A STUDY OF OPTIMIZATION FOR HYDROGEOLOGICPARAMETERS WITH HYBRID SIMPLEXPARALLELGENETIC ALGORITHM

摘要

It is a complicated nonlinear optimization problem to determine the hydrogeologic parameters with numerical methods.rnThe complexity is caused by the choice of optimization methods, the massive computing time and the abnormal parameter peak. Therntraditional methods need not only vast time but also manual intervention with experiences. Therefore, many researchers study severalrnoptimization methods to determining the parameters, including simplex methods, genetic algorithms and so on.rnRandom Simplex method is a kind of unconstrained optimization of direct research methods. There is no need to calculate thernobject function's derivative or grads to find the optimal direction and the application is very simple, the rapidity of convergence isrnvery fast and the method has a wide range of use. However, the simulating capability of the random simplex method depends on thernchoices of the starting values, and the determinism searching mechanism is chosen. Then it would be difficult to disengage when thernarithmetic had been trapped into local minima.rnGenetic algorithms are a kind of global adaptive optimization probability research algorithms. They are not restricted by thernmaterial form of the model, and they are also not tied by the condition and the amount of the model parameter. This methods,rnhowever, have some bugs which performs as the premature phenomenon of the genetic algorithm, low local optimization ability andrnslow convergence speed.rnTo prevent premature convergence effectively and improve the estimation precision and computation efficiency, a new parallelrngenetic algorithm combining the random simplex method with the parallel genetic algorithm is proposed. As shown by the calculationrnresults, the new algorithm is applied to the inverse simulation of a water source area and satisfactory results are obtained comparedrnwith results of other methods.