Pond water level simulation by applying the Hybrid Genetic Evolutionary Artificial Neural Network method

The appropriate design of the pond coast structures and the reservoir supplement management entails an accurate pond water surface simulation. A hybrid Genetic Algorithm - Artificial Neural Network (GA-ANN) is presented and applied in this research to estimate the next 3- and 5-day water surfaces. Training and validation of the GAANN is performed using the 4-year daily water surface measurements performed on Chahnimeh reservoir located on the eastern side of Iran. Various input combinations are applied to the GA-ANN method. According to the results, for both the next 3- and 5-day estimation models, the input combination, consisting of the past 2-day water surface data, contributes to the optimal yield. Root Mean Squared Error of the optimum next 3- and 5-day prediction GA-ANN models were obtained to be 0.1798 and 0.3102, respectively. This paper found that in modelling the 3- and 5-day ahead pond water level, the best input variables are the information on the one and two previous days.