Plant−Model Mismatch Analysis in Deethyleniser Simulation. 1. Methodology

Abstract

Confidence in the accuracy of a simulation model and by implication the thermodynamic data model(s) used, particularly for separation equipment such as a deethyleniser, requires validation against actual plant operating data. A methodology has been developed that quantifies plantmodel mismatch with a Calculator module appended to a PRO/II (Simulation Sciences, Inc.) Column module to facilitate an assessment of the relative accuracy of alternative simulation model formulations. The methodology is based on a manual stepwise optimization procedure using a preferred formulation of the simulation model with the key adjustable modeling parameter defined as the effective average equilibrium tray efficiency η i eff . The methodology seeks to minimize an Error Objective Function that calculates the sum of the weighted and variance-normalized least-squares errors around the deethyleniser involving as many plant measurements as practical taking into account the wide range of accuracy inherent in the various measurements.

Plant−Model Mismatch Analysis in Deethyleniser Simulation. 2. Application to Data-Model Selection and Tuning

“…To gain some perspective as to what the EOF values indicate, it should be remembered that each component EOF value is equal to the {(Simulated Value -Measured Value)/Standard Deviation} 2 . 7 Therefore, a spread of 0.07 units per variable is equal to a spread of (total EOF/10) ) (0.07) ) 0.25 standard deviations for each significant variable. This is hardly sufficient to differentiate the relative accuracy of the k ij values used in optimized A2 simulation models.…”

“…A brief account of the two-stage data reconciliation procedure is given by Kamphausen and O'Donnell. 7 Details of the data reconciliation procedure and results are available in work by Kamphausen. 12 A summary of selected measurements for the deethyleniser is presented in Tables 1 and 2.…”

“…The methodology that is used is based on a manual stepwise optimization procedure labeled as the "Basic" optimization path as previously described by Kamphausen and O'Donnell. 7 This methodology is an approximation that acknowledges the limitations of the sequential modular nature of the PRO/II 8 simulation program and makes numerous assumptions that have previously been tested through either the use of alternative simulation models and alternative optimization paths or the use of spot simulation checks. The methodology adopted is based on a compromise between accuracy and time constraints.…”

“…The final 6 categories encompassed ethylene, ethane, propylene, propane, butanes, and pentanes plus, where butanes and, in particular, pentanes plus are a large number of components lumped together under the headings 1-butene and 1,3-cyclopentadiene, respectively. A brief account of the two-stage data reconciliation procedure is given by Kamphausen and O'Donnell . Details of the data reconciliation procedure and results are available in work by Kamphausen .…”

“…The major objective in this work then was to apply the methodology developed in a previous paper by Kamphausen and O'Donnell 7 for evaluating the relative accuracy of deethyleniser simulation models in terms of different definitions of the key K-value thermodynamic data model used. A comparative study was initiated where the most relevant commercially available PRO/II 8 EOS data models for K-value generation were to be compared for applicability in deethyleniser simulation, as determined by the relative values of the minimum deethyleniser plant-model mismatch.…”