Methods tested for estimating VOC losses from floating roof turnovers

Dec. 10, 2001
A study of methods to determine the saturation level and saturation rate of organic vapors in the vapor space of floating roof tanks during tank turnover examined oxygen-combustible (O2/LEL) analyzers and canister sampling.

A study of methods to determine the saturation level and saturation rate of organic vapors in the vapor space of floating roof tanks during tank turnover examined oxygen-combustible (O2/LEL) analyzers and canister sampling.

Field tests conducted at various US pipeline breakout stations determined the percentage concentration of petroleum vapors in the space under the floating roof with respect to time (vapor saturation rate).

O2/LEL analyzers were used for all four tanks to monitor oxygen and lower explosive limit (LEL) levels continuously and to calculate vapor concentrations in the tank's headspace using the measured oxygen levels.

With the exception of the Tank 1 test, canister samples were also collected from three tanks. Two canister samples were collected for each tank: one at the beginning of the test and one at the end of the test. A commercial laboratory analyzed the canister samples by gas chromatograph (GC) to determine concentrations of hydrocarbons and permanent atmospheric gases.

Volatile organic compound (VOC) concentrations determined by the canister samples were more accurate than such concentrations derived from the O2 data. Oxygen analyzers, however, provided a trend of the VOC concentrations in the tank vapor spaces in a reasonable, safe, and cost-effective manner.

Vapor-space creation

Petroleum storage tank floating roofs are landed on tank legs and re-floated during a change of product service or certain tank maintenance operations. When the tank roof is landed on its legs, a vapor space is created between the floating roof, floor, and tank shell.

This vapor space could be saturated with petroleum vapors from both tank sludge (product clingage) and product pools remaining in the tank. As a result of diurnal temperature effects, the vapor may breathe through vacuum breakers on the floating roof.

At petroleum terminals, the same tank may be used to store different products; the contents of the tank may need to be completely drained to maintain product quality control. The frequency of these events depends on the number of product type changeovers and other operational concerns.

The emissions from refilling an empty floating roof tank could be greater than other types of tank air emissions (that is, breathing losses, rim losses, etc.). Therefore, for tank operations that routinely empty and refill tanks, the emission estimation for the event of refilling any empty tank may be a significant portion of the annual air emissions.

Determining these emissions hinges on the VOC concentration in the vapor space under the floating roof.

This study was conducted to develop a methodology for estimating product evaporation loss associated with landing floating roofs when emptying and refilling pipeline and terminal storage tanks. The study involved field tests to determine the percentage concentration of petroleum vapors in the space under the floating roof with respect to time (vapor saturation rate), which is the key parameter used to estimate emissions from floating roof tanks during a period between emptying and refilling.

In the study, the VOC concentrations were measured directly by collecting canister samples and indirectly by measuring oxygen concentrations. All data collected were analyzed to develop a series of vapor saturation levels.

Testing methods

The main objective of the field test study was to determine the saturation level of VOC inside a tank. For each tank, the vapor space temperature and oxygen level inside the tank was monitored as soon as the liquid stock was drained and the tank was safe to allow access. Also, a stock liquid sample for each tank was collected and tested to determine the saturated vapor pressure.

The research nature of this study dictated that the tank testing program be divided into two phases. Phase I was conducted as a pilot test on one tank, which allowed the study group to evaluate the testing methodology for the Phase II tests. The key testing parameter of concern was the VOC concentration in the vapor space.

The first method applied by the study to measure VOC concentrations inside a tank was the use of oxygen-combustible (O2/LEL) analyzers.1 These analyzers are commonly applied by safety professionals to ensure that VOC concentrations in a confined space are below lower explosive limits (LEL) for any entrance.

O2/LEL analyzers are selected with the following considerations:

  • AC power is not allowed in the tank area for safety reasons. This limits the selection of analyzers to the ones that are intrinsically safe. EPA Method 25 for VOC testing was ruled out because Method 25 sampling train requires AC power.2 O2/LEL analyzers meet this critical requirement.
  • O2/LEL analyzers can provide continuous measurement except for the time required for battery changes. Continuous measurement or a sufficiently large number of data points is highly desirable for the purpose of this study.
  • VOC concentrations inside tanks were expected to be as high as 30%. Organic vapor analyzers (OVA),3 although also intrinsically safe, cannot measure such high concentrations. An OVA's maximum measurement range is 10,000 ppm (that is, 1%).

Even within an OVA's measurable range, its readings only approximate VOC concentrations because different organic compounds in the tank's vapor space have different responses (similar problem as LEL measurement). Consequently, the OVA can be calibrated with only one compound. At its high measurement range (near 10,000 ppm), these errors are magnified.

The accuracy of an OVA also suffers in an oxygen-deficient environment. On the other hand, its measurement of oxygen is accurate. The problem of different responses for different compounds does not exist. Oxygen deficiency does not interfere with O2/LEL analyzers.

