Application of Adaptive Neuro-fuzzy Inference System for Dam Inflow Prediction using Longrange Weather Forecast

Dam inflow forecast plays an important role for optimal reservoir operations. There are several techniques in use for dam inflow forecast; however, accurate long-range dam inflow forecast is still a challenging task. In this study, we developed a model based on Adaptive Neuro-Fuzzy Inference System (ANFIS) for monthly dam inflow forecast. The subtractive clustering method is used to find optimum set of fuzzy rules. To obtain appropriate ANFIS structure the model is tuned with different values of cluster radius for subtractive clustering. The model is trained using dam inflow and weather data (i.e. temperature and rainfall) of preceding month and monthly normal rainfall of forecasting month as input for dam inflow forecast. To assess the significance of rainfall forecast for improvement of dam inflow prediction we attempted to incorporate Korea Meteorological Administration (KMA) monthly rainfall forecast as an input with other parameters. The use of monthly rainfall forecast showed significant improvement in the dam inflow forecast. The viability of the proposed model is demonstrated for 3 major dams of South Korea.