Optimizing Gasoline Recipes and Blending Operations Using Nonlinear Blend Models

Abstract

Gasoline is one of the largest-volume products of the oil industry that yields 60%–70% of the total refinery revenues. This work presents a novel continuous-time mixed-integer nonlinear programming (MINLP) formulation for the gasoline blend scheduling problem. It incorporates nonlinear blending correlations for an improved prediction of key blend properties, and nonlinear constraints for precisely tracking the inventory level in product tanks when multiple blenders are operated. The approach handles nonidentical blenders, multipurpose tanks, sequence-dependent changeovers, limited amounts of gasoline components, and multiperiod scenarios with component flow rates changing with the period. Operating rules for blenders and product/component tanks are also considered. A special model feature is the use of floating slots dynamically allocated to time periods while solving the problem. An approximate mixed-integer linear programming (MILP) formulation assuming ideal mixing provides a good initial point. By fix...