Using PEST for Drainage Optimization - page 02


We have presented a conversion of a drainage optimization problem to a form that can be solved using PEST software. As many groundwater modelers use PEST, a wide community will find this solution readily applicable. Since the transformation is linear, the converted problem is as smooth as the model itself. If the computed hydraulic head depends monotonically on the pumping rates as is the case in nature, the objective function does not have local minima, although it cannot be guaranteed that the objective function is convex, due to the non-linearity of the model.

The resulting minimization problem can be efficiently solved using PEST in only a few iterations. This is true as long as the model provides a solution which depends smoothly and monotonically on the input flow rates. For models which cannot converge to a high level of accuracy, the iterative process may converge slowly or not at all. In that case a heuristic method may be a better choice.

In this paper we only considered the optimization of pump flow rates. The same method can be applied to optimize pump locations. However, if the change of pump locations causes a discontinuous change of the mesh (for example if nodes are added or removed or if the connectivity is altered), then the output of the model will not depend continuously on the input parameters, and one can expect poor convergence.



The authors wish to thank the Ministry of Education, Science and Technological Development of the Republic of Serbia for financial support through the Technology Development Project TR37014.