Research of methods for design of regression models of oil and gas refinery technological units

Abstract

The problem of efficient computational models design for control and scheduling problems in terms of oil and gas refinery column distillation units is discussed in the paper. Such efficient computational models can be constructed in the form of fast static regression models supplemented with dynamic models of measurement and input channels. The effectiveness of methods for constructing fast static regression models is examined in the paper. The input parameters for such regression models are determined. It is proposed to use parametric optimization methods for such models. A preliminary study showed the possibility of using an evolutionary genetic algorithm. Numerical studies were performed using data from column distillation units. The efficiency of using the methods of additional parametric optimization is shown.