the first type mainly tends to predict the power generation power of the photovoltaic power station without using the environmental parameters of the photovoltaic power station, and directly predicts the output power of the photovoltaic power station, and commonly used prediction models mainly comprise a Markov chain model, a statistical method model, a gray scale model and the like; the other type of the photovoltaic power station is more inclined to consider the use of environmental factors related to the influence on the output power of the photovoltaic power station, and indirectly predicts the output power of the photovoltaic power station by carrying out a series of analysis modeling on the environmental factors related to the influence on the output power around the photovoltaic power station, wherein the environmental factors with higher use frequency mainly comprise the environmental temperature, the relative humidity, the irradiance and the wind speed. However, the existing photovoltaic power station output power prediction model generally has the problems of low precision, poor universality, insufficient consideration of factors influencing the output power of the photovoltaic power station, small data set and the like.
WT and K-means algorithms are introduced into the ELM network to optimize the training set precision of samples, different types of databases of photovoltaic power station environments are built, the accuracy of output power prediction of the photovoltaic power station of the ELM network is improved, and the prediction time is shortened.
At present, no study for simultaneously introducing WT and K-means algorithms into an extreme learning machine to predict the output power of a photovoltaic power station is found in publicly published documents and patents.
Disclosure of Invention
The invention aims to provide a photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine, so as to overcome the defects in the prior art.
In order to achieve the purpose, the technical scheme of the invention is as follows: a photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine comprises the following steps:
step S1, extracting the output power, horizontal irradiance, diffused irradiance, environmental humidity and environmental temperature data of the last year from the historical power generation and meteorological environment parameter monitoring data set of the photovoltaic power station as a prediction data set of the photovoltaic power station;
s2, performing statistical analysis and normalization pretreatment on the prediction data set of the photovoltaic power station to establish sample data;
step S3, extracting the smoothness characteristic of the photovoltaic output power curve of the photovoltaic power station historical output power data by adopting a Principal Component Analysis (PCA) method;
s4, according to the characteristics extracted in the step S3, carrying out secondary classification on the obtained characteristic data by using a K-means classifier, namely respectively storing and generating a training set according to a smooth type and a fluctuating type;
step S5, acquiring meteorological parameters of a day to be predicted through a numerical weather forecast center NWP, cleaning and normalizing the meteorological parameters to generate a test set;
step S6, judging the type of the day to be tested according to the Euclidean distance formula through processing the step S3 and the step S5, and selecting the most appropriate training set through traversing the Pearson correlation coefficient and the Euclidean distance of the historical power and the irradiance of the testing set;
s7, if the type of the day to be measured belongs to a smooth type, directly establishing a training model for the output power of the photovoltaic power station by using an ELM network, and performing inverse normalization processing on the output result to obtain a predicted value; if the type of the day to be measured belongs to a fluctuation type, performing characteristic decomposition on the training set historical data and the data of the day to be predicted by adopting a wavelet transform algorithm WT to obtain various decomposed characteristic quantities, respectively and correspondingly establishing a training model for the output power of the photovoltaic power station by utilizing an ELM network, performing wavelet reconstruction processing on the obtained components to obtain an output power sequence, and finally performing inverse normalization processing to obtain a predicted value.
In an embodiment of the present invention, in step S2, the prediction data set of the extracted photovoltaic power plant is subjected to statistical analysis and normalization preprocessing in the following manners:
s21, removing data which do not accord with actual conditions from the prediction data set of the photovoltaic power station through statistical processing, and cleaning incomplete data generated by problems of a data acquisition system;
step S22, normalizing the data processed in step S21 to be within the [0,1] interval.
In an embodiment of the present invention, the step S3 is specifically implemented as follows:
assuming that X is an m X n matrix representing data of m features of n objects, i.e. each column represents an object and each row represents a feature; reducing the feature to d dimension, wherein d is far less than m; outputting a matrix with the output result of Y, wherein Y is d x n;
let X be ═ X 1 ,x 2 ...x n ]Calculating an average value of each object point
To pair
Performing singular value decomposition X-X 0 ＝UΛV T (ii) a X is then 0 The point of origin of the new coordinate system is obtained, and the front d column of U is the new coordinate system after decentralization and is marked as W; then, some points are represented by Y ═ W in the new coordinate system T *(X-x 0 ) (ii) a Similarly, to restore the new proxel y to the original coordinate system, it can be written as x0+ W; here, the implementation is implemented by using an MATLAB tool box primary () function, which is specifically represented as:
[coeff score latent]＝princomp(X)
wherein coeff is a principal component, i.e., an eigenvector of the sample covariance matrix; score principal component, which is the expression form of the sample X in the low-dimensional space, i.e. the projection of the sample X on the principal component coeff, if the k-dimension needs to be reduced, only the principal component components of the first k columns need to be taken; latient is a vector containing the eigenvalues of the covariance matrix of the sample;
in the process, X is 4 meteorological values of historical output power data of the photovoltaic power station, and the extracted features are latient features.
