EPA-450/3-76-036
November 1976
EVALUATION OF METHODS
FOR MEASURING
AND CONTROLLING
HYDROCARBON EMISSIONS
FROM PETROLEUM
STORAGE TANKS
~
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
-------
EPA COMMENT
"Evaluation of Methods for Measuring and
Controlling Hydrocarbon Emissions from
Petroleum Storage Tanks"
EPA 450/3-76-036
November 1976
A major objective of this study, initiated by EPA with Battelle in
March 1976, was to develop testing procedures to conduct short-term
hydrocarbon emissions tests on floating roof storage tanks either by
modifying existing test methods or developing new ones.
The most promising procedure appeared to be the use of a local
enclosure over the entire seal space wherein hydrocarbon emissions would
be calculated from the inlet and outlet hydrocarbon concentration in an
air stream of measured volume flowing into and out of the enclosure.
This technique was based in part on a theoretical study1 by Chicago
Bridge and Iron Company (CBI) which hypothesized that the principal
mechanisms causing loss were; (1) air solution and dissolution caused
by diurnal product temperature change, and ,(2) barometric pressure change.
Data-subsequently collected by CBI during August-November 1976 in
an experimental test program^ indicates that wind flow over a tank may be
the dominant factor causing emissions through tank seals rather than the
mechanisms stated above. The loss mechanisms caused by wind flow over a
tank may not necessarily be the same as those caused by continuous steady-
state air flow directly over an enclosed seal. Users of the enclosure method
should give consideration to the fact that no experimental data exist that
substantiate emissions are the same in either case.
References
1. Chicago Bridge and Iron Company, SOHIO/CBI Floating Roof Emission
Test Program, Preliminary Information, August 27, 1976.
2. Chicago Bridge and Iron Company, SOHIO/CBI Floating Roof Emission
Test Program, Final Report, November 18, 1976.
-------
EPA-450/3-76-036
EVALUATION OF METHODS
FOR MEASURING AND CONTROLLING
HYDROCARBON EMISSIONS
FROM PETROLEUM STORAGE TANKS
by
D.A. Ball, A.A. Putman, and R.G. Luce
BATTELLE
Columbus Laboratories
505 King Avenue
Columbus, Ohio 43201
Contract No. 68-01-3159
Task No. 3
EPA Project Officer: Richard K. Burr
Prepared for
ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Waste Management
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
November 1976
-------
This report is issued by the Environmental Protection Agency to report
technical data of interest to a limited number of readers. Copies are
available free of charge to Federal employees, current contractors and
grantees, and nonprofit organizations - in limited quantities - from the
Library Services Office (MD-35) , Research Triangle Park, North Carolina
27711; or, for a fee, from the National Technical Information Service,
5285 Port Royal Road, Springfield, Virginia 22161.
This report was furnished to the Environmental Protection Agency
by BATTELLE, Columbus Laboratories, 505 King Avenue, Columbus,
Ohio 43201, in fulfillment of Contract No. 68-01-3159, Task No. 3.
The contents of this report are reproduced herein as received from
BATTELLE, Columbus Laboratories. The opinions, findings, and
conclusions expressed are those of the author and not necessarily
those of the Environmental Protection Agency. Mention of company
or product names is not to be considered as an endorsement by the
Environmental Protection Agency.
Publication No. EPA-450/3-76-036
11
-------
Abstract
The purpose of this study was to determine advances made in
petroleum storage tank design for controlling hydrocarbon emissions,
evaluate the validity of the API correlations and test procedures when
applied to modern tanks, and develop new test procedures or modify ex-
isting procedures to conduct short-term emissions tests on modern tanks.
The results of the study showed that the floating roof is the most common
means of controlling emissions from storage tanks and that most advances
in design have been directed at increased safety, durability, and lower
cost. No actual experimental data were found to indicate that any par-
ticular tank emission control system was better at controlling emissions
than any other. It was also found that the existing API data base and
correlations are of questionable accuracy, especially when applied to
stocks with true vapor pressure less than 2.0 psia or tanks with diameters
greater than 150 feet. The API emissions test procedures were found to
be technically sound although they require extensive test times to obtain
significant results. Little chance was seen to increase accuracy or re-
duce the test time of these procedures. Several new test procedures were
analyzed and one was felt to have promise of achieving accurate emissions
results in substantially reduced test times.
iii
-------
TABLE OF CONTENTS
Page
LIST OF SYMBOLS . vii
SUMMARY 1
INTRODUCTION 3
OBJECTIVES 5
TECHNIQUES USED FOR CONTROLLING HYDROCARBON EMISSIONS FROM
PETROLEUM STORAGE TANKS 6
Floating Roofs 6
External Floating Roofs . 7
Internal Floating Roofs 15
Floating Roof Seals 17
Comments on Advances in Floating Roof Design 23
Plastic Foam Coverings 25
Variable Vapor Space Systems 25
Vapor Recovery 28
ANALYSIS OF THE VALIDITY OF CURRENT API CORRELATIONS FOR
PREDICTING HYDROCARBON LOSSES FROM FIXED ROOF AND FLOATING
ROOF STORAGE TANKS 31
Evaporation Loss from Fixed Roof Tanks 31
Breathing Loss 32
Working Loss (Theoretical Relations) . 41
Evaporation Loss from Floating Roof Tanks 43
Standing Loss 44
Withdrawal Loss 49
ANALYSIS OF EXISTING API APPROACHES TO MEASURING EVAPORATION LOSS. . 50
Stock Volume Decrease Approaches 50
Stock Property Change Approaches 51
Vented Vapors Measurement . 51
Special Techniques 52
iv
-------
TABLE OF CONTENTS
(Continued)
Page
METHODS FOR MEASURING HYDROCARBON EMISSIONS FROM PETROLEUM
STORAGE TANKS :. . . 53
Enclosure Methods ........ 54
General Engineering Analyses 54
Enclosure Techniques .64
Rate of Change Method 65
Steady-State Method. ...... 67
Windage Simulation 69
Additional Data Requirements 71
Tracer Methods 73
Optical Techniques for Measuring Hydrocarbon Emissions 74
General Discussion ... 74
Details of Apparatus 75
Application to Storage Tank Emission Analysis 76
CONCLUSIONS 81
RECOMMENDATIONS. , 85
REFERENCES . . 86
APPENDIX A
LIST OF VENDORS CONTACTED FOR INFORMATION ON TANK
EMISSION CONTROL .............. A-l
LIST OF TABLES
Table 1. Range of Variables 35
Table 2. Definition of Terms for Gasoline Breathing Loss
Equation for Fixed Roof Tanks . 36
Table 3. Paint Factors . 37
Table 4. Distribution of 60 Acceptable Tests by Type of
Construction 45
Table 5. Definition of Terms for Standing-Storage Loss Equation . 47
-------
LIST OF FIGURES
Figure 1. Floating Roof Tank Designs 8
Figure 2. View of Floating Roof From Top of Tank Wall Showing
Seal Area 10
Figure 3. Fabric Floating Roof Seal with Weather Guard Retracted . 11
Figure 4. Top of Floating Roof Showing Leg Supports and
Access Hatch 12
Figure 5. Access Hatch Showing Hatch Cover Removed 13
Figure 6. Top of Floating Roof Showing Bags Over Leg Supports. . . 14
Figure 7. Top of Floating Roof Showing Anti-Rotation Column and
Sampling Ports 14
Figure 8. Internal Floating Roof Configurations 16
Figure 9. Typical Mechanical Seal 19
Figure 10. Mechanical Seal with Helper Springs to Improve Sealing . 20
Figure 11. Mechanical Seal 21
Figure 12. Fabric Seals 22
Figure 13. Fabric Seal with Metal Wear Plates to Improve Durability 24
Figure 14. Separate Variable Vapor Space Tank . .26
Figure 15. Integrated Variable Vapor Space Tank 27
Figure 16. Simplified Schematic of a Typical Vapor Recovery System. 29
Figure 17. Adjustment Factor for Small-Diameter Tanks 38
Figure 18. Tank Enclosure Parameters 56
Figure 19. Enclosure Volume as a Function of Tank Diameter 57
Figure 20. Estimated Time to Reach Various Concentration Levels
as a Function of Volume and Emission Rate 58
Figure 21. Volume Flow of Air Required and Turnover Times for
Various Enclosed Volumes as a Function of HC Emission
Rate 60
Figure 22. Minimum Volume Flow for Turbulent Mixing to Exist. ... 63
Figure 23. Schematic of Flow System 66
Figure 24. Enclosure System Operating Envelope 70
Figure 25. Mean Flow Velocity in Various Size Enclosures 72
Figure 26. Block Diagram of Differential Absorption Apparatus ... 77
Figure 27. Above the Seal Optical Detection System 78
Figure 28. Below the Seal Optical Detection System. 80
VI
-------
LIST OF SYMBOLS
A Frequency factor
C Compressibility factor
Ca Adjustment factor for small tanks
D Tank diameter (feet in empirical equation)
F Working loss (barrels in empirical equation)
Fp Paint factor
_AG Volume of liquid lost from breathing
H Average outage (feet in empirical equatibn)
K. Turnover factor
K Tank type factor
K Factor depending on fuel
k Recommended seal factor
s
k Recommended fuel factor
c
k Recommended paint factor
P
L Standing storage loss of floating roof tank (barrels per year)
z
L Breathing loss of fixed roof tank (barrels per year)
M Molecular weight
m Mass
N Turnover per year
n Exponent
p Vapor pressure
p- Vapor pressure at minimum liquid surface temperature
p« Vapor pressure at maximum liquid surface temperature
P Ambient pressure
P. Gage pressure at which tank vacuum vent opens
P Gage pressure at which tank pressure vent opens
AP Weekly amplitude of barometric pressure
Q Volume of liquid pumped into tank (barrels in empirical equation)
R Universal gas constant
s Standard deviation
T Absolute minimum average vapor space temperature
vii
-------
1 Absolute maximum average vapor space temperature
AT Ambient temperature change ( F in empirical equation)
V Minimum volume of vapor space
V Maximum volume of vapor space
AV . Volume of vapor loss by breathing
V Tank volume
V Average vind velocity (miles per hour in empirical equation)
p Density of liquid
LJ
V General enclosed volume above seal in enclosure methods
Vp Volume contained under a portable roof
VT Volume contained under a local enclosure
L Distance from floating roof to top of tank
b Distance from tank wall to point of local enclosure on floating roof
h Distance from floating roof to point of local enclosure on tank wall
c Rate of hydrocarbon concentration buildup
Qup Assumed hydrocarbon emission rate
Re Reynolds number
p Density of gas in enclosed volume above seal
d Hydraulic diameter
H
«
QT Minimum volume flow in enclosed space above seal for turbulent flow
t Volume turnover time for enclosed space above seal
XQN Wavelength of laser light in absorption range for desired species
Wavelength outside of absorption range of desired species
IQ« Intensity of laser beam in absorption range
I0FF Intensity of laser beam outside of absorption range
U Velocity of gas flow through enclosed space above seal
a Ratio of change in fluid surface temperature to corresponding
change in ambient temperature
6 Ratio of change in vapor temperature to corresponding change in
ambient temperature
0 Activation energy
viii
-------
SUMMARY
The current methods used by EPA to estimate hydrocarbon emissions
from petroleum storage tanks have recently come under criticism by
various segments of the petroleum industry. The EPA emissions estimation
methods are based on API correlations which were developed around
emissions data obtained in the late 1930's and early 1940's. This data
base was limited in coverage and has been claimed to be antiquated or
not representative of current technology, but collecting new data using
approved API test procedures would require many months to possibly
several years.
The purpose of this study was to determine advances made in
tank design towards controlling hydrocarbon emissions, evaluate the
validity of the API correlations and test procedures when applied to
modern tanks, and develop new test procedures or modify existing proce-
dures to conduct short-term emissions tests on modern petroleum'storage
tanks.
The results of the study showed that the floating roof was the
most common means of controlling emissions from storage tanks and that
most advances in design were directed at increased safety, durability,
and lower cost. No actual experimental data were found to indicate that
any particular tank emission control system was better at controlling
emissions than any other. It was also found that the existing API data
base and correlations are of questionable accuracy, especially when
applied to stocks with true vapor pressure less than 2.0 psia or tanks
with diameters greater than 150 feet. The API emissions test procedures
were found to be technically sound although they require extensive test
times to obtain significant results. Little chance was seen to increase
accuracy or reduce the test time of these procedures. Several new test
-------
procedures were analyzed and one was felt to have promise of achieving
accurate emissions results in substantially reduced test times. It was
estimated that emission tests could be performed in a few hours,
although the actual time a tank would be out of service for testing
would be on the order of several weeks or longer to allow for equipment
setup.
-------
INTRODUCTION
On the basis of present predictive methods, hydrocarbons emitted
from petroleum storage tanks appear to contribute significantly to the
oxidant burden in many Air Quality Control Regions. The procedures
currently used by EPA in estimating these emissions are based on cor-
relations developed by the American Petroleum Institute which were com-
piled using emissions data taken in the late 1930's and early 1940's.
The correlations, which are primarily based on a symposium on
evaporation loss held in 1952(1), are described in detail in API Bulletins
2517(2), 2518<3), and 2519<4>. They cover emissions from fixed roof
storage tanks and storage tanks with various types of floating covers.
The data base for all these correlations consists of a compila-
tion of tests on individual tanks compiled by various petroleum companies.
These tests reportedly were conducted following the general test procedures
described in API Bulletin 2512(5). However, because all the tests were
run prior to publishing of API 2512 (July, 1957) strict adherence to these
procedures cannot in all cases be guaranteed.
Various segments of the petroleum industry have charged that
the correlations are inaccurate and do not account for improvements in
storage tank design that have been made since the basic data basje was
developed. Also, the data base was insufficient in dealing with liquids
with low true vapor pressures and large tanks over 150 feet in diameter.
