Document Type : Research Paper
Authors
Department of Agricultural Economics, University of Sistan and Baluchestan, Zahedan, Iran
Abstract
Optimal efficiency of dam reservoirs is considered as one of the important problems in water resources management. Evolution algorithms have been presented as an intelligent technique to establish a suitable trade-off between reservoir of water resources and their release rates. In this study, a new modified approach, called “Learning automata-based differential evolution algorithm (LADE)”, has been proposed for modeling a single reservoir system in Iran with the aim of evaluating development efforts to optimize allocation of water resources. To evaluate the efficiency of LADE, first, a set of mathematical benchmark functions has been tested successfully, and then, LADE has been applied to the problem of optimization of the reservoir allocation of Golestan Dam (Iran). In addition, different versions of LADE have been offered which only vary in the type of mutation strategy to explore the search space. Moreover, the best version of LADE (LADE/curr2rand) has been compared with some of state-of-the-art algorithms, including artificial bee colony (ABC) algorithm, differential evolution (DE), imperialist competitive algorithm (ICA), genetic algorithm (GA), and particle swarm optimization (PSO). The best of numerical experimental results regarding the cases of mean and standard deviation (SD) of errors value, as well as the average runtime were associated with LADE/curr2rand. Furthermore, based on several tests including the evaluation of reliability and vulnerability indexes, release values relative to the total demand, the difference between release and demand values in an annual average, LADE/curr2rand yielded the highest performance compared to other algorithms.
Keywords
)