Nonlinear model predictive control for the ALSTOM gasifier benchmark problem

Model predictive control has become a first choice control strategy in industry because it is intuitive and can explicitly handle MIMO linear and nonlinear systems with the presence of variable constraints and interactions. In this work a nonlinear state-space model has been developed and used as the internal model in predictive control for the ALSTOM gasifier. A linear model of the plant at 0% load is adopted as a base model for prediction. Secondly, a static nonlinear neural network model has been created for a particular output channel, fuel gas pressure, to compensate its strong nonlinear behaviour observed in open-loop simulation. By linearizing the neural network model at each sampling time, the static nonlinear model provides certain adaptation to the linear base model. Noticeable performance improvement is observed when compared with pure linear model based predictive control