An evaluation of the simulated annealing algorithm for solving the area-restricted harvest-scheduling model against optimal benchmarks.

摘要

A common approach for incorporating opening constraints into harvest scheduling is through the area-restricted model. This model is used to select which stands to include in each opening while simultaneously determining an optimal harvest schedule over multiple time periods. We use optimal benchmarks from a range of harvest scheduling problem instances to test a metaheuristic algorithm, simulated annealing, that is commonly used to solve these problems. Spatial data from seven forests in British Columbia, Canada were used. Performance of the simulated annealing algorithm was assessed over a range of problem attributes such as the number of forest polygons, age-class distribution, and opening size. In total, 29 problem instances were used, ranging in size from 1269 to 36 270 binary decision variables. Overall, the mean objective function values found with simulated annealing ranged from approximately 87 to 99% of the optima after 30 minutes of computing time, and a moderate downward trend of the relationship between problem size and solution quality was observed..