Computational study of network-based mixed-integer programming approaches for chemical production scheduling

The goal of this paper is to discuss the modeling aspects and study the computational performance of scheduling approaches for batch process networks that are based on discrete-time and continuous-time representations. First, we compare the above two modeling approaches in terms of formulation size and modeling capabilities; we briefly review their main characteristics and outline their advantages and disadvantages. Second, we perform an extensive computational comparison between the two methods using a collection of more than 100 problem instances and 800 optimization runs covering five different process networks, various objective functions, different scheduling horizons, and a wide range of features (fixed and variable processing times, utilities, holding and backlog costs, intermediate shipments, and setups). We show that the computational requirements of discrete-time models increase moderately with the incorporation of these additional features, something that cannot be said for continuous-time models. We close with a number of conclusions that we believe will lead to fruitful discussions in the area and foster further development of modeling and solution methods for chemical production scheduling problems.