Global Optimization of Refinery – petrochemical Operations via Process Clustering Decomposition

We consider the short-term planning of an integrated refinery and petrochemical complex using a mixed-integer nonlinear optimization. The process network is represented by input-output relationships based on bilinear and trilinear expressions to estimate yields and stream properties, fuels blending indices and cost functions. Binary variables select the operating modes for the process units. Our global optimization algorithm decomposes the network into small clusters according to their functionality. For the constraints inside a given cluster, we formulate a mixed-integer linear relaxation based on piecewise McCormick envelopes. The partitions for the variables are updated dynamically and their domain is reduced applying optimality-based bound tightening. For the constraints outside the cluster, we use the standard McCormick envelopes. Our approach is demonstrated on an industrial-size case study representing a typical planning scenario in Colombia. Results show that it outperforms the state-of-the-art commercial solvers ANTIGONE and BARON not only in terms of optimality gap (8 vs. 58 and 48%, respectively) but also the quality of the solution itself.