DURAND Guillermo Andres
congresos y reuniones científicas
Resource Task Network MILP Formulation for Short-Term Scheduling of a Sugar House
Austin (EEUU)
Congreso; AIChE Annual Meeting 2004; 2004
Institución organizadora:
American Institute of Chemical Engineers
The objective of the present work is to develop and apply a scheduling model for the operation of a sugarhouse. The model is a MILP formulation based on a Resource Task Network strategy. A sugarhouse is the process house that concentrates, crystallizes and dries the sugar. From the multi-effect evaporators, the sugarhouse receives a thick juice of sugar in water (aprox 70% w/w) plus impurities (called non-sugars). This juice is treated in two crystallizing and drying stations, A and B. Each one of these stations has a battery of batch crystallizers. The flow from the evaporators is continuous, so, the crystallizers have to work in parallel and out-of-phase cycles, with an intermediate storage tank between the continuous and batch units. The crystallizers are large vacuum evaporator pans (60 tons each), heated by steam. They concentrate the thick juice until the sugar crystallizes. Then, the sugar is taken to a battery of batch centrifuges, where it is washed and dried. The final product is sugar ready to be packed. The operation of this house is complex and difficult, not only it is necessary to comply with the several set-points in each stage, but also there is the transition from continuous to batch operation and vice versa, leading to potential accumulation problems. In addition, the existence of long recycles within the sugarhouse makes necessary to plan (schedule) the house operation at least 48 hours in advance. To find the optimal schedule is one of the well-known NP-complete problems in the subject of operations research. To face the problem of scheduling a continuous/batch plant, the operation is often modeled dividing the horizon time in periods of equal duration, and using discrete variables to assign units to each period. This results in a MILP formulation for the problem. The concept was introduced by Kondili et al. (1993) and developed by several authors in the previous decade (Shah et al., 1993; Pantelides, 1994). Pantelides developed an alternative formulation, Resource Task Network (RTN), later used by Schiling and Pantelides (1996) to model the time line in a continuous form. Ierapetritou and Floudas (1998) proposed a new RTN-MILP formulation, based on events, for the scheduling of continuous/batch plants. This formulation generates simpler and smaller mathematical models, with less binary and continuous variables. It requires less restrictions and LP relaxations, allowing to find the solution in a significant lower computational time. Results are shown for the optimization of the sugarhouse under different operation scenarios: start-up and shut-down, maximization of production, maximization of profits, break-down and reparation of equipment, and strong restrictions in resource levels. Two commercial MILP solvers will be tested: OSL and CPLEX (under GAMS environment). Future work includes introduction of uncertainty.