|
|
|
Preparation & Validation of One fundamental purpose of sampling is to verify
online electronic instruments are providing accurate indications
of water content. The method to determine water content must be
carefully considered by observing the types of crude oil present,
the installation of the equipment (both sampling and electronic
measurement installations) and the temperature, pressure, and
level of water present. SAMPLE PORT DESIGN: The key points are to have a design that will remove the sample in a representative cross section of the main line at the same velocity as the main line. The best will sample across the area of the main line and place the sample into a closed chamber. Most well testing systems open directly to atmospheric pressure when taking the sample into an open jar. Higher velocities such as opening a port to atmospheric pressures from the main line which is at a higher pressure can bias the sample due to the change in velocity at the point where the sample is being pulled. If the port is open to the atmosphere when pulling a sample the bias may come in due to the water slipping around the exit favoring the oil. If the sample was pulled at the same velocity as the flowing line, the "bow wave" effect is eliminated and water cannot slip past. The port design should take a cross section of the main line without creating a bias for oil or water. If the slipstream is pulled at the center of the main line, the liquids need to be well conditioned just upstream. A large sample port size will create several problems. A large sample port will force the operator to pinch down on the flow so that the oil will not spray all over. If the stub between the liquids in the main line is large, the pipe between the liquids and the valve becomes a "dead leg" where water may start to accumulate when pulling the sample. A recommended sample port will be a 3/8" or a maximum of 1/ 2" tubing with a through swage fitting into a weldolet in the main pipe that will be sampled. Figure 1 gives an example of a 1/ 2" type of system using a 45 degree cut on the tubing in the middle of the stream. A mark is made on the tubing on the front side of the 45 degree cut so that after it is inserted the technician can make certain that it is facing upstream. A static mixer is shown just upstream to mix the fluids. METHODS OF HANDLING THE SAMPLE AFTER IT IS PULLED: Emulsions will "age" as they cool down and time has passed from its removal from the line. This "aging" makes it difficult to process in the lab. If the sample is pulled using a gate type of valve or a needle valve, these operate slowly with the opening of the valve, the velocity will change depending on the valve opening. This will change the apparent water cut as different operators open the valve to different port sizes. The amount of sample pulled should be just enough to process using the analysis method of choice. In the case of centrifuge, this should not be more than about 100 to 200 ml. The entire sample should be centrifuged using the ASTM API method called out in IP359/82. Karl Fischer (KF) should only be used for 0-5% water cuts and the sample pulled must be processed soon after obtaining the sample. Pulling a sample for KF is difficult due to the very small size that is processed – typically only 1ml. If the sample is left to "age" the result will not be representative of the true water cut. Mercaptan and some sulfurs will affect KF depending upon the chemicals used. LABORATORY METHODS: Most preferred is the distillation procedure although it is the most difficult and time consuming. Two problems seen in distillation methods are in not having dry, analytical grade zxylene and not using all of the sample obtained. Centrifuge is the typical method for well testing applications (0-100% water cuts) but, it is often used for 0-5% water in oil. ASTM/ API Chapter 10.4 (Field Method) and IP359/82 (Laboratory Method) are standard test methods for centrifuge which need to be followed to obtain reasonable results. One overlooked issue is that of the use of water saturated toluene as a solvent. This is of particular importance when the water cut is lower than 0.5%. Kerosene and Stoddard solvent do not suffer from absorption of water. Several of the centrifuge vendors sell tubes that do not have any graduations between major marks, these are very hard to use. The procedure does not mention using the complete sample but, many instances this is the only way to achieve reasonable results. In this case multiple tubes are filled and all of the numbers are added together to obtain the average results. Notice the repeatability and reproducibility numbers at the end of the laboratory method standard. For 0-3% water cut the repeatability is 0.12% and the reproducibility is 0.28%. This is not between separate samples pulled at the same time and run separately or for higher water cuts. In addition, many cases of laboratory bias have been seen. One way to quantify the reproducibility is to obtain a 200 to 500 ml sample from a well mixed stream where an on line analyzer is available to make certain that the water cut is not changing during the sample pulling. Immediately after pulling the sample shake well and separate it into three or four samples. Mark each with a different well name and submit to the laboratory. The results can be very interesting but, help to shed light on the problem.
