Jul-2017

APC in a mild hydrocracker fractionator

Advanced process control of a mild hydrocracker can save millions of dollars annually by maximising kerosene and/or diesel

ARNOLD KLEINE BÃœNING, Bayernoil
STEPHEN FINLAYSON, AMT
Y ZAK FRIEDMAN, Petrocontrol

Article Summary

Bayernoil Neustadt refinery is built for a high yield of diesel and jet fuel, equipped with a mild hydrocracker (MHC), hydrogen plant, plus sulphur removal and recovery units (see Figure 1). Among many units, the ones directly associated with the MHC are:
• Three crude distillation units (CDU)
• Two vacuum distillation units (VDU) taking feed from the CDUs
• An MHC taking feed from the VDUs
• Two fluid catalytic crackers (FCC) taking unconverted oil (UCO) from the MHC.

Figure 2 shows the MHC configuration with reaction and separation sections. Reactor effluent is a wide boiling range material that must be separated into narrow cut products. First, light naphtha is separated out in a stripper. Stripper bottoms are taken into the fractionator, which separates naphtha at the top, kerosene and diesel as side draws, and heavy UCO at the bottom. Naphtha is further separated into light and heavy naphtha. Product values differ significantly and specifications vary by season and type of operation. Kerosene is sometimes produced as jet fuel and at other times blended into diesel.

Advanced process control (APC) of a typical MHC process can potentially recover benefits in the order of millions of dollars per annum by maximising kerosene and/or diesel. The capture of these benefits is contingent on reliable control of product qualities at targets while nudging the unit against physical constraints.

Product qualities are typically not measured but inferred, whereas such inferences are the ‘Achilles heel’ of our industry. It takes knowledge and skill to obtain reliable product quality inferences. Neustadt is actually blessed with the ability to maintain on-stream analysers, an art that by and large has been abandoned by our industry. Even so, when compared against inference models, analyser dead time in the order of 90 minutes would negatively affect APC control performance. Furthermore, even high reliability analysers occasionally give erroneous readings, which may cause the multivariable predictive control (MVPC) to drive products off specification.

Desiring to control product qualities precisely, Neustadt has chosen the Petrocontrol/AMT generalised cutpoint calculation (GCC) inferential package. GCC employs first principles calculation methods to estimate a fractionator feed true boiling point (TBP) curve, before inferring cutpoints and other product properties. GCC was originally invented to deal with crude fractionators, and the theory and performance have been documented in several papers.^{6,8,10,11,12,13,14} Adaptation of GCC to other types of refinery hydrocarbon fractionators has also been addressed in the literature.^{1, 2, 3, 4, 5, 7, 9, 10} This is the first article detailing the application of GCC to MHC fractionators.

GCC features
GCC theory has been documented in many publications and hence this article only describes the main concepts. Fractionator temperature measurements reflect tray composition at vapour/liquid equilibrium at a given hydrocarbon partial pressure. The opposite is also possible: estimate tray compositions from column conditions. GCC begins with estimating partial pressure. That is a function of total pressure (measured), steam flows (measured), and vapour traffic (calculated from measurements and heat balances). Once partial pressures are estimated, GCC corrects temperature readings from actual conditions to atmospheric pressure. Pressure corrected temperature (PCT) formulae are well known. These PCT temperatures, corrected to atmospheric pressure, now reflect the bubble points or dew points of products, sometimes a mix of products, whereas dew points and bubble points are functions of product cutpoint temperatures. Calculation of those cutpoints becomes a simple arithmetic GCC procedure. 

From cutpoints and yields, GCC next constructs the TBP curve of fractionator feed material. A TBP curve is convenient because it describes ideal fractionation of the feed. An example of a TBP curve is illustrated in Figure 3. The heavy continuous line is column feed boiling curve, and the cutpoints define ideal product yields. Three products are shown: naphtha, kerosene and diesel. The fourth cut is called overflash and is not a real product, but material to be evaporated in the flash zone, then refluxed back down to the bottoms. Overflash is an important operating parameter in that it determines the separation between diesel and UCO. Many fractionators are designed with special flow meters attempting to measure the overflash. However by and large those measurements are not successful and the ability of GCC to infer overflash is an asset in itself.

The red lines of Figure 3 show typical product TBP curves. Had we experienced ideal fractionation, product TBP curves would coincide with the feed curve. The heavy and light ‘tails’ on product curves are due to imperfect fractionation causing boiling range inter-mixing.

Cutpoint is a theoretical concept used to estimate product properties, for example product 90% point. GCC must be validated against lab tests, which use an ASTM D86 apparatus, a simple distillation machine, but not remotely a TBP machine. The GCC D86 prediction is a function of both front and back cutpoints, as well as internal reflux. Internal reflux is typically of secondary importance in the GCC D86 boiling point estimation. The following example shows the form of GCC diesel 90% point calculation based on cutpoints and internal reflux:

Diesel 90% point = K1 * KCP + K2 * DCP + K3 * [FDSL / (FDSL + FDSLIR)]

K1, K2 = known coefficients
KCP = kerosene cutpoint
DCP = diesel cutpoint
FDSL = volume flow of diesel
FDSLIR = internal reflux below the diesel draw tray

Shown in Figure 4, the term [FDSL / (FDSL + FDSLIR)] is a number between 0 and 1, 1 when there is no reflux. K3 can be viewed as the heavy tail penalty for no reflux. The penalty function decays quickly as internal reflux increases, and beyond an internal reflux ratio of 1:1 further improvement in separation is small. Well designed fractionators operate at about 1:1.