Methodology of data reconciliation and parameter estimation for process systems with multi-operating conditions

Abstract

The economic performance of the real time optimization and process control is influenced by the accuracy of the process model. Data reconciliation and parameter estimation (DRPE) is a crucial technique to obtain reliable process models. In real industrial processes with multi-operating conditions, the effects of contaminated measured data, nonlinear characteristics of model parameters with operating conditions and different types of gross errors increase the difficulty to tune the process models. This paper focuses on the influence of those factors on DRPE problems. A practical DRPE methodology is proposed for the process system with multi-operating conditions to decrease the impact of those factors. The methodology contains the principal component analysis (PCA) based steady state detection, the clustering of multi-operating conditions and the maximum-correntropy estimate based DRPE with multiple data sets. PCA based steady state detection, which is a novel method for steady state detection, is used to choose useful and reliable measured data for DRPE. Clustering partitions the steady state data sets into multi-operating conditions. Maximum-correntropy estimate based DRPE for the data of each operating condition is used to reconcile the measured process data. The proposed methodology is finally applied to a typical real industrial process with multi-operating conditions: the air separation process. The effectiveness of the proposed methodology can be demonstrated by the results of DRPE.