Prediction of salt water intrusion using BP-RAGA coupled neural network model

Tidal river is often intruded by salt water in dry season. The deterioration of water quality will lower the reliability of drinking water. Raw water systems, which include reservoirs, pump stations and pipelines, were constructed in costal cities. In order to guarantee the safety of drinking water, precise plans of raw water system must be drawn up. So it is necessary to predict salt water intrusion in water sources precisely. The period that water source can not provide raw water is selected as indicator of salt water intrusion. Artificial neural network is applied to develop a salt water intrusion prediction model. Real coding based accelerating genetic algorithm (RAGA) and Back propagation (BP) algorithm are applied to optimize the weights of neural network. Tidal range of water resource, the period that water source can not supply raw water and observed flows in upstream hydrological station in last day are selected as influencing factors. So a BP-RAGA prediction model is developed in this paper. And the model was applied to predict salt water intrusion in PingGang water source of Zhuhai city in this paper. The BP-RAGA coupled neural network model proved to be superior to BP neural network model in precision.