A distributed state estimation and control algorithm for plantwide processes

A multirate distributed and decentralized approach to state estimation and control is developed for large-scale processes. The decentralized and scalable form of the Kalman filter algorithm is formulated for multirate sampled-data systems. The multirate formulation is necessitated by differing sample times for distributed computation and process measurement availability. The distributed linear quadratic Gaussian control methodology is implemented in a simulation environment using appropriate tools for distributed computation. The issues involved in the model decomposition for employing the distributed control algorithm are examined. Heuristic guidelines are proposed to balance the computational load and communication overhead across the distributed control network. This methodology is demonstrated in a case study involving a simulated large-scale industrial reaction-separation system.