Active Combustion Control: Theory and Practice

Professor Anuradha Annaswamy

Department of Mechanical Engineering

Massachusetts Institute of Technology

One of the most successful applications of control technology in fluid systems is in continuous combustion processes, examples of which can be found in power generation, propulsion, and heating systems. The objectives in these systems are the realization of high thermal outputs, high efficiency, and reduced emissions, all of which are often accompanied by thermoacoustic resonances that tend to increase nominal pressure levels by more than 20-30%, thus compromising safety and reliability. Over the past few years, active control technology has been increasingly sought after to achieve the desired objectives without encountering dynamic instabilities. Recent demonstrations, the most recent of which is a 20 db pressure reduction in a 260 MW power generation unit with active control, have shown unequivocally that active control is not only a feasible technology for reducing the unsteady pressure but also that the approach can be scaled up to large-scale industrial rigs.

In order to provide guranteed and uniform performance over a large range of flow-rates and loads in the presence of various system uncertainties, the overall closed-loop combustor needs to be optimized, in terms of achieving better quality, reliability, and repeatability at smaller cost, and better speeds. Towards this end, a systems approach has been deployed in the Adaptive Control Laboratory at MIT to develop model-based combustion control methods. This approach consists of the modeling of the linear instability mechanisms and nonlinear limit-cycle behavior that lead to the pressure oscillations and control of these oscillations using linear, nonlinear, and adaptive control methods. Using this approach, linear time-delay models of instability and nonlinear models of limit cycles and bifurcation have been derived using both physically-based and system-identification principles. Linear strategies based on H2 and oo control, open-loop methods based on slow-switching, and time-delay control methods have been demonstrated to lead to pressure minimization at rates that are significantly faster than those previously reported. Nonlinear control methods based on adaptation as well as neural computation have been shown to result in improved performance over an increased range of operation. In this talk, highlights of the systems approach and the impact of the resulting models and control algorithms on a number of combustors ranging from bench top models to large-scale rigs will be presented.

Friday, January 26, 2001

3:30 - 5:00 p.m.

1500 EECS