Terkos Gölüne Gelen Aylık Debinin Çeşitli Metotlarla Tahmini

dc.contributor.advisor Özger, Mehmet tr_TR
dc.contributor.author Türkoğlu, Halil İbrahim tr_TR
dc.contributor.authorID 10012954 tr_TR
dc.contributor.department Hidrolik ve Su Kaynakları Mühendisliği tr_TR
dc.contributor.department Hydraulics and Water Resources Engineerin en_US
dc.date 2013 tr_TR
dc.date.accessioned 2013-08-26 tr_TR
dc.date.accessioned 2015-07-15T13:41:33Z
dc.date.available 2015-07-15T13:41:33Z
dc.date.issued 2013-08-26 tr_TR
dc.description Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 2013 tr_TR
dc.description Thesis (M.Sc.) -- İstanbul Technical University, Institute of Science and Technology, 2013 en_US
dc.description.abstract Su yapılarının projelendirilmesinde, bazı bilgi ve verilere ihtiyaç duyulur. Bunlarında başında da su miktarının gelecekte belli bir tarihte ne olacağının tahminidir. Bu tahminlerin doğru yapılması, taşkın kontrolü amaçlı haznelerin işletilmesinde, akarsudaki su potansiyelinin belirlenmesinde, bir hidroelektrik santral için kurak dönemlerde üretimin nasıl etkileneceğinin bilinmesinde, içme ve sulama suyunun dağıtımında ve akarsulardaki ulaşımın planlanmasında büyük önemi vardır. Gelecekteki su miktarlarının belirlenmesi ile ilgili olarak çeşitli tahmin yöntemleri kullanabilinmektedir. Geleneksel olarak otoregresif (AR), hareketli ortalamalar (MA) ve otoregresif hareketli ortalama (ARMA) gibi bir dizi zaman serisi yaklaşımlarıyla tahmin yapılabildiği gibi yapay zekâ teknikleri ile de yapılabilmektedir. Bu çalışmada, İSKİ Genel Müdürlüğü’nden temin edilen Terkos Gölüne gelen aylık bazdaki veriler kullanılarak; otoregresif (AR), otoregresif hareketli ortalama (ARMA), yapay sinir ağları (YSA) ve Dalgacık Dönüşümü (WAVELET) yöntemleri ile aylık bazda sentetik diziler türetilmiş ve hangi yöntemin daha hassas sonuç verdiğini tespit edebilmek için birbirleriyle karşılaştırılmıştır. Çalışmanın Giriş bölümünde, tezin amacı açıklanmaktadır. Daha önce yapılan benzer çalışmalar araştırılıp çalışmalar hakkında bilgi verilmektedir. Çalışmanın ikinci bölümünde, çalışma alanının konumu, iklim yapısı, hidrolojik, jeolojik, jeomorfolojik, topografik, toprak yapısı ve bitki örtüsü hakkında bilgi verilmiş ve İstanbul için önemi açıklanmaktadır. Çalışmanın üçüncü bölümünde, 1995-2012 yılları arasında Terkos Barajına gelen aylık bazdaki debiler veri olarak kullanılmıştır. Terkos Barajına gelen aylık ortamla debilerin temel istatistiksel özellikleri verilmektedir. Çalışmada kullanılan yöntemler hakkında bilgi verilmektedir. En iyi model determinasyon katsayısı R2, ortalama karesel hata MSE ve kolerasyon katsayısı değerlerine göre belirlenmektedir. Çalışmanın dördüncü bölümünde, İSKİ Genel Müdürlüğü’nden temin edilen Terkos Gölüne gelen aylık bazdaki veriler kullanılarak; otoregresif (AR), otoregresif hareketli ortalama (ARMA), yapay sinir ağları (YSA) ve Dalgacık Dönüşümü-YSA (WAVELET-YSA) yöntemleri ile aylık bazda sentetik diziler türetilmiş ve hangi yöntemin daha hassas sonuç verdiğini tespit edebilmek için birbirleriyle karşılaştırılmıştır. Çalışmanın beşinci ve sonuç bölümünde, çalışılan mevcut metotlara göre alternatif olarak geliştirilen debi tahmin modellerinin, diğer modellere göre iyi performans göstermelerinden dolayı debi tahmini hesaplarında kullanılabilirliği görülmüştür. Geliştirilen modeller irdelendiğinde, Dalgacık-YSA modellerinin diğerlerine göre daha iyi performans gösterdikleri sonucu elde edilmiştir. Bu tür bir model debi tahmini gerektiren hidrolojik çalışmalarda, eksik verilerin tamamlanması ve ileriye yönelik debi tahmininde kullanılabilir.Bununla birlikte Terkos Barajı için geliştirilen modeller Türkiye’nin diğer barajlara gelen debi tahmininde de havza özellikleri dikkate alınarak kullanılabilir. tr_TR
dc.description.abstract Throughout human history, water has been one of the most important factors that played a major role not only during natural disasters (floods, drought, desertification) but also in the development of human thought. Today, it has the same importance. It is for sure that no artificial matter can be used instead of water in the future, which makes water more significant. For this reason, it is a must to pay more attention to the balance of use and delivery of water. The utilizable amount of water is running low due to the development of technology along with increasing human population . The use of water must be planned smartly according to the uses of water, which is one of the most important substances for a society’s economical, social and industrial development. It