NL2033884B1 - A Photovoltaic Power Prediction Method, System and Medium - Google Patents
A Photovoltaic Power Prediction Method, System and Medium Download PDFInfo
- Publication number
- NL2033884B1 NL2033884B1 NL2033884A NL2033884A NL2033884B1 NL 2033884 B1 NL2033884 B1 NL 2033884B1 NL 2033884 A NL2033884 A NL 2033884A NL 2033884 A NL2033884 A NL 2033884A NL 2033884 B1 NL2033884 B1 NL 2033884B1
- Authority
- NL
- Netherlands
- Prior art keywords
- data
- prediction
- model
- photovoltaic power
- feature
- Prior art date
Links
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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
Abstract
The invention belongs to the technical field of photovoltaic power generation prediction and discloses a photovoltaic power prediction method, system and medium. The driving numerical forecast model is determined based on atmospheric reanalysis meteorological data and global weather prediction field. After preprocessing meteorological data, feature selection is carried out. The prediction method based on LSTM-CNN and the regression prediction method based on Lasso are combined with the field observation data to train the model and finally output the combined prediction results. The present invention takes the weather prediction result of ERA5 driven numerical weather prediction model as the input of model learning for model training. When the weather prediction results of ECMWF driven numerical weather prediction model are input into a trained model and verified, the results are more satisfactory than those of a single model. The invention combines the advantages of linear and nonlinear prediction models, improves the generalization ability of prediction models, and improves the accuracy and stability of short-term power prediction of photovoltaic power generation.
Description
State Grid Gansu Electric Power Company, and
Gansu Tongxing Intelligent Technology Development CO., LTD. 22/112 PDNL
A Photovoltaic Power Prediction Method, System and Medium
The invention belongs to the technical field of photovoltaic power generation prediction, in particular to a photovoltaic power prediction method, system and medium.
Background Technology
As solar energy has the characteristics of large reserves of resources, clean, sustainable and direct access, photovoltaic power generation is an effective way to realize the transformation of sustainable energy structure and reduce carbon emissions in the process of power generation. The actual solar radiation received by photovoltaic system has the characteristics of large fluctuation range and strong uncertainty, which leads to the obvious random fluctuation and intermittent characteristics of its output power. With the rapid development of photovoltaic power generation, its high permeability also brings many new challenges to the operation of the existing power grid system. Improving the accuracy of photovoltaic power prediction is an effective solution to overcome these challenges.
Short-term PV power prediction algorithms mostly take Numerical Weather Prediction (NWP) as the input of meteorological data, and build the mapping model between NWP and
PV output by multiple regression or traditional machine learning methods. However, because of the uncertainty of NWP, The prediction error has a direct influence on the accuracy of power prediction. The model established by this kind of method can not accurately describe the relationship between weather and power.
Through the above analysis, the existing problems and defects of the existing technology are as follows: (1) The accuracy of photovoltaic power prediction model based on traditional methods is affected by too many features. (2) The traditional photovoltaic power prediction method only considers the mapping relationship between numerical weather prediction data and photovoltaic output data, and the simple time analysis ignores the non-stationality of spatio-temporal data. (3) Classical machine learning methods tend to fall into local optimal and overfitting states in the process of model training, so it is difficult for traditional photovoltaic power prediction methods to improve the prediction accuracy of photovoltaic power.
Aiming at the problems existing in the prior art, the invention provides a photovoltaic power prediction method, system and medium, in particular a photovoltaic power prediction method, system and medium based on atmospheric reanalysis data.
The invention relates to a photovoltaic power prediction method. The photovoltaic power prediction method includes: determining the driving numerical prediction model according to atmospheric reanalysis meteorological data and global weather prediction field; After preprocessing meteorological data, feature selection is carried out. The prediction method based on LSTM-CNN and the regression prediction method based on Lasso are combined with the field observation data to train the model and finally output the combined prediction results.
Furthermore, the photovoltaic power prediction method includes the following steps:
Step 1: Download the atmospheric reanalysis meteorological data and global numerical weather forecast data of the target area, covering the historical operation time of the photovoltaic power station;
Step 2: Atmospheric reanalysis meteorological data and global numerical weather forecast data are used to drive the mesoscale numerical weather forecast model to conduct downscale forecast respectively.
