Gaussian Process Regression Based Multi-Objective Bayesian Optimization for Power System Design

Within a disruptively changing environment, design of power systems becomes complex task. Meeting multi-criteria requirements with increasing degrees freedom in and simultaneously decreasing technical expertise strengthens the need for multi-objective optimization (MOO) making use algorithms virtual prototyping. In this context, we present Gaussian Process Regression based Multi-Objective Bayesian Optimization (GPR-MOBO) special emphasis on its profound theoretical background. A detailed mathematical framework is provided to derive GPR-MOBO computer implementable algorithm. We quantify effectiveness efficiency by hypervolume number required computationally expensive simulations identify Pareto-optimal solutions, respectively. For validation purposes, benchmark our implementation test function analytically known Pareto front compare results those well-known NSGA-II pure Latin Hyper Cube Sampling. To rule out effects randomness, include statistical evaluations. turnes as an effective efficient approach superior character versus state-of-the art approaches value-add when are high. Finally, provide example system that demonstrates both methodology itself performance benefits.