is necessary to apply scientific methods while using water both individullay and in common in order to derive benefits from it. Today, the need for water is increasing due to explosion in population and development of industries. People need water and energy for living and running factories and industries. As the demand for water increases, there emerges an urgent need for the use of water sources in an optimum way. In our country, water might be a big source of considirable income if water sources systems are operated well. It is necessary to make decisions and analyze water sources systematically in advance to benefit efficiently from the projects which are in the planning process. Thus, potential problems can easily be predicted and necessary precautions can be taken. Moreover, different alternative ways could be followed and the best solutions could be searced. Although we are in the beginning of 21st century, there isn’t accurate data about water volume, inflow and conditions in many countries. As a result of this, it is very hard to define and calculate total water reserve of the world. As 97 % of water is salty in the world, it is unusable for agriculture and daily consumption. The remaining 3 % is fresh water. Unfortunately, this amount is not delivered equally all around the world. For this reason, a lot of regional and personal difficulties occur. Eventually, there is a need for planning and building infrastructures, and getting the water to the desired regions. Hydraulic events are on a large scale and they have many uncertainties. As a result, it is impossible to carry out experiments in laboratories. For this reason, it is fundamental to record hydraulic variables frequently in nature as historical time series. In engineering water structures, incoming water data amount is needed. Predicting the amount of water in the future at a specific time plays an important role in operating catching areas which controls water floods, determining potential of rivers and understanding how the production will be affected during dry seasons for hydroelectric dams. Different prediction methods can be used to determine the amount of water in the future. The estimation might be fulfilled either conventionally through various time-series analysis methods such as Autoregressive (AR), Moving Averages (MA) and Autoregressive Moving Averages (ARMA) or through artificial intelligence techniques. During application of hydrology, there might be some problems concerning analysisof time series which has widespread usage. These problems can be defined as the abundance of parametres to determine, and loss of time during this process. In this study, in order to neglect the stated problems, considering the historical flow data records and using montly flow income of Terkos Lake, it is aimed to determine the method which could best predict the flow accurately at a certain time in the future. According to the related literature, autoregressive and artificial intelligence models were used both individually and together as hybrid models. For the prediction of flow, generally stochastic models were used, but recently hybrid models which can give more sensetive results can also be preffered. On the other hand, there aren’t enough researches about prediction with hybrid models. In this study, to predict the incoming flow of Terkos Lake, autoregressive (AR), Autoregressive Moving Averages (ARMA), Artificial Neural Networks (ANN) and WANN methods and montly senthetic series were invented, and they were compared with each other to determine the method that could give more sensetive results. This model consists of combination of WAVELET and Artificial Neural Network (ANN) models. While the wavelet system decomposes time series into its bands, Artificial Neural Network (ANN) method forms a relationship between input and output variables which are seperated into bands. In this study, it is tried to predict t- month ahead flow values using WAVELET-ANN models. Continuous wavelet is used to divide original time series into its bands. To deteremine the distribution of energy over the bands, avarage wavelet spectrum was used. For wavelet analize, Morlet wavelet was used. After the seperation of incoming flow series into subbands, each band was predicted according to the input variable’s bands. The prediction method was Artificial Neural Network (ANN). Finally every predicted band was gathered and time series prediction result was achieved. WAVELET-ANN model can be summarized in 3 steps. 