In an embodiment of the present invention, the specific implementation process of the steps S5 to S7 is as follows:
from the Euclidean distance formula
The method comprises the steps of judging the attribution type of historical output power data of a photovoltaic power station and a predicted value provided by a numerical weather forecast center NWP (non-Newton-per-unit), and introducing a Peachno correlation coefficient function
Combining the training data with the training data to select an optimal training set;
when the type of the day to be predicted is judged to be smooth, the weather parameters of the day to be predicted and the optimal training set data are directly put into an ELM prediction network for power prediction, and the output result of the prediction model is subjected to inverse normalization processing to obtain the predicted value of the output power of the predicted day;
when it is determined that the type of day to be predicted belongs to the fluctuation type, wavelets are usedThe transformation algorithm WT carries out 3-layer feature extraction on the historical power and the historical meteorological parameters of the training set and the meteorological parameters of the day to be predicted respectively, wherein a dwt function and an idwt function are selected as a kernel function of the WT algorithm to carry out 3-layer decomposition on the historical power and the historical meteorological parameters of the training set and the meteorological parameters of the day to be predicted, the dwt function is a single-layer one-dimensional wavelet decomposition performed on a low-frequency signal Lo _ D and a high-frequency signal Hi _ D of a specific wavelet decomposition filter, and the idwt function is used for reconstructing the decomposed signals; the historical power and the high-low frequency characteristics of each historical meteorological parameter of the training set are as follows: PD (photo diode) 3 、PD 2 、PD 1 、MD 3 、MD 2 、MD 1 、PA 3 (ii) a The weather parameter high-low frequency characteristics of the day to be predicted are as follows: TMD 3 、TMD 2 、TMD 1 、TMA 3 (ii) a Putting each component into an ELM prediction network for component prediction to obtain A 3 、D 3 、D 2 、D 1 Using MATLAB tool box wavelet reconstruction function X ═ waverec (A, D, 'wname'); and performing inverse normalization processing on the reconstruction result to obtain the predicted value of the output power of the prediction day.
Compared with the prior art, the invention has the following beneficial effects: according to the photovoltaic power station output power prediction method based on wavelet transformation and the extreme learning machine, the simulation and example verification analysis results show that the root Mean Square Error (MSE) and the R are used 2 Introducing evaluation as a prediction result, and combining PCA algorithm extraction and K-means classification algorithm with WT and ELM network;
the experimental result shows that the result predicted by the method is very high in precision. By integrating and optimizing the input characteristic vectors, the training time of the neural network is greatly shortened, and the training precision and the testing precision are greatly improved. The method can accurately predict the output power of the photovoltaic power station under various working conditions.
Drawings
FIG. 1 is a flow chart of a photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine in the invention.
Fig. 2 is a flowchart of a specific implementation of the method for predicting the output power of the photovoltaic power station based on the wavelet transform and the extreme learning machine according to an embodiment of the present invention.
FIG. 3 is a diagram of the effect of the PCA and K-means algorithms on two classes (10-day data per class drawn randomly) in accordance with an embodiment of the present invention.
Fig. 4 is a diagram of prediction results under different training set types based on wavelet transformation and extreme learning machine prediction models in an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is specifically explained below with reference to the accompanying drawings.
