DETERMINING RIVER FLOW USING ASPECTS OF SOFT COMPUTING & ARTIFICIAL NEURAL NETWORKS

Abstract

Successful forecasting of river flow is the main goal and an important procedure necessary for planning and managing water resources. Depending on the features of hydrological phenomena in the river basin, the flow of rivers varies unevenly. This paper determines the river flow using aspects of soft computing (SC). We have developed a technique to distinguish between the sets of forecast data into subsets before training with a series of neural networks. Genetic algorithm by optimizing fuzzy theoretical models based on rules by recombining these networks this method is demonstrated using historical time-series data from the River OUSE catchment area in northern England. The prediction should be evaluated based on global performance statistics and more specific flood-related indicators, and compared with statistical models and naive forecast benchmarks. The results showed that the development of the best methodology to provide affordable solutions that can be easily connected to an operational and reliable flood warning system.