Step 3: Obtain power generation data of photovoltaic power station and remove outliers as the target value of model training;
Step 4: Extract the elements of PV station meteorological prediction from the downscaling meteorological data of reanalysis as the training set of PV power prediction model;
Step 5: Extract the same meteorological prediction elements of photovoltaic power station from downscaling meteorological data forecast, and use them as the test set of photovoltaic power prediction model.
Step 6: Extract the features and add them to the feature sets of the training set and the test set respectively, The feature information of PV power prediction was screened from the feature set of training set by random forest.
Step 7: The feature information set of the selected training set is used as the input of the LSTM-CNN and Lasso models respectively to conduct model training and cross- validation, and obtain the well-trained LSTM-CNN and Lasso models.
Step 8: The feature information of the test set is matched according to the filtered training set set. The trained LSTM-CNN model and Lasso model are called on the test set to obtain two groups of initial prediction results, respectively. The Stacking fusion strategy is used to fuse the two prediction results and obtain the final prediction results.
Furthermore, in Step 2, the same parameterization scheme and spatial resolution of model time were adopted for the numerical model to obtain two sets of refined target range meteorological forecast data, namely reanalysis downscaling meteorological data and downscaling meteorological forecast data respectively.
Furthermore, the meteorological prediction factors of photovoltaic power station in Step 4 include total solar level radiation, direct solar radiation, solar scattered radiation, low, middle and high cloud cover, wind speed, wind direction, temperature, humidity and pressure.
Furthermore, feature extraction in Step 6 includes: (1) Convert the direct and scattered radiation of the ground to the total solar radiation of the inclined plane according to the latitude and longitude coordinates of the photovoltaic power station and the atmospheric radiation theory; (2) The wind direction of surface wind and upper wind is characterized by wind direction cosine and wind direction sine; (3) Add labels for month, week, day and hour; (4) According to the dip angle, longitude and latitude information of the photovoltaic square, add the sine of the solar altitude angle and the cosine of the solar incidence angle feature labels.
Furthermore, in Step 6, the set of meteorological feature information was input into the random forest model as a training set to calculate the degree of influence of different feature information on the prediction results, and the optimal feature subset was selected.
When there is nee in the random forest, according to the bootstrap resampling technology of random forest, 1/3 data of each sub-decision tree is not extracted in the process of training sample extraction, which is called out-of-pocket data. The out-of-pocket data of the j'th decision tree is predicted, and the error of prediction results is calculated.
Where, the error of the prediction result is the out-of-pocket data error OOBerror; of the jth decision tree:
OOBerror; = a Zia Oe — Yip)”:
Wherein, N; is the total number of data samples outside the bag, y;- is the actual value of data samples outside the bag, and y is the predicted value of data samples outside the bag.
The out-of-pocket data error of the entire random forest is:
OOBerror = Lyre OOBerror; ;
Ntree J=1 J
The importance measure of the feature information is calculated by using the data outside the bag, so as to provide reference for selecting the feature information. For out-of- pocket data error OOBerror; recording of the j'th decision tree, change one of the feature variables x;(i=1,2,3,...,M) in the feature set X = {}, M is the total number of features, other feature variables remain unchanged, the out-of-pocket data of the feature variable is randomly scrambled, and the error OOBerrorj of the changed out-of-pocket data is recalculated. The difference of the error of the out-of-pocket data before and after the changed feature variable is taken as the importance measurement of the feature x;:
MDA = 3.tree(00Berror; — OOBerror{);
Input factors filtered by random forest algorithm include: (1) Using bootstrap sampling method, K sample sets were randomly generated in the original training data set, and the features with the most classification ability were selected from the m features for node splitting. Multiple decision trees were generated to form a random forest, and the new data were classified by tree classifier voting method. (2) The importance MDA of each characteristic variable was calculated and sorted in descending order; (3) Determine the proportion to be deleted, and delete the unimportant indicators in the corresponding proportion from the current feature variables, so as to obtain a new feature set; (4) Establish a new random forest with a new feature set, calculate the importance of each feature in the feature set, and rank; (5) Repeat Step (2) to Step (4) until m features remain; (6) According to each obtained feature set and random forest, the corresponding out- of-pocket error OOBerror is calculated, and the feature set with the lowest out-of-pocket error is taken as the final selected feature set Q.