As shown in fig. 1, the invention provides a photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine, comprising the following steps:
step S1, extracting the output power, horizontal irradiance, diffused irradiance, environmental humidity and environmental temperature data of the last year from the historical power generation and meteorological environment parameter monitoring data set of the photovoltaic power station as a prediction data set of the photovoltaic power station;
s2, performing statistical analysis and normalization pretreatment on the prediction data set of the photovoltaic power station to establish sample data;
step S3, extracting the smoothness characteristic of the photovoltaic output power curve of the photovoltaic power station historical output power data by adopting a Principal Component Analysis (PCA) method;
s4, according to the characteristics extracted in the step S3, carrying out secondary classification on the obtained characteristic data by using a K-means classifier, namely respectively storing and generating a training set according to a smooth type and a fluctuating type;
step S5, acquiring meteorological parameters of a day to be predicted through a numerical weather forecast center NWP, cleaning and normalizing the meteorological parameters to generate a test set;
step S6, judging the type of the day to be tested according to the Euclidean distance formula through processing the step S3 and the step S5, and selecting the most appropriate training set through traversing the Pearson correlation coefficient and the Euclidean distance of the historical power and the irradiance of the testing set;
s7, if the type of the day to be measured belongs to a smooth type, directly establishing a training model for the output power of the photovoltaic power station by using an ELM network, and performing inverse normalization processing on the output result to obtain a predicted value; if the type of the day to be measured belongs to a fluctuation type, performing characteristic decomposition on the training set historical data and the data of the day to be predicted by adopting a wavelet transform algorithm WT to obtain various decomposed characteristic quantities, respectively and correspondingly establishing a training model for the output power of the photovoltaic power station by utilizing an ELM network, performing wavelet reconstruction processing on the obtained components to obtain an output power sequence, and finally performing inverse normalization processing to obtain a predicted value.
In step S2, the prediction data set of the extracted photovoltaic power plant is subjected to statistical analysis and normalization preprocessing in the following ways:
s21, removing data which do not accord with actual conditions from the prediction data set of the photovoltaic power station through statistical processing, and cleaning incomplete data generated by problems of a data acquisition system;
step S22, normalizing the data processed in step S21 to be within the [0,1] interval.
In the step S3, the specific implementation process is as follows:
assuming that X is an m X n matrix representing data of m features of n objects, i.e. each column represents an object and each row represents a feature; reducing the feature to d dimension, wherein d is far less than m; outputting a matrix with the output result of Y, wherein Y is d x n;
let X be ═ X 1 ,x 2 ...x n ]Calculating an average value of each object point
To pair
Performing singular value decomposition X-X 0 ＝UΛV T (ii) a X is then 0 The point of origin of the new coordinate system is obtained, and the front d column of U is the new coordinate system after decentralization and is marked as W; then, some points are represented by Y ═ W in the new coordinate system T *(X-x 0 ) (ii) a Similarly, to restore the new proxel y to the original coordinate system, it can be written as x0+ W; here, the implementation is implemented by using an MATLAB tool box primary () function, which is specifically represented as:
[coeff score latent]＝princomp(X)
wherein coeff is a principal component, i.e., an eigenvector of the sample covariance matrix; score principal component, which is the expression form of the sample X in the low-dimensional space, i.e. the projection of the sample X on the principal component coeff, if the k-dimension needs to be reduced, only the principal component components in the first k columns need to be taken; a later is a vector containing the eigenvalues of the covariance matrix of the samples;
in the process, X is 4 meteorological values of historical output power data of the photovoltaic power station, and the extracted features are latient features.
The specific implementation process of the steps S5 to S7 is as follows:
from the Euclidean distance formula
The method comprises the steps of judging the attribution type of historical output power data of a photovoltaic power station and a predicted value provided by a numerical weather forecast center NWP (non-Newton-per-unit), and introducing a Peachno correlation coefficient function
Combining the training data with the training data to select an optimal training set;
when the type of the day to be predicted is judged to be smooth, the weather parameters of the day to be predicted and the optimal training set data are directly put into an ELM prediction network for power prediction, and the output result of the prediction model is subjected to inverse normalization processing to obtain the predicted value of the output power of the predicted day;
when the type of the day to be predicted is judged to be a fluctuation type, 3-layer feature extraction is carried out on the historical power and each historical meteorological parameter of a training set and each meteorological parameter of the day to be predicted respectively by using a wavelet transform algorithm WT, wherein a dwt function and an idwt function are selected as WT algorithm kernel functions to carry out 3-layer decomposition on the historical power and each historical meteorological parameter of the training set and each meteorological parameter of the day to be predicted, the dwt function is that a specific wavelet decomposition filter low-frequency signal Lo _ D and a high-frequency signal Hi _ D execute single-layer one-dimensional wavelet decomposition, and the idwt function is that decomposed signals are reconstructed; the historical power and the historical meteorological parameters of the training set are highThe low frequency characteristic is: PD (photo diode) 3 、PD 2 、PD 1 、MD 3 、MD 2 、MD 1 、PA 3 (ii) a The weather parameter high-low frequency characteristics of the day to be predicted are as follows: TMD 3 、TMD 2 、TMD 1 、TMA 3 (ii) a Putting each component into an ELM prediction network for component prediction to obtain A 3 、D 3 、D 2 、D 1 Using MATLAB tool box wavelet reconstruction function X ═ waverec (A, D, 'wname'); and performing inverse normalization processing on the reconstruction result to obtain the predicted value of the output power of the prediction day.
