Environmental Integrity Project
1303 San Antonio Street, Suite 200
Austin, Texas 78701
512-637-9477 (phone)
512-584-8019(facsimile)
May 8, 2009
Via Electronic Submission: siprules@tceq.state.tx.us
Lindley Anderson
MC-206
Air Quality Division, Chief Engineer's Office
Texas Commission on Environmental Quality
P.O. Box 13087
Austin, Texas 78711-3087
Re:
Flare Task Force Stakeholder Group Public Comments
Dear Ms. Anderson:
On behalf of the Environmental Integrity Project (EIP), I appreciate the opportunity to
submit these comments to the Flare Task Force Stakeholder Group. In addition, I request
to be added as a member to the Flare Task Force Stakeholder Group and would like to be
notified of any additional Stakeholder Group meetings. Due to the ongoing and serious
public health consequences that result from the underreporting of flaring emissions, EIP
supports the agency’s goals to improve its understanding and regulation of flares.
Attached to this comment letter, please find the Data Quality Act Petition (Petition),
submitted to the U.S. EPA by the City of Houston which sets out in detail (technical and
legal) the need for revisions to the way that emission estimates are currently calculated at
refineries and chemical manufacturing plants. An April 7, 2009 response from EPA
headquarters is also attached. TCEQ’s effort to undertake a “comprehensive evaluation
of all aspects of flares” is an important step towards achieving more accurate
measurements and more appropriate permit limits in the refinery and chemical
manufacturing plant sectors. While the attached documentation addresses multiple
systemic flaws resulting from of the use of inaccurate emission factors at refineries and
chemical manufacturing plants, much of the documentation specifically addresses flares
and information relevant to the goals of the Flare Task Force Stakeholder Group.
As TCEQ continues with its evaluation, EIP urges the agency to pay particular attention
to significant known problems with estimating emissions based on current emission
factors. These factors are used as a basis to calculate emissions from flares in the
permitting process addressed by rules at 30 Tex. Admin. Code § 106.492 and 30 Tex.
Admin. Code Chapters 115 and 116. When assumptions that underestimate emissions
from flaring have been incorporated into individual permits, the state implementation
plan (SIP) then also suffers from invalid assumptions.
Two examples of significant problems with TCEQ’s use of current emission factors are
that (1) the factors incorporate an erroneous assumption that equipment is new and
operating under normal conditions and (2) emission factors do not account for
environmental variables that significantly impact emissions. With regard to the
assumption about operating conditions, EPA studies conducted in the 1980s used to
develop the emission factors specifically “excluded abnormal flaring conditions which
might represent large hydrocarbon releases during process upsets, start-ups and
shutdowns.”1 This is significant because the VOC emissions released from flares at
refineries and chemical plants during a single SSM event may actually exceed the
permitted annual average emissions. With regard to environmental variables, it is known
that flares become less efficient and destroy less VOCs, as wind speeds increase.2 Yet,
the emission factors for industrial flares were developed based on the assumption that 9899% of VOCs sent to the flare are destroyed.3 Specifically, it has been shown that the
ability of flares to destroy VOCs (i.e. the destruction efficiency) decreases rapidly as
wind speed increases from one to six meters per second.4 A study published in the
Journal of the Air and Waste Management Association (JAWMA) found that “[a]s wind
speeds increased beyond six meters per second, combustion efficiencies tended to level
off at values between 10 and 15%.5 The study further noted that “[t]heoretical
considerations and observational evidence suggest that flare combustion efficiency
typically may be at ~70% at low wind speeds (U ≤ 3.5 m/s). They should be even less at
higher wind speeds.”6
These are just two examples set out in the attached Petition. To the extent that staff has
not already reviewed the Petition, EIP urges the staff to carefully review the City of
1
See Robert E. Levy et al., Indus. Prof. for Clean Air, Reducing Emissions from Plate Flares (No. 61) 10
(Apr. 24, 2006) and pp. 11-12 of the attached Petition.
2
EPA, VOC Fugitive Losses, at viii (noting that “the emission factor for flare estimation is based on a
flare operating in still air conditions).
3
Douglas M. Leahey et al., Theoretical and Observational Assessment of Flare Efficiency, 51 J. Air &
Waste Mgmt. 1610, 1611 (2001).
4
Leahy et al., supra note 77, at 1611.
5
6
Id.
Id. at 1615. Houston’s Data Quality Act Petition, its exhibits A – E and the April 7, 2009 response
from EPA.
Sincerely,
/s/
Layla Mansuri
Attorney, Environmental Integrity Project
Enclosures
Attachments
This page intentionally left blank.
~jED STqT
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
a
y ~\l/gig(
C
02
~
~~~rqG PRO'1t!:S
APR 7 2009
OFFICE OF
AIR AND RADIATION
The Honorable Bill White
Mayor of Houston
Office of the Mayor
901 Bagby, 3rd Floor
Post Office Box 1562
Houston, Texas 77251-1562
Dear Mayor White :
for Correction (RFC 08003)
Thank you for your letter of July 9, 2008, filing a Request
Information Quality Guidelines (EPA IQG) . In
under the Environmental Protection Agency's
22, 2009, you cite concerns about the
that letter and your subsequent letter dated January
refineries and chemical plants . You
objectivity and utility of the emission factors pertaining to
to revise the emission factors subject
request that EPA: (1) immediately establish firm deadlines
data from direct observation and
to your petition, based upon reliable, accurate and unbiased
emission inventories; (2) require the use
other accurate measurements, in order to create valid
plants of cost-effective remote sensing
annually by large refineries and chemical manufacturing
verify emissions; and (3) require
technologies and installation of fenceline monitoring to
to document emissions reductions
refineries and chemical plants undergoing modification
installing pollution control
through the use of direct measurement if they wish to avoid
concerns about the accuracy of
equipment required under the Clean Air Act. We share your
stakeholders to improve emission
emissions estimates and hope to work with you and other
inventories at refineries and chemical plants .
a number of initiatives
As you are aware and as outlined in your request, we have
(including fenceline monitoring) and
designed to advance the use of remote sensing technologies
chemical plants . In addition, as a
better characterize emissions from petroleum refineries and
are planning to undertake a number of
direct result of the concerns outlined in your request, we
additional initiatives.
Ongoing and Planned Initiatives
measurement and analysis of
1) A grant was awarded in July 2008 to the City of Houston for
Objectivity, Utility, and Integrity of Information
' Guidelines for Ensuring and Maximizing the Quality,
Agency,
EPA, 2002 (67 FR 63657) .
Disseminated by the Environmental Protection
InfoQualityGuidelines .adf
http //www.epa :;ov/quality/informationguidelines/documents/EPA
Internet Address (URL) e http://www.epa.gov
Recycled/Recyclable 0 Printed with Vegetable Oil Based Inks on 100% Postconsumer, Process Chlorine Free Recycled Paper
volatile organic compound (VOC) and air toxics emissions in the Houston Ship Channel area
using DIAL (Differential Absorption LIDAR (Light Detection and Ranging)) technology .
This grant demonstrates EPA's support for additional data which Houston area stakeholders
can consider in making decisions to achieve improved local air quality. Additionally, the
data collected will help our understanding of these emissions nationwide . We look forward
to working with you in this effort to prioritize sources for assessment, to ensure the sources
are well characterized during the assessment, and to understand the results of the effort .
Finally, upon completion of this study (estimated to be in 2010), we will evaluate how best to
incorporate these results into future projects and ultimately into future emission estimation
guidance .
2) Prior to receipt of your Request for Correction, we had begun the development of a protocol
handbook (with detailed examples and case studies of previous projects) that would include
all essential aspects of undertaking a project using remote-sensing technologies for emissions
measurements including data quality objectives, quality assurance plans,
validation/verification, and data interpretation. Your request confirms the importance of
developing this type of handbook and we are committed to issuing a draft by the end of 2010.
Further information on this initiative can be obtained by contacting Dennis Mikel at (919)
541-5511 .
3) Subsequent to the completion of the DIAL remote sensing study that was conducted at the
BP Amoco facility in Texas City, Texas, we began evaluating the emission estimates from
the test data that was collected during that study. In addition, we will also evaluate data from
any future remote sensing studies. We believe these data are the appropriate data to review
as we improve emissions estimation methods rather than examining past remote sensing data
studies conducted at foreign petroleum refineries, where the refining practices may or may
not reflect the practices of domestic refineries and the emission sources were not well
characterized. We intend to provide a draft analysis of the BP Amoco data to the public for
review within the next 6 months. We plan to accomplish this by following the same
established procedures that we follow for soliciting public comments on draft emissions
factors. Specifically, we will post the draft analysis to our emissions factors web site
(http ://www.epa.gov/ttn/chief/efpac/abefpac .html) and notify individuals of the opportunity
to comment through our CHIEF Listserv service. Further information on this initiative can
be obtained by contacting Brenda Shine at (919)541-3608.
4) In direct response to your requests, in January 2009, we began the development of a
comprehensive protocol for the estimation of VOC and air toxics emissions from petroleum
refineries and chemical plants . This protocol will address all emissions sources and will
include startup, shutdown, and malfunction events . In developing the protocol, we will
review existing emission factors, including, but not limited to tanks, flares, and cooling
towers, and to refine or revise the emission factors as necessary. We plan to make a draft of
this protocol available for public review by following the same established procedures that
were explained in item number 3 above. In the future, we plan to use data derived from this
protocol to: a) evaluate risks to exposed populations; b) conduct comparisons to existing
emissions estimates (e.g., TRI) for specific facilities ; and c) better characterize the cost
effectiveness of controls . In addition, we will develop additional factors and methodologies
for additional emission sources including delayed cokers . This protocol will improve the
consistency, transparency and accuracy of future emission estimates for these facilities .
Further information on this initiative can be obtained by contacting Brenda Shine at (919)
541-3608 .
5) As part of our corrective action strategy to the 2006 EPA Office of Inspector General
Report,2 we have already developed tools such as the Electronic Reporting Tool (ERT) to
assist in improving the quality of our emissions factors . In addition, we will continue our
efforts to develop a self-sustaining emissions factors program that produces high quality
emission factors, quantifies the uncertainty of emissions factors, ensures the appropriate use
of emissions factors, considers stakeholder input appropriately, and improves emissions
quantification through the use of better tools and knowledge of uncertainty. More
information on the ERT can be obtained by visiting
http ://www.epa .gov/ttn/chief/ert/ert_tool .html, and more information on our efforts to
redesign our emissions factor program can be obtained by contacting Bob Schell at (919)
541-4116 .
Background
I believe our rationale for undertaking the initiatives outlined above is best explained by
first providing some background information on the purpose and intended use of AP-42
emissions factors. These factors are designed to be representative values relating the quantity of
a pollutant released to the atmosphere under normal operating conditions with an activity
associated with the release of that pollutant. By their nature, these factors are indicative of
situations that have broad applicability and, as such, were originally intended as a tool for use in
developing national, regional, state, and local emissions inventories . The idea of developing
emission factors to account for site-specific conditions such as upsets, start-ups and shutdowns is
counter to the definition of an emissions factor . We do not believe that updating emissions
factors to account for such site-specific events is the solution for improving emissions estimates
at refineries and chemical plants . We believe the issue is larger than just the quality and
coverage of specific emission factors and speaks to the need for a comprehensive protocol for
developing emission inventories. The protocol will combine emissions factors (to account for
emissions during periods of normal routine operations) with other engineering calculations (to
account for emissions during non-routine conditions) to allow for the estimation of facility-wide
emissions during any stages of operation at a facility . Ultimately, we believe the lack of such a
protocol can lead to omission of emission sources, improper characterization of process data and
subsequent emissions data, and inconsistent reporting from one facility to the next.
To illustrate our point, consider some of the more common emission sources at
petrochemical and petroleum refining facilities, such as storage tanks and flares . While AP-42
emission estimation equations exist for calculating working and standing losses from tanks, the
estimates resulting from these equations depend on whether the user has accurately characterized
the material stored in the tanks, the conditions of the fittings and seals, and the ambient
conditions surrounding the tanks . If these site-specific conditions are not properly characterized,
2 EPA Can Improve Emissions Factors Development and Management, U .S . EPA Office of Inspector General,
Report No . 2006-P-00017, March 22, 2006 . http ://www .epa .gov/oig/reports/2006/20060322-2006-P-00017 .pdf
the resulting emissions estimates will not be representative. Further, if short term inputs
resulting in short-term emission rates are then extrapolated to long term or annual emissions
without consideration of variability in operations or other conditions, resulting long term
emissions will not be representative . Even if we undertake a study to improve the emissions
equations, the inputs to these equations will always be site specific and will always affect the
quality and accuracy of the emissions estimates. Similarly, a VOC destruction efficiency of 98
percent is often used for flares . While this efficiency may not be achieved in practice under all
conditions (and this is an area where newer, state-of-the-art measurement techniques can inform
this debate), other factors, such as flow and concentration and variability over time, are just as
important to the emission estimate for a flare. Developing better flare emission factors will not
address these site-specific variables that are crucial to the overall estimates.
