Forecasting of solar radiation using different machine learning approaches

In this study, monthly solar radiation (SR) estimation was performed using five different machine learning-based approaches. The models used are support vector machine regression (SVMR), long short-term memory (LSTM), Gaussian process regression (GPR), extreme learning machines (ELM) and K-nearest neighbors (KNN). Modeling of these approaches was carried out in two stages. In the first stage, VIF analysis was carried out to develop the model. Thus, the input parameters that decrease the performance of the model are removed. In the second stage, remaining input parameters such as meteorological data, station location data and spatial and temporal information were used in the forecasting modeling according to the correlation SR. In this study, the data set is divided into two parts as test and training. 30% was used in the testing phase, and 70% of the data was used in the training phase. When comparing models, the following error statistics were used: Nash–Sutcliffe efficiency coefficient (NSE), mean absolute error (MAE), mean absolute relative error (MARE), root-mean-square error (RMSE) and coefficient of determination (R2 ). In addition, Taylor diagrams, violin plots, box error, spider plot and Kruskal–Wallis (KW) and ANOVA test were utilized to determine robustness of model’s forecast. As a result of the study, the KW test and ANOVA test results showed that the data of many models were from the same population with observations, and it has proved that LSTM and GPR algorithms are applicable, valid and an alternative for SR forecasting in Turkey, which has arid and semi-arid climatic regions.

Görüntülenme
2
22.03.2024 tarihinden bu yana
İndirme
1
22.03.2024 tarihinden bu yana
Son Erişim Tarihi
10 Nisan 2024 01:13
Google Kontrol
Tıklayınız
Tam Metin
Tam Metin İndirmek için tıklayın Ön izleme
Detaylı Görünüm
Eser Adı
(dc.title)
Forecasting of solar radiation using different machine learning approaches
Yayın Türü
(dc.type)
Makale
Yazar/lar
(dc.contributor.author)
DEMİR, Vahdettin
Yazar/lar
(dc.contributor.author)
ÇITAKOĞLU, Hatice
Atıf Dizini
(dc.source.database)
Wos
Atıf Dizini
(dc.source.database)
Scopus
Konu Başlıkları
(dc.subject)
Long Short-Term Memory
Konu Başlıkları
(dc.subject)
Gaussian Process Tegression
Konu Başlıkları
(dc.subject)
Support Vector Machine Regression
Konu Başlıkları
(dc.subject)
Extreme Learning Machines and K-Nearest Neighbors
Konu Başlıkları
(dc.subject)
Turkey
Yayıncı
(dc.publisher)
Neural Computing and Applications
Yayın Tarihi
(dc.date.issued)
2022
Kayıt Giriş Tarihi
(dc.date.accessioned)
2023-03-09T08:36:45Z
Açık Erişim tarihi
(dc.date.available)
2023-03-09T08:36:45Z
Özet
(dc.description.abstract)
In this study, monthly solar radiation (SR) estimation was performed using five different machine learning-based approaches. The models used are support vector machine regression (SVMR), long short-term memory (LSTM), Gaussian process regression (GPR), extreme learning machines (ELM) and K-nearest neighbors (KNN). Modeling of these approaches was carried out in two stages. In the first stage, VIF analysis was carried out to develop the model. Thus, the input parameters that decrease the performance of the model are removed. In the second stage, remaining input parameters such as meteorological data, station location data and spatial and temporal information were used in the forecasting modeling according to the correlation SR. In this study, the data set is divided into two parts as test and training. 30% was used in the testing phase, and 70% of the data was used in the training phase. When comparing models, the following error statistics were used: Nash–Sutcliffe efficiency coefficient (NSE), mean absolute error (MAE), mean absolute relative error (MARE), root-mean-square error (RMSE) and coefficient of determination (R2 ). In addition, Taylor diagrams, violin plots, box error, spider plot and Kruskal–Wallis (KW) and ANOVA test were utilized to determine robustness of model’s forecast. As a result of the study, the KW test and ANOVA test results showed that the data of many models were from the same population with observations, and it has proved that LSTM and GPR algorithms are applicable, valid and an alternative for SR forecasting in Turkey, which has arid and semi-arid climatic regions.
Yayın Dili
(dc.language.iso)
en
Haklar
(dc.rights)
Attribution-NonCommercial 3.0 United States
Tek Biçim Adres
(dc.identifier.uri)
http://hdl.handle.net/20.500.12498/5946
Analizler
Yayın Görüntülenme
Yayın Görüntülenme
Erişilen ülkeler
Erişilen şehirler
6698 sayılı Kişisel Verilerin Korunması Kanunu kapsamında yükümlülüklerimiz ve cerez politikamız hakkında bilgi sahibi olmak için alttaki bağlantıyı kullanabilirsiniz.
Tamam

creativecommons
Bu site altında yer alan tüm kaynaklar Creative Commons Alıntı-GayriTicari-Türetilemez 4.0 Uluslararası Lisansı ile lisanslanmıştır.
Platforms