Another purpose of the invention is to provide a photovoltaic power prediction system using the photovoltaic power prediction method, including:
Driver numerical forecast module, which is used to drive the numerical forecast model based on atmospheric reanalysis of meteorological data and global weather forecast field respectively;
Data preprocessing module, which is used for meteorological data preprocessing and then feature selection;
Model training module, which is used for the prediction method based on LSTM-CNN and the regression prediction method based on Lasso, and combines the electric field observation data to train the model, and finally outputs the combined prediction results.
Another purpose of the invention is to provide a computer device, which comprises a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the photovoltaic power prediction method. 5 Another purpose of the invention is to provide a computer-readable storage medium containing a computer program which, when executed by a processor, enables the processor to perform the steps of the photovoltaic power prediction method.
Another purpose of the invention is to provide an information data processing terminal for realizing the photovoltaic power prediction system.
Combined with the above technical scheme and solved technical problems, the advantages and positive effects of the technical scheme to be protected by the invention are as follows:
Firstly, in view of the technical problems existing in the above-mentioned prior art and the difficulty of solving the problem, closely combined with the technical scheme to be protected by the invention and the results and data in the process of research and development, detailed and profound analysis of the technical problem solved by the technical scheme of the invention, and some creative technical effects after solving the problem. Specific description is as follows:
Reanalysis of meteorological data is a kind of meteorological data obtained after quality control and fusion of numerical forecast products and atmospheric observation data (including ground observation, satellite, radar, sounding, etc.) through data assimilation method. It has the characteristics of long time series and wide spatial distribution, and is more accurate than numerical forecast. Taking it as the background driving field can improve the simulation quality of meteorological model well. The noise and uncertainty of model data can be greatly reduced by downscaling the regional numerical weather prediction model again, and the prediction accuracy of the model can be effectively improved in mining the mapping relationship between meteorological prediction factors and power fluctuations.
The reanalysis meteorological data and meteorological forecast data based on the same model have consistency in model error. Therefore, the invention can take the meteorological prediction results of ERAS driven numerical weather forecast model as the input of model learning for model training, and input the meteorological prediction results of
ECMWF driven numerical weather forecast model into the trained model. The prediction verification can effectively improve the prediction accuracy of the model.
Secondly, taking the technical scheme as a whole or from the perspective of product, the technical effect and advantages of the technical scheme to be protected by the invention are described as follows:
The invention extracts the time characteristic information of historical data through
LSTM neural network, takes the output of LSTM neural network as the input of CNN neural network, extracts the spatial characteristic information of data, and can well deal with the timing problem of photovoltaic power itself and the nonlinear relationship between various influencing factors and photovoltaic power. It has a good effect on the power prediction of photovoltaic power generation system. Then, by combining the prediction results of Lasso linear regression model and LSTM-CNN nonlinear neural network, more satisfactory results can be obtained than that of a single model. The advantages of linear and nonlinear prediction models are integrated to further improve the generalization ability of the prediction model, and improve the accuracy and stability of short-term photovoltaic power prediction.
Thirdly, as creative auxiliary evidence of the claims of the invention, it is also embodied in the following important aspects:
The expected income and commercial value after the conversion of the technical scheme of the invention are:
For the first time, the invention applies the historical reanalysis meteorological data to the short-term forecast of photovoltaic power by using the downscaling technology of meteorological model, and reduces the meteorological error. At the same time, it gives full play to the unique advantages of deep learning neural network in adapting to the processing of meteorological data and time series, and obtains the characteristic quantity from the massive input information of multiple hidden layers. After training a large amount of data to complete the model training and learning, and finally use the training and learning model to complete the photovoltaic power prediction, which has great advantages compared with other methods.
In order to more clearly explain the technical scheme of the implementation method of the invention, the following is a brief introduction of the attached drawings required in the implementation method of the invention. It is obvious that the attached drawings described below are only some implementation methods of the invention. For ordinary technical personnel in the field, other attached drawings can be obtained according to these drawings without creative labor.
Figure 1 is a flow chart of the photovoltaic power prediction method provided by the implementation method of the invention;
Figure 2 is a schematic diagram of the photovoltaic power prediction method provided by the implementation method of the invention.