Therefore, in addition to improving specific emission factors for selected processes (e.g.,
emissions from delayed cokers), we believe that a more comprehensive approach to addressing
how facility-wide emissions estimates are conducted is needed to improve the overall accuracy
of future emission estimates. This approach, or protocol, would provide a consistent method for
selecting and applying emission factors, where available and appropriate, but also would provide
guidance on the use of other emission estimation methodologies that do not rely on emission
factors. It would address, among other things, minimum data quality objectives for process
inputs, coverage of emissions sources, calculation of non-routine events such as startups,
shutdowns and malfunctions, and inclusion of other information that would inform the estimates
such as temporal variability in processing operations .
We are committed to developing such a protocol for petroleum refineries and
petrochemical plants . As part of this effort, we would also review specific emission factors and
initiate work to refine, revise and develop additional factors and methodologies for emission
sources, including but not limited to tanks, flares, delayed cokers, and cooling towers . This
effort could include the use of optical remote sensing techniques to quantify emission sources as
well as startup, shutdown, and malfunction events that have been difficult to quantify . It will
also include a critical review of available remote sensing data, conclusions drawn from the
assessment, and an assessment/prioritization of sources for further study. Finally, we will also
attempt to validate any protocol with actual measurement data . We plan to work with you and
other stakeholders to undertake this project.
Finally, as noted in item number 5 above, we have embarked upon an effort to redesign
our current emissions factor program for both criteria and air toxics pollutants to (1) make the
development of emissions factors more self supporting and open to fuller participation by
external organizations; (2) increase the use of electronic means to standardize the development
process, quantify the quality components, and streamline all aspects of emissions factors
development and use; (3) make the emissions factors uncertainties and emissions quantification
methodologies more transparent to users; and (4) provide direction on the proper application of
emissions factors consistent with non-inventory program goals including clearer guidance and
direction on use of more direct quantification tools (e .g., emissions monitoring) in lieu of
emissions factors. We believe this effort will provide the foundation that will result in high
quality emissions factors based on a significant amount of data for many industrial sectors,
including the petroleum refining and chemical industry sectors.
We believe that the efforts we have initiated, especially the development of an emissions
protocol document, will allow for more accurate estimation of emissions from these types of
facilities . Although we have not provided firm deadlines for revising the emission factors for
petroleum refineries and chemical plants, this letter provides a status update and a timeline for
the completion of key tasks for each initiative . With respect to your request to require large
refineries and chemical manufacturing plants to change their current procedures, federal agencies
can not add additional requirements without a formal rulemaking. Before considering this
option, EPA would like to evaluate the data from the initiatives outlined in this letter to
determine the most effective way to enhance the estimation of emissions from large refineries
and *chemical manufacturing plants . In closing, we look forward to working with you to further
address this important issue, including establishing milestones and priorities for the development
of solutions to these important emissions estimation issues .
If you are dissatisfied with this response, you may submit a Request for Reconsideration
(RFR). The EPA requests that any such RFR be submitted within 90 days of the date of EPA's
response. If you choose to submit a RFR, please send a written request to the EPA Information
Quality Guidelines Processing Staff via mail (Information Quality Guidelines Processing Staff,
Mail Code 2811R-, U.S . EPA, 1200 Pennsylvania Avenue, NW, Washington, DC 20460);
electronic mail (quality@epa .gov); or fax [(202) 565-2441] . If you submit a RFR, please
reference the request number assigned to the original Request for Correction (RFC #08003).
Additional information about how to submit an RFR is listed on the EPA Information Quality
Guidelines website at httn://www.epa . ov/ uality/informationjauidelines/ .
Again, thank you for your letter . If you have additional questions, or require further
information on the IQG process, please contact Reggie Cheatham at (202) 564-7713 .
Sincerely,
Elizabeth CWig
Acting Assistant Administrator
This page intentionally left blank.
This page intentionally left blank.
Reducing Emissions From Plant Flares
Paper #61 – Revised April 24, 2006
Prepared by Robert E. Levy, Lucy Randel, Meg Healy and Don Weaver
Industry Professionals for Clean Air, 3911 Arnold St., Houston, TX 77005
ABSTRACT
Regulation of emissions from plant flares in Texas is based on flare efficiency studies
conducted by the US Environmental Protection Agency (EPA) in the early 1980’s, which
concluded that flare combustion efficiencies of 98 or 99 percent are achieved when
critical operating variables are controlled appropriately. However, recent studies suggest
that, even when well-controlled, flares may operate with efficiencies appreciably lower
than 98 percent due to crosswinds and other factors. Lower than assumed flare
combustion efficiencies, particularly during emission events, could account for a
significant portion of previously unrecognized emissions from refineries and chemical
plants and help to explain Houston’s high ozone levels. This paper discusses the state of
the art in understanding flare emissions and examines the specific shortcomings of the
current Texas flare regulations, including new regulations on highly reactive volatile
organic compounds (HRVOCs). In addition, it considers steps that could mitigate flare
emissions, and finally provides a list of recommendations for industry and regulators.
Recommendations include expanding research on factors affecting flare combustion
efficiency; improving monitoring and reporting of flare operating parameters, such as
steam assist and flare gas mass ratios; minimizing the volume of waste gases routed to
elevated, unenclosed flares; and encouraging the use of flare gas recovery systems or
wind-protected ground flares and thermal oxidizers.
INTRODUCTION
Houston is classified by the EPA as being
in "severe" nonattainment of the one-hour
ozone standard and in "moderate"
nonattainment of the eight-hour standard.
The Texas Commission on Environmental
Quality (TCEQ) has recognized a link
between episodic emissions of the type
associated with flaring and sudden
exceedances of the one-hour ozone
standard by enacting a new short-term
limit on highly-reactive volatile organic
compound (VOC) emissions. Ozone and
smog result from the reaction of VOCs with nitrous oxides in sunlight. Significant
quantities of VOCs are released from elevated flares, which burn waste hydrocarbons
primarily during emergencies and upset conditions.
1
In a 2000 annual summary of emissions, the TCEQ estimated that flares were responsible
for 12 percent of total emissions of volatile hydrocarbons in the Houston-Gulf Coast area,
based on an assumed 98 or 99 percent flare combustion efficiency.1 However, flare
burning efficiencies are not readily measured. Rather, VOC destruction efficiencies of 98
or 99 percent are assumed by the TCEQ2,3 and industry, based on experimental studies
completed by the EPA in the early 1980’s.
In 1986, EPA used the data from these studies to codify the requirements for flares under
the New Source Performance Standards (NSPS) in 40 CFR 60.18. The NSPS rule
specifies limits of critical flare operating variables that must be controlled to obtain 98
percent or higher combustion efficiency. These critical operating variables include heat
content of the flare fuel mixture, the ratio of fuel gas to assist gas (air or steam) and
burner tip velocity. In 1994, similar control device requirements were added to the
National Emissions Standards for Air Pollutants (NESHAP) in 40 CFR 63.11. Other than
the addition of a provision for hydrogen fueled flares in 1998,4 the requirements have
remained essentially unchanged for 20 years.
The TCEQ has not required reporting of operating data, except weight of total
hydrocarbon burned and "engineering estimates" of stream composition. With
inadequate operating data, 98 to 99 percent combustion efficiency cannot be realistically
assumed. Another operating variable, crosswind velocity, was not addressed in the EPA
studies, and more recent experimental work suggests crosswinds reduce flare combustion
efficiency. Although some independent research has recently been initiated by the
International Flare Consortium5, neither EPA nor TCEQ has undertaken significant largescale experimental work since the early 1980’s.
In this paper, we review the literature evaluating effects of operating parameters on flare
efficiency, as well as recent approaches in both industry and government to quantify and
reduce hydrocarbon emissions from flares. The authors believe serious attention to these
issues with enforceable goals is imperative if the Houston-Galveston area (HGA) is to
reduce its “smog day count.” Recycling of waste gases, rather than flaring, must be
seriously considered and flares should be reserved for essential use during unavoidable
emergency events.
The authors represent Industry Professionals for Clean Air (IPCA), whose members have
been affiliated with the petroleum or petrochemical and are concerned about the air
pollution in the Houston-Galveston region. Based on our experience and research, we
believe elevated flares present the most significant problems for controlling emissions of
VOCs and toxic air pollutants in our region. Our purpose is to make realistic
recommendations for reducing flare emissions that will encourage industry and the
regulators to take action.
2
EMISSIONS FROM PLANT FLARES
The Texas Commission on Environmental Quality (TCEQ) uses high destruction
efficiencies, based on combustion efficiencies established in the early 1980’s by the EPA
to establish regulatory requirements, calculate permit limits, monitor compliance, enforce
control requirements and plan for attainment of air quality standards. The TCEQ
presumes that flares destroy 99% of ethylene and propylene, and 98% of other VOCs,
except for certain compounds with less than 3 carbons, as long as continuous monitoring
data for the flare inlet demonstrates compliance with the EPA’s minimum heating value
and maximum exit velocity requirements specified in 40 CFR 60.18.6 Findings from the
EPA 1983 Flare Study generally reflect use of high-efficiency flares burning simple
chemicals at natural gas processing plants under optimal operating parameters and wind
speeds less than five miles per hour. 7 The TCEQ’s approach, therefore, makes no
allowance for real world operating variables. Specifically, it is based on the unrealistic
assumptions that:
•
•
•
plants are consistently operated according to the parameters necessary to optimize
flare destruction efficiency;
crosswinds have minimal effect on combustion efficiency; and
flares perpetually operate at high destruction efficiency.
In the following discussion we will examine these assumptions and develop suggestions
for adoption of more realistic ones.
Because flares are designed and used for control of emission spikes, the hourly emission
rate permitted8 and experienced by a flare is likely to be the highest of any unit at a
facility, even assuming a 98% to 99% VOC destruction efficiency. If realistic
efficiencies were applied, then the emission rates would be dramatically higher and might
account for much of the discrepancy between measured and model-predicted air pollution
in the Houston region.
Determine More Realistic Flare Destruction Efficiencies
Operating Parameters
As stated earlier, EPA work in the 1980’s established the basis for current federal and
Texas flare regulations. 40 CFR 60.18 and corresponding state regulations require that
flares operate:
•
•
“with a flame present at all times”,9 and
“with no visible emissions …, except for periods not to exceed a total of 5
minutes during any 2 consecutive hours.”10
The waste stream routed to the flare either burns on its own or, if it has low heating value
(less than 300 Btu/scf), with the assistance of a high-energy (more than 1000 Btu/scf)
fuel gas, like natural gas or propane, to facilitate complete combustion.11 Typically,
operators use fuel gas, or some other purge gas, to keep slow flowing emissions moving
3
toward the flare.12 With or without additional fuel, the combustion of many waste streams
produces smoke – i.e., visible emissions.13 For smokeless combustion, operators typically
inject steam or air to “achieve more complete combustion.”14 The injection of steam or
air (assist gas) “at the flare tip [also] increases the mixing of waste gas with air, as well as
the residence time of the waste gas constituents into the flame zone, thereby increasing
combustion efficiency.”15
Operators must maintain a delicate, but essential, balance between smokeless and
oversteamed emissions. Studies in the 1980s “demonstrated that assist gas to waste gas
mass ratios between 0.4 and 4 were effective in reducing soot while ratios between 0.2
and 0.6 achieved the highest hydrocarbon destruction efficiency.”16 Too much assist gas
(over steaming or over aerating) “may … reduce the overall combustion efficiency by
cooling the flame to below optimum temperatures for destruction of some waste gas
constituents, and in severe cases may even snuff the flame, thus significantly reducing
combustion efficiency and significantly increasing flare exhaust gas emissions.”17 The
EPA 1983 Flare Study noted: “Combustion efficiencies were observed to decline under
conditions of excessive steam (steam quenching) and high exit velocities of low Btu
gases.”18 Thus, EPA regulations establish parameters for heat content and exit velocity.19
The EPA 1983 Flare Study also demonstrated that separation of the flame from the
burner tip results in a serious drop in burning efficiency.20 This flame separation has
been observed during emergency flaring events under high winds and during addition of
excess steam. The reported loss of efficiency occurs because, under these conditions,
some of the gases do not remain in the combustion zone long enough for complete
conversion to carbon oxides. Some of the gases have the opportunity to partially or
totally bypass the combustion zone, with the result that unburned VOCs are emitted to
the atmosphere.
