Dual Adaptive and Predictive Control of Nonlinear and Distributed Systems
Model predictive control (MPC) is the most widely used multivariable control scheme in industrial control. Success of any MPC scheme critically depends on prediction models. The closed loop performance of an MPC scheme can deteriorate over a period of time if the prediction model is not updated to account for the changing operating conditions. The adaptive (self-learning) control schemes that solve both identification and control problems simultaneously provide an attractive option to alleviate this problem. Since the learning in the conventional adaptive control is passive, the online estimates of model parameters can exhibit undesirable drifts due to insufficient excitation. This problem can be effectively handled using a dual control scheme. The dual-controller is an optimal control scheme that finds the best trade-off between the control action and the investigative action. The control action tries to bring the system state to the desired target value while optimizing a given performance index. The investigative action injects probing signals into the plant to collect information-rich data leading to a better model. Thus, adaptive dual control is an attractive way to handle the uncertainty in the system and maintain the controller performance over an extended period. This work aims at developing dual adaptive control schemes that are based on control-relevant nonlinear dynamic models. Control of moderately large scale systems using distributed MPC framework is also of interest. This work is an extension of the ongoing research in the area of dual adaptive MPC.