ILC-based two-layer economic performance assessment and improvement strategy for large-scale industrial distributed MPC systems

Abstract

When applied to large-scale industrial plants, the traditional linear quadratic gaussian (LQG) benchmark performance assessment method usually brings about unreachable economics, and its LQG curve dimension increases with the expansion of the system scale, which significantly aggravates the computation burden. To address various problems in the LQG benchmark method, an ILC-based two-layer economic performance assessment and improvement strategy are proposed and applied in large-scale distributed model predictive control (DMPC) systems. The presented strategy separates the whole operation time into multiple intervals during which the economic performance will be gradually improved and finally achieves its optimal. In each interval, the economic performance acquires its first promotion by the fixed variance obtained from the lower DMPC layer. The distributed ILC (DILC) method then provides the tuning parameters of each DMPC controller in the next period with the updating principle based on sensitivity analysis. The effectiveness of the presented strategy is verified via an improved Alumina continuous carbonation decomposition process compared to the former one.