Modelling, validation, and control of an industrial fuel gas blending system

In industrial fuel gas preparation, several compositional properties must be controlled within specified limits. This allows client plants to burn the gas safely and with consistent heat production. The variables to be controlled are the higher heating value (HHV), Wobbe index (WI), flame speed index (FSI), and header pressure. A plant in which six feed gasses are blended is considered. Four of the feeds are well-defined makeup streams (costly but always available) and the other two are byproducts that would otherwise be flared. The six feed rates comprise the manipulated variables (MVs) used to regulate the four controlled variables (CVs) while minimising the cost of the gas blend. The control system must compensate for feed composition and fuel gas demand variability. The development and validation of an industrial fuel gas header model is described, followed by a simulation study comparing three Model Predictive Control (MPC) strategies. It is shown that when iterative linearisation is used to update the prediction model and real-time optimisation (RTO) is used to update the CV and MV targets used in the MPC cost function, the plant is driven reliably to the optimal steady-state.

We derive a first principle model of an industrial fuel gas blending system.
► We validate the model against plant data.
► We apply three different Model Predictive Control (MPC) strategies. Two employ successive linearization.
► MPC with periodic real-time optimization of steady-state targets provides near-optimal economic benefits.
► We hypothesize that alternative approaches are unlikely to improve performance significantly.