Design and implementation of an improved linear quadratic regulation control for oxygen content in a coke furnace

In view of the performance deterioration of conventional model predictive control strategies under model/plant mismatches, this study first proposes an improved linear quadratic regulation control (LQR) strategy and then tests it on the oxygen content in an industrial coke furnace. The control system design consists of two steps. Based on the direct use of the measured input and output variables and the formulation of output error, a state space model is first developed. Then, by incorporating the dynamics of process states into the cost function of the controller design, a new LQR is proposed to obtain an improved performance for closed-loop control systems. Comparisons with conventional state space model predictive control are also given in terms of both regulatory and servo performance. The case study on the oxygen content in the coke furnace shows that the proposed improves control performance under model/plant mismatches.