The following is a specific implementation of the present invention.
The invention provides a photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine, and a flow diagram is shown in figure 2. Fig. 3 is a classification effect graph of the PCA and K-means algorithms in an embodiment of the present invention, and different data sets are selected according to the historical output power variation trend and the irradiance variation trend of the day to be predicted, which specifically includes the following steps:
step S1: extracting the data of output power generation power, horizontal irradiance, diffused irradiance, environmental humidity and environmental temperature of the last year from the historical power generation and meteorological environment parameter monitoring data set of the photovoltaic power station as a prediction data set of the photovoltaic power station;
step S2: preprocessing a photovoltaic power station prediction data set, such as statistical analysis, normalization and the like, and establishing sample data;
step S3: for historical output power data of the photovoltaic power station, extracting the smoothness characteristic of a photovoltaic output power curve by adopting a Principal Component Analysis (PCA) method;
step S4: according to the features extracted in the step S3, performing secondary classification on the obtained feature data by using a K-means classifier, namely storing a smooth type and a fluctuating type to generate a training set;
step S5: acquiring meteorological parameters of a day to be predicted through a numerical weather forecast center (NWP), and cleaning and normalizing the meteorological parameters to generate a test set;
step S6: processing the working conditions of the test set through the step S3 and the step S5, judging the type of the day to be tested according to an Euclidean distance formula, and selecting the most appropriate training set by traversing the Pearson correlation coefficient and the Euclidean distance of the historical power and the irradiance of the test set;
step S7: and establishing a training model for the output power of the smooth photovoltaic power station by directly utilizing the ELM network, and performing inverse normalization processing on the output result to obtain a predicted value. And for the fluctuation type, performing characteristic decomposition on historical data of a training set and data of a day to be predicted by using a wavelet transform algorithm (WT) to obtain various decomposed characteristic quantities, respectively and correspondingly establishing a training model for the output power of the photovoltaic power station by using the ELM network, performing wavelet reconstruction processing on the obtained components to obtain an output power sequence, and finally performing inverse normalization processing to obtain a predicted value.
Preferably, in this embodiment, the collected data is from the australian profit solar center.
Further, in the present embodiment, the five parameters of the photovoltaic power station in step S1 include horizontal irradiance, diffused irradiance, humidity, temperature data, and actual photovoltaic power generation power.
Further, in the present embodiment, PCA and K-means algorithms are used to perform two-classification preprocessing on the historical data, including smoothness and volatility.
Preferably, in this embodiment, for the smooth training set, the ELM network modeling prediction is directly used, and for the fluctuating training set, we use the WT algorithm to perform the following steps on the historical power and the high-low frequency characteristics of the historical meteorological parameters: PD (photo diode) 3 ,PD 2 ,PD 1 ,MD 3 ,MD 2 ,MD 1 ,PA 3 The high-low frequency characteristics of each meteorological parameter of the day to be predicted are as follows: TMD 3 ,TMD 2 ,TMD 1 ,TMA 3 . Putting each component into an ELM prediction network for component prediction to obtain A 3 ,D 3 ,D 2 ,D 1 The MATLAB toolkit wavelet reconstruction function X is used for waverec (A, D, 'wonmen'). And performing inverse normalization processing on the reconstruction result to obtain the predicted value of the output power of the prediction day.
In this example, the total number of samples collected was 3872 sets using the processed 352 day data for 2017 years.
Further, in this embodiment, the highest accuracy of the photovoltaic power station output power prediction model based on the extreme learning machine can reach more than 99.7%, and table 1 shows the prediction test results of effective monitoring periods of 4 consecutive days in different seasons.
TABLE 1 prediction results under different conditions
The above are preferred embodiments of the present invention, and all changes made according to the technical scheme of the present invention that produce functional effects do not exceed the scope of the technical scheme of the present invention belong to the protection scope of the present invention.
