Application of Neural Networks and Optimization Model in Conjunctive Use of Surface Water and Groundwater

Abstract

This study develops an optimization model for the large-scale conjunctive use of surface water and groundwater resources. The aim is to maximize public and irrigation water supplies subject to groundwater-level drawdown constraints. Linear programming is used to create the optimization model, which is formulated as a linear constrained objective function. An artificial neural network is trained by a flow modeling program at specific observation wells, and the network is then incorporated into the optimization model. The proposed methodology is applied to the Chou-Shui alluvial fan system, located in central Taiwan. People living in this region rely on large quantities of pumped water for their public and irrigation demands. This considerable dependency on groundwater has resulted in severe land subsidence in many coastal regions of the alluvial fan. Consequently, an efficient means of implementing large-scale conjunctive use of surface water and groundwater is needed to prevent further overuse of groundwater. Two different optimization scenarios are considered. The results given by the proposed model show that water-usage can be balanced with a stable groundwater level. Our findings may assist officials and researchers in establishing plans to alleviate land subsidence problems.