A New Simple Effective Continuous-time Model for Scheduling of General Batch Plants.

Abstract

Batch processes are widely used in producing high-value chemical products for its good flexibility. The problem of short-term scheduling of batch plants is to determine the optimal utilization of available resources over a given time horizon. Models of most previous works in this domain rely on the definition of time slots or time events so that they involve many binary variables and constraints. In this paper, a new MILP model for scheduling of both multi-product and multi-purpose batch plants with unit-depended changeovers is introduced. Problems are described based on the concept of task in order to deal with complex instances. The proposed formulation uses a continuous time domain representation that does not rely on the definition of time slots or time events. Binary variables for sequencing are defined to indicate whether one task is prior to another in the processing sequence without concerning their accurate position. The model is also simplified by considering the specialty of symmetrical and complemental feature of the binary variables. In a result the number of both binary variables and constraints has been reduced significantly. Moreover, different heuristic rules for preordering can be flexibly embedded into the new model and directly assign the value of binary variables. Therefore much CPU time is saved even maybe without loss of rigorous optimality. Several case studies are presented that illustrate the efficiency of the new model. Comparisons with some previous works are also provided.