Claims (3)
1. A photovoltaic power station output power prediction method based on wavelet transformation and an extreme learning machine is characterized by comprising the following steps:
step S1, extracting historical output power, horizontal irradiance, diffused irradiance, environmental humidity and environmental temperature data of the photovoltaic power station in the last year from the historical power generation and meteorological parameter historical monitoring data set of the photovoltaic power station, and using the historical output power, horizontal irradiance, diffused irradiance, environmental humidity and environmental temperature data as a prediction data set of the photovoltaic power station;
s2, performing statistical analysis and normalization pretreatment on the prediction data set of the photovoltaic power station to establish sample data;
step S3, extracting the smoothness degree characteristic of the historical output power curve of the photovoltaic power station by adopting a Principal Component Analysis (PCA) method for the historical output power data of the photovoltaic power station;
s4, according to the characteristics extracted in the step S3, carrying out secondary classification on the obtained characteristic data by using a K-means classifier, namely respectively storing and generating a training set according to a smooth type and a fluctuating type;
s5, acquiring meteorological parameters of a day to be tested through a numerical weather forecast center NWP, and cleaning and normalizing to generate a test set;
step S6, judging the type of the day to be tested according to an Euclidean distance formula through processing the step S3 and the step S5, and selecting an optimal training set by traversing the Pearson correlation coefficient of the historical output power of the photovoltaic power station for training and the irradiance of the testing set and combining the optimal training set with the Euclidean distance;
s7, if the type of the day to be tested belongs to a smooth type, directly establishing a training model for the historical output power of the photovoltaic power station by using an ELM network of an extreme learning machine, and performing inverse normalization processing on the output result to obtain a predicted value; if the type of the day to be measured belongs to a fluctuation type, performing characteristic decomposition on historical data of a training set and data of the day to be measured by adopting a wavelet transform algorithm WT to obtain various decomposed characteristic quantities, then establishing a training model for historical output power of the photovoltaic power station by utilizing an ELM (extreme learning machine) network, performing wavelet reconstruction processing on each obtained component to obtain an output power sequence, and finally performing inverse normalization processing to obtain a predicted value;
the specific implementation process of the steps S6 to S7 is as follows:
from the Euclidean distance formula
The attribution type judgment is carried out on the historical output power data of the photovoltaic power station and the predicted value provided by a numerical weather forecast center NWP, and then a Pearson correlation coefficient function is introduced
Selecting an optimal training set by combining with the Euclidean distance;
when the type of the day to be measured is judged to be smooth, the weather parameters of the day to be measured and the optimal training set data are directly put into an extreme learning machine ELM network for power prediction, and the output result of the prediction model is subjected to inverse normalization processing to obtain the predicted value of the output power of the day to be measured;
when the type of the day to be tested is judged to be fluctuation type, 3-layer feature extraction is respectively carried out on the historical output power and the historical meteorological parameters of the photovoltaic power station of the training set and the meteorological parameters of the day to be tested by using a wavelet transform algorithm WT, wherein the dwt function and the idwt function are selected as the kernel function of the WT algorithm to the photovoltaic of the training setPerforming 3-layer decomposition on the historical output power and each historical meteorological parameter of the power station and each meteorological parameter of a day to be measured, wherein the dwt function is that a specific wavelet decomposition filter low-frequency signal Lo _ D and a high-frequency signal Hi _ D execute single-layer one-dimensional wavelet decomposition, and the idwt function is used for reconstructing the decomposed signals; the historical output power of the photovoltaic power station and the high-low frequency characteristics of the historical meteorological parameters of the training set are as follows: PD (photo diode) 3 、PD 2 、PD 1 、MD 3 、MD 2 、MD 1 、PA 3 、MA 3 (ii) a The high-low frequency characteristics of each meteorological parameter of the day to be measured are as follows: TMD 3 、TMD 2 、TMD 1 、TMA 3 (ii) a Putting each component into an extreme learning machine ELM network for component prediction to obtain A 3 、D 3 、D 2 、D 1 Using MATLAB tool box wavelet reconstruction function X ═ waverec (A, D, 'wname'); and performing inverse normalization processing on the reconstruction result to obtain the predicted value of the output power of the day to be measured.
2. The method for predicting the output power of the photovoltaic power plant based on the wavelet transform and the extreme learning machine as claimed in claim 1, wherein in the step S2, the statistical analysis and normalization preprocessing is performed on the prediction data set of the extracted photovoltaic power plant by the following method:
s21, removing data which do not accord with actual conditions from the prediction data set of the photovoltaic power station through statistical processing, and cleaning incomplete data generated by problems of a data acquisition system;
step S22, normalizing the data processed in step S21 to be within the [0,1] interval.