  • Sampling with SUMMA canisters is technically feasible. (SUMMA canisters are stainless steel cylinders that have been treated by the SUMMA passivation process and evacuated to a high vacuum condition.) The evacuated canisters could be brought to the site for sampling and no power is required. The canister samples could then be sent to a laboratory for a chemical composition analysis.

But the cost-benefit of this method is unacceptable for this study because each canister sample can only provide one data point. This method was not selected as the primary testing method because a large number of testing data points is needed to determine the VOC saturation "rate" as a function of time after a tank roof is landed on its legs.

Click here to enlarge image

Table 1 provides a comparison between O2/LEL measurements and canister samples.

Although the canister sampling method was not applied for the first tank testing, the study group decided to collect two canister samples in parallel with the O2/LEL measurements for the later tank tests.

A commercial laboratory analyzed the canister samples for permanent atmospheric gases, light hydrocarbons (C1-C6), and heavy hydrocarbons (C7-C12) with a gas chromatograph (GC) method. The canister sample data provided accurate composition data at a certain time, while O2/LEL analyzers generated continuous trend data.

Using O2 /LEL analyzer

With the analyzer, the sampling probes were inserted downward into the tank vapor space through openings on valves-ports affixed to the floating roof of the tank. The pump inside the oxygen meter drew air samples from the vapor space between the floor and the roof of the tank through its sampling probe.

The air sample was analyzed by the oxygen sensors and the LEL sensor. When the VOC concentration was high (in percentage), it was determined indirectly by use of the oxygen level in the air sample. This was because a high organic vapor level in the space would displace air and cause measurable oxygen deficiency.

The relationship between measured oxygen level and organic vapor level can be expressed as shown in Equation 1 in the accompanying equations box. The organic concentration can also be calculated by Equation 2 using 100% as a volume basis, where Vorg, VO2, and VN2 are volume percentage of organic vapor, oxygen (O2), and nitrogen (N2), respectively.

Equation 3 calculates the theoretical air N2 and O2 ratio; Equation 4 calculates the organic concentration.

Equation 4 shows that a measurement error in oxygen can be magnified by 4.785 times for VOC concentrations. When the oxygen readings are close to the level of normal ambient air (20.9%), this method of estimating organic vapor levels becomes unreliable because small fluctuation in the oxygen measurement can cause a relatively large error indicating unrealistic changes in organic vapor concentration.

In this study, it was expected that the oxygen levels would remain less than 20.5% and would be considered significantly low enough to perform the calculation of vapor concentration with the oxygen data. If the readings are higher than 20.5%, they are considered to be fluctuating and therefore unreliable.

When oxygen in the tank was deficient, organic vapor concentration was very high. High organic vapor concentration caused high % LEL readings that were out of the analyzers' range (higher than 100% LEL). When the oxygen level was close to the ambient level, however, LEL became measurable.

Because LEL cannot be used accurately to represent concentrations because of analyzer response issues (different compounds and different LELs), estimates were made based on an effective LEL (that is, a weighted average of the LELs). With this simplification, Equation 5 can calculate the organic vapor concentrations based on LEL readings of the analyzers.

Sampled tanks

Click here to enlarge image

This study completed tank tests for three external floating roof (EFR) tanks and one internal floating roof (IFR) tank. Table 2 summarizes tank locations, types, and services for each tested tank.

The following limitations were observed in this study:

  • For all tested tanks, the liquid levels were too low to determine liquid volume remaining in the tank during the tests.
  • No liquid temperatures were measured because the liquid levels were too low. Therefore, the saturation vapor pressures were estimated based on tank vapor-space temperatures instead of liquid temperatures.
  • Saturated vapor pressures were not adjusted to account for weathering of the liquid stock remaining in the tank.
  • All test data were measured when the tanks were standing idle. No data were collected during tank filling operations.
  • Due to limited access to the IFR tank (Tank 2 test), only one sampling port was available. Therefore, stratification of the IFR tank cannot be addressed.
  • Although wind speeds may have significant effects of tank vapor space ventilation, these effects were out of the study scope.

Results

Click here to enlarge image

Table 3 summarizes the monitoring time period, average temperature, saturated vapor pressure (based on liquid sample), and average VOC concentration derived from O2/LEL readings for each tested tank.

For Tank 1 (a drained dry tank), the tank VOC concentration decreased rapidly after the first day of the testing. Therefore, VOC concentrations are calculated for two periods; one for the first day of the test using the O2 readings and one for the remaining testing period based on the LEL readings.

Although Tank 4 (another drained dry tank) results also show that the VOC concentrations in the vapor space decreased with time, the VOC rate of decrease for Tank 4 was not as rapid as for Tank 1. Therefore, an overall average VOC saturation level was calculated for this tank.

