Parameter structure identification using tabu search and simulated annealing

Abstract

In groundwater modeling the identification of an optimal flow or transport parameter that varies spatially should include both the values and structure of the parameter. However, most existing techniques for parameter identification only consider the parameter values. In this study, the problem of identifying optimal parameter structure is treated as a large combinatorial optimization problem. Two recently developed heuristic search techniques, simulated annealing and tabu search, are used to solve the large combinatorial optimization problem. The effectiveness and flexibility of these two techniques are evaluated and compared with simple grid search and descent search, using preliminary results from one-dimensional examples. Among the techniques examined in this paper, tabu search performs extremely well in terms of the total number of function evaluations required.