Models and techniques for scheduling farm operations: A comparison

Abstract

Scheduling of farm operations is formulated as a dynamic programming model (a precise model) and as a linear programming model (a global or relaxed model). The precise model can be solved with the dynamic programming technique at considerable cost, but in this study it is solved with simulation and a heuristic scheduling procedure, aimed at reasonable solutions amongst those feasible. The selection of operations at decision moments is done with an urgency strategy which assigns an urgency to each combination or set of gangs operating simultaneously and searches for the combination with maximum urgency. The urgency is based on the timeliness function of the material processed by a gang. Historical hourly weather data and moisture content data for grain, straw and soil for twelve grain harvest seasons are used to evaluate the urgency strategy. The resulting schedule shows a fair use of workable time. From this subjective criterion it is concluded that the urgency strategy is acceptable. Reordering the workable time within a day results in a significantly lower estimate of variable costs for drying wet grain, overtime of men and timeliness loss of grain, straw in the field and bales. A global or relaxed model with five weekly periods is solved by linear programming. Models with and without nesting of workable time for grain, straw and soil continue the tendency of lower costs found with reordering workable time. These lower costs suggest that the resulting schedule is not executable within the real sequence of workable and unworkable time for materials. Handling tighter constraints for workable time and excluding the time workable only for bale handling results in a cost estimate at the level of the simulation and so perhaps in an executable schedule. On the other hand, the lower cost may suggest that the urgency strategy is a poor heuristic procedure. This is not demonstrated, however, by the use of workable time. The performance of the strategy still has to be tested by finding the optimal strategy.