Data reconciliation and bias estimation in on-line optimisation

The reliability of measured data, which can be subject to both gross and random errors, is of great importance for the monitoring and evaluation of process performance and the determination of control action. This paper assesses bias estimation (as a type of gross error) technique and data reconciliation methods for the detection, estimation and elimination of iases and random errors respectively. It is shown how these methods can be successfully employed within an on line Integrated System Optimisation and Parameter Estimation (ISOPE) scheme for the determination of the process optimum, despite the existence of model-reality differences. The performance of the resulting scheme is demonstrated by application to a two tank CSTR system.