(374a) Multiscale Optimization Strategies for the Integrated Design, Planning and Scheduling of Process Systems

Grossmann, I., Carnegie Mellon University
Brunaud, B., Johnson & Johnson
Lara, C. L., Carnegie Mellon University
Zhang, Q., Carnegie Mellon University
Chris Floudas devoted the last decade of his life in the multiscale optimization of energy systems. Multiscale optimization is a very challenging problem that requires novel representations and effective solution strategies. We explore in this presentation several strategies for the integration of design, planning and scheduling in multiscale process systems.

In this presentation, we first give a general overview of multilevel decision making in the chemical process industry, in which strategic, tactical and operational decisions must be integrated. We discuss a network structure that helps to identify both the material and information flows that interconnect models associated at each level to facilitate their integration. We first start with integration and planning in batch multiproduct plants with sequence dependent changeovers. We show how the incorporation of traveling salesman constraints in an MILP planning model can capture the main effects of changeover at the detailed scheduling level. We next address the long term planning of electric power systems in which the challenge lies in the spatial and temporal multiscale nature of the problem. We describe strategies based on spatial aggregation and time sampling that allow effective long term strategic planning through a multiperiod MILP model for the selection of coal, natural gas, nuclear, wind and solar technologies for power generation over a very long term horizon. Finally, we address the multiscale production routing problem, which considers the coordination of production, inventory, distribution, and routing decisions in multicommodity supply chains with complex production facilities. We describe an iterative MILP model involving two different time grids, one at the finer level for the detailed mode-based production scheduling model, and one for a restricted vehicle routing subproblem that is considered in each time period of the coarse time grid. For each of the three cases, we report computational results that illustrate the effectiveness of the proposed multiscale optimization strategies.