Suboptimal nonlinear predictive control with MIMO neural hammerstein models

Abstract

This paper describes a computationally efficient (suboptimal) nonlinear Model Predictive Control (MPC) algorithm with neural Hammerstein models. The Multi-Input Multi-Output (MIMO) dynamic model contains a steady-state nonlinear part realised by a set of neural networks in series with a linear dynamic part. The model is linearised on-line, as a result the MPC algorithm solves a quadratic programming problem. The algorithm gives control performance similar to that obtained in nonlinear MPC, which hinges on non-convex optimisation. © 2008 Springer-Verlag Berlin Heidelberg.