In addition, the test methods used to develop the data base for the
correlations are time consuming (requiring many months to sometimes
several years) and may hot have sufficient accuracy to measure emissions
_a_L the low levels now claimed by tank operators and vendors.
Two segments of the petroleum industry - one represented by the;
consortium of Chicago Bridge and Iron (CBI) and Standard Oil of Ohio
(SOHIO)> and the other by the Western Oil and Gas Association (WOGA) -
have been conducting their own programs to better define the characteristics
and magnitudes of hydrocarbon emissions from petroleum storage tanks.
-------
The CBI/SOHIO program involves tests on tanks 20 feet and 8 feet in diameter
which were specially constructed to allow evaluation of specific variables
(such as temperature, windage, and degree of air stabilization) on hydro-
carbon emissions. The WOGA program involves field emission tests on a
number of floating and fixed roof storage tanks in the Los Angeles basin
area using the density change test method (with a thin layer of hydrocarbon
over water to speed results) essentially as described in API Bulletin 2512^'
for floating roof tanks and the vapor metering method as described in
API 2512 for fixed roof tanks.
These programs represent ambitious undertakings and the results
of each will undoubtedly contribute useful knowledge to the overall under-
standing of hydrocarbon emissions from storage tanks. EPA is following
the conduct and results of each program closely and is exchanging informa-
tion with both sponsoring organizations.
However, the lack of applicable test data and basic understanding
of hydrocarbon emissions from petroleum storage tanks coupled with the high
intensity of interest on the part of the industry and government organi-
zations have pointed up the need for this study. The purpose of this
report is to serve as a survey of the state of the art of petroleum
storage tank design, an analysis of the accuracy and applicability of the
existing API emissions correlations and test procedures when applied to
modern storage tanks, and an evaluation of the potential for developing
a short-term field test procedure which would allow accurate measurement
of hydrocarbon emissions.
-------
OBJECTIVES
The objectives of this study were three-fold:
(1) Conduct a survey to define the state of the art of
petroleum storage tank design relative to controlling
emissions of hydrocarbons.
(2) Evaluate the validity of the API correlations used by
EPA in estimating hydrocarbon emissions from fixed roof
tanks and tanks with floating roofs. Determine whether
these procedures can be extrapolated to stocks with true
vapor pressures less than 2.0 psia and to tanks with
diameters greater than 150 feet in diameter. Also,
evaluate the API test procedures used in developing
the data base for the correlations and determine whether
they might be modified to conduct future short-term
tests on modern tanks.
(3) Develop short-term test procedures for accurately
measuring hydrocarbon emissions from modern tanks with
floating roof tanks.
-------
TECHNIQUES USED FOR CONTROLLING HYDROCARBON
EMISSIONS FROM PETROLEUM STORAGE TANKS
Loss of hydrocarbon vapors through evaporation from petroleum
storage tanks has been of interest to the petroleum industry for many years.
Because these lost hydrocarbons -represent a salable product the industry
has employed various types of control devices to minimize this loss. Coin—
cidentally these devices also reduce the amount of unwanted hydrocarbon
emissions to the atmosphere providing a double benefit to both the petro-
leum industry and the environment.
Devices for limiting losses of hydrocarbons from petroleum stor-
age tanks can generally be classed into three groups: (1) those that are
designed to prevent vapor formation (such as floating roofs and plastic
foams) by limiting the amount of hydrocarbon liquid surface area exposed
to the atmosphere thus preventing evaporation, (2) variable vapor space
devices to allow for vapor expansion and contraction with flexible roofs
and diaphragms, and (3) vapor recovery systems that are designed to
collect the hydrocarbon vapor such that it can be used as a fuel, incin-
erated, or condensed back into liquid form. Within each of the three
groups a variety of types of systems are available depending on the ven-
dor and the application.
Floating Roofs
The most common type of device found in group 1 consists of a
roof or cover that floats on the surface of the hydrocarbon liquid in a
cylindrical storage tank. These roofs are made to cover as much of the
liquid surface area as possible while leaving sufficient space between
the edge of the roof and the tank shell to allow them to move freely up
and down within the tank without binding on the sides. This annular rim
space is usually filled with a flexible seal arrangement that hugs the
inside surface of the tank shell thus further limiting the amount of liquid
surface area exposed to the atmosphere.
-------
Floating roofs are by far the most common type of hydrocarbon
loss control device used for petroleum storage tanks. It has been esti-
mated by a large manufacturer that there are on the order of 10,000float-
ing roof storage tanks in use today. They can generally be classed as
external — where the floating roof acts as the tank roof serving as the
only barrier between the stored liquid and the environment — and, inter-
nal — where a fixed roof covering the top of the tank shields the float-
ing roof from the elements. As part of this project a variety of vendors
of floating roof equipment and seals were contacted to determine specific
features and variations in design. A list of the specific vendors con-
tacted is included in Appendix A.
External Floating Roofs
External floating roofs are most common in areas where weather
conditions are not severe such as the southwest and gulf coast regions of
the country. Because they are continually exposed to the elements they
are generally of heavier construction and more rugged than internal floating
roofs. Designs are fairly standard and have not changed much over the
years. Figure 1 shows the three basic configurations used—the pan, pon-
toon, and double deck types.
External floating roofs are made of a flat section of sheet steel
welded or riveted together on the edges to form a continuous membrane-like
cover. The pontoon and double deck types have leak proof flotation
sections with numerous bulkheads to ensure against their sinking in the
hydrocarbon liquid during upsets or periods of rapid filling. Rain water
is usually drained from the center of the roof through a flexible drain
pipe which extends under the roof through the hydrocarbon liquid and
exits at the base of the tank.
The roof itself forms a vapor tight structure, however, there
are various points around the roof where hydrocarbon leakage can occur.
The major point of leakage is from the annular space between the roof and
-------
Pan Floating Roof
c
Pontoon Floating Roof
r i ' • ' » i i
Double-Deck Floating Roof
FIGURE 1. FLOATING ROOF TANK DESIGNS
-------
tank shell. This space is usually about 1/2 to 1 foot in radial dimension
and at all times must be great enough so that the roof does not bind on
the tank walls as it travels up and down with the liquid level. This .
annular space is usually filled with a flexible seal which is designed
to rub against the tank wall thus centering the roof and keeping the ex-
.posed surface area to a minimum. Figure 2 shows the annular rim space
between the floating roof and shell of a large gasoline storage tank look-
ing down from the top of the tank. Figure 3 shows a closeup of the seal
with the weather guard (a metal shield to keep rain and sun off the seal)
retracted. Seals are normally sold separately from the roof and do not
necessarily depend on roof design. A more complete discussion of types
of seals used is presented later in this section.
Other points of potential hydrocarbon leakage are access hatches,
gage and sampling ports, anti-rotation columns, and roof supports. Figure 4
shows various, roof support legs and a typical access hatch on a new
floating roof tank not yet in service. The roof support legs project down
through the roof and maintain the roof a certain distance off the base of
the tank when the tank is not in service. These legs are adjustable with
usually a high and low setting, the high setting allowing for maintenance
crews to work under the roof when necessary, and a low setting which allows
the tank to be nearly emptied without damaging inlet and outlet pipes or
other equipment on the floor of the tank inside. The access hatch allows
for entrance below the roof and is normally sealed with a vapor tight cover
as shown in Figure 5 . The roof support legs can also be sealed against
vapor leakage in several ways such as a simple placement of plastic bags
over each support as shown in Figure 6.
Figure 7 shows an anti-rotation column that also serves OH n
gaging port. This hollow pipe is fixed at the top and bottom of the tnnk
and extends through a hole in the roof thus preventing rotation. Beomifle
of the sliding contact between the roof and column it cannot be positively
sealed against vapor leaks. Sliding seals can and are being placed around the
columns to limit vapor emissions such as was observed in field visits. However,
-------
FIGURE 2. VIEW OF FLOATING ROOF FROM TOP OF
TANK WALL SHOWING SEAL AREA
10
-------
FIGURE 3. FABRIC FLOATING ROOF SEAL WITH
WEATHER GUARD RETRACTED
11
-------
FIGURE 4. TOP OF FLOATING ROOF SHOWING LEG
SUPPORTS AND ACCESS HATCH
12
-------
FIGURE 5. ACCESS HATCH SHOWING HATCH
COVER REMOVED
13
-------
FIGURE 6. TOP OF FLOATING ROOF SHOWING
BAGS OVER LEG SUPPORTS
FIGURE 7. TOP OF FLOATING ROOF SHOWING ANTI-ROTATION
COLUMN AND SAMPLING PORTS
-------
this point of emissions is considered small (due to the small amount of
exposed liquid surface involved) compared to potential emissions coming
through the annular rim space around roof seals.
Internal Floating Roofs
Internal floating roofs float on the surface of the hydrocarbon
liquid inside a conventional fixed roof type tank. The fixed roof is
vented to allow sufficient air to circulate inside the tank to maintain
the hydrocarbon vapor above the floating roof at a concentration below
the lower explosive limit (LEL). The method in which they prevent emis-
sions of hydrocarbons to the atmosphere is analogous to that of external
floating roofs. The major differences are those of methods and materials
of construction.
Internal floating roofs need not be as rugged and sturdy as
external floating roofs because they are not directly exposed to the ele-
ments and seldom have to support any weight except that of the roof it-
self. Also, all internal floating roof designs studied are such that
they can be retrofitted to existing fixed roof tanks. In most designs
all materials of construction can be passed through existing access hatches
in the tank. This places unique restrictions on methods and materials
of construction which have been overcome by the various vendors in a
variety of inventive ways.
Figure 8 shows the three basic types of internal floating roof
designed used. The designs found in Figures 8a through 8d are those
actively promoted by most vendors today. The designs in Figures 8a and
8b use an aluminum skin laid up on a light weight aluminum framework all
of which floats on a series of aluminum or polyurethane filled pontoons.
The vapor space under the roof is usually sealed off from the annular
rim space by a continuous metal skirt around the circumference of the roof
extending 4 to 5 inches into the hydrocarbon liquid. Usually a flexible
type wiper seal such as that shown in Figure 8 is used to close the space
between the roof and tank wall.
15
-------
b)
c)
d) .
e)
ELASTOMER WIPER-SEAL OR
FOAM-FILLED COATED FABRIC SEAL
e
\/ \
FOAM-FILLED COATED FABRIC
l_
FOAM-FILLED COATED FABRIC
TANK SHELL
PONTOON
7
BUOYANT PANEL
7
7
7
STEEL PAN
FIGURE 8. INTERNAL FLOATING ROOF CONFIGURATIONS
(Courtesy of the American Petroleum Institute)
DECK
16
-------
Figures 8c and 8d show another common type design employing a
flat bouyant roof that rests directly on the liquid surface. The roof
can be made of sheet aluminum with a honeycomb core, fiberglass over
polyurethane, or other combinations of metal and plastics. Usually a
pillow seal such as that shown is employed.
The flat pan roof shown in Figure 8e has also been employed
in the past as an internal floating cover. A variety of types of seals
can also be used with this type roof.
Because of the similarity to external floating roofs, internal
floating roofs tend to have the same points of emissions and the same prob-
lems limiting these emissions. For all the designs studied, the basic
roof itself, if constructed and maintained properly, should be impervious to
hydrocarbon vapors. Therefore, the principal points of emissions are
the annular rim space between the roof arid tank wall and places where
the roof must seal around fixed vertical sections such as gaging wells,
support columns for the fixed roof, anti-rotation columns or cables, and
internal access ladders. Other potential emission points are access
hatches and roof support legs or suspension cables for maintaining the
roof a desired distance off the tank floor when the tank is emptied.
Various types of sliding seal arrangements have been devised
for sealing the roof around the fixed vertical elements in the tank.
Access hatches and roof support members can be positively sealed off in
much the same manner as those on external floating roofs.
As was the case with external floating roofs the major potential
point of emissions from internal floating roofs appears to be the seal
area. Seals used vary considerably but are similar in many respects to
those used on external floating roofs. Seals are discussed in more de-
tail in the next section.
Floating Roof Seals
A great deal of interest has recently been focused on the design
of seals used to fill in the gap between the floating roof and tank shell.
17
-------
Basically, seals can be categorized into two groups (1) mechanical or
shoe type seals and (2) flexible resilient fabric or tube type seals.
(1) Mechanical or shoe seals have been used since 1920. They
consist of a rigid metal strip which is forced out against the tank
shell usually by weights attached to a series of lever arms. Figure 9
shows a cross section view of a typical mechanical seal. A fabric
section connects the top of the seal near the tank wall to the edge of
the floating roof thus closing off the annular space to hydrocarbon vapors.
These types of seals are durable and resist wear from constant sliding
against the tank walls. However, because they are rigid they don't con-
form well to irregularities in the tank wall or to protrusions such as
welds or rivet heads. In some cases helper springs have been used to pro-
vide additional force against the top of the seal to get more consistent
contact with the tank wall. An example of such an application is shown in
Figure 10. Another means of improving the sealing effectiveness of these
seals is to install a secondary flexible seal above the primary mechanical
seal. Two examples of such applications are shown in Figure 11. Figure lla
shows a tube type wiper seal consisting of a polyurethane core with a
canvas covering. Figure lib shows a Maloney type rubber seal that is
shown partially installed.
(2) Flexible fabric type seals are a relatively recent innova-
tion in seal design and are becoming more popular as increased emphasis
has been placed on obtaining tight, consistent contact. Figure 12 shows
two common examples of fabric seals installed on an external double deck
floating roof. Figure 12a is a resilient foam seal which has a urethane
foam core covered by a tough flexible synthetic material that is
attached to the floating roof. Figure 12b has a liquid filled core which
provides the sealing force against the tank wall.
