Extraction of Optimal Operation Rules in an Aquifer-Dam System: Genetic Programming Approach

Abstract

Surface water and groundwater are two important resources that can be used simultaneously to meet demand. To sustain these resources in a planning horizon, different conditions of supply and demand can be balanced by operational rules. Those rules can be extracted from prior experience while operating a conjunctive water system and then applied in future periods. Genetic algorithm (GA) is an evolutionary algorithm that has been extensively used to determine optimal operation rules. Commonly, the operation rule pattern is fixed by a linear equation, and GA calculates optimal coefficients in a defined pattern. In this paper, to compute a more effective operation rule that yields a better objective function value compared to GA for the same conditions, a fixed-length gene genetic programming (FLGGP) rule is developed and extracted based on genetic programming (GP). The FLGGP rules are employed in an aquifer-dam system with two subsystems, in which each subsystem can supply parts of the demand. Results show that the obtained objective function value using the FLGGP rules is 26 percent better compared to the common linear rule by GA. The developed operational rule uses more functions and mathematical operators than numerical variables. Thus, FLGGP is more flexible and effective in determining optimal rule curves for a conjunctive aquifer-dam system.