Publication: Küresel İklim Değişikliğinin Orta Karadeniz Havzasında Hidroelektrik Santraller ve Enerji Üretimi Üzerine Olan Etkisi
Abstract
İklim değişikliği tüm dünyada kritik zorluklardan biri haline gelmektedir. Yenilenebilir enerji esas olarak yağış, sıcaklık ve yağış-akış oranları gibi yerel çevresel koşullara bağlıdır. Hidroelektrik enerji, temiz enerji sağlamak için birincil yenilenebilir kaynaktır ve gelecekteki katkısının önemli ölçüde artması beklenmektedir. Gerekli yağış, debi, göl seviyesi, debi ve enerji üretim verileri Meteoroloji Genel Müdürlüğü (MGM), Devlet Su İşleri (DSİ), Türkiye Elektrik Üretim İletim A.Ş. (TEİAŞ) kurumlarından elde edilmiştir. Hidroelektrik üretiminin iki önemli senaryosu kullanılmış ve çok kullanılan iki iklim değişikliği senaryosu; RCP 8.5 ve RCP 4.5 temel alınmıştır. Küresel Sirkülasyon Modelleri (GCM'ler) yağış ve ortalama sıcaklık verileri, hidroelektrik santrallerin (HES'ler) enerji üretimini tahmin etmek için kullanılmıştır. Tez, çalışma alanı olarak Yeşilırmak ve Kızılırmak Havzalarına odaklanmıştır. Bölgede yer alan altı adet Hidroelektrik Santral üzerinde çalışılmıştır. Tahmin değerleri, Makine Öğrenimi (ML) teknikleri arasındaki bağıl hata ve korelasyon değerlerine dayalı olarak hesaplanmıştır. 1971'den 2018'e kadar aylık hidroelektrik enerji üretim verileri kullanılarak enerji üretimini tahmin etmek için beş model (Derin Öğrenme, Karar Ağacı, Genelleştirilmiş Doğrusal, Rastgele Orman ve Gradyan destekli ağaçlar (GBT)) kullanılmıştır. Tahmin ve değerlendirme çalışmaları için Makine Öğrenme teknikleri ve GCM kullanılmaya başlanmıştır. Çalışmada, GCM'lerin sıcaklık ve yağış değerlerine göre, 2018'den 2080'e kadar HES'lerin üreteceği enerji miktarının tahmininde her model için ML Teknikleri ile elde edilen sonuçlar sunulmuştur. Korelasyon ve bağıl hata değerleri, GBT modelinin altı ana HES için daha doğru sonuçlar verdiğini doğrulamıştır. Göreceli Hata için GBT yüzdeleri Almus, Hasan Uğurlu, Suat Uğurlu, Hirfanlı, Kesikköprü ve Kapulukaya için sırasıyla %31, %29, %15, %22, %28,6 ve %23 olarak bulunmuştur. GBT'nin korelasyonu ise Almus, Hasan Uğurlu, Suat Uğurlu, Hirfanlı, Kesikköprü ve Kapulukaya için sırasıyla 0.717, 0.602, 0.729, 0.76, 0.623 ve 0.801 olarak bulunmuştur. Bu sonuçlara göre, elektrik üretimini tahmin etmek için GBT modeli seçilmiştir. Sonuçlar, modeller arasında küçük farklılıklar olduğunu göstermekte, bu da tüm bu modellerde tahminlerin benzer olduğu ifade etmektedir. Anahtar Sözcükler: İklim değişikliği, Hidroelektrik enerji, Küresel Dolaşım Modeli, Makine Öğrenimi, Enerji Üretimi, Temsili Konsantrasyon Rotaları, Türkiye
Climate change becomes one of the critical challenges all over the world. Renewable energy mainly depends on many criteria, like precipitation, temperature, and rainfall-runoff ratios. Hydropower is the primary renewable source for supplying clean energy, and its future contribution is anticipated to increase significantly. Necessary rainfall, flow, lake level, flow and energy production data have been provided from relevant institutions such as Turkish State of Meteorological Services (TSMS), Turkish State of State Hydraulic Works (TSHW), Turkish Electricity Generation Company (TEGC) and Turkish Electricity Transmission Corporation (TETC). Two scenarios of hydropower generation were used and developed based on two climate change scenarios of climate change (namely RCP 8.5 and RCP 4.5). The Global Circulation Models (GCMs) data of precipitation and average temperature are used for predicting the energy production of the hydroelectrical power plants (HEPPs). The thesis focused on Yesilirmak and Kizilirmak Basins as study area. It included six main HEPPs. The prediction step was calculated based on relative error and correlation values between the Machine Learning (ML) techniques. Five techniques were used to predict the energy production (Deep Learning, Decision Tree, Generalized Linear, Random Forest and Gradient boosted trees (GBT)) using monthly hydroelectric power generation data from 1971 to 2018. Using Machine