Optimal control of a swirl-stabilized spray combustor using system identification approach

Abstract

An optimal controller using the system identification (SI) method was developed for a swirl-stabilized spray combustor operating between 30 and 114 kW. The efficacy of the controller was tested with two different nozzle configurations. The first consisted of a dual-feed nozzle whose primary fuel stream was utilized to sustain combustion, while the secondarystreamwasused for active control. The second configuration used a single-feed nozzle with two different swirling air streams. An LQG-LTR (linear quadratic Gaussian-loop transfer recovery) controller was designed using the SI-based model to determine the active control input, which was in turn used to modulate the secondary fuel stream. Using this controller, the thermoacoustic oscillations, which occurred under lean operating conditions, were reduced to the background noise level. A time-delay controller was also implemented for comparison purposes. The results showed that the LQG-LTR controller yielded an additional pressure reduction of 14 dB compared to the time-delay controller in both configurations. This improvement can be attributed to the added degrees of freedom of the LQG-LTR controller that allow an optimal shaping of the gain and phase of the controlled combustor over a range of frequencies in the neighborhood of the unstable mode. This leads to the extra reduction of the pressure amplitude at the unstable frequency while avoiding generation of secondary peaks.