Influência dos parâmetros da metaheurística simulated annealing em um problema de planejamento da produção florestal

Abstract

The excessive quantity of variables in the forest production process requires the use of tools that provide answers with a high degree of accuracy in a shortest possible time. Thus, this work aimed to evaluate simulated annealing metaheuristic performance under different combinations of values for its parameters. The study considered a forest production planning optimization problem with horizon plan of 16 years, 120 stands, harvest age between 5 e 7 years and volumetric demand between 140.000 m³ and 160.000 m³. Three different values were evaluated for the initial temperature, temperature decay rate and amount of neighbors, totaling 27 combinations. For each one, it was considered 30 repetitions with stopping criterion equals to 1000 iterations. The process was performed using MeP software and for the statistic evaluation, it was applied the Kruskal-Wallis test. The solution that showed the highest net present value (R$ 31.054.933) was found using an initial temperature equal to 106, the decay rate of 0,01 and 30 neighbors. All the solutions obtained had a satisfactory efficacy and it ranged from 94,97% to 94,16%. The results found showed a statistically significant difference between the best and the worst combination of parameters. It is possible to conclude that the set of parameters is important to find good solutions but it is not determinant to find viable solutions.