Groundwater Level Simulation with Combined Grey Neural Networks and Modflow Models

摘要

The prediction of groundwater level is one of the main work of hydraulic government, which is predicted based on the history data and the relative influence factors. By modflow model, the groundwater level is simulated and the results are given. Because of Boundary types and geologic conditions, which possess random and obscure characteristics, groundwater heads vary with the conditions. Therefore, prediction precision depends on the accuracy of history data. Data mining has provided a new method for analyzing massive, complex and noisy data. According to the complexity and ambiguity of groundwater system, a new integration of grey with neural network model is built to forecast groundwater heads, which were used to judge whether future groundwater heads were extraordinarily over the history range or not. This method overcomes the disadvantages which the grey method only predict the linear trend. And the reults predicted are compared with the results of Modflow model. The methods were used to analyze the random characteristics of groundwater heads in anyang city and comparison with Modflow. The results indicate that the method is reliable, and reasonable.