The Use of a Genetic Algorithm in the Calibration of Estuary Models

This paper describes an artificial intelligence (AI) system for estuarine model design. It is created by the combination of case-based reasoning and genetic algorithm techniques. This application aims to make the utilisation of complicated and expensive hydrodynamic models flexible, cost-effective and accessible to non-specialists. By organising the available knowledge of estuarine modelling into an interactive and dynamic framework, the AI system provides the user with the necessary guidance and information for numerically solving hydroenvironmental problems related to estuaries. As soon as a new problem is given to the system, the case-based module for estuarine modelling (CBEM) is activated. This module accesses information about estuarine models and estuaries to which numerical solutions have been previously applied. After comparison and evaluation, the case-based search engine returns from its memory the most effective modelling scheme available for the solution of the new problem. The system then calls the genetic algorithm (GA) module which optimises the physical parameters of the selected modelling procedure to suit the new application. The main focus of this paper is on the description of the GA module. This module is developed by combining the classical evolutionary approach with problemspecific information to carry out the required parameter optimisation. The effectiveness of this procedure is illustrated using a one–dimensional hydrodynamic model for the Upper Milford Haven estuary in UK. The comparison between manual and genetic algorithm based calibrations for this specific case suggests that the GA routine can very effectively calibrate estuarine models under realistic situations. This means a significant reduction in the time normally necessary for the implementation of a numerical modelling scheme.