Real Time Forecasting Using Neural Networks

摘要: Forecasting hydrological variables, like river flows, water levels and rainfall is necessary in planning, design, maintenance operation of resources systems. Depending on the lead time, forecast can be real (i.e., near-real time or online), short term long term. Real forecasting has applications operational flood as well drought management. It forewarn extreme conditions, help optimum reservoirs power plants. Over last few decades many approaches have been presented to make forecasts real-time. They are deterministic stochastic nature involve conceptual statistical understanding. Of late, techniques based modelling data, rather than those underlying physical process, seem become popular following advent computational methods, Kitanidis Bras, (1980 a b), Georgakakos (1986 Ambrus Forward (1990), Garrote Bras (1995a, Nalbantis (1995). Most these models, however, distributed type where made at several locations within catchment. also require considerable exogenous information. Distributed often cost accuracy specific locations. In situations information needed only sites basin adequate meteorological topographic not available, site-specific simple neural networks (NNs) attractive alternatives apply.