Robust parameter estimation and output prediction for reactive carrier-load nonlinear dynamic networks

In this paper an extension of on-line model simplification technique for a class of networked systems, namely reactive carrier-load nonlinear dynamic networked system (RCLNDNS), kept within point-parametric model (PPM) framework is addressed. The PPM is utilised to acquire a piece wise constant time-varying parameter linear structure for the RCLNDNS suitable for the on-line one step ahead prediction that may be applied to monitoring and model predictive control purposes. The advantageous structure of PPM offers a considerable simplification of those algorithms in terms of computational burden in comparison to the nonlinear structures. Moreover, the availability of technical tools makes the analysis of the final form algorithms straightforward. An application to water quality estimation/prediction in drinking water distribution system illustrates the technology.