Model Predictive Control and Adaptive Kalman Filter Design for an Underground Coal Gasification Process

The control of a nonlinear system such as an underground coal gasification (UCG) process is a challenging task. Several nonlinear design approaches are implemented to improve the tracking performance of the UCG process, however, the nonlinear techniques make implementation complex and computationally inefficient. In this work, a constrained linear model predictive control (MPC) is designed for the UCG process to track the desired trajectory of the heating value, while satisfying actuator constraints pertaining to UCG. The unknown states required for MPC design are reconstructed by using linear adaptive Kalman filter (AKF) and unscented Kalman filter (UKF). The design of MPC and AKF is based on the quasi-linear model of the UCG process. A fair comparison between different control strategies is conducted which include MPC– AKF, MPC– UKF, MPC– gain scheduled modified Utkin observer (GSMUO) and dynamic integral sliding mode control (DISMC)–GSMUO. The quantitative analysis and simulation results show that MPC- AKF outperforms its counterparts by yielding the least tracking error and average control energy. This conclusion holds, even in the presence of an external disturbance, parametric variations, and measurement and process noises. Moreover, MPC- AKF yields 51%, 44% and 46% improvement in absolute relative root-mean-squared error with reference to MPC– UKF, MPC– GSMUO and DISMC–GSMUO, respectively. A quantitative analysis has also been carried for AKF and UKF, which shows that the performance of AKF is more robust against changes in the initial values of measurement and process covariances.