Robust distributed economic model predictive control based on differential dissipativity

Abstract In this article, a robust distributed economic model predictive control (DEMPC) approach is developed for plant‐wide chemical processes. The proposed approach achieves arbitrary feasible setpoints that may vary frequently, attenuates the plant‐wide effects of unknown disturbances and minimizes a plant‐wide economic cost. In this approach, a plant‐wide process is represented as a network of process units and each process unit is controlled by an individual controller which shares a plant‐wide optimization economic objective and stability conditions through the network. To ensure the convergence of process variables to arbitrary setpoints, a contraction condition is developed for the DEMPC, based on the contraction theory. To deal with the effects of interactions among process units, the concept of dissipativity is adopted. Using sum‐separable control contraction metrics, a reference‐independent robust stability condition is developed to ensure the plant‐wide disturbance effects (under interactions among process units) to be attenuated in terms of differential ℒ 2 gain and represented by a plant‐wide differential dissipativity condition, which is converted into the differential dissipativity conditions that individual controllers need to satisfy. This approach facilitates the optimization of plant‐wide economic costs with global constraints in a distributed way, allowing efficient implementation of alternating direction method of multipliers (ADMM). The proposed approach is illustrated using a reactor‐separator process.