Improving accuracy of refinery optimization by the on-line characterization of crude oil

The simulation of crude oil distillation units, necessary for the optimization and the online performance assessment is very sensitive to the characterization of the feed. The conventional on line characterization technique is based on considering that the crude oil feed is the combination of known crude oils present in a database combined by volume average. Alternatively, a technique based on near-infrared spectroscopy that is able to characterize the petroleum on-line was implemented in order to characterize the crude oil. Prediction models for the main physico-chemical properties have been developed using chemometrics. For Real Time Optimization (RTO) a reference model must be frequently adjusted to fresh operation data. The adaptation of the model is necessary in order to ensure that it represents the system even if there are changes in its constitutive parameters. This feature also ensures that the mathematical optimization problem to be solved in order to obtain the set points of a new operating mode for the process does represent an optimal solution, or at least is close to it. Both the characterization techniques were applied to petroleum samples available for a very long interval of operation (a year) and their result are introduced into the detailed model that is adjusted to the operating data available for this same time range. The accuracy of the predictions shows that the new on-line technique would drive the system to better operating modes.