1. _3000# Threadolet Figure 1. Sample Port Configuration
HANDLING OF THE DATA: Data should be plotted with the lab results on the x-axis and the water cut analyzer on the y-axis. Be certain that the data from the analyzer is read at the same time when the sample is being pulled and that it is a fairly steady reading. If it is varying the sample becomes the average of the analyzer readings and this is difficult to track. Differences in DCS or PLC timing of data collection usually means that writing down the time of day and then looking up the analyzer reading at that time does not give useful information. A scatter plot of the data for a given pipeline or well vs time will help in observing the probable error in sampling and method. Fitting a best fit line through the data will show the offset from the analyzer to the laboratory methods. After obtaining 10-30 points the pattern should be uniformly distributed about the 45 degree line and have correlation between the data. During the data taking, no adjustments should be performed on the analyzer or these should be noted on the record so that the plots will be meaningful. The scatter of the points will give an indication of the deviation between the two sets of data. If a constant offset in either direction can be seen then it is time to place this offset into the analyzer. Constantly changing the offset in the analyzer is not productive because this is most likely due to the reproducibility of pulling and processing the sample. Once the graphed data shows a clear offset this should be entered once and more data collected. If records are kept and each change is noted the performance of the analyzer vs time will be seen.
Figure 2. Low Water Cut Data EXAMPLE OF DATA: Figure 2 shows real data from a pipeline application where the data was obtained and compared to a centrifuge. The middle dashed line is the best fit line and the two outer lines are the +/- 0.28% lines which is the API standard for reproducibility. Reproducibility is defined as the difference between different operators using the same method with different equipment on the same material. Each data point above may have been a different operator but, with the same equipment, and different samples. This data set has a few "outliers" which are data points that have a problem in the data collection. Other than these few points the reproducibility of the information appears to be close to the +/-0.28%. In this case the mean or best fit line goes through zero intercept and therefore the Cal Factor in the analyzer is correct. It is easily seen from this plot that at each point if the operator changed the Cal Factor to make the analyzer match the lab, they would be changing the baseline without reason. If this was done the data could not be used to determine process variance unless each change of Cal Factor was known. SUMMARY: Sampling is always a very difficult problem. Flow dynamics, physical properties of the fluids, emulsion properties, temperature, pressure, physical pipe layout, sample method and analysis procedures all make dramatic changes in the resultant data. Data plots vs time comparing the instrument value vs the chosen validation method usually show trend and provide information that cannot be obtained in any other manner. If only a random correlation exists between the two data sets there is something very wrong with one or both systems of measurement.
Figure 3. Correct Installation of Sampler, Courtesy Welker Engineering Co., Sugar Land, Texas Figure 3 shows some of the issues in the sampling of crude oils. Notice the oil at the bottom of the pipe, after the mixer, after the sample loop, and before the sampler. These problems exist in many installations with the best of intentions to eliminate them foiled by the product’s characteristics as mentioned earlier. At best sampling is an art and the water cut analyzers are making a measurement which will reduce the difficulties in obtaining good data. The problem is in the validation of the water cut analyzer due to the sample and laboratory issues.