Specific Implementation Methods
In order to make the purpose, technical scheme and advantages of the invention more clear, the invention is further explained in detail in combination with implementation methods. It should be understood that the specific implementation methods described herein are intended only to explain the invention and are not intended to qualify it.
Aiming at the problems existing in the prior art, the invention provides a photovoltaic power prediction method, system and medium. The invention is described in detail in combination with the attached drawings below.
Atmospheric forecast data: The Medium-Range Weather Forecasts products of the
European Centre for Medium-Range Weather Forecasts (ECMWF).
As shown in Drawing 1, the photovoltaic power prediction method provided by the implementation method of the invention includes the following steps:
S101. The numerical prediction model is driven by atmospheric reanalysis meteorological data and global weather prediction field respectively;
S102. Preprocess meteorological data and then proceed the feature selection;
S103. The prediction method based on LSTM-CNN and the regression prediction method based on Lasso, combined with the field observation data to train the model, and finally output the combined prediction results.
As a preferred implementation method, the photovoltaic power prediction method provided by the implementation method of the invention, as shown in Drawing 2, specifically includes the following steps: 1. Download the atmospheric reanalysis meteorological data and global meteorological forecast data of the target area, covering the historical operation time of the photovoltaic power station; 2. The ERAS atmospheric reanalysis data and the ECMWF atmospheric prediction data are respectively used to drive the mesoscale numerical weather prediction model to conduct downscale prediction. The numerical model adopts the same parameterization scheme and spatial resolution of the model time, and two sets of target range refined weather forecast data can be obtained. Are ERA5 downscaling meteorological data and ECMWF downscaling meteorological data, respectively. 3. Obtain power generation data of photovoltaic power station and remove outliers as the target value of model training; 4. Extract the meteorological prediction elements of photovoltaic power station from the downscaling meteorological data of ERAS, such as total solar horizontal radiation, direct solar radiation, solar scattered radiation, low, middle and high cloud cover, wind speed, wind direction, temperature, humidity, pressure, etc., as the training set of photovoltaic power prediction model; 5. In the same way as the above steps, the same meteorological prediction elements at the photovoltaic power station were extracted from the downscaling meteorological data of
ECMWF, and used as the test set of the photovoltaic power prediction model; 6. Since photovoltaic power generation is related to time, local latitude and longitude, solar radiation amount, cloud cover, wind speed and wind direction on the square array, extraction and representation of key features are very important for discovering the periodicity rule between data and improving the convergence rate of model training.
Features are extracted in the following ways and added into the feature set of training set and test set respectively: (1) The direct and scattered radiation of the ground is converted into the total solar radiation of the inclined plane according to the atmospheric radiation theory; (2) The wind direction of surface wind and upper wind is characterized by wind direction cosine and wind direction sine; (3) Add labels for month, week, day and hour; (4) According to the dip angle, longitude and latitude information of the photovoltaic square, add the sine of the solar altitude angle and the cosine of the solar incidence angle feature labels; 7. Random forest was used to screen out the feature information of photovoltaic power prediction from the feature set of the above training set. First, the above meteorological feature information set was input into the random forest model as the training set, and the influence degree of different feature information on the prediction results was calculated, and the optimal feature subset was selected based on this.
When there is ntree in the random forest, according to the bootstrap resampling technology of random forest, 1/3 data of each sub-decision tree is not extracted in the process of training sample extraction, which is called out-of-pocket data. The out-of-pocket data of the j'th decision tree is predicted, and the error of prediction results is calculated.
Where, the error of the prediction result is the out-of-pocket data error OOBerror; of the j'th decision tree;
Wherein, N; is the total number of data samples outside the bag, y;- is the actual value of data samples outside the bag, and y; is the predicted value of data samples outside the bag.
The out-of-pocket data error of the entire random forest is:
OOBerror = Lyle OOBerror; ;
Niree J=1 J
The importance measure of the feature information is calculated by using the data outside the bag, so as to provide reference for selecting the feature information. For out-of- pocket data error OOBerror; recording of the j'th decision tree, change one of the feature variables x;(i=1,2,3,...,M) in the feature set X = {}, M is the total number of features, other feature variables remain unchanged, the out-of-pocket data of the feature variable is randomly scrambled, and the error OOBerrorj of the changed out-of-pocket data is recalculated. The difference of the error of the out-of-pocket data before and after the changed feature variable is taken as the importance measurement of the feature x;:
MDA = 3.tree(00Berror; — OOBerror{);
It can be seen that if the value of a certain feature is added to the random noise data, the larger the error of the out-of-pocket data is, the more important the feature is.