Flexible fabric seals are used extensively on internal floating
roofs in a variety of configurations. Figure 8 (see section on Internal
Floating Roofs) shows three different examples as they might be employed
on the three basic types of internal floating roofs. Figure 8a shows a
18
-------
FLEXURE ClOSUfil
FIGURE 9. TYPICAL MECHANICAL SEAL
(Courtesy Chicago Bridge and Iron)
19
-------
FIGURE 10. MECHANICAL SEAL WITH HELPER
SPRINGS TO IMPROVE SEALING
20
-------
a. With helper springs and tube
type secondary seal
b. With Maloney type secondary seal
FIGURE 11. MECHANICAL SEAL
21
-------
K>
WEATHER SHIELD
HANGER BAR
CURTAIN SEAL
SEAL ENVELOPE
SEAL SUPPORT
RING
RESILIENT
URETHANE FOA;
WEATHER SHIELD -
SEALING BAND
SEALING LIQUID
NORMAL
PRODUCT LEVEL
ADAPTABLE
SEAL SUPPORT
a. Urethane foam
b. Liquid filled
FIGURE 12. FABRIC SEALS
(Courtesy Chicago Bridge and Iron)
-------
wiper type seal consisting of either an elastomer type material or foam
filled coated fabric. These seals contact the wall above the liquid sur-
face and flop back and forth as the roof goes up and down. Figures 8b
and 8c show urethane foam filled seals with coated fabric covers which
ride with the roof either at the liquid level or just above it as shown.
The flexibility of fabric seals allows them to conform more
readily to irregularities in the tank wall thus providing a tighter more
consistent fit than mechanical seals. In areas of severe weather they
are usually protected by weather guards (see Figure 3 ), or by fixed roofs
over internal floating roofs. In some cases metal wear plates have been
installed to increase life as shown in Figure 13. However, rigid wear
plates have many of the same disadvantages of mechanical seals in that
they are inflexible and also some gaps will occur around the edges.
Comments on Advances in Floating Roof Design
The design and construction of floating roofs for reducing
hydrocarbon losses from petroleum storage tanks has evolved over the years
in a very competitive market place. As such, many innovations and unique
features have been incorporated into the various vendor's designs to
provide them with an edge over the competition. However, the floating roof
is generally considered inherently effective in reducing hydrocarbon
losses and most vendors claim reductions of around 95 percent over fixed
roof tanks. Such a reduction is also predicted by comparing calculations
made in API Bulletin 928(6) based on the API correlations of Bulletins 2517
and 2518'2*3'. Therefore, most all design improvements over the years
have been directed at reduced time and cost of construction and maintenance,
greater safety and reliability—especially in trying to eliminate the possi-
bility of sinking—and in improving the life and durability of the structure.
The same is also true of seals in that most design features concentrate
on durability, ability to conform to the tank wall, and cost. No specific
experimental data were found in contacts with vendors showing that any
particular floating roof or seal arrangement is superior to others in
terms of controlling hydrocarbon emissions.
23
-------
FIGURE 13. FABRIC SEAL WITH METAL WEAR
PLATES TO IMPROVE DURABILITY
-------
Plastic Foam Coverings
Another method for controlling hydrocarbon emissions by preventing
vapor formation is the use of plastic foam coverings. These are described
in detail in API Bulletin 2515(?) .
The use of plastic foam involves spraying a foam like blanket
over the surface of the stored hydrocarbon liquid, thus preventing the
formation of hydrocarbon vapors. According to API 2515, use of foam is
limited to crude oil. Its use on gasoline is not effective and causes
operating problems. Reductions in emissions of from 70 to 90 percent are
reportedly possible over fixed roof tanks with vent-valve settings of
0.5 ounce per square inch of pressure.
Variable Vapor Space Systems
These devices consist of basically two types —lifter roofs
and flexible diaphragms. They can be either separate tanks such as that
shown in Figure 14 or integral with a fixed roof tank as that shown in
Figure 15.
These systems are primarily designed to limit breathing loss
from fixed roof tanks. They are generally used for gasoline storage where
tank throughput is low (less than 6 to 12 turnovers per year).
Usually a series of fixed roof tanks are connected to a variable
vapor space tank by a series of manifolds. Vapors evolving from the
products stored in the fixed roof tanks during periods of thermal expan-
sion or reduced barometric pressure are temporarily stored in the variable
vapor space tank. During periods when the vapors are contracting, such
as at night, they are transferred back to the storage tanks. In this
manner, normal breathing losses are effectively controlled.
Filling losses are also controlled up to the point where the
expelled vapor exceeds the capacity of the variable vapor space system.
In some cases "balanced pumping" (filling some tanks while other are emptying
25
-------
FIGURE 14. SEPARATE VARIABLE VAPOR SPACE TANK
Taken from API Bulletin 2520 "Use of Variable-
Vapor-Space Systems to Reduce Evaporation Loss"
September, 1964
26
-------
FIGURE 15. INTEGRATED VARIABLE VAPOR SPACE TANK
Taken from API Bulletin 2520 "Use of Variable-
Vapor-Space Systems to Reduce Evaporation Loss"
September, 1964
27
-------
on a common manifold) can be used to further control emissions. However,
problems with scheduling and preventing contamination of dissimilar stocks
limit the applicability of this technique.
Potential points of emissions from these systems are from leak-
age through manifolds seals and fabrics used for diaphragms, and from ex-
ceeding the capacity of the variable vapor space system during filling
periods.
Vapor Recovery
Vapor recovery is in effect an extension of a variable vapor
space hydrocarbon control system discussed earlier. A key difference,
however, is that vapor recovery systems are designed to handle both breath-
ing and working losses while variable vapor space systems alone are de-
signed only to handle breathing loss.
Figure 16 shows a simplified schematic diagram of a modern
vapor recovery system design.for refinery use, As in a variable vapor
space system a battery of cone (fixed) roof storage tanks are connected
by a series of gas manifolds to a gas holder (usually a low pressure
diaphragm type design).
Provision is made for removing excess gas from the gas holder
with a compressor. To make up for this lost gas a blanket gas is allowed
to flow into the vapor space of the cone roof tanks during periods of
vapor contraction (due to cooling or barometric pressure changes) or
during periods of liquid withdrawl. The blanket gas usually consists of
either natural gas or boiler flue gas and its purpose is to maintain a
non-explosive mixture in the vapor space of the cone roof tanks.
If natural gas is used as a blanket then the vapors recovered
from the gas holder represent a high grade combustible fuel gas. This gas
may or may not be processed through various stages of refinement such as
cooling to condense low volatile liquids separating them from the gas
stream as shown in Figure 16. The fuel gas can be used as an energy
source for the various gas-fired furnaces in a refinery.
28
-------
Blanket
Gas
NJ
VO
Battery of Cone Roof
Storage Tanks
Hydrocarbon Vapors plus
Blanket Gas from Storage Tanks
Evaporative
Condenser
Fuel Gas or Hydro-
carbon Vapors to
Furnaces or
Incineration
Separation Column
Variable Vapor
Space Gas
Holder
•J Compressor
Liquid
Recovery
Emergency
Vents
I
FIGURE 16. SIMPLIFIED SCHEMATIC OF A TYPICAL VAPOR RECOVERY SYSTEM
-------
The Valdez crude oil storage terminal in Alaska uses boiler
flue gas as an inert blanket. This flue gas is available from large
boilers supplying steam to operate pipeline compressors. When an inert
gas such as flue gas is used as a blanket the recovered vapors from the
gas holder are not high enough in combustible content to represent a
fuel gas. However, the gas contains significant hydrocarbon content and
therefore must be incinerated either in the boilers supplying the blanket
gas or in separate auxiliary fuel fired incinerators.
The major points of emissions from vapor recovery systems during
normal operation are leaks in manifolds, valves, compressors, and gas
holders. These leaks can be kept very small with proper maintenance.
However, because vapor recovery units are relatively complex mechanical
systems, normal operation may be difficult to insure 100 percent of the
time. When upsets or equipment failures occur the storage tanks are
operated in the same manner as conventional fixed roof tanks with excess
vapors being vented through an emergency pressure vent valve.
Vapor recovery systems with secondary processing offer a posi-
tive means of hydrocarbon emission control and under normal operation
with proper maintenance could be more effective in controlling emissions
than floating roofs. Natural gas blanketed systems allow for a form of
product conservation in that vapors from the storage tanks can be used
as a fuel gas. Inert gas blanketed systems do not allow for the same de^
gree of conservation but do provide for positive control of vapor emissions,
30
-------
ANALYSIS OF THE VALIDITY OF CURRENT APT CORRELATIONS
FOR PREDICTING HYDROCARBON LOSSES FROM FIXED ROOF
AND FLOATING ROOF STORAGE TANKS
Industry concern for evaporation losses from petroleum storage
tanks resulted in a symposium on the subject on November 10, 1952, at the
32nd Annual Meeting of the American Petroleum Institute ' (Symposium on
Evaporation Loss, Proceedings, American Petroleum Institute, Vol. 32, 1952,
pp 212-281). The papers prepared by the various committees for this sym-
posium culminated eventually in a series of API bulletins formalizing
analytical and computational procedures for estimating hydrocarbon losses
from fixed and floating roof type storage tanks containing gasoline and crude
(3 2)
oilv ' '. Some modifications based on observations (generally proprietary)
were made to computed results to allow for improvements in design and
maintenance. The purpose of this discussion is to evaluate the present day
validity of using these equations to estimate hydrocarbon emissions to the
atmosphere in modern equipment.
Evaporation Loss From Fixed Roof Tanks
(API Bulletin 2518)
There are two losses that were dealt with in the scope of this
study ~ the breathing loss and the working loss. Breathing loss is defined
as "vapor expelled from a tank because of the thermal expansion of existing
vapors, and/or expansion caused by barometric pressure changes, and/or an
increase in the amount of vapor from added vaporization in the absence of
(•\\
liquid-level change, except that resulting from boiling . Working
losses are defined as "vapor expelled from a tank as a result of liquid
(3)
pumping into and out of the tank" . The history of the derivation of the
empirical relations for both these types of losses will be discussed in
turn, along with comments on the validity of the empirical relations as
applied to modern technology.
31
-------
Breathing Loss
A theoretical breathing loss equation as discussed in several
references ' ' was developed to foster a better understanding of the
mechanism of hydrocarbon losses from fixed-roof tanks. Following these .
references, we consider a closed container partly filled with a volatile
liquid. Then, assuming the perfect gas law for a given mass of air
*
(1)
If the pressure exceeds the vent opening pressure, the loss of volume is
given by
' T,
AV = V,
- P
P2 -
(2)
Since the liquid volume of the loss is proportional to the vapor pressure
of the fuel in the discharged gas, one may put this in the form
AG = K
>1 + P2
Pa + ?2 - P2
(3)
where K is a function of the compressibility factor and composition of the
lost fuel. While a certain factor is recommended for gasoline, as of 1952,
the factors would have to be reevaluated for modern gasoline as well as for
other fuels.
If |P1|«Pa;iP2|«Pa;|p1-p2|«Pa;|T1-T2|«(T1 + T,,) = 2 T; PI +
P = 2 p; it is found that
- T1)/T
- P1)/(Pa - P) -
(4)
* Definition of symbols used in various equations are listed at the
beginning of this report,
** Within the accuracy of the assumptions this equation corresponds to
the equivalent equation in Reference 8.
32
-------
—0/RT
If p = Ae , then to the same degree of approximation as above, and
assuming that the variation of vapor temperature is 3 AT, where AT is the
ambient temperature change, and the variation in temperature of the fluid
surface is a AT,
AV/V = IAT/TJ
0
8 + a
P - p / \RT/
a
- (P2 - ^/(^ - P ) (5)
Since a may be about 1/10 and B about 5/4, P2 ~ ?1 = 1~1/2~inches water»
and AT is the order of 10 C, it is clear that the last term may be dropped,
but the first two terms may be about the same order of magnitude,
It should be noted that there are also barometric pressure changes
that occur about once a week, with an amplitude of AP /P of about 1/30,
This term should be added directly to AV/V, , giving a value per day of, about
(1/210)(P /(P -p)); this is a small term compared to two of the previously
cl cl
mentioned terms, but somewhat larger than the relief valve term. Since, one
is subtraction and the other addition, they roughly cancel,
In Reference 1, it was concluded that there was no hope of using
Equation (5) because of the large number of variables involved. As a result
the basic equation was dropped and resort was made to an empirical analysis
of available experimental data.
The basic relationships expressed in the derivation of; Equation
(5) could possibly be used for guidance in developing a more basic correlation
around experimental data.
The next step in developing a basis for analyzing the experimental
data would be to derive an equation relating diurnal ambient air temperature
*
variation, solar input , and wind velocity to the temperature variation of
the vapor and the liquid in the tank. If one only writes out the terms for
* We note that Appendix 4 of Reference 3 (API 2518) presents a detailed
treatment of some data on the relation between total reflectance of a
tank surface and with bulk liquid temperature relation to that with a
white tank and loss relative to that with a white tank. However, no
theory is presented.
33
-------
a steady state transfer of heat to the tank fluid and vapor from the environ-
ment, it is clear that convective transfer results in two terms (a constant
plus a term depending on the Reynolds number) and solar radiation results
in a third term, which depends, among other things, on the finish, of the
tank; For the diurnal cycle, the heat capacity of the fluid and vapor in
the tank must also be considered to determine the loss per day, Obviously,
then, the results must be averaged over solar conditions, wind conditions,
and temperature. The form for the final equation could have been set up
at the time of the formulation of Reference 3 (API 2518), and terms con-
densed on the basis of qualitative evaluation of terms to permit the pre-
sentation of a form of empirical equation to use in correlating the data,
Failure of the correlation form would then indicate either a deficiency in
understanding the phenomena, the presence of an unknown variable, or a deficiency
in the method of determining the data. Mathematically, there is little
probability that the loss should be expressed as the first term only in
the Buckingham IT-theorem derivation' '' as is typified by the technique
used in API 2518 . Because reasonable cost computer capacity is now
available that far exceeds that when API 2518 was prepared, sample compu-
tations could now be carried out over the range of variables to evaluate
the significance of various terms and result in a greatly improved
empirical relation.
