(149u) Advanced Model Predictive Control Strategies for Large-Scale Dynamic Systems Based on Data-Driven Artificial Neural Networks

Model Predictive Control (MPC) is a promising control methodology which was developed and applied in the process industries since the 1990s. Models in the form of closed equation sets are normally needed for MPC, but it is often difficult to obtain such formulations for large nonlinear systems. To extend MPC application to nonlinear distributed parameter systems (DPS) with unknown dynamics, advanced MPC control strategies based on model reduction technique and artificial neural network have been developed. An off-line model reduction technique, the proper orthogonal decomposition (POD) method, is first applied to extract accurate non-linear low-order models from the non-linear dynamic large-scale distributed system. Then a series of successive ANNs are trained based on the time coefficients of POD basis functions to obtain the model for the system. There are ANN (artificial neural network) –POD (proper orthogonal decomposition) – nonlinear MPC (model predictive control) and ANN-POD-TPWL (Trajectory Piecewise-Linear) - MPC, and both of them can be applied to highly non-linear dynamic systems. The novelty of our methodology lies in the application of POD's highly efficient linear decomposition for the consequent conversion of any distributed multi-dimensional space-state model to a reduced 1-dimensional model, dependent only on time, which can be handled effectively as a black-box through ANNs. It is potential to apply them for process optimization and control in wider process industries.

A benchmark case study to use cooling zones to stabilize a tubular reactor with recycle, as shown in Figure 1, will be used as an illustrative example to show those control strategies.