In addition, the TCEQ learned from a contractor’s evaluation of flare gas flow rate and
composition measurement methodologies that although “data on destruction efficiency
versus assist gas ratio obtained under controlled conditions would suggest that poor assist
gas control might negatively impact destruction efficiencies, there are little or no data
available on the impact of assist gas ratio control on destruction efficiency of operating
flares.”21 Thus, “the effect of assist gas to waste gas ratio on flare combustion efficiency,
as well as destruction efficiency, requires further investigation.”22 Based on a review of
some 50 refinery and petrochemical plant flares, and discussions with petrochemical
plant operators, the TCEQ learned that the assist gas injection rate for 90% of the flares is
controlled manually “by the operator based on [visual] flare observations (either directly
or on a video monitor).”23 Nevertheless, neither the EPA’s nor the TCEQ’s regulations
adequately address the critical role that steam content plays in flare combustion, and
apparently neither agency is actively investigating steam content control for flares in the
Gulf Coast region.
Furthermore, because the EPA 1983 Flare Study focused on simple hydrocarbons,
subsequent analyses may not take into account the possibility that while the original
compound may be destroyed, large hydrocarbons could simply be broken down into
smaller hydrocarbons and other compounds, some of which may be toxic as well.
4
An independent group, the International Flare Consortium, has initiated research focused
on exactly these issues in their project: "The effect of flare gas flow & composition;
steam assist & flare gas mass ratio; wind & flare gas momentum flux ratio; and wind
turbulence structure on the combustion efficiency of flare flames focusing on speciated
emissions of the highly reactive volatile organic compounds (ethylene, propylene,
butadiene) and the class archetypal hazardous air pollutant carcinogens (formaldehyde,
benzene, benzo(a)pyrene)."24
Upsets present even more of an operations problem. An evaluation of emission events in
the Houston-Galveston area between January 31 and December 31, 2003 “shows that
HRVOC events and possibly VOC emissions events have the potential to contribute
significantly to ozone formation in HGA.”25 A 2002 TCEQ toxicological evaluation of
VOC monitoring data collected downwind of three Harris County plants noted that
“exposure to recurrent elevated short-term levels of 1,3-butadiene may increase the risk
of reproductive and developmental effects.”26
Consider this specific example in which a large chemical complex reported 304 tons of
VOC emissions due to upsets and 622 tons
of VOC emissions total for the year 2000.
The applicable permit allowed only 124
tons of VOC emissions. Among other
emission events in 2000, this company
reported an upset, shutdown and startup
from July 17, 2000 through August 18,
2000. As part of the response to this upset,
the plant operator “maximized steam flow
to the flares to optimize combustion and
minimize smoke.”27
As noted above, too much steam can reduce combustion efficiency by cooling the flame.
A TCEQ study determined that an “assist gas to waste gas mass ratio between 0.2 and 0.6
achieved the highest hydrocarbon destruction efficiency.”28 The company cited above
reported that “[t]he hydrocarbon stream being flared during the July upset most likely
required a steam to hydrocarbon ratio of 0.7.” We do not have enough information to
accurately calculate the destruction efficiency of this company’s flare during the July
2000 upset, but experience suggests it is likely that the heat content was too low and the
exit velocity too high for the efficiency to be 98+%, as assumed in most of the
Upset/Maintenance Notification Forms filed regarding the incident.
The TCEQ’s new regulations regarding flares that burn HRVOCs assign 93% destruction
efficiency to flares not meeting the EPA’s standards for minimum heat content and
maximum exit velocity based on continuous monitoring.29 During the above-cited July
2000 upset, if a flare destruction efficiency of 93% is assumed, rather than 98%, the 304
tons of VOC emissions would become 1064 tons of VOC emissions. This represents 1.7
times the 622 tons of total VOC emissions reported at this plant during the entire year
2000. Moreover, reductions in residence time during startup and shutdown operations,
5
when flares operate at high rates for extended periods, may reduce combustion efficiency
substantially below the 93% provided for in the new regulations.
Crosswinds
The TCEQ’s assumed flare destruction efficiencies of 98+% also do not take into account
routine, yet less than ideal, weather conditions, such as crosswinds. An open flame, in the
absence of a crosswind, assumes a symmetrical shape of maximum volume having an
equilibrium flame temperature dependent upon operating conditions. Crosswinds distort
the flame, reducing flame volume and flame temperature. High combustion efficiency
requires that the combustible material be present in the high temperature region of the
flame for a significant period. Crosswinds in excess of 5 miles per hour, however, may
significantly degrade combustion efficiency because they shorten the residence time of
the combustible material in the flame.
The EPA 1983 Flare Study only conducted tests on flares at wind speeds up to 5 miles
per hour because flame instability made it impossible to obtain proper samples at higher
wind speeds.30 Consequently, there is a significant gap in the EPA field data, but labscale data suggests potentially significant reduction in combustion efficiency at high
wind speeds.31,32
Ongoing studies by the Engineering Department of the University of Alberta and the
Alberta Resource Council also demonstrate the need to consider the effects of crosswinds
on flares. The University of Alberta studies not only confirm findings in the EPA 1983
Flare Study regarding flame separation, they also conclusively demonstrate that
crosswinds can have a serious deleterious effect on the combustion efficiency of an open
flame.
Since significant crosswinds are usually
present along the Texas Gulf Coast,33 these
wind effects must be accounted for. Yet, the
TCEQ inappropriately dismissed the findings
from the University of Alberta research when
they reviewed the data in 2001 and 2002. We
requested internal documents from the TCEQ
relating to this review and found that the
TCEQ dismissed the entire body of research
from the University of Alberta based
primarily on the TCEQ Staff’s review of only
one 2001 study.34 In analyzing this study, the
TCEQ Staff concluded:
•
questionable simplifying assumptions
were made in the development of a
mathematical model from the
experimental work on a pilot-scale facility; and
6
•
poor flare destruction efficiency results obtained with field studies of a simple oil
field flare could not be extrapolated to more sophisticated plant flares “with
engineered burners and good liquid knockout systems.” 35
The University of Alberta researchers did not directly investigate commercial plant flares
with engineered flare tips, but the basic findings of this study indicate that crosswinds
affect combustion efficiency under a variety of circumstances. Thus, while we agree with
TCEQ’s specific critiques, it is inappropriate for them to exclude the basic research by
the University of Alberta on the basis that results of a field study of an oil field flare
could not be directly applied to Gulf Coast flares because of design differences.
Baylor University collected some samples in canisters during flyovers it conducted in
2001 for TCEQ, but apparently there has been no follow-up to this work. We have found
no documentation indicating that the EPA or the TCEQ subsequently considered the
effects of crosswinds on flares in policies or guidelines related to flares.
In the TCEQ Emissions Inventory Guidelines, in the technical supplement on flares
revised in 2004, TCEQ does acknowledge the potential for unstable flames in developing
the 93% destruction efficiency to be used when 40 CFR 60.18 requirements are not met36.
Nonetheless, neither the EPA nor the TCEQ routinely consider the critical variable of
wind speed in permit reviews, compliance investigations or emission reduction planning.
The entire question of crosswind impact on flare combustion efficiency appears to have
disappeared from their deliberations, without explanation, for more than two decades.
Research being undertaken by the International Flare Consortium37 is intended to directly
address the issue of crosswind effects on industrial flares and needs to be followed
closely by the EPA and TCEQ.
Performance Testing
The absence of further study or testing by the regulatory authorities is particularly
perplexing, since the TCEQ and the EPA acknowledge problems with accurately
estimating air emissions generally, and from flares in particular. The TCEQ “has
determined that [VOC] emissions may be underestimated in air shed emission
inventories.”38 These deficiencies are important because emission inventories are the
foundation for effectively controlling air pollution.39 And, since flare emissions represent
a significant portion of an industrial plant’s ozone-forming emissions,40 undercounting of
flare emissions could represent a significant portion of underestimated emission
inventories.
Flare emissions, however, are much more difficult to measure than those of other
pollution control devices. According to the EPA 1983 Flare Study, “Flare emission
measurement problems include: the effects of high temperatures and radiant heat on test
equipment, the meandering and irregular nature of flare flames due to external winds and
intrinsic turbulence, the undefined dilution of flare emission plume with ambient air, and
the lack of suitable sampling locations due to flare and/or flare heights, especially during
process upsets when safety problems would predominate.”41 In addition, the EPA 1983
7
Flare Study specifically “excluded abnormal flaring conditions which might represent
large hydrocarbon releases during process upsets, start-ups and shutdowns.”41
This, however, does not justify excusing the monitoring of flare emissions. Without
proper monitoring it is impossible to know whether flares are performing as expected.
The TCEQ expects “that emissions from flares would be better estimated if they were
based on waste gas flow rate and composition measurements. … The overall objective of
the [TCEQ] studies on flare emissions is to obtain performance specifications that ensure
quality assured sampling, testing, monitoring, measurement and monitoring systems for
waste gas flow rate, waste gas composition, and assist gas flow rate.”42 Modern insertion
meters can measure mass flow within +1%, and continuous composition analyzers are
readily available. However, measuring flows within an uncertainty of + 5% to 10% “in
flare systems with highly variable compositions or where the meter cannot be located in a
section of pipe with a representative flow profile will be a challenge.”43
Accordingly, the TCEQ now requires that operators of flares that burn HRVOCs – 1,3butadiene, butenes, ethylene and propylene – continuously monitor compliance with
“maximum tip velocity and minimum heat content requirements to ensure proper
combustion by the flare.”44 These new regulations do not adequately reduce flare
emissions, however, because:
•
•
•
•
•
In setting the appropriate assist gas flow rates and aggregate flow velocity, it is
important to know the composition of the flow. The TCEQ, however, does not
require continuous composition monitoring.
Most operators control assist gas injections manually, based on the visual
evaluation of the flame’s smokiness by the operator. Thus, depending on the skill
and attention of the operator, significant fluctuations in heating value and exit
velocities can occur over the course of an hour, such that substantial short-term
fluctuations in heating value could offset each other. One study notes that the ratio
of assist gas to waste gas with manual control varied from about 2 to more than
50.45 In this way, oversteaming can significantly reduce combustion efficiency
without violating the minimum heat value requirement for the one-hour average.
Although most flares are designed to be most efficient at the high volumes
experienced during non-routine operations, many are routinely used for disposal
of low-flow emissions.
The TCEQ presumes that “because many of these flares are also used for nonHRVOC streams, the regulations will result in better combustion of other VOC
streams as well. This improved combustion will reduce emissions of less-reactive
VOCs.”46 The TCEQ, however, did not make the continuous monitoring
requirement applicable to waste gas streams of other VOCs. So there is no quality
control on flares that burn only other VOCs and air toxics, which could represent
a significant volume of VOC emissions in the Houston-Galveston area.
The results of industry monitoring are not readily accessible to the public.
Although the San Francisco Bay Area has far fewer industrial flares emitting
much lower volumes of pollutants, the Bay Area Air Quality Management District
(BAAQMD) in California requires all refinery operators with elevated flares to
submit monthly reports of daily quantities (and species) of releases during the
8
•
•
period reported.47 The BAAQMD posts these reports, complete with graphs
illustrating daily spikes in emissions, on its website.48
Historically, TCEQ enforcement of monitoring requirements, if any, generally
comprised only minor recordkeeping violations.
The monitoring requirements on many flares with the potential for substantial
emissions are significantly weaker. Generally, these relaxed regulations require
only a combination of calorimeter, engineering calculations and process
knowledge for monitoring flares used for abatement of emissions from loading
operations, maintenance, startup and shutdown activities, emergencies, temporary
service, liquid or dual phase streams, and metal alkyl production processes.49
In addition, the type of continuous monitoring required by the TCEQ may not be
adequate. Flow measurement devices typically “calculate volumetric flow by sensing a
velocity in the pipe and multiplying that velocity by the cross sectional area of the pipe in
which the velocity is being sensed.”50 The accuracy of these measurements, however, is
based on assumptions that:
•
•
velocity is uniform across the cross section; and
the gas is of a known composition.
Thus, frequent changes in the waste gas composition could significantly marginalize the
quality of flare performance assessments.
