Statistical method used to estimate refining tankage

Feb. 28, 2000
A statistical risk simulator program can determine adequate tank-storage capacity for a grassroots refinery or oil terminal, removing the risk of building too much capacity.

A statistical risk simulator program can determine adequate tank-storage capacity for a grassroots refinery or oil terminal, removing the risk of building too much capacity.

One such program, developed by ABB Lummus and used in a project for a grassroots Taiwan refinery, reduced the proposed refinery tankage capacity by 20%-a savings of $58 million in capital.

Proper estimation of needed storage capacity results in significant savings in capital expenditures for building new tanks and lines. It also increases the effective throughput of an existing facility by maximizing the use of existing facilities.

Wildly swinging oil prices have a dramatic impact on refinery and bulk terminal profitability because they tie up an enormous amount of oil and capital in inventory. In 1996, EIG Inc., Washington, DC, estimated that there were about 7 billion bbl of worldwide oil inventory, which is equivalent to 2.5 years of Saudi Aramco production.1

Despite this huge inventory, there are still times when refineries almost run out of feed, are in danger of overflowing a tank, or are late with product shipments. Each of these situations has an economic impact:

  • The cost of oil stocks tied up in unnecessary inventory.
  • The cost of delayed shipments because of insufficient stocks.
  • The cost of running process units at less than the optimum capacity.
  • The cost of unnecessary capital investment in building additional tanks, lines, pumps, and loading or unloading facilities.

Options for sizing storage

What is adequate tank storage capacity?

A literature search uncovered old (ca. 1959 and 1973) "rules of thumb" that were out of date and so of limited use.2 3

It also uncovered a relatively new technique of statistical-risk, tankage-inventory modeling.4

The search failed to show any commercially available tools to perform the modeling.

Thus, ABB Lummus developed its own statistical risk simulator program that models the refinery off sites tank-farm operation and accounts for such common uncertainties as a ship delay, lifting schedules, hurricanes, power failures, and unit-stream factor. The program also accommodates predictable events, such as unit or plant turnarounds and scheduled tank maintenance.

Program runs for a new grassroots Taiwan refinery indicated a reduction in the refinery proposed tankage by about 20%, with a 95% confidence. This simulation saved the operator $58 million in capital investment. Besides new refineries, these simulations also apply to old refineries and old and new oil terminals.

Similar analyses for existing tank farms showed similar orders of magnitude-reduction possibilities.4 Plant designers can obtain additional 2-5% savings through use of software for in-line blending, oil movements and storage automation, multitime period off sites planning and scheduling, and Internet-based financial instruments. These are documented elsewhere.5 6

Savings are in terms of lower tankage requirements, lower inventories, and faster shipments.

The initial tool for the Taiwan refinery was programmed in Microsoft Excel. It modeled uncertainties, such as ship delays, by using Excel's random number generator to perform a Monte Carlo simulation. Excel's built-in solver linear program optimized the tankage.

Currently, ABB is developing a more-powerful tool with a commercial software package with built-in Monte Carlo simulation coupled with a nonlinear genetic algorithm optimizer.

The Monte Carlo simulation imposes a variability on receipts, shipments, shutdowns, and flow rates. The variability is computed based on preselected probabilities of each occurrence.

Off sites statistical model

The model has six parts:

  1. A refinery steady-state simulator, which shows the flows of major process units and feed, intermediate and finished product tankage, and their interconnectivity.
  2. Input-output disturbances (for example, delivery of crude by ship) and their estimated statistics.
  3. An initial state from which to start the simulation.
  4. A parametric simulator that changes the receipts, shipments, operation of the units, and flow perturbations using a Monte Carlo simulation for the disturbance statistics.
  5. A user-interface panel, which allows the user to input initial data or change the simulation data in a user-friendly fashion with point and click "buttons."
    There are buttons for inputting crude deliveries, chemicals or intermediate products, product shipments, process-unit flow rates, unit-yield vectors, unit interrupts (such as a scheduled shutdown), and initialization values.
  6. Tankage-estimate results.

The model runs using a daily basis time granularity and executes 360 days of operation per pass, with the option of iterating in successive 360-day periods.

The results are either the tankage (volumes) or inventory estimates.

Click here to enlarge image

The refinery steady-state model is a simple statistical material-balance model using the input-output tankage segregated by material types, the process units with their yield vectors and design flow rates, and their interconnectivity network (Fig. 1). Line fills are not taken into account.

The model takes several factors into consideration:3

  • Crude-delivery methods (such as tankers, pipeline, railcars).
  • Product shipment methods (such as tankers, pipeline, railcars).
  • Refining complexity (more storage needed for more products).
  • Seasonal demand, which requires tankage to accumulate products throughout the year.
  • Turnarounds, catalyst regenerations, and other scheduled maintenance.
  • Blocked operations using multiple feedstocks.
  • Availability of in-line blending.
  • Optimal levels of running and closed tanks to supply sales and address emergencies.
  • Availability of rental tank storage.

