Model Predictive Control and Genetic Algorithms for Optimization of Continuous Stirred Tank Reactors

Problem statement: The Continuous Stirred Tank Reactor is a classic example of a chemical reactor with nonlinear dynamics that is extensively studied and modeled in process control literature. However it is often considered using linearizations and other simplifications of a model of the object of control. Purpose of research: The material presents a sophisticated approach to controlling systems with inherently nonlinear dynamics (the Continuous Stirred Tank Reactor), through the advanced integration of a Genetic Algorithm (GA) with Model Predictive Control in order to avoid linearizations and other simplifications. While the application of Genetic Algorithm within the realm of Model Predictive Control is not novel and has been explored across a range of applications, the research under consideration stands out by offering a unique perspective: we optimize a consistent set of control parameters across the entire control horizon, as opposed to the more conventional approach of adjusting control inputs incrementally over the prediction horizon or merely optimizing for the immediate subsequent step. Results: The proposed innovative methodology is underpinned by a carefully crafted cost function, augmented with a discount factor that effectively addresses the limitations associated with optimizing solely for immediate future steps. By employing our approach, the process of determining control inputs is significantly simplified, which, in turn, markedly improves computational efficiency. The experimental findings robustly confirm that the proposed GA-enhanced Model Predictive Control strategy adeptly navigates the Continuous Stirred Tank Reactor towards the desired operational state, achieving superior performance over traditional control methods. Practical significance: The study underscores the adaptability and applicability of Genetic Algorithm within the framework of Model Predictive Control with superior efficacy in managing intricate control scenarios, thus presenting a sophisticated solution in situations where conventional methods do not suffice.