estimation of dam reservoir volume fluctuations using artificial neural network and support vector regression

Estimation of reservoir volume fluctuation is important for the operation of dam reservoir, design of hydraulic structures; determine pollution in reservoir and the safety of dams. Artificial Neural Networks (ANN) and support vector regression (SVR) approach provides a common basis for quantitative modeling in this respect. In this study, reservoir volume was estimated using average monthly precipitation, monthly total volume of evaporation, dam spillway discharge volume, released irrigation water amount and periodicity. The data were collected on a monthly basis over the 29 years for Tahtaköprü Dam in the southeast Mediterranean region of Turkey. For this purpose, three well known methods, artificial neural networks, support vector and multiple linear regressions were employed for estimating the reservoir volume. In this paper, a multi layer perception (MLP) methodology is used as the ANN approach. Levenberg-Marquardt training algorithm is used for optimization of the network. MLP and SVR results are compared to multi-linear regression (MLP) model results. The results show that reservoir volume was successfully estimated using ANN and SVR with low mean square error and high correlation coefficients.