Although safety concerns may preclude direct monitoring of emissions, parametric
monitoring and remote sensing techniques do exist which would provide data more
indicative of actual flare performance and emissions. For example, Open Path Fourier
Transformation Infrared (FTIR) technology “can identify, measure, and speciate over 100
compounds” from a distance of more than 100 meters.51 FTIR is particularly suited for
VOC identification and quantification because VOCs present strong absorption spectra in
the infrared region.52
In the near term, the TCEQ could follow the lead of California regulators in requiring
more extensive reporting of flare operations and emissions as a means to identify
priorities in reducing flare emissions and motivating operators to undertake emission
reduction projects sooner rather than later. Even before the BAAQMD issued its Flare
Monitoring Rule, its staff reported that flaring dropped dramatically because of increased
industry attention to flaring and flare monitoring.53
Similar observations were made in Southern California. Their monitoring rule, Rule
1118 – Emissions from Refinery Flares, was promulgated by South Coast Air Quality
Management District (SCAQMD) in 1998 and amended in November 2005. During the
period from 2000 to 2003, SOx emissions were reduced from 2633 tons to 735 tons with
only a fraction attributed to new equipment and the rest to expanded use of “ best
management practices.”54
These same data showed 79% of emissions were from unknown causes or nonrecordable
events. In response SCAQMD amended Rule 1118 to require a “Specific Cause
9
Analysis” of significant flaring events as defined by 1118 (c)(D), or an analysis of the
relative cause of “any other flare events where more than 5,000 standard cubic feet of
vent gas are combusted. (Rule 1118 (c)(E)). The revised rule also incorporates other
provisions to further reduce flaring emissions, such as mitigation fees and flare
management plans (1118 (d)).
Require Alternatives to Elevated Flares
For more consistent reductions in flare emissions over the long term, the TCEQ could
require alternatives to elevated flares. It is common practice for industry to use elevated
flares for routine destruction of vent gases or off-spec hydrocarbons, not just for
emergency or short-term releases. Most flares are built for non-routine events, such as
upsets, startup and shutdown, so they are not designed for optimal efficiency at low
temperatures and low flow rates.55 Consequently, routine flaring may result in
unnecessary emissions of HRVOCs, VOCs and toxic materials.
The TCEQ appropriately requires that many vent and relief valve emissions be
controlled, rather than vented to the atmosphere. Ideally, these routine emissions should
be recovered in a flare gas recovery system,56 which recycles the valuable components of
the waste stream, using an elevated flare only as a backup system.
Where gas recovery is impractical, we believe TCEQ should require operators to install
high efficiency combustion devices to handle all predictable demand. Enclosed ground
flares, incinerators and thermal oxidizers are acceptable alternatives because they can
consistently achieve high combustion efficiencies as a result of the enclosed firebox,
longer residence times at high temperature and negligible wind effects.
But high-efficiency combustion devices themselves need further attention from the
TCEQ as well. Like owners of motorized vehicles, operators should be required to
demonstrate the emission control performance of each device on an annual basis. After
the TCEQ gains experience with the results of such testing, the frequency for specific
classes of equipment, or particular companies, could be adjusted to ensure that testing
occurs at appropriate intervals.
While avoiding flaring of routine vent gases is important, minimizing episodic emissions
may be even more critical in reducing emissions of combustion byproducts, carbon
monoxide (CO), carbon dioxide (CO2) and nitrogen oxides (NOx). As demonstrated by
the example cited earlier, emissions from a single episodic event can exceed annual
average emissions. In reviewing emission events occurring during 2003, the University of
Texas’ Center for Energy and Environmental Resources found that the Houston
Galveston Area averaged more than one emission event per week: “Over an 11-month
period there are 58 times (affecting 395 hours) when ethylene event emissions exceed the
2000 annual average of 586 lbs/hr and 7 times (affecting 44 hours) when event emissions
exceed 5 times the annual average.”57 Unlike in the rest of Texas, and the rest of the
United States, emissions in Houston “change all the time,” and “[p]oor air quality [is] due
mostly to days with both ozone conducive meteorology and high emissions.”58 Hence
10
preventing unnecessary releases may provide the greatest decrease in overall VOC
emissions while also reducing emission of combustion byproducts, CO, CO2, and NOx.
In an effort to reduce such variable emissions, EPA Region 6, the Texas Natural
Resources Conservation Commission (TNRCC, predecessor to the TCEQ), the Louisiana
Department of Environmental Quality, and 13 petrochemical facilities in Louisiana and
Texas, participated in the Episodic Release Reduction Initiative. In 1999 and 2000, the
Initiative participants evaluated “the causes of releases to the air associated with
startups/shutdowns, equipment failures, and process upsets.”59
In the Technical Exchange on Startup/Shutdown practices, petrochemical facilities shared
case studies and examples of methods used to reduce flaring. Participants noted that
changes to procedures and training as well as design improvements could be used to
reduce emissions.60 Key findings on ways to reduce emissions include:
•
•
•
using flare gas recovery systems for routine venting and planned shutdowns;
improving training of operators, better documentation of procedures highlighting
environmental impacts, and allowing additional time for startup and shutdown;
and
reducing flaring among ethylene producers by recycling off-spec streams to
furnace feed, augmenting the plant’s steam capacity, and using a ground flare to
handle off-spec and startup loads.
Since that time, individual facilities in Texas have implemented site-specific programs to
reduce flaring. In 2001, the Dow Chemical Plant in Freeport, TX initiated a flare
minimization project at the Light Hydrocarbons plant. Before project implementation,
nearly all off-spec hydrocarbons at the unit, which includes an ethane/propane cracking
process, were flared. By optimizing equipment and procedures related to plant start-up,
shutdown, upsets and plant trips, including improving overall plant reliability, the plant
had an “89% reduction in overall upset flaring – using a two year running average.”
Further, from 2001 to the end of 2003, the plant achieved documented savings of $2.5
million.61
Also in Texas, Shell Chemicals developed a “parking mode” to reduce feed rates during
upset conditions in order to minimize flaring at its two ethylene units in Deer Park.
Implementation resulted in a 50% reduction in flaring between 2002 and 2003.62
In the San Francisco Bay Area, flare minimization projects and studies such as these are
now required of refineries regulated by BAAQMD under Regulation 12, Rule 12: “Flares
at Petroleum Refineries”, adopted July 20, 2005. This rule builds on their 2003 rule,
Regulation 12, Rule 11: “Flare Monitoring at Petroleum Refineries”. Flare minimization
plans submitted under Rule 12 must be approved by the Air District and “must include:
•
•
Detailed information about equipment and operating practices related to
flares,
Steps the refinery has taken and will take to minimize the frequency and
duration of flaring, and
11
•
A schedule of implementation of all feasible flare prevention measures.”63
TCEQ should consider implementing regulations similar to BAAQMD Rule 12 that
would encourage other facilities in Texas to follow the examples of Dow and Shell cited
above.
More extensive testing and reporting by plant operators on the operating parameters and
performance of flares and other waste gas combustion devices also would help the TCEQ
enforce existing regulations and identify priorities for reducing the use of elevated flare
stacks as emission control devices.
CONCLUSION AND RECOMMENDATIONS
We conclude that the TCEQ must take action
to determine more realistic flare destruction
efficiencies, minimize the volume of
emissions routed to elevated, unenclosed
flares, and encourage the use of flare gas
recovery systems, or wind-protected ground
flares and thermal oxidizers. Specific
recommendations are as follows:
1. Enforce existing requirements for flare
operations rigorously and consistently.
2. Expand and accelerate TCEQ, EPA and
others’ research on the factors affecting
combustion efficiency of flares, alternatives to flares and flare monitoring
technologies.
3. Revise TCEQ policies and guidelines for estimating flare emissions. At a minimum,
the effects of steam and crosswinds should be factored into emission estimates for
rulemaking, permitting, enforcement, reporting and planning activities. These effects
must be based on best available data rather than assumed values.
4. Conduct a rulemaking proceeding for regulations requiring more extensive
monitoring and reporting of flare emissions. At a minimum, operators should be
required to report daily emissions each month, and the TCEQ should post these
reports on its website.
5. Develop a strategy to increase the use of flare gas recovery systems or, where
impractical, use of more effective destruction technologies, such as enclosed ground
flares or thermal oxidizers, rather than elevated flare stacks, as emission control
devices.
6. Use elevated flare stacks only for release of combustibles in emergencies, for safety
reasons, or as necessary during planned startups or shutdowns of equipment.
12
7. Divert uncontrolled emissions from vents and relief valves to vapor recovery systems
and other alternatives to flares, with flares serving only as a backup system. The
TCEQ should set a goal for eliminating uncontrolled, authorized VOC emissions by a
specified date, and systematically review its regulations and permitting policies to
identify steps towards that goal.
8. Test high efficiency combustion devices, such as enclosed ground flares and thermal
oxidizers, regularly to demonstrate emission control performance.
REFERENCES
1. Gabriel Cantu, TCEQ, 2000 Houston - Galveston Speciated Point Source Modeling
Inventory, October 2003, Slide 17.
2. TCEQ publication RG-109 (Draft) Air Permit Technical Guidance for Chemical
Sources: Flares and Vapor Oxidizers, October 2000, pp.19, 35.
3. TCEQ,“Technical Justification For 99% Flare Efficiency,” attached as Appendix L to
Revisions to the SIP for the Control of Ozone Air Pollution, HGB Ozone
Nonattainment Area (HGB 2004 SIP Revisions), October 2004.
4. Federal Register May 4, 1998, pp. 24436-24437, Standards of Performance for New
Stationary Sources: General Provisions; National Emission Standards for Hazardous
Air Pollutants for Source Categories: General Provisions
5. James Seebold, Peter Gogolek, John Pohl, & Robert Schwartz, “Practical
Implications of Prior Research on Today's Outstanding Flare Emissions: Questions
and a Research Program to Answer Them”, Presented at AFRC-JFRC 2004 Joint
International Combustion Symposium, Environmental Control of Combustion
Processes: Innovative Technology for the 21st Century, October 10 – 13, 2004, Maui,
HI.
6. TCEQ, RG-109, pg. 19.
7. Flare Efficiency Study, EPA-600/2-83/052,. USEPA, Cincinnati, OH July 1983 (EPA
1983 Flare Study) Table 1.Flare Efficiency Test Results, p. 4.
8. URS Corp., Extraction of Allowable VOC Release Levels From TCEQ permits,
prepared for Houston Advanced Research Center Texas Environmental Research
Consortium, April 15, 2004.
9. 40 CFR §60.18(c)(2).
10. 40 CFR §60.18(c)(1).
11. TCEQ Work Assignment 5 Draft Flare Gas Flow Gas Rate and Composition
Measurement, Methodologies Evaluation Document, prepared by Shell Global
Solutions (US), Inc., p. 5-1. (Measurement Methodologies Evaluation).
12. Measurement Methodologies Evaluation, p. 1-6.
13. John F. Straitz, III, “Clearing the Air About Flare Systems,” Chemical Engineering,
September 1996, reprint, p. 5.
14. Straitz, p. 5.
15. Measurement Methodologies Evaluation, p. 5-1.
16. Measurement Methodologies Evaluation, p. 5-5.
17. Measurement Methodologies Evaluation, p. 5-2.
18. EPA 1983 Flare Study, p. ii.
13
19. 40 CFR §60.18(c)(3) and (4).
20. EPA 1983 Flare Study, Table 1, p. 4.
21. Measurement Methodologies Evaluation, p. 5-6.
22. Measurement Methodologies Evaluation, p. 5-2.
23. Measurement Methodologies Evaluation, p. 5-3.
24. Seebold, et al.
25. Cynthia Folsom Murphy and David T. Allen, “Event Emissions in the Houston
Galveston Area” (HGA), January 14, 2004 (Event Emissions in HGA), p. A-31,
available at www.harc.edu/harc/Projects/AirQuality/Projects/Status/H13.aspx.
26. Joseph T. Haney, Jr., and Laura Carlisle, Toxicology & Risk Assessment, Office of
Permitting, Remediation & Registration, TNRCC Interoffice memorandum to Dan
Thompson, Director, Region 12, Houston, July 31, 2002, p. 3.
27. Reference omitted to protect the company.
28. Measurement Methodologies Evaluation, p. 5-5.
29. 30 TAC §115.725(d)(7).
30. EPA 1983 Flare Study, p. 19.
31. M.R. Johnson, O. Zastavniuk, J.D. Dale and L.W. Kostiuk, “The Combustion
Efficiency of Jet Diffusion Flames in Cross-flow,” presented at the Joint Meeting of
the United States Sections – The Combustion Institute, Washington, D.C., March 1517, 1999.
32. Matthew R. Johnson, Adrian J. Majeski, David J. Wilson and Larry W. Kostiuk, “The
Combustion Efficiency of a Propane Jet Diffusion Flame in Cross Flow,” presented at
the Fall meeting of the Western State Section of the Combustion Institute,
Washington, October 26-27, 1998 (Paper #98F-38).
