Multi-Objective Bayesian Optimization of NGL Fractionators Using Process Simulation

After remaining relatively constant for over 30 years, domestic natural gas liquids production has tripled since 2010 as a direct result of the U.S. Shale Revolution. This work presents an approach to optimize natural gas liquids fractionator throughput from existing plants using multi-objective optimization. The method maximizes operating net revenue while simultaneously minimizing direct and indirect emission rates by applying a custom multi-objective variation of the efficient global optimization (EGO) algorithm. This algorithm optimizes kriging surrogate models of the objective functions to determine which points to sample in the process simulation before the next iteration. The Pareto front obtained from multi-objective EGO is compared with results from other optimization methods to recommend both a formulation and a solution method to economically and environmentally optimize a natural gas liquids fractionator.