Model-Based and Model-Free Optimal Control of a Gas Liquid Cylindrical Cyclone

Abstract

Optimal control is important in the oil and gas industry, where small changes of process variables may have significant economical, environmental and safety impacts. Finding the optimal setpoint for a controlled variable may not be straight-forward for an operator, motivating the use of automatic optimization methods. A widely used optimal control method is the model predictive control (MPC) method which requires a dynamic model to predict the future behaviour of the system. The MPC satisfies constraints and finds the optimal input to a plant. However, the requirement of a model can make the method difficult to implement and computation time might be too high for real-time implementation. An alternative to MPC is the Extremum Seeking (ES) method. This method aims to find the input that optimizes the output by slowly perturbing the controlled variable. It does not require a model and is less computationally expensive. In this paper we apply an MPC and an ES optimization scheme to a gas liquid cylindrical cyclone in order to optimize the purity of the gas outlet. Simulations show that both methods are able to find a liquid level setpoint that ensures high quality of the gas outlet stream under varying inlet conditions.