Tank Water Management System for Sri Lanka Cultivation

Abstract

There are considerable numbers of classical reservoir operational models exists in irrigation m management, But they are inadequate and use advanced optimization techniques with stochastic hydrologic variables. Applicability of Artificial Neural Network is utilized predicting level of water for issuing from a water tank is considered. Neural Network has been applied to a range of problems covering a variety of sectors. Historical data of water tank in Sri Lanka are used to train neural networks and the system produces predictions and recommendations in order to manage the irrigation in advance. All tank data cannot be used to model a neural network because some datasets even don’t have a target field data (amount of water issued for irrigation). However, there are tanks which have sufficient information for simulating the prediction system. Users want decide on target water levels for neural network heuristically. Accurate results are obtained with a neural network model with more hidden layers and a feasible selection of levels of water discharge.