Experimental data on breathing losses of gasoline were contained
in 256 individual tests which were screened down to 178 acceptable tests.
Of these, 86 were for nonworking tanks and the 64 of these tests for tank
diameters of 20 ft or more (see Table 1) were used to determine the co-
efficients in the empirical relation for breathing loss.
24 f p \ 0.68
y ~ 1000 P - p D1'73 H°t51 AT°'5° F C.
\ a I P
The definitions of terms in this equation are given in Table 2.
34
-------
TABLE 1.
RANGE OF VARIABLES
(3)
OJ
D,
feet
10 to
20 to
40 to
60 to
100
110 to
20
30
55
70
120
Number of
Tests
_
-
7
3
4
37
AT,
degrees
Fahrenheit
5 to 10
>10 to 15
>15 to 20
>20 to 25
>25 to 34
P,
pounds per
Number of
Tests
6
12
36
28
4
—
H,
feet
1
> 2
> 5
>10
>20
>30
to
to
to
to
to
to
2
5
10
20
30
40
Number of
Tests
4
10
18
26
16
12
square
inch
absolute
1.0 to
3.0 to
4.0 to
5.0 to
6.0 to
9.0 to
2.9
3.9
4.9
5.9
6.9
9.3
Number of
Tests
22
19
14
13
16
2
-------
TABLE 2. DEFINITION OF TERMS FOR GASOLINE BREATHING
LOSS EQUATION FOR FIXED ROOF TANKS(3>
L = breathing loss, in barrels per year.
P = true vapor pressure at bulk liquid temperature,
in pounds per square inch absolute. Average
liquid body temperature usually is available
from gaging records; however, if this informa-
tion is not available, it may be estimated by
adding 5 F to the average ambient temperature
from meteorological records. (Meteorological
data for several locations in the United States
and Canada are presented in API Bulletin 2513,
Appendix VI.)
D = tank diameter, in feet.
H = average outage, in feet, including correction
for roof volume. (A cone roof is equivalent in
volume to a cylinder which has the same base
diameter as the cone and is one-third the
height of the cone.)
T = average daily ambient temperature change, in
degrees fahrenheit (difference between U.S.
Weather Bureau average daily recorded maximum
and minimum temperatures).
Fp = paint factor, determined from field tests (see
Table 3).
*
C = adjustment factor for small-diameter tanks.
* The height to diameter ratios of the smaller tanks
are ordinarily considerably greater than for the
large tanks, and the multiplying shape factor (Cp)
was derived to compensate for the lack of corre-
lation of these data from the smaller tanks
(Figure 17).
36
-------
TABLE 3. PAINT FACTORS
(3)
Paint Factor.
Tank Color
Roof
Shell
Paint in
Good Condition
Paint in
Poor Condition*
White
Aluminum (specular)
White
Aluminum (specular)
White
Aluminum (diffuse)
White
Light gray
Medium gray
White
White
Aluminum (specular)
Aluminum (specular)
Aluminum (diffuse)
Aluminum (diffuse)
Gray
Light gray
Medium gray
1.00
1.04
1.16
1.20
1.30
1.39
1.30
1.33
1.46
1.15
1.18
1.24
1.29
1.38
1.46
1.38
Values for poor condition provided only as a guide.
as outlined in Appendix IV of ^Reference 3.
Factor may be obtained
37
-------
-o
1.00
.80
.60
.40
.30
10 20 30
Tank Diameter In Feet
FIGURE 17. ADJUSTMENT FACTOR FOR SMALL-DIAMETER TANKS
38
-------
Considering the ratio of the observed to calculated value for
the 86 tests reveals a Gaussian type distribution on a logarithmic basis,
*
with 68 percent of the points falling within + 34 percent of the mean .
This does not appear to agree with the implication of a rather vague state-
ment in Reference 3 that "the calculated breathing loss will deviate in the
order of 10 percent from the actual breathing loss".
A close examination of individual sets of data indicated that
the factor of 1/2 on H appeared quite good (the significance of 0.51 as
compared to 0.5 is highly questionable). The data seemed to indicate that
the effect of AT should be considered to be linear at low values, and fall
off at higher values. With the adjustment factor, the effect of diameter
appears to be well represented; however, the need for an adjustment factor
shows that the basic phenomena are not all included in the result and make
one question the validity of extrapolation to larger sizes. The basic
analysis of data discussed in reference to the paint factor (Appendix IV,
(3)
API 2518 ) appears to be excellent. However, comments on the actual
factors recommended in the text of API 2518 do not reflect the more in-
depth analysis of Appendix IV.
The term which includes the true vapor pressure is quite puzzling.
The relation (p/(P -p)) shows up in the theoretical derivation, and appears
3.
quite logical to include, however, one notes that in deriving a relation
for the liquid volume of fuel loss, as contrasted to the total vapor loss,
a vapor pressure term must be included.** On the basis of the theoretical
relation, then, one would expect a variation of liquid loss with p(p/(P -p) )
(3) a
where n is between zero and unity. Furthermore, in API 2518 , breathing
losses from 15 crude oil tests were also studied. The results, to be dis-
cussed in more detail later, showed a variation in effect of vapor pressure
contrary to what would be expected, comparing gasoline and crude oil losses
* That 34 percent is half 68 percent is a coincidence and has no intrinsic
implications.
** Discussed in more detail in connection with filling losses later in
this section.
39
-------
by using the empirical equations. Using the empirical equation directly
results in higher losses for crude oil than for gasoline which is contra-
dictory to expectations. One has a great suspicion that the two contra-
dictions may be related. A cursory inspection of the 64 sets of data
reveals that the lowest values of p are associated with the smaller tank
diameter, and the highest values of p are associated with the largest tank
sizes. This could result in misleading coefficients.
For the correlation of the crude oil breathing loss, 15 sets of
data on tanks of 115 to 120 ft diameter were given. Because of the lack
of variation in tank size, it was assumed that the same empirical equation
held as for gasoline. However, it was necessary to change the coefficient
of 24 to 14 to obtain a fit. No satisfactory explanation for the need of
the change was given, though both the correlation relations used to go from
a Reid vapor pressure to a true vapor pressure curve and the probably slower
convection in the liquid fuel were mentioned as possible contributing
factors.
Reanalyzing the crude oil data on the basis of the loss divided
by the product of true vapor pressure, diameter squared (essentially constant),
and average outage, showed no further statistical variation with true vapor
pressure or outage. Except for three points, the temperature difference
varied too little to consider significant any observed effect.
A comparison of the standard deviations of this set of data, using
(3)
the recommended correlation from API 2518 , with the standard deviation
for the gasoline data, showed about equal values. However, using the
simplified correlation just presented reduced the standard deviation to about
half. Including a linear temperature effect (remembering that there were
only three points with wide deviations in temperature from the mean) increased
the standard deviation to about half way between the two extreme values just
mentioned. While this result cannot be used as an argument in favor of a
new correlation, it again shows that serious questions can be raised about
the validity of the correlating relation of API 2518.
40
-------
Working Loss (Theoretical Relations)
The calculation of a theoretical relation for working loss is
less complex than for the standing loss. As the tank is filled, the
vapor above the liquid is displaced out of the tank, the total volume of
vapor equalling the volume of liquid pumped, and the vapor carries with
it a certain amount of hydrocarbon vapor. Assuming that the vapor is
uniformly distributed on an average over the time of filling, the liquid
volume of the working loss, F, for gasoline is given by
F = 3 P vwV1Q4'
where V is the volume of liquid pumped and K is a turnover factor (to
be discussed later). For crude oil, the value 3 is replaced by 2.25. This
equation for gasoline was arrived at in the following manner. The volume
of vapor discharged from the tank is assumed equal to the volume of liquid
added to the tank. Using the gas law equation,
p V = CmRT/M,
F = m/pL = pVy M/pL CRT .
Comparing this with the previous equation for F, and neglecting for the
time factor K ,
M/pT CRT = 3 x 10~4 .
LJ
The numerical value was obtained by assuming an average molecular weight
for the fuel of 61 and concomitant density and compressibility factors, and
a temperature of 60 F. We note that if F and V are the same units, the
value of the numerical term is not affected bv these units.
41
-------
In Appendix 1 of Reference 3 (API 2518), consideration is given
to the working loss from emptying the tank. It is shown that if the tank
is emptied in rapid small increments so that equilibrium conditions are
attained after each emptying step, the working loss factor F should be
increased by the mulitplying factor (1 - p/P) . This is because there
is an inrush of air, part of which (with some fuel vapor) is driven back
out as the fuel vapor present approaches equilibrium. If the tank is
—1
emptied rapidly, the multiplying factor is (P /p) In (1- p/P ) ; relative
3. Si
to the step-like withdrawal, this factor is a fraction decreasing with
increasing vapor presssure. It is noted that with a rapid withdrawal,
followed by an immediate refilling, there is no loss due to this effect
and the factor should be unity. The same occurs for the case of slow
continual withdrawal, maintaining equilibrium conditions in the tank.
(3)
In the recommended computing procedure , an additional factor for
these effects is not included. This leads to questions on the validity
of the implied confirmation of the theory by observation, discussed
below.
The available data on turnover (123 sets) include only 6 tests
with 10 or more turnovers per year, with 30 turnovers the largest number.
It was felt that the composition in the tank would be near enough equilibrium
with turnover more than 10 days apart that the turnover factor, Kt should
be unity for these test data. It is commented for gasoline that "For turn-
overs up to 30 or 40 per year, the data substantiate a value of K^ equal
to unity". (API 2518). (From Appendix II of Reference 8 (API 2513) ,
there is an implication that equilibrium saturation may take place in the
order of an hour for gasoline.) The important implication is that thorp
was an agreement with the data, although the datn ;irc- riot pro.sfnlr.od.
The data on crude oil are conceded to be insufficient Lo pi-null
a formal correlation. Assuming that the reason for the- discrepancy be-
tween crude oil and gasoline noted with the breathing loss is a result of
a slower convective movement in the crude oil, it is assumed that the
working loss factor of 0.58 between gasoline and crude oil should be re-
vised upward for use in the working loss case. It is indicated that a re-
view of scattered data led to a selection of a value of 3/4; this resulted
in 2.25 replacing 3 in the equation for F.
42
-------
As pointed out earlier, there were no data on the effect of
large numbers* of turnovers of the tank fluid. It was felt that because
the tank vapor would not approach equilibrium rapidly, there would be
less loss when the number of turnovers was high. It is difficult to
reconcile, however, the actual value allowed, which starts to show an
effect at about one turnover every 10 days, with implications in other
parts of the development that equilibrium is often reached in a few hours.
Further, the inclusion of the factor Kt which is unity or less and exclu-
sion of a factor for fill loss discussed above does not seem consistent.
In any case, factors for K^ were suggested in Reference 1 but an incon-
sistency in the values assumed leading to the same total turnover loss
per year prediction for 40, 50, and 80 turnovers lead to a modification.
It was recommended that K^ = (180 + N)/6N for turnovers N greater than
36 per year. Thus, if Vt is the tank volume, the turnover loss per year
for gasoline was given by
F + 5 x 10~5 p Vt (180 + N), N >_ 36 .
A turnover every 2 days would thus nearly double the tank loss per year
over that obtained with 36 turnovers per year but would decrease the
loss per turnover by one-third. It might be noted that part of the
decrease that is required is a result of the overlapping of standing
losses and turnover losses as the turnover period approaches the diurnal
time.
Evaporation Loss from Floating Roof Tanks
(API Bulletin 2517)
From Reference 2 (API 2517), there are two types of loss that
are dealt with —• the standing loss and the wetting loss. Standinj-
losses are defined as "those losses other than those resulting from
breathing or change in liquid level". For floating-roof tanks, the .lar^c-.sL
potential source of standing-storage loss is attributed to "an improper Tit
* Turnover per year equals the annual throughput divided by the tank
capacity.
43
-------
of the seal and shoe to the shell" (API 2513)(8). The wetting loss, or
withdrawal loss, results from the "vaporization of liquid from a wetted
tank wall exposed when a floating roof is lowered by withdrawal of liquid"
(API 2513). The history of the derivation of the empirical relations for
both these types of losses will be discussed in turn, along with compo-
nents on. the validity of the empirical relations as applied to modern
technology.
Standing Loss
The discussion of standing loss indicates that the principal
source of evaporation loss is the annular rim space between the tank shell
and the outer rim plate. The loss, in the development of Reference 2
(API 2517), was considered to result from evaporation between the seal
and tank shell, and vapor permeation through the sealing fabric. Product
exposure results from poor fitting seals, shell irregularities, rivit
heads, and so forth. It is surmised that any wind helps clear the narrow
vapor space and promotes evaporation. The vapor permeability is an addi-
tive factor which is considered, in most instances, to be relatively small.
It should be noted that there are other throughts as to the
cause of the major portion of the vapor loss, including a diurnal theory
which would result in a form similar to that indicated by the theoretical
development for fixed roof tanks. This theory presumes that the equili-
brium air absorbed in the fuel changes with fuel temperature, as does the
vapor pressure of the fuel. As the fuel temperature varies during a day,
the periodic outflow of gas through the annular space carries fuel vapor
away to be lost.
Sixty tests on tanks were found acceptable and used for analyses
in Reference 2. These were divided into the six classes as shown in
Table 4.