33. Houston’s average annual wind speed is 7.9 miles per hour and Galveston’s is 11.0
miles per hour. See the University of Utah Department of Meteorology’s Utah and
National Climate Data at
http://www.met.utah.edu/jhorel/html/wx/climate/windavg.html.
34. Douglas M. Leahey, Katherine Preston and Mel Strosher, Theoretical and
Observational Assessment of Flare Efficiencies, 51 J. Air & Waste Mgmt., 1610,
1611 (2001)
35. Karen Olson, Email to Terry Blodgett, et al., February 27, 2002, 11:31 AM (Olson
Feb. 27 Email) (from TCEQ Response to Open Records Request, March 29, 2005
(Mar. 29 Response).
36. TCEQ publication RG-360, 2005 Emissions Inventory Guidelines, Technical
Supplement 4; Flares, January 2006, p. A-46.
37. International Flare Consortium web site: URL
http://home.earthlink.net/~international-flare-consortium/index.html. Accessed March
2006.
38. Measurement Methodologies Evaluation, p. E-1.
39. TCEQ Science Synthesis Committee, “Accelerated Science Evaluation of Ozone
Formation in the Houston-Galveston Area,” November 13, 2002, p. 4. An analysis of
scientific data on ozone formation in the Houston-Galveston area as part of the
TCEQ’s Texas Air Quality Study in the summer of 2000.
14
40. Cantu, TCEQ, 2003, Slide 17.
41. EPA 1983 Flare Study, p. 1.
42. Measurement Methodologies Evaluation, p. E-1.
43. Measurement Methodologies Evaluation, p. 6-1 to 6-2.
44. HGB 2004 SIP Revisions §1.6.2.1 Collateral VOC Reductions.
45. Measurement Methodologies Evaluation, p. 5-4.
46. HGB 2004 SIP Revisions §1.6.2.1 Collateral VOC Reductions.
47. Bay Area Air Quality District Regulation 12-11-401.
48. URL http://www.baaqmd.gov/enf/flares.
49. 30 TAC §115.725(e)-(k).
50. Measurement Methodologies Evaluation, p. 2-1.
51. Survey and Demonstration of Monitoring Technology for Houston Industrial
Emissions (Project H31.2004) ENVIRON International Corporation. Prepared for
Houston Advanced Research Center, January 12, 2005, pp. 3-12 to 3-13 (Monitoring
Technology for Houston).
52. Monitoring Technology for Houston, p. 3-16.
53. BAAQMD Staff Report, Regulation 12, Rule 11, p. 31-32.
54. SCAQMD Summary Evaluation Report on Emissions from Flaring Operations at
Refineries, Version 1, September 3, 2004.
55. Matthew R. Johnson, et al. (University of Alberta), “The Combustion Efficiency of a
Propane Jet Diffusion Flame in Cross Flow,” presented at the Fall Meeting of the
Western States Section of the Combustion Institute, Washington, October 26-27,
1998, p. 11.
56. P.W. Fisher and D. Brennan, “Minimize Flaring with Flare Gas Recovery,”
Hydrocarbon Processing, June, 2002, p. 83.
57. Event Emissions in HGA, p. A-21.
58. Harvey Jeffries, et al. Stochastic Emissions Inventories for Houston Point Sources,
Concepts and Examples, presentation to TCEQ, October 2000, Slide 2, available at
URL http://www.airchem.sph.unc.edu/Research/Projects/Texas/MCCG/ (emphasis in
original).
59. The Episodic Release Reduction Initiative, July 5, 2001 (ERRI), p. 1, URL
http://www.epa.gov/earth1r6/6en/a/erri07-5fin.pdf.
60. ERRI, Appendix F, pp.32-36.
61. Steven Krietenstein, “Flare Minimization Strategy During Plant Upsets: Freeport”
presented at 2005 AIChE Spring National Meeting, 17th Annual Ethylene Producers’
Conference, Session TA009 – Ethylene Plant Operations, Atlanta, GA, April 12,
2005.
62. Nicholas Genty and Bryce Kagay, “Development of a Parking Mode at Shell
Chemical’s Deer Park Plant Olefin Unit OP-III, presented at 2005 AIChE Spring
National Meeting, 17th Annual Ethylene Producers’ Conference, Session TA009 –
Ethylene Plant Operations, Atlanta, GA, April 12, 2005.
63. BAAQMD Press Release July 20, 2005, “Air District Board Adopts Refinery Flare
Rule”.
15
KEY WORDS
flare, combustion, emissions, combustion efficiency, destruction efficiency, air pollution,
crosswinds, ozone, ozone-forming emissions, HRVOC, VOC, elevated flare, ground
flare, thermal oxidizer, flare minimization, flare gas recovery, refinery, petrochemical,
TCEQ, BAAQMD, University of Alberta. Alberta Resource Council, HoustonGalveston, Gulf Coast, FTIR, International Flare Consortium
16
This page intentionally left blank.
FINAL REPORT
A Review of Experiences Using DIAL Technology to
Quantify Atmospheric Emissions at Petroleum Facilities
PREPARED FOR
Environment Canada
Pollution Data Division
Science and Risk Assessment Directorate
Science and Technology Branch
351 St. Joseph Blvd., 9th Floor
Gatineau, QC
K1A 0H3
Contact:
Roy McArthur
Telephone:
Facsimile:
E-mail:
(819) 953-9967
(819) 934-4158
Roy.McArthur@ec.gc.ca
PREPARED BY
Clearstone Engineering Ltd.
700, 900-6 Avenue S.W.
Calgary, Alberta, T2P 3K2
Canada
Contact:
Telephone:
Facsimile:
E-mail:
Website:
David Picard
1 (403) 215-2730
1 (403) 266-8871
david.picard@clearstone.ca
www.clearstone.ca
September 6, 2006
Final Report
EXECUTIVE SUMMARY
This report presents the results on a technical literature review of Canadian and international
experiences regarding the application of differential absorption lidar (DIAL) for the
measurement of emissions from petroleum facilities.
Preliminary results from fugitive emission measurements undertaken as part of a DIAL
demonstration project at a petroleum refinery in Western Canada indicate that these emissions
may be significantly greater than the values estimated using currently established inventory
methods. Similarly, DIAL measurement studies conducted during 2003 and 2004 in the upstream
oil and gas sector (i.e., by Alberta Research Council and Sectrasyne Ltd., working with CAPP
and PTAC) indicated that the emission estimates derived using currently established methods
may significantly under estimate volatile organic compound (VOC) emissions. The fugitive
emissions from two of the gas plants surveyed were 4 to 8 times the mass emissions estimated
based on installed equipment and standard industry emission factors, the current NPRI reporting
method. Process flares typically were the source of 10 to 15% of the methane emissions from
these sites. These were the first DIAL measurements of this type conducted in North America.
Furthermore, U.S. EPA Inspector General recently published a report stating that current
methods of estimation based on emission factors are not accurate and lead to significant
underreporting 1 .
In an attempt to facilitate the analysis of the implication of this recent information, Environment
Canada (EC) commissioned this literature review to provide a background document covering
the following topics:
1. The European Commission IPPC Bureau’s Integrated Pollution Prevention and Control
(IPPC) Reference Document on Best Available Techniques on Emissions from Storage
(draft January 2005 available) and elucidate on recommendations and limitation for the
use of DIAL to update emission factors and monitor emissions.
2. The DIAL study results for the Canadian upstream oil and gas sector and for the Western
Canada petroleum refinery.
3. The European experience with DIAL (e.g. history and rationale of DIAL development,
legal requirements to use DIAL, scope and frequency of such measurements for industrial
facilities, uncertainty of DIAL measurements, measurement protocols and data quality
assurance and control, facility level measurement results).
4. The current U.S. opinion and/or conclusions on the potential for application of the DIAL
technology and other assessments that indicate significant underreporting of emissions by
industrial facilities (e.g. magnitude, reasons for underreporting, emission sources affected
by underreporting).
5. Any outstanding technical issues that must be resolved.
6. Potential impact of all of this information on the Canadian VOC emission estimates.
1
Source: US. EPA. 2006. EPA Can Improve Emissions Factors Development and Management. Report. No. 2006P-00017. Prepared by US EPA Office of Inspector General, March 22, 2006. .pp 37.
i
Final Report
TABLE OF CONTENTS
Section
Page
1.0
INTRODUCTION.........................................................................................................................................1
2.0
AN OVERVIEW OF THE DIAL TECHNOLOGY...................................................................................2
2.1
2.2
2.3
2.3.1
2.3.2
2.3.3
2.3.4
2.3.5
2.3.6
2.3.7
2.3.8
2.4
2.5
2.6
3.0
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
3.9
4.0
4.1
4.2
5.0
BASIC METHOD ...........................................................................................................................................2
EMISSION QUANTIFICATION PROCEDURES ..................................................................................................3
FACTORS INFLUENCING DETECTION LIMITS AND ACCURACY .....................................................................3
Distance From Source ...........................................................................................................................4
Spatial Resolution..................................................................................................................................4
Interferences from Other Compounds....................................................................................................4
Optical Noise .........................................................................................................................................5
Aerosol or Particulate Distribution .......................................................................................................5
Interference from Nearby Sources .........................................................................................................5
Data Averaging......................................................................................................................................6
Extrapolation of Results.........................................................................................................................6
APPLICATIONS .............................................................................................................................................7
MANUFACTURERS .......................................................................................................................................7
ADVANTAGES, DISADVANTAGES AND LIMITIATION ....................................................................................8
EXPERIENCES WITH DIAL ...................................................................................................................10
BELGIUM ...................................................................................................................................................10
CANADA ....................................................................................................................................................10
CZECH REPUBLIC ......................................................................................................................................11
EUROPEAN COMMISSION ...........................................................................................................................12
GERMANY .................................................................................................................................................13
SWEDEN ....................................................................................................................................................13
THE EUROPEAN UNION NETWORK FOR THE IMPLEMENTATION AND ENFORCEMENT OF ENVIRONMENT
LAW (IMPEL)...........................................................................................................................................14
UNITED KINGDOM .....................................................................................................................................15
UNITED STATES .........................................................................................................................................16
CONCLUSIONS AND RECOMMENDATIONS ....................................................................................18
CONCLUSIONS ...........................................................................................................................................18
RECOMMENDATIONS .................................................................................................................................19
REFERENCES CITED ..............................................................................................................................20
ii
Final Report
LIST OF ACRYNOMS
DIAL –
DOAS FTIR IMPEL IR LASER LIDAR –
NPRI OP PI RADAR ROMT ROSE SODAR TDLAS UV VDI VOC -
Differential Absorption LIDAR
Differential Optical Absorption Spectroscopy
Fourier Transform Infrared Spectroscopy
European Network for Implementation and Enforcement of Environmental Law
(An informal Network of the environmental authorities of member States)
Infrared
Light Amplification by Stimulated Emission of Radiation
Light Detection and Ranging.
National Pollutant Release Inventory
Open Path
Path Integrating
Radio Detection And Ranging
Remote Optical Sensing Techniques
Remote Optical Sensing Evaluation
Sonic Detection and Ranging
Tunable Diode Laser Absorption Spectroscopy
Ultraviolet
Verein Deutscher Ingenieure (The Association of German Engineers)
Volatile Organic Compound
iii
Final Report
1.0
INTRODUCTION
This study presents a general overview of DIAL and the experiences in Canada and
internationally in its application for detection and quantification of atmospheric emissions at
petroleum refineries and other facilities or sources.
Section 2 delineates the DIAL method, discusses some of the factors that influence the method’s
detection limits and accuracy, lists its potential applications, highlights key advantages and
disadvantages, and lists some of the manufacturer’s of DIAL systems.
Section 3 discusses the experiences and findings of different researchers, in Canada and
internationally, applying the DIAL technology. Relevant standards, guidelines, best practices and
regulatory requirements are noted. The conclusions and recommendations of this report are
presented in Section 4 and all references that have been cited are listed in Section 5.
1
Final Report
2.0
AN OVERVIEW OF THE DIAL TECHNOLOGY
2.1
Basic Method
Differential absorption LIDAR (DIAL) is an open-path optical sensing technique used for the
remote measurement of trace gases in the atmosphere. It offers the unique ability to rapidly map
pollutant concentrations in both two and three dimensions using a single instrument (i.e., laser
sounding). A volume of several cubic kilometres surrounding the instrument location can be
mapped, and a target plume cross-section can be mapped in minutes. Moreover, DIAL allows
emissions to be monitored where physical access is difficult or hazardous, including high
elevation plumes, and there is negligible disturbance of the plume by the measurement. DIAL is
often used as a research tool to obtain detailed and fast-repeating measurements of important
plume quantities, such as plume spread, plume meandering, instant concentration profiles and
cross-sections.
