Study On The Parameters Identification Of Groundwater Systems Based On The Artificial Intelligent Technique

Efficient management through modeling of groundwater systems depends primarily on adequate knowledge of its parameters. Parameter Identification of groundwater system is difficult in water resources management fields. Unreasonable parameters of groundwater system directly affect dependability of water resources management calculation. Intelligent methods are introduced such as neural network, genetic algorithm. The question is discussed deeply and systematically through synthetic and field aquifer problems.Definition of "Groundwater system" is presented based upon analysis of existent correlative research from systemic opinion. Constitution and characteristic are analyzed in details. Mathematical model and solution of groundwater system are discussed.Intelligent methods such as neural network, genetic algorithm, tabu search, simulated annealing developed in recent years are summarized, and integration of the algorithms are primary probed into. Several methods of groundwater system parameter identification are designed. Groundwater system parameter identification of BP artificial neural networks is studied. RBF artificial neural networks are presented, and there is better effect because the algorithm has local approach ability.Nonlinear optimal model of groundwater system parameter identification is constructed including genetic algorithm, simulated annealing and genetic algorithm with simulated annealing. Procedure of algorithm is designed. A synthetic aquifer problem shows that the results are reasonable, and RBF artificial neural networks and genetic algorithm with simulated annealing has better results. Genetic algorithm with simulated annealing has better convergence than genetic algorithm.A groundwater model has been developed in Mihuashun area, Beijing, and genetic algorithm is applied into paremeters'identificaiton of the groundwater model. Application of the algorithm for different observed head data sets indicate that the model can be successfully applied for aquifer systems where data available may be sparse and with errors. Calculated groundwater heads by identification results in fourteen parameter areas are fit for observed heads in field, and flowing filed is similar. The study demonstrates the effectiveness of the GA global optimization model for parameter identification, which is an important step towards real system simulation and effective planning and management of groundwater resources.