Li Tiantian, Ha Da, Wang Shilei
In practice engineering, it is commonly using pumping tests to determine hydraulic parameters, but the previous of estimated parameters cannot be proved due to the lack of unified standard and hard to choose a suitable algorithm. Meanwhile, due the limitation of independent algorithm, in some difficult hydraulic models, the engineers have to adjust parameters continually to meet the convergence criterion which cost lots of time and the accuracy of results cannot be proved. In this study, a series of pumping tests has been conducted in Dongli district of Tianjin city, China. With considering storage effect in the pumping wells and observation wells which has important impact in the early experiment, the effective ration has been proposed to eliminate the effect of well storage. Based on the local hydrogeological environment and measured drawdown, an unsteady leaky seepage model, called Hantush-Jacob model, is used to estimate hydraulic parameters. By combination of measured drawdown and analytical solution,three methods, extend Kalman filter algorithm, LM algorithm and Genetic algorithm, are all adopted to estimate hydraulic parameters in this study. According to the result, three algorithm all shows its limitation in the calculation. Thus, a combination of Genetic algorithm and LM algorithm has been proposed in this study. Based on the result, the combination algorithm shows its superiority in searching the optimal solution accurately and avoiding the defect of the three separate algorithm. According to the comparison, the measured drawdown fit well with the calculated drawdown based on the estimated hydraulic parameter. Furthermore, compared with the single algorithms used in previously published literature, the combined GA-LM algorithm has demonstrated its unique advantages in the inversion of hydrological parameters.
