Otimização de modelos lineares dinâmicos como instrumento de melhoria nos processos de gestão do Centro de Sementes Nativas do Amazonas: CASE

Abstract

The purpose of this study is to discuss the agricultural economic activity collection of native Amazonian seeds that has a high degree of complexity, resulting in increase in the amount of physical variables to manipulate and control the problem of establishing the procedure and statement of assumptions for the model and search for a solution of the problem, control, calibration and validation of the solution. For it is a agroeconomic system such collection activity of native seed demand the establishment of robust quantitative models. In this context, the computational mathematical modeling tool that is appears as the simulation and optimization solutions to such problems, analyzing phenomena, developing quantitative models for description and developing computer code for obtaining optimizations. Since economic activity gathering Amazonian native seeds receives benefits of this modeling, we see that Agroeconomic systems require exponentially more complex responses with respect to those demanded by other business models and, based on this need, a simulation model was developed and optimization that has as main feature the incorporation of non-linearities of the processes inherent in agricultural economic activity in question. Their utility function is designed to maximize the level of phenology (harvest), compliance with the collection demands based on priority criteria established by the decision maker. Operational limitations of seed collection activity were included in the model using linear constraints (budget resource and collection costs) and nonlinear (namely Amazon, difficult collection, difficulty of processing and others) using the modeling software GAMS and simulation. Linear and non-linear constraints were represented and solved by linear programming using GAMS modeling and programming language. The results reported by the simulator show that all constraints were met, respecting their impacts. The processing time used by the quantitativecomputer model to meet the optimization requirements was relatively low.