Steps of input factor screening by random forest algorithm: (1) Using bootstrap sampling method, K sample sets were randomly generated in the original training data set, and a feature with the most classification ability was selected among the m features for node splitting. Multiple decision trees were generated to form a random forest, and the new data were classified by tree classifier voting method. (2) Calculate the importance MDA of each characteristic variable and sort it in descending order. (3) Determine the proportion to be deleted, and delete the unimportant indicators in the corresponding proportion from the current feature variables, so as to obtain a new feature set. (4) Establish a new random forest with a new feature set, calculate the importance of each feature in the feature set, and sort it. (5) Repeat Step (2) to Step (4) until m features remain. (6) According to each obtained feature set and random forest, the corresponding out- of-pocket error OOBerror is calculated, and the feature set with the lowest out-of-pocket error is taken as the final selected feature set Q. 8. The feature information set of the selected training set was used as the input of the
LSTM-CNN and Lasso models respectively to conduct model training and cross-validation, and the trained LSTM-CNN model and Lasso model were obtained; 9. The feature information of the test set is matched according to the filtered training set set, and the trained LSTM-CNN model and Lasso model are called on the test set to obtain two groups of initial prediction results, respectively. The fusion strategy is used to fuse the two prediction results and obtain the final prediction results.
The photovoltaic power prediction system provided by the implementation method of the invention includes:
Driver numerical forecast module, which is used to determine the driving numerical forecast model based on ERAS atmospheric reanalysis meteorological data and global weather forecast field.
Data preprocessing module, which is used for meteorological data preprocessing and then feature selection;
Model training module, which is used for the prediction method based on LSTM-CNN and the regression prediction method based on Lasso, and combines the electric field observation data to train the model, and finally outputs the combined prediction results.
In order to prove the creativity and technical value of the technical scheme of the invention, this part is an application implementation method of the technical scheme of the claim on a specific product or related technology.
In order to verify the effect of the proposed method, a photovoltaic power station in
Shanxi Province and Gansu Province were selected for testing. The installed capacity of
Shanxi power station is 60MW and that of Gansu power station is 59MW. The grid- connected data of two centralized photovoltaic power plants were obtained, and the ERAS historical reanalysis data and ECMWF global weather forecast data were used. The downscale numerical weather prediction was carried out, the training set and the test set were divided, the deep learning neural network LSTM-CNN and Lasso models were used to build the fusion prediction results, and the method was compared with the traditional prediction methods. The method provided in this patent can improve the average accuracy of the two electric fields in September 2022 by about 0.9%.
The calculation formula of power prediction accuracy is as follows:
CR = £ - cie ie y ì x100%
Vea GJ
Where, Pui is the actual generation power at i'th moment, Ppiis the predicted generation power at i'th moment, Ci is the startup capacity at i'th moment, n is the number of actual generation power collection times every 15 minutes during the daytime, i=1,2... n.
The implementation method of the invention has obtained some positive effects in the process of research and development or use, and indeed has great advantages compared with the current technology.
Traditional NWP-power | cNN.LSTM | Lasso | Fusion forecasting method model of model of | result of (Based on random forest . . . . . . : invention invention | invention algorithm)
A power station in 88.39% 89.15% 89.08% 89.30%
Shanxi
A power station in 88.84% 89.63% 89.37% 89.73%
Gansu 88.62% 89.39% 89.23% 89.52% value
What should be awared is that implementation methods of the invention may be realized by hardware, software or a combination of software and hardware. The hardware part can be realized by using special logic; The software part can be stored in memory and executed by the appropriate instruction system. Such as a microprocessor or specially designed hardware to perform. The general technical person in the field can understand that the above devices and methods can be implemented using computer executable instructions and/or contained in processor control code. For example, such code is provided on a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device and its module of the invention may be realized by the hardware circuit of a programmable hardware device such as a very large scale integrated circuit or gate array, a semiconductor such as a logic chip or a transistor, or a field programmable gate array or a programmable logic device. It can also be implemented by software executed by various types of processors, or by a combination of the above hardware circuits and software, such as firmwaree.
