Supply-Demand Based Algorithm for Gasoline Blend Planning Under Time-Varying Uncertainty in Demand

Abstract

In this work, we employ a three-level decomposition algorithm to solve the gasoline blend production planning problem under uncertainty in products demands. The products demand uncertainty is modelled as time-varying uncertainty as opposed to time invariant uncertainty used in most existing literature. Time-varying uncertainty is more effective in representing the real-life production planning model since uncertainty tend to increase when a model is used to predict demands far into the future. Chance constraint programming is implemented to evaluate trade-off between optimality versus reliability. The gasoline blend planning is modelled as an MINLP model where nonlinear blending rules are used to more accurately compute the products qualities. The three-level decomposition algorithm based the supply-demand pinch concept (also referred to as inventory pinch) computes the optimal solution (or solution close to the optimal solution) with remarkable reduction in computational time compared to the MINLP single model solution.