1- Related time series are divided into subbands. In this study, incoming flow time series were divided into four bands. Each band has different characteristics.For example, the first band represents change in high frequency, while the fourth band represents the low frequency. Therefore, time series can be shown as several homogenious time series. 2- Considering each produced band as a seperate time series, it could be predicted with Artificial Neural Network (ANN) using the same input variables in all band models. Consequently, we have four different band models. 3- The results were achieved after the prediction of four different bands within theirselves and their total. In this study, based upon the monthly data concerning Terkos Dam, which is issued by the General Management of İSKİ, monthly synthetic series have been derived by means of Autoregressive (AR) and Autoregressive Moving Averages (ARMA), Artificial Neural Networks (ANN) and WAVELET, which have also been compared with each other with a view to determining which method will yield more precise results. The objective of the study has been explained in the Introductory Section thereof. Again, information has been given on the earlier similar studies as a result of a thorough examination thereof. In the Second Section of the study, information has been given on the position of the operation field, climate structure, hydrologic, geologic, geomorphologic, topographic, soil structure and the flora, whereby explaining the significance thereof for Istanbul. In the Third Section of the study, the monthly flows coming to the Terkos Dam between 1995 and 2012 have been used as data, whereby giving the basic statistical features of the monthly average flows coming to the Terkos Dam and giving information on the methods used in the study. The results were divided into two groups namely training and test data. One-third of the last data were tested and the remaing data was used for the training of model parametres. For the input variables, different combinations were used. Predictions were actualized 1, 3, 6 and 9 months later. The best model determination coefficient R2 is fixed according to the Mean Square Error (MSE) and correlation coefficient values. In the Fourth of the study, based upon the monthly data concerning Terkos Dam, which is issued by the General Management of İSKİ, monthly synthetic series have been derived by means of Autoregressive (AR) and Autoregressive Moving Averages (ARMA), Artificial Neural Networks (ANN) and WAVELET; and these have also been compared with each other with a view to determining which method will yield more precise results. In he fifth and concluding part of the study, the current flow according to the methods developed as an alternative forecasting models, the estimated flow rate compared to other models because they show a very good performance has been the availability of accounts. When the developed models are examined, it is seen that the performance of wavelet-neural network models perform better than others. This kind of can be used for flow predictions in hydrological studies, completion of missing data, and forecasting of flows. The proposed models can be employed for inflow predicitons of other dams in Turkey, considering the characteristics of the basin. en_US
dc.description.degree Yüksek Lisans tr_TR
dc.description.degree M.Sc. en_US
dc.identifier.uri http://hdl.handle.net/11527/7974
dc.publisher Fen Bilimleri Enstitüsü tr_TR
dc.publisher Institute of Science and Technology en_US
dc.rights İTÜ tezleri telif hakkı ile korunmaktadır. Bunlar, bu kaynak üzerinden herhangi bir amaçla görüntülenebilir, ancak yazılı izin alınmadan herhangi bir biçimde yeniden oluşturulması veya dağıtılması yasaklanmıştır. tr_TR
dc.rights İTÜ theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. en_US
dc.subject Dalgacık dönüşüm tekniği tr_TR
dc.subject debi tr_TR
dc.subject yapay sinir ağları tr_TR
dc.subject tahmin modelleri tr_TR
dc.subject Wavelet transforms technique en_US
dc.subject flow en_US
dc.subject Artificial neural networks en_US
dc.subject Estimation models en_US
dc.title Terkos Gölüne Gelen Aylık Debinin Çeşitli Metotlarla Tahmini tr_TR
dc.title.alternative Montly Inflow Prediction For Terkos Lake By Various Methods en_US
dc.type Thesis en_US
dc.type Tez tr_TR
Dosyalar
Orijinal seri
Şimdi gösteriliyor 1 - 1 / 1
thumbnail.default.alt
Ad:
13987.pdf
Boyut:
2.79 MB
Format:
Adobe Portable Document Format
Açıklama
Lisanslı seri
Şimdi gösteriliyor 1 - 1 / 1
thumbnail.default.placeholder
Ad:
license.txt
Boyut:
3.14 KB
Format:
Plain Text
Açıklama