Jan-2022

Hydroprocessing routes for flexible residue upgrading

Efficient and flexible resid processing options maintain profitability and reduce risk when price volatility is high.

Kevin Knob and Goutam Biswas, Chevron Lummus Global
Steven X. Song, Chevron U.S.A. Inc.

Article Summary

Starting 1 January 2020, the international community began enforcing the 0.5 wt% sulphur limit on fuel oil used by ocean-going vessels as specified in the regulation known as IMO 2020. Predicted price spreads of up to $300/t for very low sulphur fuel oil (VLSFO) relative to high sulphur fuel oil (HSFO) were reached briefly in early 2020 but then largely disappeared in the wake of the COVID-19 pandemic. As the world recovers from the pandemic-induced economic slowdown, current forecasts are predicting gradually improved spreads from the average of $50/t in 2020 up to $100/t in the coming years.1 Although the price of HSFO has stabilised compared to the beginning of 2020 due to various factors (pandemic-induced imbalance in demand for transportation fuel, reduced throughput, HSFO purchased as coker feed), the bunker market has seen the expected significant drop in demand for HSFO compared to 2019. As much as 75% of HSFO has disappeared from the marine bunker fuel market; some of the surplus HSFO will need to be upgraded to higher value products for better economic returns.

The key takeaway for refiners is the importance of having efficient and flexible processing options to maintain profitability and reduce risk in an environment of rapidly changing and unpredictable price volatility. Decisions around processing the heavy distillation residues known as ‘resid’ play an especially important role.

There are three pathways for resid disposals: 1. do nothing, 2. reject carbon, and 3. add hydrogen. The ‘do nothing’ approach — typical of a low complexity topping-reforming or hydroskimming type refinery — has the lowest capital and operating costs but the least flexibility. A refiner without resid upgrading capabilities must either purchase more expensive light crude to meet new IMO requirements for VLSFO or sell low-value HSFO. ‘Reject carbon’ processes include resid fluid catalytic cracking (RFCC), delayed coking (DCU), and solvent deasphalting (SDA) units. RFCC requires high-quality resid with low metals and high hydrogen content. Coking and SDA processes can upgrade more difficult and heavier, less expensive crudes with higher contaminant levels, however they require further upgrading to make finished products and produce various amounts of lower value byproducts: heavy cycle oils and decant oil from RFCC; petcoke from the DCU; asphalt pitch from SDA. The ‘add hydrogen’ approach employs resid hydroprocessing technologies such as those offered by Chevron Lummus Global (CLG), widely accepted by the industry as the preferred choice to maximise high-value product yields.

CLG hydroprocessing technology options for maximising product value by combining carbon rejecting and hydrogen adding processes are well covered in a recent review.2 In this article we discuss three specific applications of CLG’s fixed bed resid hydrodesulphurisation (RDS) technology. These units are designated as ARDS or VRDS depending upon whether the predominant feed is atmospheric (AR) or vacuum (VR) resid. The first case involves the installation of a new fixed bed VRDS unit to produce on specification VLSFO product directly from VR with an option to produce RFCC feed by varying the ratio of AR to VR. In a second case, resid feed capacity and run length were improved at low incremental capital cost by adding an upflow reactor (UFR) to an existing fixed bed RDS unit. Finally, among the wide range of retrofit options available to owners of existing LC-Fining units is the newly developed LC-LSFO process in which a fixed bed RDS system is employed to upgrade unconverted oil (UCO) to VLSFO.

Case 1:
SK Energy grassroots VRDS unit
In 2017, SK Energy, South Korea’s leading refiner, approached CLG with various potential ideas for upgrading an anticipated excess of VR at the Ulsan refinery. SK has enjoyed a history of success with CLG RDS technology and has added RDS capacity over the years as the refinery has expanded. As of 2017, the company was operating three RDS units at a peak capacity of 200000 b/d of resid feed, compared to an original nameplate capacity of 170 000 b/d.

A 2017 process study by CLG supported SK’s decision to build a new grassroots facility to process up to 30 000 b/d of additional VR. The #2 VRDS unit is configured with two parallel trains of multiple downflow fixed bed reactors. A simplified PFD of the VRDS unit reaction and low-pressure separation section is shown in Figure 1. The composition of SK’s design VR feed (see Table 1) was challenging but well within the range of previous designs when combined in a suitable ratio with AR. The final design feed was composed of 75% VR and 25% AR.

SK acted quickly through an affiliated engineering company to complete the detailed design and execute the project. The new VRDS unit, including expansion of supporting units such as sulphur recovery and hydrogen manufacturing, was up and running at full capacity within about two and a half years of initiating work on the CLG Engineering Design Package and just three months after the new IMO 2020 regulations went into effect. The total project investment of $837 million is expected to return a gross profit of between $167 million and $250 million annually.³

SK elected to install Isomix-e flow distribution nozzles in the VRDS unit reactors. These nozzles create a conical spray pattern that overlaps with adjacent nozzles, resulting in complete wetting at the top of the bed and reactor wall. Fixed bed RDS units had been slow to adopt this nozzle design over concerns of fouling inside the relatively smaller internal clearances. An initial successful commercial application proved that fouling tendency is mitigated by the swirling motion of liquid inside the nozzle and high gas/liquid two-phase velocity through the nozzle throat.

CLG developed the Isomix-e internals over 50 years of experience designing hydroprocessing reactors. All tray assemblies use wedge pin fasteners for easy maintenance. The improvement of flow distribution quality over conventional chimney distributor elements is illustrated in Figure 2. Performance of conventional chimneys vs Isomix-e flow nozzles is compared under identical conditions of liquid and gas flow rates with distribution device B placed 6 mm higher than A to simulate tray out-of-levelness. Test results demonstrate greatly improved tolerance to a typical deviation in tray levelness. In the Isomix-e nozzle, in contrast to liquid head-controlled flow in a conventional chimney, the primary driving force to draw liquid into the nozzle is a low pressure region created as gas flow accelerates through the upper orifice in the nozzle.

But uniform wetting at the top of the catalyst bed is not enough; long experience with trickle bed reactor hydrodynamic behaviour has resulted in design guidelines that specify a favourable range of liquid mass flux and two-phase flow pressure drop per unit length for successful operation. In this case, feed viscosity is controlled to meet these requirements by blending AR into the feed. Maximum VR:AR ratio was also constrained by the target run length of 340 days, discussed below.

The additional constraint arises from considerations of desulphurised resid product stability. As processing severity increases, reactions that tend to reduce the solubility of the asphaltenes are accelerated, increasing the tendency for asphaltenes to precipitate and form sediment. In addition to increasing the total sediment potential (TSP) of the product (maximum value 1000 ppm), deposits of dry sediment can cause high pressure drops in the catalyst bed and fouling of downstream equipment. Sediment forming tendency correlates well with the conversion of material boiling above 1000°F (538°C). Therefore, a limit is placed on 1000°F+ conversion, which translates to an end-of-run (EOR) temperature constraint. Calculations using CLG’s proprietary kinetic ageing model determined that it would not be feasible to attain a target run length of 340 days while processing 100% VR feed. In addition to having an unfavourable impact on project economics, simply adding catalyst would have resulted in a risky design with respect to VLSFO product instability at end-of-run. A maximum ratio of 3:1 VR:AR was adopted for the design basis to meet both feed viscosity and catalyst ageing rate constraints to achieve the target run length of 340 days.