Long-range prediction for poorly-known systems

Abstract

Predictive control of the outputs or states of a system relies on the quality of the predictions over a certain time horizon into the future. The conventional ways of obtaining these predictions suffice when the system characteristics allow either the development of a relatively accurate model or the use of a relatively short time horizon. Such is the case for many continuous processes, where the system dynamics render short prediction horizons adequate, or where relatively simple, possibly on-line adapted, low-order, input-output models yield accurate predictions through integration. However, conventional methods cannot yield reliable predictions for processes that are poorly modelled and require long prediction horizons for good control performance. Batch processes especially belong in this category, due to their high-order time-varying characteristics. Their changing nonlinear behaviour allows only inadequate state-space modelling from first principles, and the significant long-range future effect of their current manipulated input dictates the use of long prediction horizons for satisfactory control. This paper presents a reliable strategy for long-range prediction in poorly-modelled systems. A state-space-based combination of two neural networks in series is employed to predict system outputs and unmeasured states over a long horizon into the future. A simulation example of a poorly modelled, exothermic, semi-batch reaction system demonstrates the excellent prediction capability of the new structure. The results establish that the new strategy solves a realistic prediction problem where the previously available alternatives fail.