Performance Evaluation of Control Methods on the Water Side of Drum Boilers

Abstract

This thesis evaluates control strategies for a drum boiler unit. Drum Boilers are a highly nonlinear system, as there are non-minimum phase shrink-and-swell effects to account for. A more complex control strategy may prove to be a better option than what is used in industry today. The goal is to showcase different control strategies on the nonlinear system given specified design constraints, from three element cascade control with a feed forward, to using a Linear Quadratic Regulator (LQR), to the more contemporary Model Predictive Control (MPC). The process is built around the Åström-Bell non-linear complex drum-boiler model, and is extended with super-heater and turbine dynamics using other known results. The model involves simplification of controlling heat flux instead of modeling the heat transfer and fuel combustion from the air side of the boiler. The implementation of the complete system is carried out in MatLab. Simulation results are presented for the three element control method, the linear quadratic regulator (LQR) applied to a nonlinear system, and a model predictive control (MPC) algorithm to use on nonlinear systems. The simulation results are focused on automatic control operation and finding satisfactory response behaviors. The LQR and MPC approach assume full state feedback without the use of an observer. The research shows that all of the controllers can meet the design criterion, however secondary effects cause both the Three element cascade PID controller and the LQR controllers to be less desirable than the MPC approach. A heuristic trial and error approach to tuning was used in all methods due to the highly coupled nature of the system. This evaluation of the types of controllers showcasing tuning to a specified design criterion proves that the controller type is more important than optimal tuning.