Neural-network-based approximate predictive control for the start-up of a steam generator

This paper presents a neural-network-based approximate predictive control scheme, which has demonstrated via simulations, the feasibility of its application as a supervisory scheme for the steam generator startup process at a fossil electric power plant. Appropriate methodology is offered to carry out the implementation of this scheme. It includes the modelling of the process to be controlled using experimental data, the predictive control algorithm, the design of a state observer, as well as the linearization process of a dynamic nonlinear model based on a MLP network.