Empirical data modeling and state estimation for a steam boiler system

This paper presents an empirical data modeling approach for a class of systems called quasi-linear. In addition, state estimation based on the identified system and the design of a robust feedback controller for a complex industrial application are included. A drum boiler steam generator system is selected as a prototype example of this class of nonlinear dynamic systems to be identified and to be controlled in a deterministic and stochastic environment. Simulation results indicate that the resulting controller shows a significant improvement over existing proportional-integral (PI) control, including reduced transient response, time delay tolerance and also robustness against full load range (large steam flow changes), for different characteristics exhibited by the input (feedwater) valve. The prominent feature of this method is that no specific model information is initially required throughout the identification and control of the nonlinear plant, and hence the method is suitable for both identifying and controlling of the critical industrial processes.