44
-------
TABLE 4. DISTRIBUTION OF 60 ACCEPTABLE TESTS
BY TYPE OF CONSTRUCTION
Number of
Tests
8
13
9
15
7
8
"60
Tank
Shell
Riveted
Riveted
Riveted
Riveted
Welded
Welded
Type of
Roof
Pan
Pan
Pontoon
Pontoon
Pontoon
Pontoon
Seal
Single
Double
Single
Double
Single
Double
Paint Color
White (3) , gray (5)
Gray or aluminum
Gray or aluminum
Gray or aluminum
White (1), gray (6)
Gray or aluminum
Unfortunately, as in the tests on the fixed roof tanks, the
assumed condition of each tank is not given, Further, while it was possible
to pick out the two cases in which, crude oil was used (rather than gasoline) ,
it was not possible to identify the remaining two white tanks that were used.
Further, no wind velocity data were given for several tanks in which no loss
was found; it is not clear whether the values were not available or were
too low to measure.
The analysis of the data apparently was based on a progressive
consideration of the influence of various factors, culminating in the
empirical expression
(p/(P -p))°'7 .V °'7
I a / w
L = K A p/(P -p)' .V ' k k k
scp
45
-------
where
A = D1'5, D < 150 ft
A = 150°'5 D, D > 150 ft
and the other terms are defined in Table 5.
It was noted that outages and daily atmospheric temperature
variations showed no trend; however, if these factors did not cover a
wide range of variation, within the accuracy of the data, this observation
is not unexpected.
It has been noted previously that the factor (p/(P -p)J enters
the reasoning from a consideration of diurnal temperature variations, but
a factor p to the first power should be present to account for the change
from a gas volume to an equivalent liquid volume basis-. Noting as before,
that (p/(P -p) ) ' is not too different from p over the range of variables
considered, it is seen that such a substitution would be quite logical.
In Reference 2 (API 2517), it is indicated that "twelve tests on
similar tanks, 117 ft in diameter with pan roofs and double seals, sub-
*
stantiate the selection of an exponent of 0.7 for the wind velocity" .
This afforded a good test of the significance of this factor, so,
;l
defined as
assuming that K,., k , k , and k were all constant, a standard deviation
r s c p
2
s
= ( log ((L/p*)/(L/p*) ) - n log (Vw/Vw)) 2 /(12-1)
was inspected for poth p' = p, and p1 = (p/(P -p)) " , for a range of
3.
values of n about the minimizing value. The two minima were different
by less than 1 percent, indicating no preference. But more important,
allowing a 5 percent increase in the value of s, i.e., the range of
* Immediately following is a comment that for velocities loss rh;in l\ mph,
the values at 4 mph should he used. Actually, no dat;i are tahu l;il etl
for wind velocities less than 5.0 mph unless the occasional dashes
indicate such values.
46
-------
TABLE 5. DEFINITION OF TERMS FOR STANDING-STORAGE LOSS EQUATION
\0.7 v 0,7k k k
w s c p
L7 = standing-storage evaporation loss, in barrels per year
LJ
Kf = a tank-type factor which changes as follows:
K = 0.045 for welded tank with pan or pontoon roof, single or
double seal
Kf = 0.11 for riveted tank with pontoon roof, double seal
K,. = 0.13 for riveted tank with pontoon roof, single seal
K,. = 0.13 for riveted tank with pan roof, double seal
K,. = 0.14 for riveted tank with pan roof, single seal
D = tank diameter, in feet. For tanks 150 ft or less in diameter,
D ",for tanks larger than 150 ft in diameter, use 1501'5
150
p = true vapor pressure of the stock at its average storage tempera-
ture, in pounds per square inch absolute. (This may be determined
from the Reid vapor pressure.)
Vw = average wind velocity, in miles per hour
k = a recommended seal factor :
s
ks = 1.00 for tight-fitting seals (typical of modern metallic
and nonmetallic seals)
k = 1.33 for loose-fitting seals (typical of seals bult prior
to 1942)
kg = a recommended factor distinguishing between gasoline and crude
oil storage:
k =1.00 for gasoline
k =0.75 for crude oil
e
k = a recommended paint factor for color of shell and roof:
k = 1.00 for light gray or aluminum
k = 0.90 for shite.
P
47
-------
value of n for p' = p was 0.20 to 0.97, which included the recommended
0.7, while the range of values of n for p' defined in accordance with
the suggested equation was -0.10 to 0.62. At the most, this exercise
indicates a preference for the simpler pressure term if one is convinced
the effect of wind is of the order of 0.7. More, likely, it only indi-
cates that the scatter of data is sufficiently large and the range of
values of wind velocity sufficiently small, that no positive, conclusion
can be drawn.
A close inspection of the data indicates that there appears
to be more than a linear effect of diameter. This could result from
a shape factor, as assumed with the fixed roof tanks, but more probably
relates to the ratio of the distance of the floating roof below the
top of the tank to the tank diameter (especially if there is a real
wind effect). It is clear that, with no data on tanks over 145 ft in
diameter, the assumption made for tanks larger than 150 ft in diameter
has no confirmation.
The values of Kf, relative to each other (see Table 5) are
quite reasonable. The imposition of these factors in a plot of the data
brings the data closer together, especially in the case of welded tanks
relative to the riveted tanks.
The value of k for crude oil relative to gasoline is based
principally on the value assumed for the working loss for fixed roof
tanks, but the two crude oil test results for the pan roof riveted tank,
with single seal, appear to confirm the value for this one type of tank.
The value of k (a seal factor depending on the snugness of fit of the
s
seal) could not be checked on the basis of the data presented in
Reference 2 (API 2517) . The value of k , a point factor normalized
on the basis of light gray or aluminum, is given as 0,9 for white tanks.
This does not check in magnitude with that deduced from the fixed roof
* Some laboratory tests on losses were discussed briefly, in which
loss rates relate to seal gaps were given as 1/8 inches, 1.0;
1/4 inches, 2.6; 1/2 inches, 3.7.
-------
studies, which could certainly be a result of a different balance of
phenomena involved in the evaporation in the two types of tanks.
Confirmation of a value less than 1 is .apparently clear from the data,
but only two of the four tests on white tanks are identified, those
containing crude oil wherein another difference is also present.
Withdrawal Loss
Quoting from API 2517
"Withdrawal loss resulting from evaporation of stock
which fails to drain down the shell as the roof descends
is negligible under most conditions. Laboratory data
determining the clingage of gasoline, to a rusty steel
surface support this conclusion. The tests show that
clingage ranges from 0.02 to 0.10 bbl of gasoline per
1,000 sq ft of steel surface which has a scale, ranging
from light rust to dense tight rust,
The loss is represented by the equation;
W = 22,400 ~
Where:
W = withdrawal loss, in barrels per, million barrels
throughput
C = 0.02 (based on barrels of clingage per 1,000 sq-
ft of shell surface)
D = tank diameter, in feet."
For C = .02, the. equation that is given is W = 448/D.
"Withdrawal loss for gunite-lined tanks used for
storing gasoline can be significant and should be con-
sidered. The laboratory clingage data indicate a factor
C = 2,0, which is hundredfold greater than for light
rust scale. Since withdrawal loss suppresses standing-
storage loss, it is recommended that a factor C = 1.0
be used for gunite-lined tanks used for storing gasoline.
49
-------
ANALYSIS OF EXISTING API APPROACHES TO MEASURING
EVAPORATION LOSS (API BULLETIN 2512)
Accepted methods and measurement techniques to determine evap-
oration losses from petroleum tanks and transportation equipment are
presented in API Bulletin 2512^ '. The bulletin was published July, 1957,
by the Subcommittee I of the American Petroleum Institute's Evaporation
Loss Committee for general information and is not considered an API standard.
The approaches to measuring evaporation loss can be character-
ized by any one of the three effects of evaporation: (1) decrease in
stock volume, (2) change in stock properties, and (3) measurement of the
vented vapors. The approach selected usually depends on whether the tank
is in static storage or working service, on the time available for obtain-
ing data, and on the available instrumentation and skill of the personnel
performing the test.
Stock Volume Decrease Approaches
Three approaches are discussed for gaging the contents of static
and working-service tanks in this section. These approaches are (1) direct
manual gaging,(2) liquid metering, and (3) hydrostatic weighing. Direct
gaging requires accurate measurements of the liquid level and the average
temperature of the stored product. This technique is useful for static
storage, no flow to or from the tank. Under working tank conditions when
material is either flowing into or from a tank, liquid meters are used to
measure the inflow and outflow of material and manual gaging is used to
determine the stock inventory. The difference in the metered stock mate-
rial and manual gaging is attributed to the combined breathing and working
loss. The third method uses an auxiliary liquid column or manometer to hydro-
statically weigh the contents of the storage container. This method had not
been used to any degree at the time of this report, and it was indicated that
further development was required. This technique could have many advantages:
since weight changes can be magnified with the use of proper instrumentation,
shorter test periods can be used, depending on overall sensitivity of the
50
-------
manometer or pressure measurements instrumentation; and volume corrections
for temperature variations are eliminated.
Stock Property Change Approaches
Two methods associated with stock property changes are discussed
in API Bulletin 2512 for determining evaporation loss. The two methods are
referred to as the "vapor pressure change method" and the "density change
method". In principle, the vapor pressure change method involves taking a
series of samples of the stored stock from a tank at two different times.
From the samples, vapor pressure-evaporation curves can be established. The
initial and final pressure-evaporation curves will be displaced from each
other because when evaporation occurs (assuming evaporation losses occur
between sampling) the more volatile components evaporate at a greater
rate than the less volatile. The displacement of the pressure-evaporation
curves is a direct measure of the evaporation loss.
The density change method involves taking a series of stock
samples from the tank at the beginning and end of a test period. From
the initial samples, a density-evaporation curve can be established as a
function of the percentage of stock evaporated. The final series of
samples taken from the tank are used to determine a final density. The
percentage evaporation loss is determined directly from the density-evapo-
ration curve previously established from the initial samples. Both methods
require special laboratory apparatus, highly skilled personnel, and truly
representative samples of the tank contents.
Vented Vapors Measurement
These methods are applicable only.to fixed roof tanks where
vapors entering and leaving the tank do so through a small number of well-
defined points such as pressure vent valves. Two methods are available
for measuring and analyzing the vented vapors. The simplest method is
to measure the air flowing in and the vapors flowing out with positive-
51
-------
displacement (PD) meters. The difference in mass represents hydrocarbon
vapors lost. Liquid loss can be calculated accurately with a rough
knowledge of the molecular weight of the vented hydrocarbon.
Another method of measuring evaporation loss by analyzing the
vented vapors is conducted by measuring the total mass flow of vented
vapor and the total hydrocarbon content of the vapor using a hydrocarbon
analyzer. The hydrocarbon vapor loss is computed from the meter and
analyzer data after correcting for temperature and vapor density.
Special Techniques
Two methods are suggested under the heading of special techniques
and are referred to as (1) subscale testing and (2) repeated-transfer
technique.
(1) When performing subscale tests, it is essential that proper
means be used in scaling the full-scale mechanisms of importance. It is
very easy to obtain results from subscale models that are in error when
scaled up to the full-scale situations.
Sectional modeling of certain components of the full-scale sys-
tem in the laboratory can be performed to determine the operational
characteristics under various conditions. Care must be exercised in setting
up these laboratory tests, also, to ensure that all aspects of importance
related to the specific full-scale component are properly accounted for in
the laboratory setup,
(2) The repeated-transfer method is a technique to magnify
working losses to a measurable level over a relatively short period of
time. In this technique, filling and emptying cycles are repeated a number
of times to give sufficient stock loss, due to working losses, so that the
stock loss is large enough to measure. The use of this technique requires
additional equipment for pumping material in and out of the storage con-
tainer and using one of the techniques specified above to measure the loss
of stock material. Corrections must be made for reductions in stock vola-
tility which occur with stock loss sufficient to yield measurable change.
These losses, when large enough to measure, would normally and eventually
affect evaportation rates.
52
-------
METHODS FOR MEASURING HYDROCARBON EMISSIONS .FROM
PETROLEUM STORAGE TANKS
Presented in this section are descriptions of experimental
procedures that can be used to measure hydrocarbon (HC) evaporation losses
from petroleum storage tanks. Storage evaporation losses have been
characterized as breathing losses, standing-storage losses, filling
and emptying losses, wetting losses, and boiling losses. A detailed
description of each of these loss mechanisms is given in Reference 8.
A comprehensive description of a number of methods that have been used.or
proposed for measuring evaporation losses from petroleum storage tanks is
presented in Reference 5. As pointed out in Reference 5, evaporation
losses can be determined by measuring; (1) decreases in stock volume, (2)
changes in stock properties, and (3) vapors escaping to the atmosphere.
The first two measuring techniques, mentioned above, require several months
to several years to obtain significant results. This is because of the
limitations in accurately measuring small changes that occur in large stock
volumes and stock properties with time. The third measuring technique
(direct vapor measurements) appears to be an acceptable technique for
making accurate short term measurements of emissions from fixed roof tanks
but has little applicability to tanks with floating roofs. The only methods
used extensively on floating roof tanks in the past for measuring evapora-
tion loss are those described under (1) and (2) above.
A prime objective of this study is to define test procedures for
accurately measuring HC emissions from storage tanks that require appre-
ciably less test time than those test procedures that have been used in the
past. The primary reason for this objective is: (1) increased ability
to correlate important parameters (such as atmospheric effects) with emis-
sion rates, (2) reduced time tanks have to be out of service because of
testing, and (3) the overall test accuracy can be increased by reducing
the test time using the appropriate Instrumentation.