DIAL systems are available as a truck mounted mobile laboratory, and have also been installed
in aircraft.
DIAL can measure simultaneously in the infrared (IR), visible and ultra-violet (UV) spectral
regions and provide real-time data for any gaseous species with characteristic absorption in these
spectral regions including: SO2 , NO2 , NO, Ozone, Benzene, Toluene, Xylene and higher
aromatics, Alkanes, Alkynes, petroleum and diesel vapours, Hg, HCl, N2O, HF and H2S. Other
uses include the measurement of ambient concentrations of aerosols and opacity measurements.
DIAL is an important advance on the more conventional optical line monitoring systems such as
differential optical absorption spectroscopy (DOAS) and fourier transform IR (FTIR)
spectroscopy in which a retro-reflector, which must be re-positioned after each measurement, is
used to return the laser beam to the detector. In these conventional systems an average
concentration of the species to be measured is obtained and range resolution is not possible,
which is a significant limitation. DIAL also uses a coherent light source to measure not just
contents of a direct path or line, but full 3D volumetric data. The downside is that the pulse has
to be strong and the receiver large to cover the typical target ranges of several kilometers.
DIAL relies on back-scattered laser light using a general method known as light detection and
ranging (LIDAR). LIDAR is like RADAR but instead of microwaves it uses light in the infrared
(IR), visible and ultraviolet (UV) ranges. A pulsed laser beam is sent out into the atmosphere and
small proportions of the light are backscattered by particles along the beam path to a sensitive
detector (or optical telescope). The dust particles and aerosols present in the atmosphere serve as
reflectors. The laser light is in short pulses and time resolution of the backscattered light (along
with the speed of light) gives range resolution.
DIAL relies on the unique "fingerprint" absorption spectrum of each molecule and measurements
are usually made on a single compound at a time. The particle backscatter light is measured for
two wavelengths where the target absorbs strongly and weakly, respectively. The selection of
more than two wavelengths is a mathematical necessity for simultaneous measurement of
multiple species or for resolving interference effects between a target compound and a
background gas such as water vapour or carbon dioxide (Weibring et al, 2004). This is especially
2
Final Report
true in the mid IR region, where many hydrocarbon compounds have overlapping spectral
features.
The concentration of the target substance is determined based on the size of the differential
return signal at different distances along the laser beam path. The time history of the return
signals provides the range from the transmitter/receiver.
The strength of the signal received by the DIAL system depends on the distribution both of the
target gas and of aerosol. These vary depending upon the nature of the source being investigated.
The ability to range resolve DIAL to measure the concentration of gaseous species is determined
by both hardware and data processing considerations (Warren, 1989). The latter must perform a
number of functions, including signal averaging, transmit energy normalization, plus shape
deconvolution (if needed), path-integrated concentration estimation by the familiar log-ratio
DIAL algorithm, and, finally, numerical differentiation to produce the concentration estimate and
its uncertainty as a function of range. Because raw concentration estimates are intrinsically
noisy, the algorithm chosen to perform the differentiation is of critical importance. This is
particularly true in a dynamic environment, where only limited pulse averaging can be performed
prior to the estimation, either because a large volume must be monitored quickly or because the
concentration of the target species changes rapidly.
2.2
Emission Quantification Procedures
The mass emissions of a target substance from a process or fugitive source of interest may be
determined by making a series of DIAL scans vertically at a right angle to the wind to locate a
the plume and obtain the concentration profile across the plume cross-section, while at the same
time measuring local meteorological conditions. Normally wind speed and direction
measurements are taken with equipment located on the ground. Some researchers (e.g.,
Weibring, 1998) have developed a remote sensing technique (wind videography) and combined
it with DIAL measurements.
The compiled concentration and wind speed data are combined to produce a mass emission
profile for a whole site; for instance, for fugitive emissions from an oil refinery. A representative
“upwind” or “clean-air” flux from the recorded downwind data is then subtracted from the
results to determine the final emissions rate. If there are no potential sources upwind of the plant
being surveyed, it is sufficient to subtract a single clean-air column to allow for system offsets.
Otherwise, a further correction can be applied by subtracting a measured upwind flux. In this
case, care is needed to ensure that only the relevant portion of the upwind mass flow rate is
subtracted.
2.3
Factors Influencing Detection Limits and Accuracy
DIAL is capable of measuring gas concentrations of a few ppm per metre. Thus, the minimum
detection limit is several ppm for spacial mapping at a resolution of 1 m. At a coarser resolution
of 100 m, the minimum detection limit is on the order of a few tens of ppb.
3
Final Report
A typical DIAL measurement has an accuracy better than 10 percent and <5 mg/m3·m. However,
the accuracy is very much determined by the weather conditions and other atmospheric
parameters. The determination of emission rates using DIAL is less accurate since uncertainties
in wind profiles and source variability are also introduced. For example, Egeback et al (1984)
report uncertainties of 30 percent in their results due mainly to uncertainties in the wind velocity
determinations.
The following sections delineate some of the key factors that influence DIAL detection limits
and the accuracy of emission rate determinations, namely:
•
•
•
•
•
•
•
•
Distance form the source.
Spatial resolution applied.
Interference from other compounds.
Optical noise.
Aerosol or particulate distribution.
Interference from nearby sources.
Data averaging.
Extrapolation of results.
2.3.1
Distance From Source
The plume is usually measured sufficiently far downwind that mixing within it is fairly
uniform and recirculation and other wake effects have died away. However, a
compromise must be made between accuracy, which improves with distance from the
source, and sensitivity which decreases with distance from the source. Walmsley and
O’Connor (1998) report that: depending on the compromise, and conditions at the time,
the uncertainty in the emission rate measurement may vary from 20 percent or better
associated with controlled release experiments in un-congested conditions to a factor of
four associated with the use of oversimplified wind data in congested areas. For large
emissions (i.e., tens of kg/h and above) it is normally possible to make measurements at
the accurate end of this range by measuring at a large distance from the source. For
smaller emissions, where measurements must be made relatively close to the source, the
achievable accuracy is often less favourable.
2.3.2
Spatial Resolution
The final accuracy of a measurement depends greatly on the number of measurement
lines. Walmsley and O’Connor (1998) recommend operating with a 10 m resolution; it is
usually best to avoid 2.5 m to reduce noise and 30 m or 100 m because of the poor
localization of gas and the inability to recover quickly from disturbances. The latter is
important because disturbances due to steam leaks or hard-target returns from pipes,
cables, etc. are often unavoidable and recovery takes more than three times the spatial
resolution. The extended response to disturbances has usually prevented good quality
measurements at 30 or 100 m in plant areas.
2.3.3
Interferences from Other Compounds
4
Final Report
There are significant overlaps in the absorption spectra of the different hydrocarbons that
may be detected by DIAL, as well as interference effects from water vapour (Weibring et
al., 2004). Such interferences or cross-sensitivities may compromise the accuracy of the
measurement results when making measurements on unknown mixtures such as the
cocktail of fugitive hydrocarbons from a refinery. Walmsley and O’Connor (1998) have
dealt with this by making measurements using the butane absorption coefficient and then
correcting the results using the species ratios measured by absorption tubes and gas
chromatography together with the absorption coefficients in the DIAL system’s spectral
database. For a typical refinery mixture the correction factor for total alkanes relative to a
simple as-butane interpolation has been determined to be about ±5 percent.
2.3.4
Optical Noise
The accuracy is greater for a nighttime recording in an atmospherically stable area. At the
other extreme, measurements are not at all possible if the visibility is dramatically limited
by fog or rain. Increasing the laser pulse power improves the accuracy somewhat and
allows the measurement range to be increased.
For a given concentration of gas, the detectable range reportedly improves by more than
50 percent during the night due to the reduction in background optical noise.
2.3.5
Aerosol or Particulate Distribution
The signal received from a DIAL system depends on the distribution both of the target
gas and of aerosol. For simplification purposes, it is often assumed that a uniform
distribution of ambient aerosol exists. With variable aerosol concentrations resulting in
variable backscatter, DIAL will tend to overestimate peak concentrations in the plume
(Bennett, 1998).
According to Walmsley and O’Connor (1998), fluctuations in the backscatter coefficients
are often the main noise source. These fluctuations are most likely to occur around
process units and water treatment areas where steam condensation can produce strong
local increases in backscatter well beyond the boundaries of visible steam plumes.
Significant local increases in backscatter have also been observed in association with dust
from active work areas or roads or squally showers of rain or particularly snow.
Conversely, heat inputs from fin-fan coolers or furnaces have sometimes been found to
eliminate most of the backscatter, presumably by evaporation of atmospheric aerosols.
Ansmann (1985) reports that great care must be taken in the analysis of H2O DIAL
measurements when layers with high aerosol concentration, clouds or strong temperature
inversion exist.
2.3.6
Interference from Nearby Sources
Clearly, the more congested an area and the more nearby sources there are, the more
difficult it is isolate the emission contributions for a particular source within a facility.
This is true for any remote sensing technique.
5
Final Report
For DIAL measurements, the noise on both the clean-air line and the individual
measurement lines is an important factor. Since the clean line is subtracted from every
measurement line, optimal accuracy is obtained by spending as much measurement time
establishing the single clean-air column as is spent in total on all the measurement
columns from which it is subtracted (Walmsley and O’Connor, 1998).
2.3.7
Data Averaging
A difficulty with the DIAL technique arises from its sensitivity to noise in the received
signals. A DIAL system estimates gas concentrations from subtle variations between
shots and as a function of range. DIAL typically requires the averaging of many shots to
obtain an acceptable signal to noise ratio. Depending on the desired sensitivity and the
range, this may lead to temporal and spatial resolutions of tens of seconds and 50 to 100
m (Bennett, 1998).
The amount of gas can be underestimated when measuring large fluctuating gas
concentrations, because of the bias introduced by averaging the raw signals before
deriving concentrations. Under practical conditions; however, the degree of
underestimation is likely to be small.
2.3.8
Extrapolation of Results
An extrapolation from the measurement results is needed to determine annual emissions.
This requirement is not unique to DIAL measurements. Any measurements that are costly
or labour-intensive, either to operate equipment or in subsequent analysis, are usually
only deployed for short-term measurements, and these are then usually only made during
the day. Dry conditions are preferred for some equipment, and most remote sensing
techniques require a minimum wind speed to guarantee a well defined plume downwind.
All of these factors mean a simple extrapolation on a time basis is subject to considerable
uncertainty.
While it is desirable for the measurement to be as accurate as possible (within practicable
limits), there is little point in making a highly accurate measurement over a short period,
if there are much larger uncertainties regarding the extrapolation to cover all the
unmeasured periods (Richardson and Phillips, 2001). These uncertainties arise primarily
due to operational factors (change in working practice, changes in equipment, changes in
feedstock), and due to the weather (effects of temperature, rain, frost, snow, calm days,
and high winds).
According to Richardson and Phillips (2001) there is a tendency to compile inventories
without regard for the uncertainty in the estimates, and to set targets for improvement as
if it were a simple accounting exercise. Interestingly, their work shows that the nature of
uncertainties skews estimates towards under-estimation. The result is that improved
methods of estimation often result in higher emission estimates which are unwelcome to
all parties involved, especially when money has been invested to meet reduction targets.
6
Final Report
2.4
Applications
The primary applications of DIAL and DIAL in combination with wind profiling (e.g., using
SODAR) include the following:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
The monitoring and charting of diffuse and source emissions in industrial areas.
Mapping of hidden sources and estimation of their contribution to the total air pollution
over a given area.
Studies of the spreading of gas from a source and its effects on air quality in surrounding
areas are also important.
The estimation of fluxes of fugitive emissions.
Detection of plumes and monitoring of their propagation.
Monitoring of pollutant dispersion and distribution above a complex relief and during
smog episodes.
Study of the creation and propagation of ozone smog.
Acquisition of the input, calibration and verification data for air pollution modeling.
Remote measurements into inaccessible, hazardous or elevated areas.
Wide area surveys of ambient air quality.
Measurement of total industrial site emissions.
Boundary fence monitoring.
Identification and quantification of leaks, storage losses, and other fugitive and
engineered sources of emissions.
Plume tracking and source identification from complex industrial plants.
Environmental impact assessments.
Validation of emission estimates or modeling techniques.
The need for such measurements to control emissions from an industrial area is evident.
DIAL is also one of a variety of tools that can be used to screen for significant cost-effective
emission control opportunities at facilities, and has, in some cases, resulted in significant
savings due to avoid product losses. The technique might also be of use to study the transport
of pollutants across the borders. Not least, DIAL is a remote measuring technique for
research on air pollution problems.
