Oct-2005

Deep-cut vacuum unit design

Process modelling and equipment know-how are needed to design deep-cut vacuum units that operate reliably at high temperature and low pressure

Tony Barletta and Scott W Golden, Process Consulting Services

Article Summary

With high crude prices and historically high light-sour heavy differential prices, processing lower-cost heavy crude oil from Canada, Mexico, Venezuela and the deep water of the Gulf of Mexico looks very attractive. However, when processing these types of crudes, they produce 20–35% of the whole crude as vacuum tower bottoms (VTB). As a result, vacuum unit investment is needed; otherwise, the heavy vacuum gas oil (HVGO) product TBP cutpoint will drop as crude gravity decreases. Reducing the crude API gravity by six to eight numbers has reduced the HVGO product TBP cutpoint by 100°F (56°C) without any design and operating changes. Feeding gas oil to the coker unit consumes capacity, converts a portion of the vacuum gas oil (VGO) in the coker charge to coke and ultimately reduces the quality of the feed to the FCCU. If only 10 wt% of the VGO feeding the coker unit is converted to coke, the profit loss is huge. Coke has virtually no value, while FCCUs convert each barrel of feed into approximately 1.1 barrels of liquid product.

Vacuum unit operation
Processing heavier crudes reliably while meeting 1050°F (565°C)-plus TBP cutpoints requires a high heater outlet temperature (790–800°F; 421–427°C), a low transfer line pressure drop (140 mmHg), a minimum column flash zone pressure (15–20 mmHg absolute), steam use in the heater to minimise oil residence time and VTB steam stripping to raise the HVGO product yield. Unit design is challenging, because high temperature can cause coking in the heater and vacuum column, and maintaining a low pressure demands reliable ejector system performance to sustain low flash zone pressure even during the summer months. Common ejector problems, such as breaking in the summer time, can easily raise the operating pressure by 10 mmHg or more, significantly reducing the HVGO product cutpoint and increasing the coker charge rate. When switching to heavy crude, it is not unusual to produce less HVGO, make more coker charge, yield HVGO product containing a high microcarbon residue (MCR) and vanadium, and have short run lengths due to rapid coking.

Process and equipment modelling
Twenty-five years ago, a simple process model took weeks to develop, with engineers using punch cards to create input files. These simple models would take several hours to run on large mainframe computers. Today, complicated rigorous models are created using graphical user interfaces with pull-down menus in less than an hour and are run in less than a minute on a desktop or laptop PC. But then, as now, the model results must represent the true performance of the installed equipment. When cutting deep into the VTB, there is little room for error.

Some modelling errors are academic, while others have real economic consequences. In post-revamp evaluation of a vacuum unit processing light North Sea crude oil, there were large differences between model predictions and plant performance. Table 1 shows a comparison of total VGO yield as a percentage of the vacuum unit feed, wash oil flow rate and overflash rate between the process model results and plant data. Actual VGO yield was lower by 3 wt% of feed and overflash was very low. The wash bed plugged with coke in less than six months, because the spray header installed could not provide sufficient wetting of the packing to prevent coke forming. An unscheduled shutdown was needed to replace the packing and wash header. In this case, as in several others, model results and plant data did not match. Process models need to predict real operating performance, and the only way to ensure results match operation is to check them against accurately measured plant data.
When designing deep-cut vacuum units for heavy crudes, it is not unusual for the process engineers’ model to over-predict VGO yields by 2–3 vol% on whole crude and under-predict the wash rate needed to prevent coking by 300%. In one case, the process model predicted 7Mbpd of wash oil flow rate, whereas the actual flow rate required to prevent coking was 21Mbpd. When packing becomes plugged with coke, the HVGO product metals and MCR rapidly increase due to VTB entrainment. In many instances, shutdowns have been required in less than 12 months from start-up. When margins are high, these outages are extremely costly.

Process models need to characterise feed properly, yet vacuum unit feeds consist of thousands of distinct compounds that cannot be identified. Composition is approximated through lumping these molecules into pseudo-components based on TBP and gravity curves. Correct TBP and gravity curves are essential, and no process modelling should be done until accurate feed composition data is developed. Since models have become so easy to develop and run, many users have forgotten the basics.

Conventional models treat the vacuum column as a straightforward fractionator with equilibrium stages (Figure 1). Feed entering the column is assumed to be in equilibrium. Table 1 shows the results of making this assumption. In this case, to meet the model-predicted VGO yield and the wash rate needed to prevent coking, the heater outlet temperature had to be increased by more than 45°F (25°C) to 804°F (429°C). This temperature could only be maintained for a very short period of time, because it caused excessive thermal cracking and would have coked the heater. Assuming that the liquid and vapour entering a vacuum column flash zone are in equilibrium is a critical mistake.

Structuring the model properly requires an understanding of what happens within the equipment system. Vacuum unit feed is pumped through the charge heater, heat is added and pressure decreases as oil flows through the tubes and transfer line to the vacuum column. Pressure profile throughout this system must be estimated using equipment models that can adequately deal with two-phase critical flow. Heater outlet tubes are relatively small, so vapour and liquid are probably in or close enough to equilibrium to make this assumption valid. Since the transfer line consists of large-diameter piping, liquid and vapour separate (Figure 2) in the horizontal section of the transfer line, vapour flows along the top of the pipe and liquid across the bottom. Describing the heat, mass and component balances at all points in the transfer line is not practical; however, estimating it is important. As long as the model approximates what actually happens in the equipment, as opposed to being theoretically perfect, it is a useful tool for process design engineers.

A practical approach to modelling transfer lines and vacuum columns that better predicts yields and other critical operating parameters requires the model to be segmented into a number of operations prior to the vapour entering the column wash section. Using multiple unit operations allows the non-equilibrium nature of the system to be estimated. This technique, although an approximation of the actual non-equilibrium process, more accurately predicts VGO and VTB product yields and better estimates the wash flow rate (Figure 3) needed to avoid coking in the wash section (Figure 4). Operating pressure at the heater outlet and in the first large section of the transfer line must also be estimated. An accurate heater-through-transfer line model is needed that can calculate operating pressure using two-phase critical velocity limits. Both process and rigorous equipment system models must be used when designing vacuum units. It is not possible to separate the design of the transfer line from the heater.