PNNL

Abstract

Review of recent literature indicates an emergence in the use of combinatorial methods such as simulated annealing in groundwater management during the past nine to ten years. While previous studies demonstrated the feasibility of using these methods, a general finding was that computational processing requirements were inordinately high relative to gradient-based methods. An enhanced annealing algorithm was developed and used to demonstrate the potential for greatly improving the computational efficiency of simulated annealing as an optimization method for groundwater management applications. The algorithm incorporates "directional search" and "memory" capabilities. Selecting search directions based on better understanding of the current neighborhood of the configuration space was shown to improve algorithm performance. Also, "memory" concepts derived from the Tabu Search Method show particular promise for improving the rate and quality of convergence. Performance of the enhanced annealing method was evaluated and the resultant management method was demonstrated using an example from the literature.