Fredriksson et al (1979) have used the LIDAR in several studies of particle emissions from
industrial smoke stacks. Measurements of relative particle distributions are easy to perform
using elastically backscattered light and neglecting weak effects of beam attenuation. If
absolute particle loads in stack effluents are to be measured, the LIDAR system should be
pointed to the plume as close as possible to the mouth of the stack as possible. This approach
avoids both influences due to wind and due to condensing water droplets. Because of the
complexity of the Mie scattering theory and the lack of detailed information on particle
characteristics, it is normally necessary to provide an in-stack calibration.
2.5
Manufacturers
A few companies, such as ORCA Photonics Systems Inc. (www.orcaphoton.com), Lockheed
Martin Coherent Technologies Inc. (http://www.lockheedmartin.com), Optech Inc.
7
Final Report
(www.optech.ca) (a Canadian company), and Elight Laser Systems GmbH (www.elight.de)
produce commercial LIDAR systems for aerosol, turbulence, and other measurements. Although
experiencing some success, LIDAR systems are not high-volume systems due to their significant
cost.
Q-Peak (www.qpeak.com) has been developing frequency-agile laser systems and other
components for defense-related LIDAR and DIAL systems.
Additionally, there are companies, including some of those listed above, and others such as
Spectrasyne Ltd. (http://www.spectrasyne.ltd.uk/) and the UK’s National Physics Laboratory
(NPL) (http://www.npl.co.uk/), that offer commercial DIAL services.
2.6
Advantages, Disadvantages and Limitiation
The key advantages of DIAL are as follows:
•
•
•
•
•
True remote sensing up to 1 kilometre or more.
Can target specific chemicals, as well as be used in a more "open" mode much like a point
source organic vapor analyzer. In the open mode a chemical family such as alkanes is
measured by picking a band that is common to many and interpreting the results as an
"average."
Rapid scanning and two- and three-dimensional mapping of emissions in near real time
allowing emissions and their atmospheric dispersion to be tracked over time.
Able to measure the emissions from very elevated sources and very complex sources.
Able to detect hidden sources and emission hot spots. With traditional fenceline monitoring
techniques it is possible that a toxic release plume could pass around, over, or below the
monitors without being fully detected.
The main disadvantages or constraints are as follows:
•
•
•
•
•
•
•
•
Significant expense for instrument costs and staff (e.g., the price is approximately $15K+ per
day and it normally takes about two weeks to complete a survey of mid to large sized sites).
Large size and weight (truck mounted mobile laboratory).
It requires experts to run the system and interpret the data.
Considerable data processing.
Susceptible to interferences.
Requires good downwind access.
Constrained by meteorological conditions which could result in standby charges if these
conditions are not appropriate at the time of the survey (all remote monitoring methods have
this same limitation).
While DIAL can provide quantification of total emissions, its ability to identify hidden
sources and emission hot spots is more of a coarse screening capability due to its inability to
access congested areas or go inside buildings. For example, knowing that a large process
building or a congested area of a plant contributes a significant amount of emissions is not
the same as knowing exactly which source or sources in these areas are causing the emissions
and need to be controlled. Qualitative methods such as handheld IR cameras and traditional
leak survey methods offer a more practicable and affordable approach for pinpointing
8
Final Report
•
•
•
emission control opportunities in these situations; but lack the ability to quantify the
emissions (e.g., as may be needed to justify control expenditures).
Not suitable for continuous monitoring.
The process of reviewing data to assure it meets quality assurance standards can be
burdensome.
While DIAL’s ability to both identify and quantify emissions has many useful benefits
compared to purely qualitative detection methods; this comes at a financial cost. At the
operations and maintenance level, the quantification of emissions is only necessary where the
practicability or need for emissions control is in question. For example, most facilities would
prefer to simply repair any detected leaks rather than go to the added cost of quantifying the
leak rate before making the repairs.
Because of the unique information that is expected to be acquired by the DIAL system, the
question of its accuracy and compatibility with air quality monitoring reference methods is of
great importance (Keder et al., 2004).
9
Final Report
3.0
EXPERIENCES WITH DIAL
The general experience reported in the literature from the application of DIAL technology to
quantify atmospheric emissions at petroleum refineries has been that, despite some limitations,
DIAL is able to accurately quantify the amount of VOC emissions occurring at the time of
measurement. The results have shown that potentially significant unaccounted for contributions
may occur at some facilities. DIAL has proven effective in quantifying hidden or missed sources
as well as sources and controls with deteriorated performance. Fugitive equipment leaks and
evaporation losses from product storage, loading and unloading are typically determined to be
the major sources of VOC emissions at petroleum facilities.
Recognition that current policies and targets governing the management of VOC emissions are
being understated by inventorying and environmental reporting initiatives is driving increasing
emphasis on measurement and improved estimation of these emissions. For example, data from
the Texas Air Quality Study (TexAQS) 2000 suggest that the VOC emissions inventory for
Texas is low by a factor of 3 to 10 (D. Allent – University of Texas). Tropospheric ozone
reduction strategies, in particular, require good VOC emissions data.
With a few exceptions, DIAL systems have been seen largely as a research tool and less as a
regular monitoring technique due to their significant costs. While DIAL is but one of a variety of
techniques that may be used to develop quantitative measurements of VOC emissions from
fugitive and process sources at petroleum refineries, it remains one of the most powerful options
available. Increasing demand will only improve its affordability.
The following sections summarize some of the specific experiences with the use of DIAL in the
different countries in which it has been applied.
3.1
Belgium
In the late 1990’s all refineries in Flanders, Belgium reported emissions of 13,000 tonnes per
year. A DIAL analysis on 2 refineries (about 10 percent of throughput of the total), found
emissions of 16,000 tonnes per year.
3.2
Canada
The most recent DIAL work done in Canada was conducted by Spectrasyne in cooperation with
Alberta Research Council. This work involved the measurement of fugitive emissions from
several gas processing plants in Alberta during 2003 and 2004 (Chambers, 2003; Chambers,
2004), and from a petroleum refinery in 2005 (Chambers and Strosher, 2006).
The basic objective of these studies was to use the DIAL method to measure the mass emissions
of methane, C hydrocarbons and benzene, apportion the measured fugitive emissions to various
2+
areas of the plants, and compare the DIAL measured rate of fugitive emissions with the emission
rates calculated using estimation methods.
At the refinery, measurements of SO2 from a tail gas incinerator and NO emissions from a gas
turbine power plant where also performed and compared to the corresponding measurements
10
Final Report
performed using the DIAL system with differences of only -11 and +1 percent respectively.
However, no verification measurements were performed on fugitive sources; consequently, it is
not clear that the DIAL’s performance would be as good on these more difficult sources. Ideally,
such checks on fugitive emission sources should involve the quantification, by DIAL, of know
releases of tracer gas in realistic fugitive emission scenarios.
The DIAL survey at the refinery was performed over a period of ten survey days. The results
were extrapolated, with some assumptions, to develop estimates of total annual emissions of C2+
hydrocarbons and were compared to VOC estimates reported by the facility to Environment
Canada’s National Pollutant Release Inventory (NPRI). The authors noted that VOCs exclude
ethane but felt that C2+ was still a reasonable proxy for VOCs. There were no significant upsets
in the plant operation or hydrocarbon spills during the survey.
The extrapolated DIAL measurement results indicated that the value of product lost due to
storage tank and process plant fugitive emissions was 15 fold greater than that determined by the
emissions estimation procedures. While this finding is consistent with the general finding noted
by other researchers that emission inventory methods tend to understate actual emissions due to a
common assumption of no deteriorated performance of sources and emission controls, it is not a
completely fair comparison. Most emission estimation methods, such as the use of emission
factors, have a statistical basis and are recognized as having large uncertainties when applied to
relatively small numbers of sources or used to estimate instantaneous emissions. Still, the
observed differences are noteworthy.
3.3
Czech Republic
An extensive field measurement campaign was performed by Keder et al (2004) in the Czech
Republic in the summer of 2001 in which ozone was measured by DIAL, aircraft and ground
monitoring stations simultaneously. Good agreement was obtained between the DIAL results and
an analyzer located near the ground. However, the comparison with the other results was less
favourable. Accordingly, Keder et al recommended that a substantial effort should be focused on
the explanation of causes of discrepancies between the concentration measurement results from
DIAL and the results from the other analyzers.
The application of combined DIAL/SODAR techniques was demonstrated in the following
cases:
•
•
•
•
•
•
Mapping of hidden sources and estimation of their contribution to the total air pollution
over a given area.
Monitoring of distribution and propagation of atmospheric pollution emitted from line
sources.
Detection of plumes and monitoring of their propagation.
Monitoring of pollutant dispersion and distribution above a complex relief and during
smog episodes.
Study of the creation and propagation of ozone smog.
Acquisition of the input, calibration and verification data for air pollution modeling.
11
Final Report
3.4
European Commission
In 2004 the European Commission funded a project entitled Remote Optical Sensing Evaluation
(ROSE) aimed at developing an improved understanding of the factors affecting the validity of
measurements made using remote optical sensing techniques (ROMTs). The project took place
as part of the Fifth Framework scheme and brought together eleven organizations from all over
Europe, and representing a wide range of expertise. The lead member of the consortium was Sira
Ltd from the UK.
The project began with a field measurement campaign conducted under genuine measurement
conditions at locations across Europe using a variety of open-path techniques including DIAL.
The team then moved on to a series of controlled tests, both laboratory-based and using a
specially-constructed test facility, the design of which was based on the experience gained during
the field test campaigns.
The experiences of the consortium members both inside and outside the project were presented
in two public documents (Sira Ltd, 2004a,b): (1) Recommendations for Best Practice in the Use
of Open-Path Instrumentation and (2) Recommendations for Performance Standards for OpenPath Instrumentation.
While much of the information presented in these two documents pertained to optical techniques
other than DIAL, the following two relevant points were made:
•
Experimental work during the field trials could be constrained by security and access
issues to the detriment of the ideal operation of the ROMTs. The instruments might be
capable of higher level performance, lower detection limits or greater sensitivity if it was
possible to set up equipment in the best locations and at optimum path lengths for the
trials. This is an important consideration for ROMT use.
•
DIAL validation is difficult as there are no other measurement techniques which can
measure, range resolved concentrations along a line, 2D concentration profiles or mass
emissions. In most cases correlations have been with only one facet of the DIAL
capability, e.g. concentration measured along a path with sorption tubes compared with a
single line range resolved DIAL concentration measurement.
In July of 2006 the Eurpoean Commission published a reference document on best available techniques
for the monitoring and control of emissions from storage tanks. The document noted that atmospheric
emissions from storage tanks and loading/unloading operations (e.g., at refineries and oil terminals) are
normally determined by calculation methodologies published by API, US EPA and CEFIC/EVCM
(European Council of Vinyl Manufacturers). At sites where significant VOC emissions are to be
expected, it was stated that BAT includes calculating the VOC emissions regularly. Because of
uncertainties in the models it was suggested that storage losses at these facilities may occasionally need to
be monitored to quantify the emissions and to give basic data for refining the calculation methods. It was
further suggested that this could be done using DIAL techniques, but the necessity and frequency of
emission monitoring should to be decided on a case-by-case basis. Notwithstanding this, no consensus
could be achieved on how to monitor VOC emissions and how to validate calculation results. DIAL is
used commonly in Sweden for monitoring emissions from tanks storing hydrocarbon products at
refineries and oil terminals, but there is not enough information on the use of DIAL at other sites and in
12
Final Report
other countries. Accordingly, it was recommended that more information be collected on the monitoring
of VOC emissions from storage tanks.
3.5
Germany
Germany is the only European country that currently has any formal standards pertaining to the
application of DIAL. These and other related standards are listed below:
•
•
•
•
•
VDI 4202 Part 1 Minimum requirements for suitability tests of automated ambient air
quality measuring systems - Point-related measurement methods of gaseous and
particulate pollutants.
VDI 4202 Part 2 (2004) Minimum requirements for suitability tests of ambient air quality
measuring systems - Optical remote sensing systems for the measurement of gaseous
pollutants.
VDI 4203 Part 4 Control planning for automatic measurement equipment proving
procedures for remote optical measurement equipment for measurement of gaseous
emissions.
VDI 4210 Part 1 (1999) Remote sensing. Atmospheric measurements with LIDAR.
Measuring gaseous air pollution with DAS LIDAR.
VDI 4280 Part 1 (1996) Planning of ambient air quality measurements: General rules.
Copies of the above standards could not be obtained for examination within the time available
for this literature review; however, according to Sira Ltd (2004a), VDI 4210 covers the
principles of the LIDAR method, characterization of performance, a little about the design,
planning and execution of measurements, calibration, and evaluation of both data and system
performance. Appendix B of the standard gives a variety of examples of the use of DAS-LIDAR
(also known as DIAL-LIDAR) in various applications.