Included in this section are descriptions of two test methods,
referred to as (1) Enclosure Methods, and (2) Radioactive Tracer Methods,
that were considered as possible candidates for making accurate measurement
of hydrocarbon losses from storage tanks. These methods are discussed in
53
-------
terms of their applicability to floating roof tanks (FRT) with external
floating covers and to fixed roof tanks with internal floating covers. They
are also discussed in terms of their ability to measure hydrocarbon losses
through the seal space between the tank roof and walls. Field trips con-
ducted on this study as well as industry contacts have established that
this seal space is the major point of hydrocarbon emissions from these
types of storage facilities and also the most difficult to control,r| Field
investigations determined that other points of emissions, such as gage
hatches and roof supports, could be effectively controlled by various means.
Nevertheless, the techniques described in this study could be modified to
measure emissions from these points also.
Enclosure Methods
One method of measuring HC emissions through floating roof tank
(FRT) seals is to collect and analyze the HC vapors that pass through
the seal. In order to collect the vapors, an enclosure must be used over
the volume of space above the seal into which the vapors are released.
The enclosure should be ventilated to eliminate excessive build-up of HC
vapors. There are two ways that can be used to enclose the volume of
space above a FRT's seal. One approach is to construct a portable roof
that would completely cover the top of the FRT. Another approach is to
locally cover the space above the seal by attaching a single piece of flex-
ible material to the roof and inside tank walls near the seal. There are
advantages and disadvantages of both covering techniques which will be
discussed in the following section.
General Engineering Analyses
Presented in this section are general engineering details for
HC emission measurements related to enclosure methods. The results of
the analyses are presented in a form such that estimates for a variety of
enclosure test configurations can be evaluated. Engineering estimates of
the time to reach various HC concentration levels, volume turn over-time,
air volume flow rate required to maintain various HC concentration levels
and critical flow conditions for good mixing are given in this section.
54
-------
Enclosure Volume and Concentration Accumulation Time. Figure 18
is a schematic that shows various tank enclosure configuration parameters
that will be used in the following analysis. The enclosed volume above
the seal for portable roof and local enclosure configurations are
Vp = f D2£ , and (6)
i V = ^ bhD , where (7)
i Lj £-
Equation (1) is for portable roof configurations and Equation (2) applies
to local enclosures located at the seal. "D" is the tank diameter, £ is
the distance between the top of the tank and the floating roof, b is the
distance between the tank sidewall and the enclosure attachment point on
the floating roof, h is the distance between the floating roof and the en-
closure attachment point on the tank inside wall.
Calculated enclosure as a function of tank diameter are presented
in Figure 19. As expected, the enclosed volume associated with portable
roof configurations are substantially larger than local seal enclosure
configurations for floating roof positions greater than 4 feet from the
top of the tank. Later in this section, it will be shown that larger
volumes are a disadvantage for this type of testing.
The time variation of HC concentration level within the enclosure
will depend on the HC emission rate, enclosure volume, and uniformity of
mixing within the volume. The rate of change of HC concentration level
within a given volume is given by
(8)
V
•
where QHr is the volume flow of HC vapor, and V is the enclosed volume.
Estimates of the time (t ) required to achieve various HC concentration
levels within an enclosure of known volume (V) are presented in Figure 20.
These estimates are based on the following assumptions: (1) no leaks in
the enclosure covering, (2) there is a uniform mixture of HC vapors and
air in the enclosure, (3) the initial HC concentration level Is approxi-
mately zero at time zero, and (4) the HC emission rate is expressed .-is .-i
time average value over the concentration build-up time. For ;i const.-ml.
volume system, the concentration accumulation times shown in Figure 20
represent estimates of overall system response times and test times. To
reduce the test time below 2 hours as shown in Figure 20 the product of the
55
-------
Local seal
enclosure
Portable roof
enclosure
T
jft
_L
FIGURE 18. TANK ENCLOSURE PARAMETERS
56
-------
10
10
ro
o>
E
TJ
0)
O
c
10
10'
—. Portable roof
Local enclosure h = b
I
5.0
100
150 200 250
Tank Diameter (D), ft
300 350
FIGURE 19. ENCLOSURE VOLUME AS A FUNCTION OF TANK DIAMETER
57
-------
10
1.0
ZJ
o
u
10
10'
c - Concentration level, volume percent
rh - Emission rate, bbl/yr
Calculation based on:
Liquid density*= 5.2 Ib/gallon
Vapor density = 0.13 Ib/ft
i i i i i
i i i i i i i
i i i i i i i
10'
cV- (ft3)
10"
FIGURE 20. ESTIMATED TIME TO REACH VARIOUS CONCENTRATION LEVELS
AS A FUNCTION OF VOLUME AND EMISSION RATE
* Density of condensed vapor.
58
-------
3
concentration level and volume must be below 10 for emission rates of
22 bbl/yr and higher. It should be noted that for a constant cV condi-
tion, the accumulation time varies approximately an order of magnitude
for emission rates from 25 to 250 bbl/yr.
The general approach of using a constant volume system to collect
the HC vapors appears quite simple and straightforward to use in deter-
mining emissions. However, there are some apparent difficulties such as:
(1) determining the system volume accurately, (2) assuring that a uniform
mixture exists, and (3) assuring sufficient control of test conditions to
achieve optimum data accuracy.
Steady State Concentration Level in Flowing Systems. The steady
state HC concentration level that will be obtained in an enclosure that
is continually fed with air will depend on air volume flow rate to the
enclosure and the HC emission rate into the enclosure. An estimate of
the HC concentration level in a flowing system can be calculated assuming
uniformity of the HC vapors and air by
%C
c = -?£- (9)
QAir
where QHC and Q.. are the HC and air volume flow rates, respectively.
Estimates of the HC concentration level as a function of air
volume flow and HC emission rates are shown plotted in Figure 21 as lines
of constant concentration from 100 to 10,000 ppm. The results presented
in Figure 21 are for a representative condensed vapor liquid density of
5.2 Ib/gal and vapor density of 0.13 lb/ft3.
The enclosure. HC concentration level can be held to less than
100.0 ppm for sustained emission rates as high as 1000 bbls/yr with an air
supply system that will deliver a maximum 10 CFH (see Figure 21) . Main-
taining the system concentration level well below the flammable mixture
limits is important relative to safety aspects of conducting the test.
The concentration level in the enclosure can be varied over a wide range
by changing the air volume flow rate. For instance at an emission rvite
of 100 bbls/yr, the concentration level can 'be varied from 100 to 5000 ppm
by changing the air volume flow rate from 10 to 4 x .10 CFH,
59
-------
Volume Turnover Time (tv), hr
i.O 10.0
100.0
\ x*^
Hydrocarbon Emission Rate, bbl/yr
FIGURE 21. VOLUME FLOW OF AIR REQUIRED AND TURNOVER
TIMES FOR VARIOUS ENCLOSED VOLUMES AS A
FUNCTION OF HC EMISSION RATE
60
-------
The time required to completely replace the air in the enclosed
volume (termed the volume turn over time) is important in determining the
total response time of the system. Volume turn over times (t^) for various
size enclosures and air volume flow rates are presented in Figure 21. It
is important that the turn over time be minimized to increase the system
response.to variations in hydrocarbon losses, but,at the same time the
concentration level must be within measurable range. The data presented
in Figure 21 can be used in the following manner to estimate turn over
times and air volume flow rates required to maintain desired concentration
level,for various emission rates. For example, assume the following
conditions exist: (1) it is desirable to maintain the concentration level
at 0.5 percent due to instrument accuracy, (2) the enclosed volume is
approximately 10,000 ft3, and (3) the HC leak is estimated at 100 bbl/yr.
Enter Figure 21 on the 0.5 percent concentration line at a point corres-
ponding to 100 bbl/yr (Point 1). An air volume flow rate of 3800 CFH
will be required to maintain 0.5 percent concentration level for an
assumed leak rate of. 100 bbl/yr. To estimate the turn over time, progress
horizontally along the line of constant volume flow corresponding to
3800 CFH, intersecting the line of constant enclosed volume of 10,000
(Point 2). The control volume turn over time can then be read off the
upper coordinate in Figure 21, for this case it would be approximately
2.7 hours.
It is obvious from Figure 21 that to maximize system response
(minimize turn over time, tv), the system should be designed with small
enclosed volumes and high air flow capability.
Critical Flow Conditions. Sampling of the enclosure gases wL.1.1
be necessary to determine the percentage of HC in the gas mixture. More
details of the sampling procedures will be discussed later in this section.
For now, it is important to establish flow conditions within the enclosure
that will promote good mixing. Since the flow in the enclosure is similar
to duct flow, it is possible to define a critical Reynolds number for
turbulent flow. Ensuring that turbulent flow conditions exist is a
61
-------
practical way of promoting the desired mixing of the HC and air. The
critical turbulent flow Reynolds number for duct flow is approximately 3000.
The Reynolds number for a noncircular flow cross section is defined as
pUd
where p is gas density, U is mean cross sectional flow velocity, djj is an
equivalent diameter of the flow cross section equal to 4 times the hydraulic
radius, and y is gas viscosity.
For estimating purposes, assume that the enclosure is triangular
and also that the height (h) and base (b) are equal, see Figure 18. The
equivalent diameter for this configuration is
du = b/1.707
ri
The ratio of the gas density to absolute viscosity (p /p ) is approximately
equal to 6.35 x 10^ sec/ft . The Reynolds number can be expressed as
Re = 7.44 x 103 , (10)
* o
where 2Q/b has been substituted for the mean cross sectional flow
velocity (U) and b/1.71 for the equivalent diameter.
Setting the Reynolds number equal to the critical value of 3000
and solving for the volume flow rate as a function of cross section dimen-
sion gives
QT = 1451 b (CFH),
j_>
where Q represents the minimum volume flow condition for turbulent flow.
L
The minimum volume flow rate conditions for various cross section
dimensions are shown plotted in Figure 22. It will be shown later that
these minimum volume flow boundaries, are not restrictive regarding opera-
tional characteristics of various enclosure configurations.
62
-------
.c
>*-
o
IS)
•*—
o
CC
U.
o>
E
"o
E
o
E
10
Cross Section Configuration
10*
10 15
Enclosure Dimension (b), ft
20
25
FIGURE 22. MINIMUM VOLUME FLOW FOR
TURBULENT MIXING TO EXIST
63
-------
Enclosure Techniques
As mentioned earlier, there are two basic ways of collecting all
the EC vapors escaping through FRT seals, (see Figure 18). Conceptually, the
two methods require the installation of a covering either over the top of
the tank (portable roof) or locally at the seal location (local enclosure).
Based on the results of the preliminary engineering estimates presented
in the previous sections, it appears that for large tanks the use of
enclosure configurations that completely seal off the space between the
floating roof and the top of the tank (portable roof design) would be more
difficult to use than local seal enclosures. Specifically, the following
disadvantages are associated with portable roof designs:
(1) Concentration accumulation times are long because of
large enclosed volumes (see Figure 20).
(2) High volume flow rates of air are required to achieve
volume turnover times less than 1 hour (see Figure 21).
(3) High volume flow rates of air result in very low
equilibrium concentration levels (see Figure 21)
that could lead to overall measurement accuracy
difficulties.
(4) Obtaining uniform mixing of the HC vapors and air
is more difficult in large volumes.
(5) Construction, installation, and handling of a
large covering of this nature would be expensive
and difficult.
For the reasons mentioned above, the test procedures outlined in the
following section are in general directed toward enclosures that are placed
locally over the seal area of the tank. It should be noted that the test
procedures discussed in the following section could also be used to esti-
mate HC emission for fixed roof storage tanks that have been fitted with
internal floating roofs, as well as for floating roof tanks discussed here.
64
-------
Preparing a tank will involve installing the fabric material over
the seal as indicated above. Specific details concerning the fabric attach-
ment to the roof and tank have not been thoroughly reviewed at this time,
based on the results of preliminary discussions with BCL staff members in the
Macromolecular Science and Technology Section, it appears that composite
flexible fabrics that have low-HC permeability characteristics are commer-
cially available and can be attached to the tank surfaces using an adhesive.
There is also the possibility of using magnetic strips as an integral
part of the fabric covering which would give additional support to hold
the fabric in place while a secondary sealing compound was applied to
make the attachment vapor tighti The details and selection of an attach-
ment procedure that can be used easily and quickly to install and remove
a covering is left for the final engineering test design analysis.
Two test methods referred to as Rate of Change and Steady-State
Methods are discussed in the following subsection for estimating HC
emissions. Both methods require measurement of HC concentration levels
in HC/air mixtures.
Rate of Change Method
The rate of change method is a direct way of measuring HC emissions
through FRT seals. This method hinges on the assumption that an effective,
leak-tight covering can be constructed over the FRT seal. It is anticipated
that the enclosure will be well ventilated using a blower system which has
sufficient flow capacity to initially maintain a relatively low HC concen-
tration level in the enclosure. Air ventilation of the enclosure is
desirable during periods when tests are not being conducted to control and
maintain HC concentration levels below a flammable limit.
Figure 23 is a schematic of a proposed flow system configuration
that could be used to perform rate of change tests. When tests are in
progress, valves (1) and (3) would be closed and gas vapors in the enclosure
65
-------
Blower
Air Control,
Metering and
Monitoring Section
Enclosure
Volume(V)
Blower
Measurement
Section
FIGURE 23. SCHEMATIC OF FLOW SYSTEM
66
-------
would be circulated by the blower to promote mixing in the enclosure.
Recycling of the enclosure gases has the additional feature that gas
sampling can be performed at one station. If air was not circulated through
the enclosure, gas sampling at a number of locations in the enclosure
would be necessary to obtain a representative time history of the HC concen-
tration buildup which would sequentially be used to estimate total HC
emission losses. If makeup air is required, valve (T) would be opened to
feed a controlled amount of air to the blower. When tests are not in
progress, valve (2) would be closed and valves (1) and (3) opened to feed
fresh air through the system.
