Surrogate model-based optimisation of a batch distillation process

Abstract

A surrogate model-based method is proposed for optimising batch distillation processes and applied to the recovery of methanol from a five-component azeotropic waste solvent mixture, where pollutants are removed in two fore-cuts and an after-cut. The objective function is the profit of a single batch, while constraints are for the purity of the main cut and composition of the second fore-cut to be recycled. Simulations are performed by a flow-sheet simulator in a set of points in the space of optimisation variables (reflux ratios of steps, stopping criteria of fore-cuts). Algebraic surrogate models are fitted by ALAMO to simulation results to describe the objective function and the constraints. The resulting optimisation problem is solved numerically. The profit obtained is by 5% higher than the one previously obtained by genetic algorithm (commonly used for optimisation of batch distillation), while the number of simulations is reduced to its third. The highest profit, previously obtained by the Nelder-Mead simplex method, is approached within 1%. Although the simplex method required fewer simulations, the new method proposed here is a global one. The process is re-optimised for different prices to investigate their influence on the profit and optimal values of operational parameters.