The testing results of Tanks 2 and 3 (not drained dry tanks) indicated that the VOC concentrations in both tanks remained at a certain level and did not decrease with time.

Click here to enlarge image

Table 4 summarizes VOC saturation levels determined by the canister samples with the comparable VOC concentrations derived from the O2/LEL analyzer readings. As shown in Table 4, the VOC vapor concentrations indirectly derived from O2 data are all significantly higher than the results directly measured by canister samples with the exception of the second canister sample collected at Tank 4, while the O2 readings within the comparable time period were higher than the ambient O2 level.

Click here to enlarge image

Factors attributable to the significant differences between the two testing methods applied for this study, O2 analyzers and canister samples, are evaluated to assess the accuracy of each method.

O2 /LEL analyzers

Safety professionals commonly use O2/LEL analyzers to ensure safe entrance to a confined space with potential flammable chemicals (VOCs). Although this instrument provides reasonable measurements for safety concerns, it is not designed for accurate VOC measurement. The O2/LEL analyzer was initially selected for this test based on manufacturer-provided information, which satisfied the following basic criteria:

  • Reasonable measurements in a high-VOC environment.
  • Intrinsically safe instrument to be used near tanks with high VOC concentrations.
  • Continuous monitoring and data recording.

During the first tank test (Phase I for this study), the O2/LEL analyzers functioned reasonably well for a drained dry tank. For each of the later three tanks, however, one or two analyzers used for this study malfunctioned during the tests. These malfunctions typically occurred after the analyzers were exposed to a high-VOC environment for more than 1-2 hr.

Each malfunctioned analyzer was repaired then reapplied for monitoring. The malfunctions may have desensitized the analyzers, which may have introduced a bias.

In addition to the operability issues of the analyzers, the accuracy of the VOC concentrations derived from O2 readings is also a concern, which has been recognized by the study group from the beginning.

Canister samples

In general, a VOC concentration directly measured by a canister sample is very reliable because the sampling procedure is very simple and the lab analysis of VOC using gas chromatograph (GC) is a well-developed method for scientific studies. For this project, the project team evaluated several possible factors when assessing the accuracy of the test.

First, if a VOC sample is collected in a very hot environment, the VOC contained in the canister may condense in a laboratory with a normal room temperature. In this study, only Tank 2 sampling (during August in Baton Rouge, La.) may have been affected by this problem. The resulting canister VOC concentration (5.46%), however, is significantly less than the corresponding saturated vapor concentration (41.4%).

Even if the gas sample temperature contained in the canister dropped from 105° F. to 70° F., it is not enough to cause condensation.

The second possible cause examined by the project team is vapor stratification inside the canister. The vapor compounds contained in the canister (C3-C8) are all heavier than O2 and N2. During lab analysis, a gas sample is extracted from the top of the canister. If the canister has been sitting for a while, the heavier chemicals may have settled to the bottom due to their densities.

Review the GC testing reports, however, revealed no evidence to indicate any stratification of the canister sample because each of the resulting GC profiles shows a well-balanced VOC distribution (not skewed to lighter VOC).

The third possible cause is normal relative accuracy associated with any GC analysis. The GC analysis of VOC is a well-developed method for scientific studies. Although the VOC samples collected in this study contain very high levels of VOC and must be diluted many times, this is not unusual for a GC analysis.

Even if a relative error of ±50% is assumed (reasonable for trace-gas analysis), the upper bound of possible true value (for example, 11.19% for Tank 3 Canister 2) would still be significantly less than the VOC concentrations derived from O2 readings (24.5% for Tank 3 Canister 2 sample time).

For future studies, the project team recommended a test be conducted with multiple canister samples (for example, one sample for every 4 hr) for a selected tank to validate the testing results.

References

  1. Passport 2000 Personal Alarm, Technical Manual, Mine Safety Appliances Co., Pittsburgh, Pa.
  2. US EPA, Method 25, Determination of Total Gaseous Non-Methane Organic Emissions as Carbon, EMTIC M-25, 1993.
  3. Passport PID II Organic Vapor Monitor, Technical Manual, Mine Safety Appliances Co., Pittsburgh, Pa.

Based on a presentation to the ILTA Annual International Operating Conference, Houston, June 11-14, 2001.

The author

Click here to enlarge image

Hung-Ming (Sue) Sung is a senior technical associate with Trinity Consultants, Dallas, which she joined in 1989. From 1994 to 1997, she was an appointed member of the North Central Texas Air Quality Advisory Committee and actively involved in regional air quality studies. Sung holds a BS from the National Chung-Hsing University, Taiwan, an MS from Marquette University, Milwaukee, Wis., and a PhD from Vanderbilt University, Nashville, Tenn., all in environmental engineering. She is a registered professional engineer in Texas, Oklahoma, and Louisiana.