چكيده به لاتين
Cellular Automata was originated from physical simulation models. Later on, in heed of its capabilities, its application within optimization problems was studied as well; achieving proper solutions, and reducing computational costs are some of those capabilities that distinguishes this method form others in optimization area. However, because of the need for a specific formulation for each problem – and as there is no unique algorithm for this – CA has been mostly used for single-objective problems, and on a limited basis for multi-objective ones.
Previously, methods called Wighted Cellular Automata (WCA), and Parallel Cellular Automata (PCA) has been proposed for solving multi-objective problems. The former, weighting objective functions of the problem, and truning it into a set of single-objective ones, solves the actual problem. As for PCA, a number of CAs works individually and with the help of interchanging their solutions an approximation of global pareto will be gained. But as mentioned, deploying CA calls for a specific formulation for each problem, and due to multi-objective CA’s just-out aspect, it has been used for numerable problems only.
In this study, the possibility of using PCA in another range of problems has been surveyed, and according to previous formulations’ incapability to handle such problems, another method called Multi-Step Parallel Cellular Automata (MSPCA) has been proposed to solve them. For two reservoir operation problems (Dez and Sefidrud dam), MSPCA was used, and in rder to evaluate it, its results were compared to that of Non-Dominated Sorting Genetic Algorithm (NSGAII).
MSPCA was able to reach higher-quality solutions, with less computational costs, compared to NSGAII. In the first problem, MSPCA dominated more area of solutions space (8.64% more), found more solutions on pareto (1674), and was finished in much less time (99.64%) compared to NSGAII. As for the second problem, the numbers above are 34.81%, 2069, 99.65% respectively. Also, MSPCA’s results were weighte up against the results of actual operation in reality, which showed MSPCA has lead to producing more optimized objective functions, for all objectives.
