Achieving better liquid measurement accuracy.

Author:Rudroff, Daniel J.

Custody transfer measurement in the oil and gas business has been described many ways. It has been called, "An accuracy in measurement that both the buyers and sellers can agree upon" and "The best that can be achieved to meet the contract conditions." But I prefer to call it, "The search for the truth."

Ever since petroleum has been bought and sold, people have searched for better ways to measure oil and petroleum products on the fly with greater accuracy. One big advancement has been the pipe prover, which compares a known volume between two switches in a pipe to the meter reading. API requires an accuracy of the prover volume of 0.02% when compared to a standard such as NIST traceable Seraphin Cans.

If we want to put 0.02% accuracy into perspective, that is 6.45 teaspoons or a little more than two tablespoons of oil in a single 42 gallon barrel. That is very good measurement even at the worst case. We all strive to exceed the 0.02% required by API. We know and understand the value increased accuracy has to our companies. Today, prover water draws are repeatable to 0.001 and 0.002%

Accuracy of measurement is important when oil is selling at $100/bbl and profits are good, but it is even more important when oil is at $80/ bbl and the margins are fight. One lost barrel becomes a much larger percentage of the profit.

When we are trying to implement a new and better method, we are burdened many times by company and industry standards and the old adage, "This is how we have always done it." But sometimes, when we are faced with a problem, it needs to be solved outside the conventions. I say, "Necessity must become the mother of invention." This is what is happening with bi-directional pipe provers.

By making the calibrated straight, repeatable detector switches, tipping the horizontal launchers, sizing the launchers properly and placing the pressure and temperature transmitters correctly, we are able to improve the bidirectional pipe prover. In some ways it is like the Alfred Hitchcock "Lamb to the Slaughter" TV episode of April 13, 1958. A wife kills her abusive chief of police husband with a frozen leg of lamb, then invites his detective friends for dinner and serves it to them. The famous line from the episode came from one of the detectives as they discussed the possible murder weapon at dinner. He said, "For all we know, it might be right under our very noses." Better practices and methods can be like that, right under our very noses.

Are pressure and temperature transmitters really needed on the inlet and outlet of a prover?

For example, using the pressure transmitter's published error span of [+ or -] 0.15% of span at 100 psi, the possible error would be 0.3 psi on just one transmitter. It takes 25 feet of pipe at full flow to cause a pressure drop of 0.3 psi. Would it be advisable to use two transmitters on 8-inch pipe, if the distance of the pipe between the meter transmitters and prover transmitter is less than 25 feet?

We used the following fairly typical flow conditions:

* Flowrate (Q) = 2,000 Barrels Per Hour

* Specific Gravity (S) = 0.88

* Viscosity ([gamma]) = 10cP

* Process temperature 700[degrees]F

* Ambient temperature 900[degrees]F

* The Velocity through an 8-inch ID line at the above conditions is 8.9 feet per second

From a pressure transmitter manufacturer's data sheet we have the following:

Pressure transmitter data:

[+ or -] 0.15% of span

Span: 100 psi

100 x 0.003 = 0.3 PSI worst-case error per transmitter

This from a pressure D-drop calculation:

It would take 25 feet of pipe at full flow to cause a pressure drop of 0.3 psi.

Query: Would it be advisable to use two transmitters if the distance of the pipe between them is less than 25 feet of 8-inch pipe?

On temperature, from the transmitter data we have a possible error of 0.02% of span on the 8-inch pipe at full 2,000 barrels per hour flow. Then using a span of-500[degrees]F to 2000[degrees]F the worst case error is 0.1450[degrees]C or 0.260[degrees]F. As above, the worse-case would be 0.520[degrees]F if one transmitter reads high and the other low. However, in this case we will split the error as above on the PT and use an error for one temperature transmitters of 0.260[degrees]F.

Temperature transmitter data:

0.02% of S-span

Normal Span: -500[degrees]F to 2000[degrees]F = 2500[degrees]F

Worst Case...

To continue reading