Surrogate modeling for nonlinear gasoline blending operations

Abstract

Abstract The application of surrogate modeling in engineering is surging recently for predicting the functional behavior of a system using analytical formulations as an alternative to complex models that often lead to non-convergence issues and not sufficiently accurate solutions in decision-making problems. The surrogate model building procedure addressed in this paper consists of four major steps to be applied in nonlinear blending of gasoline streams. The first is the input (x) dataset generation, performed using the Latin Hypercube Sampling (LHS) technique, which is coupled with a rescaling strategy, and used for evaluating the output (y) dataset. Secondly, the generated data are improved with a normalization procedure to mitigate numerical issues and to avoid biased surrogates. Thirdly, mixed integer quadratic programming (MIQP) formulation based on the least-squares regression is employed to build an optimizable surrogate function for each variable of interest. Fourthly, smaller and simpler surrogates are established and selected to be employed for gasoline blending operations by substituting the complex nonlinear and nonconvex rigorous formulation in an optimization case.