Learning techniques and GCM have started for predicting and evaluation studies. According to the temperature and precipitation values of the GCMs, the study presented the results of deploying ML Techniques in predicting the energy production which will be produced by HEPPs from 2018 to 2080. The performance criteria showed the differences between the five models and the quality of results in each model. The correlation and relative error values verified that GBT model gives more accurate results to the six main HEPPs. For Relative Error, the percantages of GBT were 31%, 29%, 15%, 22%, 28.6% and 23% for Almus, Hasan Ugurlu, Suat Ugurlu, Hirfanli, Kesikkopru and Kapulukaya, respectively. On the other hand, the results of corelation of GBT were 0.717, 0.602, 0.729, 0.76, 0.623 and 0.801 for Almus, Hasan Ugurlu, Suat Ugurlu, Hirfanli, Kesikkopru and Kapulukaya, respectively. According to that, the GBT model was used for predicting the production of electricity. The results show that there are small differences between the models which means that the predictions are going in similar directions at all these models. Keywords: Climate change, Hydropower, Global Circulation Model, Machine Learning, Energy Production, Representative Concentration Pathways, Turkey
Climate change becomes one of the critical challenges all over the world. Renewable energy mainly depends on many criteria, like precipitation, temperature, and rainfall-runoff ratios. Hydropower is the primary renewable source for supplying clean energy, and its future contribution is anticipated to increase significantly. Necessary rainfall, flow, lake level, flow and energy production data have been provided from relevant institutions such as Turkish State of Meteorological Services (TSMS), Turkish State of State Hydraulic Works (TSHW), Turkish Electricity Generation Company (TEGC) and Turkish Electricity Transmission Corporation (TETC). Two scenarios of hydropower generation were used and developed based on two climate change scenarios of climate change (namely RCP 8.5 and RCP 4.5). The Global Circulation Models (GCMs) data of precipitation and average temperature are used for predicting the energy production of the hydroelectrical power plants (HEPPs). The thesis focused on Yesilirmak and Kizilirmak Basins as study area. It included six main HEPPs. The prediction step was calculated based on relative error and correlation values between the Machine Learning (ML) techniques. Five techniques were used to predict the energy production (Deep Learning, Decision Tree, Generalized Linear, Random Forest and Gradient boosted trees (GBT)) using monthly hydroelectric power generation data from 1971 to 2018. Using Machine Learning techniques and GCM have started for predicting and evaluation studies. According to the temperature and precipitation values of the GCMs, the study presented the results of deploying ML Techniques in predicting the energy production which will be produced by HEPPs from 2018 to 2080. The performance criteria showed the differences between the five models and the quality of results in each model. The correlation and relative error values verified that GBT model gives more accurate results to the six main HEPPs. For Relative Error, the percantages of GBT were 31%, 29%, 15%, 22%, 28.6% and 23% for Almus, Hasan Ugurlu, Suat Ugurlu, Hirfanli, Kesikkopru and Kapulukaya, respectively. On the other hand, the results of corelation of GBT were 0.717, 0.602, 0.729, 0.76, 0.623 and 0.801 for Almus, Hasan Ugurlu, Suat Ugurlu, Hirfanli, Kesikkopru and Kapulukaya, respectively. According to that, the GBT model was used for predicting the production of electricity. The results show that there are small differences between the models which means that the predictions are going in similar directions at all these models. Keywords: Climate change, Hydropower, Global Circulation Model, Machine Learning, Energy Production, Representative Concentration Pathways, Turkey
Description
Keywords
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
158
