Designing an evolution program for solving integer forest management scheduling models: An application in Portugal

In this article, basic concepts of both genetic algorithms and evolution program design are presented. An evolution program is presented to solve a Model I harvest scheduling problem with 0–1 decision variables for the management alternatives for each stand, with annual constraints on harvested volume. An appropriate data structure (i.e., chromosome representation) is presented, as well as modified selection, crossover, and mutation strategies specially designed for application to large forest scheduling problems. Emphasis is on designing an efficient evolution program to address the complexity of large integer problem model solving and to provide both strategic and operational guidance to forest managers. A new stopping criterion for this iterative heuristic based on the asymptotic behavior of the evolution process is further presented. The new evolution program is applied to a two-product timber harvest scheduling problem in Portugal with a temporal horizon extending to seventy 1 yr periods. Results from 50 test computer runs are discussed for application to this large problem encompassing approximately 122,000 binary integer variables and 1,000 constraints. Statistic analysis of the convergence process suggests that the evolution program may seek optimal solutions at reasonable computational cost. Results thus suggest that evolutionary techniques may be used to confront the complexity of integer forest management scheduling model solving.