Controlling the mundane naphtha stabilizer

1 Petrocontrol Leader in inferential control technology 34 East 30th Street, New York, NY 10016 Tel: 212-481-6195 Fax: 212-447-8756 Controlling the mundane naphtha stabilizer By Y. Zak Friedman, PhD Principal Consultant September 2007 Controlling the mundane stabilizer , September 2007, Page 1 Controlling the mundane naphtha stabilizer Part 1 Consider the distillation column of figure 1. It is called stabilizer and has the task of removing LPG from naphtha . Every refinery has at least three of them, and often five or more, one each in every crude unit, reformer, FCC, hydrocracker, coker, isomerization, and possibly other units. Being simple in structure, and with a well defined objective you would think Controlling a stabilizer is a cinch. Why then we encounter so many of them being run with the main DCS controllers in manual?

2 The name stabilizer reflects a historical task of removing enough LPG to reduce the RVP (Reid Vapor Pressure) of naphtha , permitting its storage in floating roof tanks. Economics have changed however, and today the objective is to remove essentially all LPG from the naphtha , subject to a constraint on LPG purity of about 1% C5 contamination. The reasons why operating targets have changed vary from unit to unit, and usually have to do with RVP and alkylation economics. On FCC gas plants it is of value to remove butane from naphtha because olefinic FCC butane is one of the alkylation ingredients. On saturated gas plants we try to maximize isobutane recovery because it is the other alkylation ingredient. That calls for removal of all butane and processing it in a deisobutanizer tower. Furthermore, stabilized virgin naphtha is split into light and heavy naphtha , and light naphtha is processed in an isomerization unit.

3 LPG is problematic in the isomerization unit and should be removed from the feed. Even where these considerations do not apply, RVP specifications have tightened, and gasoline blending is easier when the components have low RVP. In short, today a more appropriate name for the column of figure 1 would be debutanizer , but that is in theory. Most debutanizers I have looked at leave about 2% butane in naphtha and their top C5 contamination is less than IE, we still operate them as stabilizers, losing money by not recovering all of the butane. How much money? Take typical numbers, 2% LPG in naphtha and 200 M3/Hr of naphtha (that s the approximate naphtha production of a 150,000 BPD crude unit), and a cost penalty for leaving LPG in naphtha of $50/M3. If those numbers are realistic for your refinery then the penalty for leaving LPG in naphtha adds up to $1,600,000 annually. How is a debutanizer to be controlled to maximize removal of LGP from naphtha ?

4 Figure 1 suggests a rectifying section tray temperature controller, several trays from the top, manipulating reflux, while the reboiler heat duty is set manually by the operator. The alternative to this structure is actually more common: a stripping section tray temperature controller manipulating the reboiler. Which temperature controller is better a rectifying tray or stripping tray temperature controller? The stripping section tray temperature can be viewed as an approximate inference of C4 in naphtha . If the naphtha cutpoint, column pressure and reflux ratio are approximately constant, the stripping tray temperature defines naphtha C4 content on that tray. Thus, stripping tray temperature control is very appropriate if the operational objective is to leave 2% C4 in naphtha . The tray would contain approximately 20% C4, and at constant operating conditions the ratio between tray composition and bottom composition does not very much.

5 But if the objective is to minimize C4 in naphtha , and we could expect Controlling the mundane stabilizer , September 2007, Page 2 values of C4 in the debutanizer bottom, then the stripping tray temperature loses its ability to infer bottom C4. Instead of bottom C4 it starts inferring the quantity of C5 in naphtha , perhaps of interest to process engineers but of no relevance to the control objective. Suppose we have a stripping section tray temperature controller and a rectifying section temperature indicator, should we restructure the DCS control as figure 1 suggests? The main problem with rectifying section tray temperature control is that it does not provide a very precise inference of C5 in LPG. The debutanizer feed typically comes from an overhead condenser on an upstream fractionator, where the LPG composition and C2 content vary with weather and with the number of fans being turned on the upstream air cooled condenser.

6 IE, the correct setpoint of the rectifying tray temperature controller is not constant. Part 2 Part 1 of this editorial, (see HPJ September issue), has discussed a debutanizer control strategy as shown in figure 1. Given that feed composition varies with ambient conditions one could argue that unless you have advanced control with a more precise inference of C5 in LPG Controlling rectifying section tray temperature is not a good idea. I have further investigated this issue using data from a crude unit stabilizer and Petrocontrol s GDS (generalized distillation shortcut) inferential package. The results are shown in figure 2. The figure first compares lab values of C5 in LPG against the model to illustrate the validity of this investigation. The inferential blue line model is in good agreement against lab results for a period of two month without any bias change. The inference oscillates at a frequency of 24 hours, obviously due to day night ambient temperature differences, reflecting LPG composition change with weather as discussed above.

7 And while the general correlation of purity with tray temperature is clear enough, sometimes even at a constant temperature the LPG purity can drift. This last effect cannot be blamed on pressure variations. Pressure was constant throughout the period. It must be due to non-constant reflux ratio, and crude to crude LPG composition variation. Did operator actions of changing tray temperature controller setpoint make sense? At the beginning of the period of figure 2 we see the LPG impurity trending up, and operators responded incorrectly by increasing tray temperature from to 71 C. Still, 71 C was not a bad overall decision and for about three days the C5 in LPG was kept at Then, in a series of changes over about one month operators set the tray temperature gradually lower, down to 70 C. No one noticed that it was a money loosing operation, even though the lab values of C5 in LPG were quite low.

8 Products were within specification and there were no complaints. On day 40 of the period the tray temperature went up one degree, to 71 C. Someone must have finally noticed LPG going to the wrong place. The temperature then continued to be nudged up gradually to 72 degrees. The moral of this story is simple. If a debutanizer should be operated at 1% C5 in LPG but is operated at , then that difference of is translated to a much higher LPG Controlling the mundane stabilizer , September 2007, Page 3 yield penalty, not to mention other inconveniences of too much C4 in naphtha , and that is a strong economic incentive to implement APC on the column. With or without APC, the DCS control configuration should be as shown in figure 1, with a rectifying section tray temperature controller. Without APC one would need frequent lab support, plus frequent operator training, and then it would be possible to keep the LPG C5 content at an average With the addition of closed loop reliable inference model that average could be brought up to or 1%, because besides being more precise than the operator, APC would also eliminate the day night cycles.

9 Controlling the mundane stabilizer , September 2007, Page 4 Figure 1. stabilizer control configuration \ LC FC LC TI PI PC FC TC FC Heat source FI FC LC FC Controlling the mundane stabilizer , September 2007, Page 5 Figure 2. C5 in LPG inferential model versus lab values trend monthC5 in LPG %vol5557596163656769717375 Tray 7 temperature, deg CC5_in_LPG_ModelC5_in_LPG_LabTray 7 temperature