Nonlinear predictive control of the catalytic reforming process based on a hybrid model [in Ukrainian]
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
The paper investigates the problem of improving the efficiency of process control for catalytic reforming with continuous catalyst regeneration (CCR). The relevance of this research is determined by the fact that traditional control systems based on PID controllers cannot ensure optimal operation of the unit due to the significant inertia of the process, transport delays, nonlinear characteristics, and parameter variability caused by catalyst deactivation. The aim of the study is to develop a nonlinear model predictive control (NMPC) system based on a hybrid model that combines a physical kinetic model of the process with a neural network component for correcting prediction errors. A comprehensive mathematical model of the CCR reforming process has been developed. It includes a kinetic model with eight pseudocomponents and twelve reactions, a catalyst deactivation model, and a model for calculating the octane number using an additive formula. A hybrid model architecture with sequential integration of components is proposed, where multilayer perceptron (MLP) neural network with two hidden layers (64 and 32 neurons) compensates for systematic errors of the physical model. The hybrid model provides a 25% improvement in the accuracy of octane number prediction compared with the baseline physical model. An NMPC control system has been synthesized using a quadratic objective function that minimizes the deviation of controlled variables from their setpoints while penalizing rapid changes in control inputs. The optimization problem includes constraints on temperature ranges, pressure, and the rate of change of control actions. The software implementation was carried out in Python using the CasADi library for symbolic computations and automatic differentiation, and the IPOPT solver for nonlinear programming based on the interior-point method. A model adaptation mechanism has also been implemented to compensate for drift caused by catalyst deactivation. Simulation results for a reforming unit with a capacity of 1.5 million tons per year confirm the advantages of the developed NMPC system: the average absolute error of the octane number is reduced by 57% compared with PID control and by 36% compared with linear MPC; the settling time of transient processes is reduced by a factor of 2.8; and the maximum deviation under feed composition disturbances is reduced by 63%. The practical significance of the work lies in the possibility of applying the proposed algorithms to improve the efficiency of control systems for industrial CCR reforming units.
Fields of Study & Subjects
Advanced Control Systems Optimization
Process Optimization and Integration
Control Systems and Identification
Publication details
- Journal / source
- Proceedings of the NTUU “Igor Sikorsky KPI” Series Chemical engineering ecology and resource saving
- Issue
- 1
- Pages
- 40–52
- Publication date
- 30 March 2026
- Language
- EN