The input-output disturbances considered are either of a statistical nature (for example, delays in ship arrival time or shipment times, lightning storms, and unscheduled unit downtimes), or deterministic (for example, frequency and duration of scheduled unit turnarounds, typical crude-settling times, typical shipment parcel size for each product, and tank minimum or maximum volumes).

The initial conditions (as for starting volumes) are supplied either by the refinery, using historical data, or by use of industry-wide best practice default values.

The tankage estimate results on the computer screen shows the recommended volumes sorted by material: crude, naphtha, MTBE (methyl tertiary butyl ether), etc.

An additional screen shows the economics involved. For instance, current values of low vapor pressure storage capacity are about $60/bbl and butane or propane storage is $90-120/bbl. These costs include tanks, piping, transfer pumps, dikes, fire-protection equipment, and tank car or truck loading facilities.

Additional costs account for real estate, depending on the country and specific circumstances.7

Click here to enlarge image

Table 1 shows an example of the model-results screen.

Analysis of results

Work by Nelson quantified tankage requirements based on refinery complexity.2 3 The Nelson 1959 formula based on historical data was:

S = 1.75 P + 10

where:

S = Barrel of product storage per daily barrel of crude oil processing capacity

P = Number of products produced by the plant.

For simple plants, P=8 and S=24. For complex plants, P=36 and S=73.

The Nelson 1973 formula based on historical data was:

S = C + 4.5 P + 11

where:

C = Barrel of crude storage per daily barrel of crude-oil processing capacity.

C ranged from 4 for refineries supplied by pipelines to 55 for refineries served by tankers. S and P were the same as in Equation 1.

The authors analyzed the amount of tankage required for the proposed grassroots Taiwan refinery case using four different methods: Nelson 1959, Nelson 1973, Gary 1994,4 and ABB's statistical model. The proposed refinery was to be a complex refinery that received its crude by tanker.

Click here to enlarge image

Table 2 shows that the statistical model estimated much lower storage requirements than any of the other three methods. It reduced proposed storage capacity by 415,000 cu m, which corresponds to a construction savings of $58 million.

The statistical model is conservative also. It takes into account refinery owner constraints such as crude unit daily throughput, assuming a 100% stream factor. It superimposes a variability (number of days shut down) together with an associated statistical distribution (for example, a normal distribution with 95% confidence means that it will be shut down 18 days/year, 67% confidence means that it will be down 36 days/year).

The nonstatistical methods are empirical methods based on historical data over 30-40 years. Oil prices and availability have changed since the oil embargoes of the 1970s. Only the statistical model accounts for current programs to increase profitability, such as just-in-time manufacturing and enterprise resource planning.

Acknowledgment

The authors thank Joseph Campagnolo for his assistance in implementing and running the model and test cases.

References

  1. EIG Inc., "How Much Oil Inventory Is Enough?," EIG's Oil Market Intelligence, 1997, p. 7.
  2. Nelson, W.L., "What is adequate storage capacity?," OGJ, Sept. 7, 1959, p. 184.
  3. Nelson, W.L., "How Much Refinery Storage?" OGJ, Apr. 23, 1973, p. 88.
  4. Price, J.A., Joffrion V., and Blankenship, J.L., "A stepwise approach to offsites tankage rationalization," Petrotech98 Conference Proceedings, September 1998, Vol. 2, pp. 705-14.
  5. Barsamian, J.A., and Czech, R.S., "Intelligent Oil Movements Automation Enhances Refinery Profitability," Petrotech98 Conference Proceedings, September 1998, Vol. 2, pp. 1,017-31.
  6. Intille, G.M., "How refinery inventories threaten profitability," Hydrocarbon Processing, July 1986, pp. 90-100.
  7. Gary, J.H., and Handwerk, G.E., Petroleum Refining-Technology and Economics, New York: Marcel Dekker Inc., 1994; pp. 354-55.

The Authors

Click here to enlarge image

Ara Barsamian has more than 30 years of experience in blending, oil movements, and off sites automation. Currently, he is president of Refinery Automation Institute LLC. Previously, he was vice-president, Refinery Automation, for ABB Simcon and president of 3X Corp.

Barsamian has also been group head and section head responsible for digital control technology and advanced computer control projects and off sites automation research and development at Exxon Research & Engineering Co. He holds a BS and MS in electrical engineering from City University of New York.

Click here to enlarge image

Richard Whitehead has 36 years of experience in project development, project management, and process planning. Currently, he is senior project manager with ABB Lummus Global. Previously, he worked at Rhone Poulenc, KTI-Fish Engineering, Nuclear Power Services, and Chem Systems Inc. He holds a BSc and PhD from the University of Birmingham, UK.