Finite Adaptability in Data-Driven Robust Optimization for Production Scheduling: A Case Study of the Ethylene Plant

A novel adaptive robust optimization methodology called Pareto optimal finite adaptability (POFA) is proposed for production scheduling of the ethylene plant. As an improvement to conventional robust optimization methods, POFA exhibits high efficacy dealing with multiple evolution paths caused by possible decoking decisions. With the introduction of adjustable variables, both continuous and discrete, POFA can calculate a set of solutions which is not only robustly optimal but also Pareto optimal, and each solution corresponds to an evolution path. Then, the best solution selected from the set will be implemented in each period according to the past observations about the evolution of uncertainty. The raised approach was applied in a real-world ethylene plant and showed promising scheduling results. As expected, by increasing the number of adjustable variables, POFA improved the cost-effectiveness of the fuel pre-orders without compromising the robustness.