Performance Acceptance Test Procedure The purpose of this procedure is to qualify a water cut analyzer for use in a pipeline crude oil application. It is assumed that proper sampling, laboratory and handling procedures have been implemented. 1. Obtain a Sample: 1.1 Prepare clean sample receivers not larger than 200 ml. Assure that the selected laboratory method is available and appropriate for the water cut to be seen. 1.2 Upon arrival at site of sampling, verify that the water cut analyzer is seeing a relatively constant water cut so that sampling will be over very small changes in water. If the analyzer display is not close to the sample point, have another technician observe the reading while the sample is being pulled. 1.3 Using a waste bucket, purge the sample port line of existing liquids and then grab a sample into the container without shutting off the flow in the sample line until an appropriate amount of liquid has been accumulated. 1.4 Write down the information in the attached form for the conditions at sampling. This is data for water cut, temperature, Cal Factor and known density of the crude. Cal Factor is available at the first or second menu of the Phase Dynamics, Inc. analyzer. This is to assure that the Cal Factor has not been changed between samples. If the analyzer reads 0.00% water cut at any point in the data collections, an adjustment to the Cal Factor should be made to always see a positive number on the Display. If this is not done the calculated number could be negative and this is not displayed on the Analyzer’s screen. If a change in Cal Factor is made the data should be restarted. 1.5 Pull two or three samples and label them accordingly. It would be preferred that the samples have a different water cut. Be careful that the water cut is not changing quickly and therefore the sample will not represent the analyzer reading.
2. Process the Samples: 2.1 Using the ASTM recommended procedures for centrifuge, KF or distillation process the sample noting the results on the line with respect to the sample number. Process the samples taken within _ hour of the collection. 2.2 If an on line densitometer is not available, laboratory density can be obtained by having another sample pulled at the same time which will be large enough to use a hydrometer. Keep that sample separate from the water cut sample. Recognize that this density will be different from the flowing density obtained by an on line densitometer unless it has been corrected to temperature. The Phase Dynamics, Inc. analyzer water cut can be compensated in the customer’s PLC or DCS by using FLOWING DENSITY for density correction. This is not density which has been temperature corrected to 60º F.
3. Process the Data: 3.1 Calculate the Error column on the data sheet by taking the Laboratory water cut and subtracting the observed water cut on the analyzer. 3.2 Determine the Average Error. At the bottom of the Error column calculate the Average Error by summing all of the numbers in the Error column and dividing by the number of samples. This should only be done for samples of the same crude type and density. 3.3 Plot the data on X-Y in Excel or by hand with the Laboratory water cut on the X-Axis and the Analyzer water cut on the Y-Axis. USE ONLY the data from the same source with the SAME DENSITIES for this plot. If the data encompasses more than one crude density then the density correction routine must be applied to the analyzer water cut (see density correction documents). 3.4 Calculate the best fit line through the data points using Excel or another program. 3.5 Place Error Limits around the best fit line depending upon the laboratory method used, see Figure 2 above for an example. For Centrifuge these error limits would be +/- 0.28%, Karl Fischer would be +/- 0.1% and Dean-Starke distillation would be +/-0.11%. These limits are ONLY tied to the reproducibility of a single "master" sample not between samples taken at different times. Sample to sample variations would be much larger than these numbers. The drawing of these error bounds is to establish a reasonable definition to view the sampling and laboratory issues statistically. 3.6 Observe data points that do not fall inside of the Error limits and mark these samples as Outliers which should NOT be included in the calculation of the Mean Error in Section 3.2. Recalculate this Mean Error. 3.7 If the Mean Error is added to the Phase Dynamics, Inc. water cut numbers obtained during sampling a new graph can be generated which will be Laboratory Water Cut on the X-Axis and Analyzer Water Cut + Mean Error on the Y-Axis. This will show the performance of the analyzer after the Cal Factor has been changed in the next section.
4. Analyzer Calibration: 4.1 The Mean Error can be a positive or negative number which becomes the Calibration Factor to be entered into the Phase Dynamics, Inc. Menu for this offset. Now the analyzer will be reading the mean value of the results obtained in Section 3. 4.2 If density changes the water cut will be affected The calibration performed above should be correct for all densities except for an offset calculated through the equation given in the Density Correction write up from Phase Dynamics, Inc.
Sample Data Sheet
Mean Error (%)_______
|
