Baraj haznelerinin tasarymy ve i?letilmesi a?amalarynda, güvenilir akym tahminlerinin ve akym modelleme çaly?malarynyn yapylmasy büyük önem ta?ymaktadyr. Sunulan çaly?mada, Levenberg-Marquardt optimizasyon algoritmasy tabanly bir yapay sinir a?y (LM-YSA) modeli geli?tirilerek, Keban Barajy'ny besleyen, Murat Nehri, Munzur Çayy ve Peri Suyu günlük akymlaryna uygulanmy?tyr. Hazyrlanan LM-YSA modelleri içsel ba?ymly stokastik AR(p) model yapylaryyla da kar?yla?tyrylmy?tyr. Uzun dönemde gözlenmi? ve modellenmi? akymlaryn istatistikleri kar?yla?tyryldy?ynda; kurulan tüm modellerin yakla?yk sonuçlar verdi?i ancak LM-YSA modelinin Keban akymlaryny istatistiksel açydan daha iyi temsil etti?i görülmü?tür.

It is very important to make reliable runoff predictions and runoff modeling studies when planning and operating of dam reservoirs. In the study presented, the Levenberg-Marquardt optimization algorithm based artificial neural network (LM-ANN) model was improved and applied to the daily runoff observations of Murat, Munzur and Peri Rivers which feed the Keban Dam. The LM-ANN models were also compared with autoregressive stochastic AR(p) model structures. When the statistics of the long term recorded and modeled runoff values are compared, it can be seen that all model results are approximately similar but the LM-ANN that has been developed, is more successfully represents the daily runoff values of the Keban Dam.