The above mentioned is only the specific mode of implementation of the invention, but the scope of protection of the invention is not limited to this, and any modification, equivalent substitution and improvement made by any technical person familiar with the technical field within the scope of the technology disclosed by the invention and within the spirit and principles of the invention shall be covered by the scope of protection of the invention.
Claims (8)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
NL2033884A NL2033884B1 (en) | 2022-12-30 | 2022-12-30 | A Photovoltaic Power Prediction Method, System and Medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
NL2033884A NL2033884B1 (en) | 2022-12-30 | 2022-12-30 | A Photovoltaic Power Prediction Method, System and Medium |
Publications (2)
Publication Number | Publication Date |
---|---|
NL2033884A NL2033884A (en) | 2023-03-07 |
NL2033884B1 true NL2033884B1 (en) | 2023-09-06 |
Family
ID=85414970
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
NL2033884A NL2033884B1 (en) | 2022-12-30 | 2022-12-30 | A Photovoltaic Power Prediction Method, System and Medium |
Country Status (1)
Country | Link |
---|---|
NL (1) | NL2033884B1 (en) |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115049171B (en) * | 2022-08-17 | 2022-10-28 | 国能日新科技股份有限公司 | Photovoltaic power prediction method and system based on feature migration |
-
2022
- 2022-12-30 NL NL2033884A patent/NL2033884B1/en active
Also Published As
Publication number | Publication date |
---|---|
NL2033884A (en) | 2023-03-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kumari et al. | Deep learning models for solar irradiance forecasting: A comprehensive review | |
Salcedo-Sanz et al. | An efficient neuro-evolutionary hybrid modelling mechanism for the estimation of daily global solar radiation in the Sunshine State of Australia | |
Blaga et al. | A current perspective on the accuracy of incoming solar energy forecasting | |
Ceci et al. | Predictive modeling of PV energy production: How to set up the learning task for a better prediction? | |
Heydari et al. | A novel composite neural network based method for wind and solar power forecasting in microgrids | |
Leva et al. | Analysis and validation of 24 hours ahead neural network forecasting of photovoltaic output power | |
Li et al. | Machine learning for solar irradiance forecasting of photovoltaic system | |
Eslami et al. | A data ensemble approach for real-time air quality forecasting using extremely randomized trees and deep neural networks | |
Carneiro et al. | Review on photovoltaic power and solar resource forecasting: current status and trends | |
Wang et al. | Comparative study of machine learning approaches for predicting short-term photovoltaic power output based on weather type classification | |
del Campo-Ávila et al. | Binding data mining and expert knowledge for one-day-ahead prediction of hourly global solar radiation | |
Sobolewski et al. | Gradient boosting-based approach for short-and medium-term wind turbine output power prediction | |
CN112180471A (en) | Weather forecasting method, device, equipment and storage medium | |
Cholissodin et al. | Prediction of rainfall using improved deep learning with particle swarm optimization | |
Khan et al. | Optimal photovoltaic panel direction and tilt angle prediction using stacking ensemble learning | |
De Caro et al. | Robust assessment of short-term wind power forecasting models on multiple time horizons | |
Alcántara et al. | Deep neural networks for the quantile estimation of regional renewable energy production | |
Omar et al. | Seasonal clustering forecasting technique for intelligent hourly solar irradiance systems | |
del Campo-Ávila et al. | A data mining system for predicting solar global spectral irradiance. Performance assessment in the spectral response ranges of thin-film photovoltaic modules | |
NL2033884B1 (en) | A Photovoltaic Power Prediction Method, System and Medium | |
Brahma et al. | Visualizing solar irradiance data in ArcGIS and forecasting based on a novel deep neural network mechanism | |
Sharma et al. | Deep Learning Based Prediction Of Weather Using Hybrid_stacked Bi-Long Short Term Memory | |
Joseph et al. | Short-term wind speed forecasting using an optimized three-phase convolutional neural network fused with bidirectional long short-term memory network model | |
Mudunuru et al. | SWAT watershed model calibration using deep learning | |
Widodo et al. | Modeling solar potential in Semarang, Indonesia using artificial neural networks |