SIMPLE TECHNIQUES REDUCE COKE INVENTORY, SALES MEASUREMENT ERRORS

Wilson A. ThorpePetroleum Loss Prevention Corp. Chester N.J.
Srini SivaramanExxon Research & Engineering Co. Florham Park, N.J.

The uncertainty accompanying open coke pile inventory measurements can cause wide variations in the oil loss performance reported by a refinery.

The combination of large, inherent measurement errors and large standing pile inventories can introduce significant distortions into the refinery material balance each time an adjustment is made to the book inventory.

Guidelines for selecting measurement techniques and frequency can help refiners minimize these distortions and improve the accuracy of both short and long-term reported oil loss performance.

BACKGROUND

Petroleum coke produced from delayed and fluid cokers is frequently stored in open piles. Periodic physical surveys are conducted to determine the estimated total weight of the piles based on surveyed volumes and estimated average dry density.

Upon completion of these surveys, adjustments are made to the refinery inventory accounting. These adjustments reconcile the current book inventory with the survey estimates.

Contrary to popular belief, more frequent measurements and book adjustments are not necessarily the optimum solution for minimizing the distortions resulting from open coke pile inventory errors.

An evaluation of the distortions introduced into the refinery material balance by coke Inventory measurement errors is essential in understanding and minimizing these errors. Also, following a set of recommendations will help reduce the impact of these errors on both short-term and long-term refinery oil loss performance.

It should be noted, however, that this article deals strictly with the measurement of coke piles and does not address the physical losses of coke inventory attributable to wind losses. The overall impact of wind losses on the refinery loss index is estimated to be less than 0.02%.'

MONTHLY LOSS INDEX

Coke pile inventories are not measured on a monthly basis because of the cost and difficulty of performing a physical survey. Refineries, therefore, typically use an indirect calculation to determine the monthly change in pile inventory.

The coke change-in-inventory is calculated by subtracting measured coke sales from production estimates. The total uncertainty (2 oci) introduced into the monthly refiner-N, loss index by this calculation can be expressed in simplified form as shown in Equation 1 (see equations and nomenclature).

This simplified equation assumes that the systematic errors generated by the production determination and sales measurements dominate the monthly uncertainty calculation to such a degree that random errors can be disregarded. This assumption is consistent with the authors' experience.

Based on available industry data for each measurement method, the major contributors to error, and their typical combined systematic error percentages, are:

• For volume and dry density estimate methods, 2op = 5%
• For vessel drafts and water density samples, 2oms 5%
• For weigh scale and moisture estimates, 2ots 1%.

Note that the uncertainty of weigh scales is quite small compared to the other measurement methods. The greatest contributors to uncertainty in most weigh scale transactions are the sampling practices used to determine the moisture content estimate.

Applying the typical coke measurement/estimate error percentages to Equation 1 yields Equation 2. The importance of using weigh scales to reduce sale measurement uncertainty and improve the accuracy of the change-in-inventory calculation is even more apparent in Equation 2.

Equation 2 can be used to determine the potential variability in the monthly reported refinery oil loss index caused by changes in the calculated coke inventory.

Table 1 presents the results calculated using Equation 2 for four different sale scenarios, when coke production equals 6% of refinery input. Cases 1 and 4 show that the production estimate errors dominate the calculation. The weigh scale introduces little additional uncertainty in Case 4, as compared to Case 1 (no coke sales).

In Case 2, marine sales of the entire production produces the worst case scenario because the large errors generated by both the marine measurements and production estimates are included in the change-in-inventory calculation. The situation is moderated in Case 3 because of the use of truck weigh scales for 50% of the sales.

It is important to note that the total uncertainty introduced into the oil loss determination by coke operations is actually higher than the values shown in Table 1. Table 1 shows the effect of coke change-in-inventory only, but in reality, the coke sales errors enter the material balance twice-once through the change-in-inventory calculation and again through the sales ledger.

LONG-TERM LESS INDEX

To reconcile the long-term calculated inventory, periodic (e.g., quarterly, semiannual, annual) physical surveys typically are conducted on the coke piles. The volumes of the piles are determined using survey techniques.

The average coke density is computed from a number of samples gathered during the physical survey. Based on these data, the total weight is calculated. Subsequently, the current book inventory is adjusted to reflect the results of the physical survey.

The adjustment to the coke inventory contains the uncertainty of the physical survey method. In fact, a very large error can be introduced into the book inventory, even if the adjustment is small. This is true because the physical survey error is proportional to the weight of the total storage pile.

The effect of the survey uncertainty (2osi) on the long-term refinery oil loss index may be expressed by Equation 3.

The overall uncertainty of the physical survey method of determining storage pile weight is estimated to be about 5-10%. Applying a typical error value of 7% to Equation 3, the impact on long-term refinery loss can be computed for varying survey frequencies and storage pile inventories.

The results of this analysis are presented in Table 2.

The authors believe that the target uncertainty should be 0.2% or less on long-term loss, for most refinery material balances. Hence, the boldface portion of the table represents what the authors consider acceptable uncertainties for a given Winv/Wiav ratio.

From Table 2, it is apparent that the levels of uncertainty introduced into the long-term loss index can be very significant. To reduce the impact of the high uncertainty levels, it is necessary to extend the time periods between the survey adjustments.

For example, a refinery currently performing a semi-annual survey and having a coke inventory equal to 40% of the refinery monthly throughput (Winv/Wi = 0.4) should extend the period between physical surveys to 12 months or longer.

The table also shows some interesting trends and relationships. It is evident that, for a fixed inventory, the overall uncertainty is reduced by one half for every doubling of time between surveys.

In addition, the table shows that, as coke pile inventories double, the overall uncertainty also doubles for a given period between surveys.

This trend has important and undesirable implications for refineries that are expanding their coke storage. Their ability to report and sustain accurate long-term loss indexes is dramatically compromised with increasing pile inventories, unless they choose longer periods between surveys.

This poses a dilemma because periods greater than 2 years (or even less) are usually unacceptable from audit and management standpoints.

MINIMIZING UNCERTAINTY

There are only a few options available for limiting or reducing the impact of coke pile inventory uncertainty on short-term and long-term refinery oil loss. Although a refiner's ability to implement the following options may be limited, they provide various degrees of control over the problem:

• Minimize or eliminate coke piles to the extent possible.
• Utilize weigh scales for all sales.
• Select time periods between physical surveys to limit the impact of uncertainty to 0.2% or less.
• Improve production measurements/estimates.
• Minimize water injection/carry-over on storage piles to improve dry density determinations.
• minimize physical survey errors.

ACKNOWLEDGMENT

The authors sincerely thank Esso Sapa (Argentina) refinery management for making facilities available for the study. Special acknowledgments go to A. P. Silva, C. Pollini, and H. R. Nicola for their help.

REFERENCE

1. Compilation of Past Practices and Interpretations by EPA Region VII on Air Quality Monitoring, U.S. Environmental Protection Agency, December 1979.

Copyright 1993 Oil & Gas Journal. All Rights Reserved.