3. The method for predicting the output power of the photovoltaic power station based on the wavelet transform and the extreme learning machine as claimed in claim 1, wherein the step S3 is implemented as follows:
assuming that X is an m X n matrix representing data of m features of n objects, i.e. each column represents an object and each row represents a feature; reducing the feature to d dimension, wherein d is far less than m; the output result is Y ', Y' is a matrix of d x n;
let X be ═ X 1 ,x 2 ,...,x n ]Calculating an average value of each object point
To pair
Performing singular value decomposition X-X 0 ＝UΛV T (ii) a X is then 0 The new coordinate system is the original point of the new coordinate system, the front d row of the U is the new coordinate system after the decentralization and is marked as W; then, it is represented as Y' ═ W in the new coordinate system T (X-x 0 ) (ii) a Similarly, the projection point Y' in the new coordinate system is restored to the original coordinate system and written as x 0 + W x Y'; here, the implementation is implemented by using an MATLAB tool box primary () function, which is specifically represented as:
[coeff score latent]＝princomp(X)
wherein coeff is a principal component, i.e., an eigenvector of the sample covariance matrix; score is a principal component and is a representation form of X in a low-dimensional space, namely the projection of X on a principal component coeff, and if the k dimensions are required to be reduced, the principal component components of the first k columns are taken; latient is a vector containing the eigenvalues of the covariance matrix of the sample;
in the process, X is a training set formed by historical output power data of the photovoltaic power station, and the extracted features are latient features.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811539356.7A CN109711609B (en) | 2018-12-15 | 2018-12-15 | Photovoltaic power station output power prediction method based on wavelet transformation and extreme learning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811539356.7A CN109711609B (en) | 2018-12-15 | 2018-12-15 | Photovoltaic power station output power prediction method based on wavelet transformation and extreme learning machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109711609A CN109711609A (en) | 2019-05-03 |
| CN109711609B true CN109711609B (en) | 2022-08-12 |
Family
ID=66256661
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811539356.7A Active CN109711609B (en) | 2018-12-15 | 2018-12-15 | Photovoltaic power station output power prediction method based on wavelet transformation and extreme learning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109711609B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110163437B (en) * | 2019-05-23 | 2022-05-13 | 太原理工大学 | Day-ahead photovoltaic power generation power prediction method based on DPK-means |
| CN110334870B (en) * | 2019-07-09 | 2022-05-31 | 福州大学 | Photovoltaic power station short-term power prediction method based on gated cyclic unit network |
| CN110826803A (en) * | 2019-11-06 | 2020-02-21 | 广东电力交易中心有限责任公司 | Electricity price prediction method and device for electric power spot market |
| CN111130117B (en) * | 2020-01-07 | 2021-02-19 | 杭州电子科技大学 | Probability optimal power flow calculation method based on high-dimensional data clustering |
| CN111680826A (en) * | 2020-05-14 | 2020-09-18 | 沂南力诺太阳能电力工程有限公司 | Photovoltaic power generation capacity prediction analysis method |
| CN111832825B (en) * | 2020-07-14 | 2024-04-16 | 山东大学 | Wind power prediction method and system integrating long-term memory network and extreme learning machine |
| CN111915084A (en) * | 2020-08-03 | 2020-11-10 | 国网山东省电力公司电力科学研究院 | Hybrid photovoltaic power generation power prediction method and system based on neural network |
| CN112085350A (en) * | 2020-08-20 | 2020-12-15 | 国电新能源技术研究院有限公司 | Method for evaluating photovoltaic array state in large photovoltaic power station |
| CN112149905A (en) * | 2020-09-25 | 2020-12-29 | 福州大学 | Photovoltaic power station short-term power prediction method based on wavelet transformation and wavelet neural network |
| CN112364477B (en) * | 2020-09-29 | 2022-12-06 | 中国电器科学研究院股份有限公司 | Outdoor empirical prediction model library generation method and system |
| CN112257941B (en) * | 2020-10-28 | 2023-04-18 | 福州大学 | Photovoltaic power station short-term power prediction method based on improved Bi-LSTM |
| CN112965060A (en) * | 2021-02-19 | 2021-06-15 | 加特兰微电子科技(上海)有限公司 | Detection method and device for vital sign parameters and method for detecting physical sign points |
