Simulation and prediction of saltwater intrusion based on artificial neural network

This paper regard the eastern coast plain of Laizhou bay as study area. Considering the factors that affect the saltwater intrusion, the improved Neural Network Model has been applied to construct the random model that can analog the dynamic change of saltwater intrusion. The model parameters are corrected by the data from 1998 to 2003. On this base, according to the exploitation status of the groundwater and the using plan of the water in Laizhou city, two different exploited schemes are designed. After calculation, the dynamic trend of saltwater intrusion is predicted. This research provides the scientific support for the prevention of saltwater intrusion. Saltwater intrusion is a familiar geologic disaster in coast area. 100 years ago, Holland scholar Badon-Cyben and German Scholar Herzberg pointed out Gyben-Herzberg salt-fresh water interface formula. The study of saltwater went through static phase, seep dynamics phase and seep-dispersion dynamics phase. Now great achievement has been gained about the forming mechanism, prevention measure and the transport numerical model of salt-fresh water interface of saltwater intrusion. The city of Laizhou is one of the earliest areas in our country where found the saltwater intrusion. Then people began to takes the relevant measure to prevent the phenomena from curing in that area. Since the seventies of the 20th century, we have begun to carry on a series of research work. But from 1998, the new situation has appeared in this city, namely the developing state of rollback appears for the first time in some areas. After amounts of simulating and research, FuLin Li proposed: The occurrence of saltwater intrusion rollback was mainly owing to exploitation of a large amount of underground saltwater of sea floor caused by cultivating Dalingping fish in the littoral zone ([8]). As everyone knows, the developing state of the saltwater intrusion is influenced by many kinds of factors synthetically, and there are complicated non-linear relations between every factors. So obviously, we can't use the linear parameter method to estimate the relation between the exploitation of the underground salt water in the sea floor and development of the saltwater intrusion. Artificial neural network method uses the random nonlinear function of finite subsets by the complex function of neuron function, it is effective to estimate non-linear parameter and choose the structure that routine mathematics statistical method is difficulty to solve. For the purpose of finding inherent relation between the exploitation of the underground salt water in the sea floor and the development of saltwater intrusion, this paper uses the model of improved BP artificial neural network to imitate the phenomenon of saltwater intrusion happening in the east bank of Laizhou bay and carry on macro-forecast to its dynamic change.