COMPARISON OF GENETIC ALGORITHMS WITH OTHER METHODS FOR GROUNDWATER MONITORING NETWORK PLANNING BASED ON GEOSTATISTICS

摘要

Groundwater monitoring network planning largely concerns itself with constructing a flexible and efficient monitoring network. While attempting to solve a related problem, this study presents a novel procedure that combines genetic algorithms with geostatistical theory. The proposed method is compared to other methods that also integrate geostatistical theory with other optimization schemes, including the sequential design method (SDM), branch and bound method (BBM) and non-linear programming method (NPM). Those methods are implemented and applied to a simplified field case. The finding indicates that the BBM method is computationally inf easible for practical applications even if it can obtain the global optimal solution in principle. Genetic algorithms (GAs) are characterized by their ability to obtain a set of near optimal solutions instead of a single solution. In general, the total variations among the network defined by these methods do not significantly differ. For a network design problem, the SDM method provides a computationally efficient solution for a preliminary study. On the other hand, the multiple choices given by the GAs provide decision makers with flexibility to consider factors that geostatistics can not.