| CN112949936A (en) * | 2021-03-29 | 2021-06-11 | 福州大学 | Short-term photovoltaic power prediction method based on similar-day wavelet transform and multilayer perceptron |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013074695A (en) * | 2011-09-27 | 2013-04-22 | Meiji Univ | Device, method and program for predicting photovoltaic generation |
| CN106803135A (en) * | 2017-01-17 | 2017-06-06 | 广东工业大学 | The Forecasting Methodology and device of a kind of photovoltaic power generation system output power |
| CN108280551A (en) * | 2018-02-02 | 2018-07-13 | 华北电力大学 | A kind of photovoltaic power generation power prediction method using shot and long term memory network |
| CN109002915A (en) * | 2018-07-13 | 2018-12-14 | 福州大学 | Photovoltaic plant short term power prediction technique based on Kmeans-GRA-Elman model |
-
2018
- 2018-12-15 CN CN201811539356.7A patent/CN109711609B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013074695A (en) * | 2011-09-27 | 2013-04-22 | Meiji Univ | Device, method and program for predicting photovoltaic generation |
| CN106803135A (en) * | 2017-01-17 | 2017-06-06 | 广东工业大学 | The Forecasting Methodology and device of a kind of photovoltaic power generation system output power |
| CN108280551A (en) * | 2018-02-02 | 2018-07-13 | 华北电力大学 | A kind of photovoltaic power generation power prediction method using shot and long term memory network |
| CN109002915A (en) * | 2018-07-13 | 2018-12-14 | 福州大学 | Photovoltaic plant short term power prediction technique based on Kmeans-GRA-Elman model |
Non-Patent Citations (2)
| Title |
|---|
| 基于EMD与ELM的光伏电站短期功率预测;李多等;《可再生能源》;20160220(第02期);全文 * |
| 基于小波分解与极限学习机的光伏电站辐照度短期预测研究;张俊蔚;《青海电力》;20151231(第04期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109711609A (en) | 2019-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109711609B (en) | Photovoltaic power station output power prediction method based on wavelet transformation and extreme learning machine | |
| CN106779223B (en) | Photovoltaic system power generation real-time prediction method and device | |
| CN106447098B (en) | Photovoltaic ultra-short-term power prediction method and device | |
| CN112257941B (en) | Photovoltaic power station short-term power prediction method based on improved Bi-LSTM | |
| CN110309603A (en) | A kind of short-term wind speed forecasting method and system based on wind speed characteristics | |
| CN109255728B (en) | Photovoltaic power generation power neural network prediction method based on chaotic phase space optimization reconstruction | |
| CN110766200A (en) | Method for predicting generating power of wind turbine generator based on K-means mean clustering | |
| CN110378286B (en) | DBN-ELM-based electric energy quality disturbance signal classification method | |
| CN111027775A (en) | Step hydropower station generating capacity prediction method based on long-term and short-term memory network | |
| Shi et al. | Genetic algorithm-piecewise support vector machine model for short term wind power prediction | |
| CN115829105A (en) | Photovoltaic power prediction method based on historical data feature search | |
| CN110717610A (en) | Wind power prediction method based on data mining | |
| CN112925824A (en) | Photovoltaic power prediction method and system for extreme weather type | |
| CN116151464A (en) | Photovoltaic power generation power prediction method, system and storable medium | |
| CN112149905A (en) | Photovoltaic power station short-term power prediction method based on wavelet transformation and wavelet neural network | |
| CN114021483A (en) | Ultra-short-term wind power prediction method based on time domain characteristics and XGboost | |
| CN115115090A (en) | Wind power short-term prediction method based on improved LSTM-CNN | |
| CN109921462B (en) | New energy consumption capability assessment method and system based on LSTM | |
| CN114266416A (en) | Photovoltaic power generation power short-term prediction method and device based on similar days and storage medium | |
| CN107358323A (en) | A kind of power forecasting method of short-term photovoltaic generation | |
| CN113837434A (en) | Solar photovoltaic power generation prediction method and device, electronic equipment and storage medium | |
| CN107844872B (en) | Short-term wind speed forecasting method for wind power generation | |
| CN112836876A (en) | Power distribution network line load prediction method based on deep learning | |
| CN111090679B (en) | Time sequence data representation learning method based on time sequence influence and graph embedding | |
| CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning |
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 |