Multi-model Coordinated Predictive Control for Plant-wide Operation Systems under Networked Information Mode

We propose a multi-model coordinated predictive control algorithm under networked information mode for plant-wide operation nonlinear systems. When a large-scale switching occurs, we have adopted the gap metric theory to divide the operation space of the nonlinear system into local subspaces under different operation conditions. In each subspace, we have applied a local linear model to design an infinite-horizon predictive controller. To transfer the problem of state variables in different subspaces, we have assessed the influence of the predicted state transitions in subintervals on the control performance. The optimization of N-step horizon control law sequences (instead of a single feedback control law) reduces the conservativeness caused by ignoring subspaces transition in multi-model predictive control. We have applied the algorithm to the continuous stirred tank reactor, which is widely used in chemical processes. Our simulation results show the effectiveness of the proposed algorithm.