An integrated framework for on-line supervised optimization

Abstract

In this paper, an integrated supervisory framework is envisaged for more robust on-line optimization and exception handling. This system takes advantage of the capabilities of real time evolution algorithm [Sequeira, S., Graells, M., Puigjaner, L. (2002). Real time evolution for Online Optimization of Continuous Processes. Industrial & Engineering Chemistry Research 41, 1815–1825] plus a fault diagnosis system (FDS) for managing abnormal situations. Both systems are part of a supervisory module which is responsible for handling plant incidences (faults and disturbances) by taking the appropriate corrective actions. Thus, a more satisfactory on-line performance is achieved, while reaction to incidence is enhanced by providing combined cause-effect information to plant managers. The implementation of the supervised real time evolution scheme has been performed using Matlab and the commercial simulation package HYSYS.Plant, taking advantage of their communication capabilities (COM technology). The developed fault diagnosis system comprises a detection module based on multivariate statistical techniques and an isolation module that uses an artificial neural network as a pattern classifier. The benefits of this framework are illustrated through a case study consisting in a debutanizer distillation column. The handling of various plant incidences, involving different sources and types of disturbances are considered. Finally, results of the supervised real time evolution (SRTE) are compared with those obtained using the standard real time optimization approach (RTO), showing superior performance in most of the cases.