Gas sampling measurements can be performed in either the blower's
exhaust or intake line. A section of pipe will be used with appropriate
probes positioned within the pipe to monitor the HC concentration level.
Gas composition will be determined at various times during the test using
gas chromatography methods. Depending on the quantity of gas required for
gas composition analysis, air may be vented into the recycling system in
controlled amounts.
Steady-State Method
Although the rate of change method is a fairly direct way of
measuring HC emission rates, it is necessary that the total system volume
be accurately known (refer to Equation 8). The test procedures can be
slightly modified to eliminate the requirement of determining the system
volume. The suggested test arrangement for the Rate of Change Method
shown schematically in Figure 23 can be used to perform the tests in the
proposed Steady-State Method.
In this method, it is necessary to supply atmospheric air to
the enclosed volume at known flow rates. Referring to Figure 23, this
can be accomplished by closing valve (2) and opening valves (T) and (3).
Atmospheric air to the blower is controlled, metered, and monitored for
HC concentration upstream of the blower as shown in Figure 23. The
67
-------
exhaust gases from the enclosure flow to a Measurement Section and then to
atmosphere through valve (3) . The HC vapors are collected and uniformly
mixed with the air flowing through the enclosure. At the Measurement
Section, the HC concentration level in the exhaust gases and the gas flow
rate are measured. These two measurements are the only data required to
perform a total HC emission rate calculation. It is anticipated that the
volume flow rate of air to the enclosure can be held constant with time.
There should not be any major oscillations or variations in this quantity.
Therefore, the only variable that should vary with time is the HC concen-
tration level (excluding any atmospheric variations that may occur). It
will probably be necessary to time-average the concentration data over some
increment of time (5 to 60 minutes). This can be done efficiently by
electronic means so that on-line results can be reviewed while a test is
in progress.
The quantity of air flow supplied by the blower and the HC emis-
sion rate will affect the HC concentration level in the exhaust gases from
the enclosure. Estimated HC concentration levels for various HC emission
rates and quantities of air flow are shown cross-plotted in Figure 21.
Based on published estimates of tank emission factors of approximately 10
to 300 bbl/yr and a supply of air at 10,000 cfh, the HC concentration
level will vary from approximately 100 ppm to 5000 ppm. Supplying the
measuring air flow rates of approximately 10,000 cfh and measuring HC
levels over a range of 100 to 5000 ppm during field testing is well within
state-of-the-art technology. For instance, Beckman Hydrocarbon Analyzers
are capable of accurately measuring HC concentration levels over.a range
of 1 to 250,000 ppm.
The system time constant and/or system response time is a func-
tion of the size of the enclosure and the volume flow rate of air delivered
to the system. Calculations of the time required (t ) to supply equivalent
volumes of air equal to the enclosure volume were performed and the results
are presented as cross-plots in Figure 21. The so-called turnover time (tv)
calculations were discussed in the General Engineering Detail Section. In
68
-------
order to minimize the system response time, the air flow through the system
should be maximized and the enclosure volume minimized. In the process of
minimizing the system response time, care must be exercised to ensure that
the estimated EC concentration level is in a measurable range with respect
to instrumentation that can be used during the conduct of field experiments.
Based on the data presented in Figure 21, it appears that system response
times on the order of 1 hour or less can be achieved. In comparison with
other testing techniques, this would be a dramatic reduction in the elapsed
time required to perform a FRT emission measurement. A reduction in test
time from, say, weeks to hours is substantial and would greatly influence
the accuracy of any correlations done on the test data.
Based on the estimated system response time for an enclosure
constructed over a FRT seal, it appears possible that daily, cyclic emis-
sion profiles can be obtained. It is anticipated that hourly data measure-
ments will be suitable for determining emissions losses over this period
of time. This has not been possible in the past because of the lengthy
time period required to perform a test. The data obtained using the
proposed enclosure method would also lend itself quite nicely to performing
accurate correlations with other test parameters (mostly atmospheric and
tank conditions).
Figure 24 is a compilation of the results presented in the General
Engineering Analysis Section that shows a system operating envelope for the
Steady-State Method. The boundaries of the envelope are consistent with
the following: (1) minimum HC concentration level of 100 ppm, (2) maximum
HC concentration level of 1000 ppm, (3) minimum volume flow based on enclo-
sure cross-sectional dimension and critical flow Reynolds number, and
(4) maximum blower flow capacity. Also, implicitly included in this
envelope are the conditions of system time response of .less I.li;m I hour,
and maximum enclosure volume of 50,000 ft^.
Windage Simulation
As mentioned earlier, enclosing the space above the seal isolates
the seal from flow eddies generated by natural atmospheric winds blowing
69
-------
icr
0
5
(1)
E
_o
o
10
I I I l I I I
10
10
Hydrocarbon Emission Rate, bbl/yr
10-
FIGURE 24. ENCLOSURE SYSTEM OPERATING ENVELOPE
70
-------
over the tank. The influence of flow currents at the seal on emissions
is not known at this time. The results related to windage presented in
API Bulletin 2517^' are not detailed enough to accurately establish
windage effects.
In the proposed enclosure techniques described above, it will
be possible to vary the flow velocity in the vicinity of the seal by
varying the quantity of air flow through the enclosure—Figure 25 is a
plot of average flow velocity as a function of air volume flow rate for
various size enclosures. The boundary curve shown in Figure 25, labeled
critical Reynolds number condition, represents the minimum conditions for
volume flow and flow velocity for turbulent flow to exist in various size
enclosures. The system should, therefore, be operated at flow conditions
that lie to the right of the critical Reynolds number boundary to promote
good mixing of the gases (air and HC vapors).
Volume flow rates of air over a range from 10^ to 10^ cfh with
an appropriate blower system can easily be achieved; therefore, it is
readily possible to vary the duct mean flow velocity over a range approxi-
mately 0.1 to 1.0 fps (see Figure 25) for various size enclosures.
It is anticipated that a test series would be performed to
determine if the emission rate varies with changes in mean flow velocity
through the enclosure. The results of such a test series would have an
impact on estimating the influence of eddy currents near the seal.
Additional Data Requirements
Additional data will be required during tank emission loss tests
to perform correlation analyses. Atmospheric conditions such as tempera-
ture, barometer, wind velocity and direction, and solar radiation will be
obtained simultaneously with the emission measurement. These measurements
are more of a routine nature and therefore specific details are not
included.
Data related to the tank such as type of tank, seal design,
external surface condition, and internal wall condition will also be
recorded. Specifying the contents of the tank will also be necessary.
71
-------
10
£ '-°
o
o
c
o
CO
O
O
c
ID
.O
2
10'
Critical Reynolds Number
condition
Enclosure cross section
J I
I i I I I/
I i i i I I i i
I i I i I I I I
10° 10'
Air Volume Flow Rate, cf h
10*
FIGURE 25. MEAN FLOW VELOCITY IN VARIOUS SIZE ENCLOSURES
72
-------
Appropriate temperature data from the wall and roof of the tank
and the stored product will be obtained as deemed necessary. Specifica-
tion of the actual location for thermocouples is premature at this time.
Tracer Methods
Tracer methods were also evaluated as a means of measuring
emissions of hydrocarbons from petroleum storage tanks. The particular
tracer method developed would involve injecting a known amount of tracer
gas into the vapor space below the primary seal of the floating roof cover
and then measuring the depletion rate of the tracer gas. In this way the
known amount of tracer gas satisfies the same purpose as the known volume
enclosed above the seal in the enclosure method discussed earlier.
Such techniques have certain inherent advantages such as mini-
mizing the amount of modification necessary to the source, eliminating
interferences from other emissions sources near the one being tested, and
eliminating the need for obtaining vapor samples to determine vapor compo-
sitions that are required for interpreting and analyzing measured data
taken in the enclosure method approach.
However, these techniques also involve certain inherent disadvan-
tages. These can be summarized as follows:
(1) A tracer gas must be selected that is similar to
the hydrocarbon in diffusional or transport characteris-
tics, insoluble in the liquid to ensure that all
tracer that is lost to the atmosphere, and capable
of being detected separately from the hydrocarbon
with a minimum of interference. An example of such
a gas would be krypton, although the solubility of
krypton in petroleum liquids is unknown.
(2) When several tests are to be run in a relatively
short period of time, care would have to be taken
to minimize interferences from previous runs. If a
radioactive tracer such as krypton were used the
radioactive life (reported as half-life) would have
to be selected consistent with desired test times.
73
-------
(3) Any tracer method employed would be an indirect
method of measuring emissions. Arguments are bound
to exist over the similarity of emissions of tracer
gas to those of the hydrocarbons that are the purpose
of the test. The only way to resolve these arguments
would be to compare the results of the tracer tests
to those of more basic and direct methods such as
the enclosure method described earlier.
Optical Techniques for Measuring
Hydrocarbon Emissions
General Discussion
The monitoring of storage tank emissions centers on detecting
the presence of gaseous species above the tank seal or their loss from the
vapor space above the liquid product. Both the enclosure and tracer
method discussed previously can be used with conventional gas detectors
for measuring hydrocarbon or tracer gas concentration. Such analyzers
(such as the Beckman hydrocarbon analyzer) are well accepted and are
extremely accurate over the range of concentrations expected to be encoun-
tered. Optical methods are another means of providing this detection.
The general test procedure and data reduction process would be identical
to that used in other detection methods. The difference being that output
from optical sensors rather than other type sensors would be analyzed as
a function of time.
Optical techniques, in general, have several advantages over
other types of detectors that might be used to monitor hydrocarbon or
tracer gas emissions. Often laser beams are used in optical analysis.
Using mirrors to reflect the light around in a specific space, a large t)
area can be covered using a single source and a single detector. This
type of system would offer the best possibility for long term instrumen-
tation stability compared to a multi-detector approach. Also, since the
laser can be tuned to a specific wavelength, the possibility exists for
increased detection sensitivity for a broader range of specific gaseous
74
-------
species that might be considered for hydrocarbon or tracer gas constituents.
This degree of selectivity could be quite useful when attempting to monitor
the emissions from one type of product in the presence of interference
from dissimilar product emission sources. Of course, the interference from
like product sources would still be present. However, tunable lasers could
also be used to detect a greater variety of tracer gases. This could
prove important since solubility not detectability may be the driving
force behind picking a particular tracer gas. The possible disadvantage
to using optical techniques might be cost and ease of setup. Also, inter-
ference from sunlight might occur. However, those factors would have to
be weighed against the advantages of using optical methods and the cost
and complexity of comparable systems utilizing other detector techniques.
Details of Apparatus
Several optical detection methods are applicable to tank emissions
monitoring. The more promising techniques are
• Direct adsorption of light by gaseous species
• Differential absorption
• Induced fluorescence.
Other optical techniques have possible application but would be more diffi-
cult to apply in a field environment. Of the three techniques mentioned
above, the differential absorption method offers the most probable chance
of success. In this technique a single beam can be utilized as the light
source. This beam consists of radiation at two distinct wavelengths (see
Figure 9). One line at wavelength, A™ is at the absorption wavelength of
a particular gaseous species. The other nearby wavelength, ^QFF» is not
directly absorbed by the molecule of interest. However, the second beam
is attenuated by particulate (in solid and aerosol) matter in the test
region due to gross absorption and scattering. These unwanted effects
are also present in the absorbed beam. Hence, by subtracting the two
beams (i.e., differential absorption), the desired measurement of gaseous
absorption can be accurately obtained.
75
-------
Besides correcting for particulate and molecular scattering as
well as particulate absorption, this differential absorption technique
can be somewhat self-calibrating. As Figure 26 shows, the use of a chopping
wheel with various light filters can provide four distinct frequencies
for each off/on-in/out combination. Hence, a single detector with fre-
quency tuned amplifiers can be used to monitor intensities Ini71? mTT>
Urr,Uul
•'"OFF IN' ''"ON OUT' an<^ ^ON IN* ^e amount °f absorption can be calculated
from
., _. "^N.IN IOFF,IN ,-,.
Absorption = -— '- . (11)
ON,OUT OFF,OUT
.Using the setup .in Figure 9 and Equation (6), the effects of
source intensity drift and detector sensitivity variation can be accounted
for automatically.
Application to Storage Tank
Emission Analysis
To apply the differential absorption technique to storage tanks,
two modes must be considered. These are generally separable into "above"
and "below seal" application. Above the seal, a single source consisting
of one or more tunable lasers can be used with a single detector as shown
in Figure 27. A series of mirrors could be used to traverse the beam
around the circumference of the seal. Gaseous emissions from the vapor
space will be detected the moment they occur at any point where an emission
vapor cloud intersects the laser path.
Also, the integrated absorption along the entire path will l>e
indicative of the total amount of emissions leaking by the tank seal. The
monitoring of total integrated absorption with time will yield information
on the total leak rate past the seal.
Below the seal, the optical setup shown in Figure 27 would be
difficult (if not impossible) to construct. In this case, a compact double
or multipass source detector combination would be more applicalbe as
76
-------
DM(Xon)
DM(Xoff)
wheel Ion.'" Ion.out I0ff,in T0ff,out
Light pipe
Detector.
OM - ordinary mirror
DM(XR) - dichromic mirror
(ie reflects only XR)
'
1
Frequently
tuned
amplifier
»> Ion, in
». Ion, out
»• I0ff- in
+• I0ff, OUt
To absorption
calculator
FIGURE 26. BLOCK DIAGRAM OF DIFFERENTIAL ABSORPTION APPARATUS
-------
Mirrors
Detector
FIGURE 27. ABOVE THE SEAL OPTICAL DETECTION SYSTEM
78
-------
shown in Figure 28. Due to the short path length, a lower intensity source
could be used for below seal applications. Also, daylight interference
would be nearly eliminated.
