Skip to Main Content
A high-efficient real-time optimization method for process systems - the mnemonic enhancement optimization (MEO) with incremental Delaunay triangulation and modified barycentric interpolation is developed. It improves the performance of process real-time optimization (RTO) systems further, compared with the traditional method of RTO for starting point generation and the MEO method with zero-order approximation. Based on the repetitive nature of RTO, MEO turns the problem sequence into a kind of parametric optimization problem solved repetitively when the parameters change and takes advantages of this feature to accelerate the solution process of optimization. In this work, it is proved that the approximation error of the MEO method converges to zero with probability 1 as the experience accumulates. And a modified barycentric interpolation scheme using more empirical information than that in the previous study to estimate the optimum is suggested. The multi-dimensional incremental Delaunay triangulation algorithm is hired to maintain the geometric structure of the empirical database and provide suitable nodes for MEO approximation. A numerical case study on a depropanizer and debutanizer distillation sequence validates the proposed method and shows the reductions of 65% and 26% in the solution time of optimization, compared with the traditional method in RTO and the MEO method in our previous study, respectively.