Nonlinear Structure based Artificial Neural Computing for Upstream Flow Functional Models

Most of the real world systems are nonlinear and complex and it is challenging to model these types of systems for analyzing and forecasting the hidden behaviour of the systems. In the paradigm of vague complex systems, data-based time series modeling approaches of intelligent systems showed its applicability for coping with the problems of hidden noise and dynamicity which are encapsulated in the data. Getting from nature is one of the humans’ features and they are striving to produce the intellectual schemes by coping rare features of cognitions and intellect of the brain. In this paper, nonlinear autoregressive structure based modeling of the brain (i.e. Artificial Neural Network) is the aim of this study that suggest various Dynamic Neural Network (DNN) models by using time deferred autoregressive configurations, for the stream-flow of Sukkur barrage on lower Indus river basin. The suitability of the models for training, validation and testing stages, are evinced on assessment metrics which demonstrate the accuracy and sufficiency of the models which may be beneficial for water-resource management.