A particular laser detector combination that looks promising for
tracer gas detection is a CO™ laser with line selecting capability coupled
with a liquid nitrogen cooled, gold-doped germanium detector. The tracer
gas could be sulfur hexafluoride SFg which has low solubility in hydro-
carbon fuels. The CC>2 laser has several lines which are absorbed by the
SFg and others that are not. Rapid switching from line to line is
possible using acusto-optic translator systems. All the above are off-
the-shelf components, so such a system could be put into operation with
minimal development costs.
Applications of differential laser absorption techniques to hydro-
carbons themselves is more difficult at this time. Some work is under way
on diode lasers which would be tunable in the correct wavelength region
to obtain absorption from many hydrocarbon fuels. Other laser hydrocarbon
absorption combinations are
• He-Ne 3.39y line for CH4
• C02-10.53 line for Cfa
• Iodine laser 3.93 line for C,H •
For the iodine laser at 3.93 ym, the C-H bond is measured. Therefore, the
absorption at this wavelength is indicative of the total amount of gaseous
organic material in the laser beam path. A more in-depth study of these
specific lasers should be undertaken in order to assess their applicability
for differential laser absorption measurements.
79
-------
Vapor
space
Source
Detector
FIGURE 28. BELOW THE SEAL OPTICAL
DETECTION SYSTEM
80
-------
CONCLUSIONS
• Current Techniques for Controlling Hydrocarbon
Emissions from Petroleum Storage Tanks
(1) The floating roof tank - with either external or
internal floating roof - is by far the most widely used
.device for controlling hydrocarbon emissions from
petroleum storage tanks. Advances in roof design
have been directed at improved safety, durability, ease
of construction, and low cost. No experimental data
were found to indicate that any particular design
(whether new or old) was more effective at controlling
hydrocarbon emissions than any other design.
(2) The most notable advances in floating roof design have
been in the seals between the roof and tank wall. The
trend has been to highly flexible, resilient fabric
seals that have greater ability to conform to irregulari-
ties in the tank walls than conventional mechanical seals.
These seals can be employed in single stage or in some re-
cent designs in two stage combinations. However, no ex-
perimental data were found indicating that any particular
seal was superior to any other in controlling emissions.
Claims as to seal effectiveness are based on appearance
and apparent closeness of fit between the seal and tank wall.
(3) Vapor recovery offers the potential for positive control
of hydrocarbon emissions from storage tanks. However,
these systems are complex compared to floating roofs and
their effectiveness at controlling emissions is contingent
upon their operating dependability. If the recovered
vapors can be used as a fuel gas or condensed to liquid
form vapor recovery offers the potential for product
conservation similar to floating roofs.
Analysis of Current API Correlations for
Estimating Hydrocarbon Emissions from
Fixed and Floating Roof Tanks
(1) In neither the fixed roof (API Bulletin 2518) nor the
floating roof cases (API Bulletin 2517) were basic
phenomenological or theoretical considerations used in
developing the correlations of hydrocarbon losses.
The equations developed in 2517 and 2518 are strictly
empirical representations of the data base used in
each case.
81
-------
(2) The data bases used in 2517 and 2518 were limited in
range and number of variables covered.
(a) In neither case was sufficient data available
on which to base a correlation on crude oil
losses and hence the gasoline loss equations were
used with somewhat arbitrarily chosen coeffi-
cients to correct for crude oil emissions.
(b) In neither case were diameters over 150 ft tested
nor stocks less than 1.9 psia true vapor pressure.
(c) For the fixed roof tanks (2518) the data base was
biased in that high vapor pressure stocks tended
to apply to larger tanks and low vapor pressure
stocks to small tanks. This could have led to
establishing misleading coefficients for the loss
equations.
(d) In all cases the tests used in establishing the
data bases were conducted prior to the publishing
of API Bulletin 2512 on standard test methods.
Therefore, the exact methods used in conducting
the tests are unknown, however, it was assumed
that the tests followed generally the methods
outlined in 2512.
(3) The accuracy claimed for the correlations in estimating
losses is misleading. In one case (breathing losses for
fixed roof tanks) the claim was made that the calculated
loss would be within 10 percent of the actual loss "in
most cases". The standard deviation of this equation
in reproducing its own data base was 35 percent, indi-
cating a greater error than was stated.
(4) Based on the above findings it is felt that the existing
correlations are not valid when extrapolating outside
the original data base (that is, to stocks less than
about 2.0 psia true vapor pressure, to tanks greater
than 150 feet in diameter, and to seals or tank features
that were not used in developing the test data).
Also, the accuracy of the correlations in predicting
emissions within the data base is questionable. More
accurate correlations could probably be developed without
great difficulty, however, this would probably not be
worthwhile in lieu of the inherent limitations of the
data base.
82
-------
Analysis of Current API Test Methods for Measuring
Hydrocarbon Emissions from Petroleum Storage Tanks
(1) The test methods described in API 2512are based on
first principals - that is, they involve direct
measurement of loss with basic instrumentation. As
such, the accuracy of these tests depends mostly on the
manner in which they are conducted and little chance is
seen for improving on their accuracy if the procedures
of API 2512 are strictly followed. The stated accuracy
of the tests is + 10 percent if the guidelines are
followed. No justification was given for the selection
of this degree of accuracy and a rigorous error analysis
would be necessary to substantiate it.
(2) The major drawback of the API test procedures is the
amount of time necessary for significant test results.
For floating roof tanks, test times of from 6 months
to 1 year are recommended for accurate results using
either the density change or vapor pressure change
method. These long test times require removing tanks
from service for extended periods (no stock additions
are allowed) and make it difficult if not impossible
to evaluate the effects of windage, temperature changes,
barometric pressure changes, and other variables
that change significantly many times during the test
period. Little opportunity exists for reducing the test
times required by these methods.
• Development of Short-Term Test Procedures
(1) Of the various short-term test technique possibilities
analyzed for floating roof tanks, the one that appears
most attractive for development is a local enclosure
over the seal space using either a steady state or
concentration buildup measurement technique. There
are several reasons for this conclusion.
(a) Calculations show that by using the local enclo-
sure technique significant test results can be
obtained with test times on the order of a few
hours or less for a wide range of tank sizes
and emission rates. Such short test times
would allow accurate determination of effects
of diurnal variables such as temperature and
barometric pressure.
The actual time a tank would be out of service for testing depends on
the extent of the test program (number of variables to be studied) and
setup time required.
83
-------
(b) The enclosure technique represents a direct
measurement of emissions from the total seal
area of the tank. Emissions measurements can
be performed with standard accepted flow
metering and hydrocarbon measuring instruments.
These considerations are important in establishing
the acceptability of emissions data that may be
developed in the future.
(c) Due to the presence of the enclosure, windage
effects would be eliminated. This may be an
advantage because in actual field tests windage
is an uncontrollable variable having an unknown
effect. In nonenclosure techniques this effect
would have to be correlated over a number of tests.
The enclosure technique would allow simulation
of windage as described in the body of the
report.
(d) The principals and instruments used in the
technique can also be adapted to determining
emissions from other sources besides the seal,
such as roof support legs, access hatches, and
sampling wells.
(e) The technique can also be adapted to fixed roof
tanks with internal floating covers by using
the existing fixed roof as the local enclosure.
(2) Other techniques involving relatively unproven technology
such as laser techniques or use of tracer gases should
still be considered as they may offer more versatility,
even further reduced test times, and lower test cost
than the enclosure technique. However, the results of
measurements made with unestablished technology would
probably require comparison to those with established
technology to be accepted. This further points up the
need for developing a direct, short-term test procedure
using established measuring methods.
84
-------
RECOMMENDATIONS
As a result of information gained and analyzed on this project
the following recommendations for future action are suggested for EPA
consideration. These recommendations are made in lieu of the immediacy
of the current petroleum storage regulatory problems in Southern California
and in lieu of the fact that the Western Oil and Gas Association (WOGA),
the American Petroleum Institute (API), and Chicago Bridge and Iron com-
bined with Standard Oil of Ohio (CBI/SOHIO) are all currently planning
their own emissions tests using various techniques.
(1) A program should be undertaken to develop and prove
the enclosure technique described in this report
on one or more small field tanks. It is suggested
that WOGA be approached for possible test tanks
which are already dedicated for testing. This would
also allow for comparison of emission results with
those of WOGA and CBI/SOHIO. The CBI/SOHIO mock-up
tank is also a possibility, though less attractive
than an actual field tank.
(2) Concurrent with the development of the enclosure
technique, a study should be undertaken to develop
a preliminary test plan for establishing a new data
base and emissions estimation procedure. This study
should be completed by the time the short-term test
technique is developed. The purpose of this study
would be to lay out a test schedule determining the
numbers of tanks, tank types, and test conditions
that should be evaluated. It would also outline a
procedure for correlating the data gathered in field
tests into a usable emissions estimating model.
(3) The results of the tests by WOGA and CBI/SOHIO should
be closely scrutinized by EPA. The results of these
tests will undoubtedly be compared with any future
data obtained by EPA. It is important that Kl'A have
close knowledge of the method and manner with which
the tests are conducted in order to explain any dif-
ferences that may occur and in order to evaluate the
validity of the data. This close knowledge could be
obtained by having trained observers visit the various
test sites for at least some of the tests to observe
the method of data acquisition and reduction.
85
-------
REFERENCES
(1) "Symposium on Evaporation Loss" Proceedings, American Petroleum
Institute, Vol. 32, 1952, pp 212-281.
(2) API Bulletin 2517: Evaporation Loss from Floating Roof Tanks,
American Petroleum Institute, 1962.
(3) API Bulletin 2518: Evaporation Loss from Fixed-Roof Tanks,
American Petroleum Institute, 1962.
(4) API Bulletin 2519: Use of Internal Floating Covers for Fixed-
Roof Tanks to Reduce Evaporation Loss, American Petroleum Institute,
1962
(5) API Bulletin 2512: Tentative Methods of Measuring Evaporation
Loss from Petroleum Tanks and Transportation Equipment, American
Petroleum Institute, 1957.
(6) API Bulletin 928: Hydrocarbon Emissions from Refineries, American
Petroleum Institute, 1973.
(7) API Bulletin 2515: Use of Plastic Foam to Reduce Evaporation
Loss, American Petroleum Institute, 1961.
(8) API Bulletin 2513: Evaporation Loss in the Petroleum Industry—
Causes and Control, American Petroleum Institute, 1959.
(9) Boardman, H. C., "Storage of Volatile Petroleum Products",
Petroleum Refiner, Vol. 25, No. 4, pp 109-116 [149-156], 1946.
(10) Bridgeman, P. W., "Dimensional Analyses", Harvard University
Press, 1921.
(11) Kline, S. J., "Similitude and Approximation Theory", McGraw Hill,
1965.
86
-------
APPENDIX A
LIST OF VENDORS CONTACTED FOR INFORMATION
ON TANK EMISSION CONTROL
-------
APPENDIX A
LIST OF VENDORS CONTACTED FOR INFORMATION
• ON TANK EMISSION CONTROL
Consolidated Aluminum Company
Nesquehoning, Pennsylvania
Larry Oxley 717-669-9411
Pittsburgh Des Moines Steel Co,
Neville. Island
Pittsburgh, Pennsylvania
Mike McLaughlin 412-331-3000 — internal roofs
Dean McKibben — floating roofs
Petrex
Warren, Pennsylvania
Mr. Bloomquist 814-723-2050.
Tresco
Sapulca, Oklahoma
Mark Englebrecht 918-22.4-9600
Mayflower
Little Ferry, New Jersey
Norm Plantz 201-641-0200
General American Transportation
Chicago, Illinois
Mr. O'Garrett 312-621-6200
Graver Tank and Manufacturing
East Chicago, Indiana
Robert Bodley 219-392-9300
Chicago Bridge and Iron Co,
Chicago, Illinois
Jim Mayer 312-654-7403
Robert Brown Associates
Carson California
Robert Brown 213-770-3630
ULTRAFLOTE Corporation
Houston, Texas
Ron Kern 713-461-2100
A-l
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-450/3-76-036
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Evaluation of Methods for Measuring and Controlling
Hydrocarbon Emissions from Petroleum Storage Tanks
5. REPORT DATE
November 1976
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
D. A. Ball, A. A. Putman, and R. G. Luce
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
10. PROGRAM ELEMENT NO.
BATTELLE
Columbus Laboratories
505 King Avenue
Columbus, Ohio 43201
11. CONTRACT/GRANT NO.
68-01-3159 Task No. 3
12. SPONSORING AGENCY NAME AND ADDRESS
U. S. Environmental Protection Agency
Research Triangle Park,
North Carolina 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final Report
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The purpose of this study was to determine advances made in petroleum storage
tank design for controlling hydrocarbon emissions, evaluate the validity of the API
correlations and test procedures when applied to modern tanks, and develop new test
procedures or modify existing procedures to conduct short-term emissions tests on
modern tanks. The results of the study showed that the floating roof is the most
common means of controlling emissions from storage tanks and that most advances in
design have been directed at increased safety, durability, and lower cost. No actual
experimental data were found to indicate that any particular tank emission control
system was better at controlling emissions than any other. It was also found that the
existing API data base and correlations are of questionable accuracy, especially when
applied to stocks with true vapor pressure less than 2.0 psia or tanks with diameters
greater than 150 feet. The API emissions test procedures were found to be technically
sound although they require extensive test times to obtain significant results. Little
chance was seen to increase accuracy or reduce the test time of these procedures.
Several new test procedures were analyzed and one was felt to have promise of
achieving accurate emissions results- in substantially reduced test times.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
Petroleum Storage Tanks
Hydrocarbon Emissions
I).IDENTIFIERS/OPEN ENDED TERMS
Fixed Roof Tank
Floating Roof Tank
Internal Cover
c. COSAFI Held/Croup
13, 08
8. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
20. SECURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
------- |