Image Distribution Model Based Iterative Learning Control for Combustion Processes

In combustion processes, the goals of the operation are to keep the combustor in the optimal condition and alleviate instabilities and its severe consequence. The conventional control systems based on the oxygen content usually have responses with certain delay. Most flames are turbulent and are affected by several unmeasured disturbances and continuous variation due to several types of fuels burnt simultaneously. This paper presents a closed-loop combustion control algorithm for a combustor using the distribution control theory developed recently. Using the distribution data obtained by a pilot-scale experiment, the equivalent selective B-spline wavelet approximation is proposed to construct a probability distribution function model so that the resulting output density distribution curve matches the desired distribution. In the distribution control design, an iterative learning PDF control strategy is developed. It ensures the output PDF to follow the desired distribution. The proposed method is applied to a pilot scaled combustion control system to demonstrate its effectiveness.