VDI 4280 covers what you must know in advance about the measurements you are going to
make and the capabilities of the personnel involved. There is comprehensive coverage of the
factors which must be considered, and the catalogue of questions in Appendix A makes a good
checklist for anyone contemplating a measurement campaign of this kind.
3.6
Sweden
Sweden has the most experience using DIAL to measure refinery emissions. A Swedish national
mobile LIDAR system was developed in 1979 at the Chalmers University. The construction was
based on the results and experiences from research and previous LIDAR systems. Work has also
been done in Sweden by several mobile LIDAR systems constructed by other research groups
(i.e., The Stanford Research Institute, the research institute of ENEL in Italy, and the National
Physical Laboratory in England).
Sweden has required remote sensing at refineries since the late 1980’s. Initially they also tried
differential optical absorption spectroscopy (DOAS) and other single-beam techniques, but by
1995/6 all refineries were required to use DIAL. DIAL measurements are currently performed
every 2 to 3 years. Table 1 summarizes some of the available DIAL measurement results for
petroleum refineries in Sweden.
13
Final Report
Notes
Table 1. A summary of DIAL measurement results at petroleum refineries in Sweden.
Company
Location
Contractor
Year
Estimated
%
Annual
Emitted/Rated
Emissions1
Capacity
(t/y)
AB Nynas
Gothenburg
Spectrasyne 1999
82.5
0.129
AB Nynas
Gothenburg
Spectrasyne 1995
120
0.188
Preem
Gothenburg
Spectrasyne 1999
268
0.050
OK (Preem)
Gothenburg
Spectrasyne 1995
274
0.051
OK (Preem)
Gothenburg
Spectrasyne 1992
317.4
0.059
BP (Preem)
Gothenburg
BP
1989
840
0.155
Research
BP (Preem)
Gothenburg
BP
1988
990
0.183
Research
Shell
Gothenburg
Shell
1999
157
0.0380
Global
Solutions
Shell
Gothenburg
Shell
1996
167
0.040
Global
Solutions
Scanraff
BrofjordenSpectrasyne 1999
503
0.049392548
Lysekil
Scanraff
BrofjordenSpectrasyne 1995
332
0.030999619
Lysekil
Scanraff
BrofjordenSpectrasyne 1992
691
0.0677672
Lysekil
S11
Source: Barrefors, G. (2003) and a PowerPoint presentation by A. Cuclis and D. Byun from the University of
Houston.
1
Based on extrapolations from DIAL measurements.
3.7
The European Union Network for the Implementation and Enforcement of
Environment Law (IMPEL)
In 2000, IMPEL, the environmental inspectors network for the European Union (EU)
commissioned a project to review diffuse VOC emissions estimation methods and measures in
the EU and to propose guidelines to improve the monitoring, licensing and inspection of
industrial activities.
The project focused on the VOC emissions of diffuse sources of large process installations
(primarily refineries and petro-chemical plants), and considered both fugitive emissions (leakage
from equipment) and emissions from storage tanks, loading and unloading facilities. Emissions
resulting from the use of solvents and from petrol filling stations were excluded as they were
already regulated by existing directives.
14
Final Report
At the time it was determined that specific standards for process equipment with respect to
diffuse VOC emissions did not exist; although, a few general guidance documents such as the
German TA-Luft & VDI-3479/3790 and the British ETBPP documents existed.
The study made a number of general recommendations regarding emission targets, control
requirements, emissions monitoring and reporting and non-compliance actions. It was further
recommended that the IMPEL set up an EU-wide information exchange programme on the
licensing and enforcement practice in relation to diffuse VOC emissions. Such a programme
could include a bench marking on subjects like estimation methods and measures.
It was also suggested that supporting activities may be considered by the authorities, such as:
•
•
•
•
organizing an information and training programme in regions where the subject is
relatively new (targeting both companies and licensing & enforcing bodies),
establishing national guidelines,
performing an eco-audits of the industrial plants,
establishing a helpdesk to assist both companies and licensing and enforcing bodies .
While the study examined the merits of DIAL and other measurement technologies, it did not
present any specific recommendations on a preferred method.
3.8
United Kingdom
There have been three mobile DIAL systems in the UK. Spectrasyne, a private company formed
by a management buyout from British Petroleum operates the only commercially available DIAL
system in the UK. Much of their work is described throughout this report.
For many years (beginning in 1995) Shell Research operated a one-third share of an infrared
DIAL system along with SESL (Siemens Environmental Systems Ltd.) and BG (Walmsley and
O’Connor, 1998; Richardson and Phillips, 2001). That system was built by SESL and NPL (the
UK National Physics Laboratory) using technology developed by NPL. It could measure
concentrations well below 1 ppm at ranges up to 1 km. Shell used the system to measure the
emissions of methane, ethane, and heavier alkanes from a range of their petroleum industry sites;
both as a research tool and in locations where DIAL is preferred by the regulators (e.g. at oil
refineries and the harbour in Gothenburg, Sweden). However, it is understood that Shell, along
with SESL, have since discontinued their involvement in this technology due to the limited
market and regulatory demand.
Some of the work and noteworthy findings published by Shell regarding DIAL and its
application at petroleum facilities are as follows:
•
•
Walmsley and O’Connor (1998) recommended that future tests with more comprehensive
sets of anemometry (e.g., SODAR) be conducted to define the errors incurred by the use
of relatively limited wind data sets.
The National Physical Laboratory (NPL), the European oil company’s organization for
environment, health, and safety (CONCAWE), and Shell, all performed studies of
emissions from storage tanks using the DIAL technique (Richardson and Phillips, 2001;
15
Final Report
•
•
3.9
CONCAWE, 1995). One of the major conclusions from that work was that the API
models for estimating annual VOC emissions from storage tanks are appropriate for tanks
in first class condition, but do not allow for the increased emissions from tanks in poor
condition. According to Richards and Phillips (2001), it was rather like assuming
emissions from private cars could be based on the assumption that they were all brand
new and running to specification. The few worst tanks account for a major proportion of
the emissions. On a broader scale, Richards and Phillips also note that improved
estimation and the discovery of overlooked sources can result in upward revision of the
emission estimates, and they go on to state that this is both awkward to explain to the
public at large, and hides the real improvements that will normally have taken place.
Shell’s study of floating roof storage tanks also showed that the emission flux varied with
the position of the roof in the tank. This behavior was also noted by CONCAWE (1995).
The greatest flux occurred when the tank was full and the roof was high relative to the
walls of the tank. When the tank was half full, a recirculation air pattern formed within
the tank that tended to keep the hydrocarbon escape rate down. O'Conner et al (1998)
concluded that the model being used to predict fugitive emission flux from tank farms
might underestimate the actual amount escaping. In another project conducted by Shell,
the DIAL system was used to monitor the emissions from numerous tank facilities
located at a port. The DIAL was able to image the emissions from these facilities and
provided overall flux estimates. The study identified a small number of tanks that were
responsible for a majority of the emissions.
Richardson and Phillips (2001) report, based on their experiences in locating and
quantifying emission sources at petrochemical plants, that conventional open-path
measurement techniques give large coverage at a more modest cost than DIAL, and are
more readily shipped around the world. They suggest using upwind/dowind monitoring
combined with dispersion modeling to back-calculate the source strength. However, they
go on to point out that the difficulty with such methods for source location and emission
rate estimation is in measuring or modeling the vertical extent of the plume, especially
for process plants where there may be a large heat input leading to complicated heat
island effects, and especially under low wind conditions. The actual accuracy of the
emission estimate will depend on a variety of factors including the reliability of the
dispersion modelling, the quality of the measurements performed, the detection limits
achieved, the representativeness of the compiled data, meteorological conditions,
background noise and interferences. Accordingly, the true accuracy is never really
known unless appropriate confirmation measurements are performed which may be
difficult and costly to do on large, complex sources.
United States
Most of the work in the US with LIDAR has been done for, or by, the US Department of
Defense. However, Active Imaging Solutions of ITT Industries Space Systems Division has
developed a commercial airborne DIAL system for detection and measurement of fugitive
emissions at oil and gas facilities (Brake, 2005). This system provides 2-dimension concentration
profiles of the emissions from a facility when looking down on the facility from an aerial
position, but does not provide quantification of emission rates. Demonstrations have been
conducted on tank batteries and a gathering pipeline segment being repaired with gas release
16
Final Report
rates as low as 0.6 m3 per minute being readily detected. It is claimed that the system can survey
up to 1600 km of pipeline per day and can operate day or night.
Additionally, US EPA (2006) recently developed a protocol for characterizing gaseous emissions
from non-point pollutant sources. The protocol is specific to the use of open-path, PathIntegrated Optical Remote Sensing (PI-ORS) systems in multiple beam configurations to directly
identify “hot spots” and measure emission fluxes. PI-ORS systems include scanning open-path
FTIR, UV-DOAS, TDLAS, and PI-DIAL, The choice of PI-ORS system to be used for the
collection of measurement data (and subsequent calculation of PIC) is left to the discretion of the
user. Basic user knowledge of a PI-ORS system and the ability to obtain quality path-integrated
concentration (PIC) data is assumed.
17
Final Report
4.0
CONCLUSIONS AND RECOMMENDATIONS
The conclusions and recommendations of this study are presented in the following subsections:
4.1
Conclusions
The DIAL technology is unique in its ability to rapidly develop near real-time two- and threedimensional mapping of the atmospheric emissions plume from point, line and complex area or
volume sources. Subject to proper quality control/quality assurance (QA/QC) measures, suitable
meteorological conditions and downwind access, DIAL can provide quite accurate quantification
of emission rates and provide coarse screening for hidden sources and emission hot spots.
Moreover, it is an invaluable research tool for developing an improved understanding of fugitive
and other complex emission sources, and of the atmospheric dispersion of these emissions.
Its significant cost is the primary reason DIAL has not seen widespread use as a frequent
monitoring technology for use at industrial facilities. Even in Sweden where refineries are
required to conduct regular DIAL surveys, these surveys are only conducted for typically a two
week period once every two to three years. Still, as the technology gains increasing acceptance
and demand, costs are likely to decrease making it a more practicable choice.
The validity of taking snapshot emission measurement results from a DIAL survey and
extrapolating them to determine annual emissions is a potential issue that requires careful
consideration of the characteristics of the sources being considered and the operating conditions
at the time. However, there are really no low-cost approaches that can be used to accurately
quantify total VOC emissions from a single facility or process area except for point sources with
continuous emission monitoring systems in place. Traditional inventory estimation methods
remain the most practical means of developing emission estimates for regional or national issues.
Although, the current literature indicates that these inventory methods may often introduce a
significant negative bias due to inadequate consideration of the deteriorated performance of
emission sources and controls with time. Furthermore, indications are that the unaccounted for
emissions from such effects are not normally distributed. Rather, they are characterized by more
of a skewed distribution where only a few sources in each category are contributing most of the
unaccounted emissions at a facility, and only a few facilities are contributing most of the
unaccounted for emissions by the industry.
A quantitative measurement approach is really the only option for developing an accurate
assessment of an individual facility’s total VOC emissions, identifying the primary sources of
these emissions and potential emission reduction opportunities (e.g., to address local air emission
issues). DIAL is one of various measurement options that could be considered, each having its
own advantages and disadvantages. The best option should be determined on a case-by-case
basis giving consideration to the accuracy of the emission estimates needed to facilitate sound
decisions in the final environmental analysis to be performed. The uncertainty contributions of
all elements of the analysis should be considered, not just those of the emission estimates, and a
practicable approach taken in managing these uncertainties.
18
Final Report
4.2
Recommendations
Clear guidelines should be established that set out specific accuracy targets for the various
emission reporting requirements imposed on industry. These targets should be science-based
values that consider potential local, regional and national environmental decision-making needs,
and reflect a practicable approach to managing the uncertainty in the final environmental
analyses to be preformed using the emissions data. These targets may be different for different
pollutants. Alternatively, approved technologies or estimation methods should be identified,
which, when applied in accordance with good practice, may be deemed to comply with such
objectives. At a minimum, current VOC inventorying methods, guidelines and emission factors
should be reviewed to identify opportunities for improvements.
19
Final Report
5.0
REFERENCES CITED
Ansmann, A. 1985. Errors in Ground-Based Water-Vapor DIAL Measurements Due to DopplerBroadened Rayleigh Backscattering. Applied Optics. v 24, n 21. November 1985. pp. 34763480(5).
Barrefors, G. 2003. Fugitive VOC-emissions Measured at Oil Refineries in the Province of
Vastra Gotaland in South West Sweden (Development and Results 1986 to 2001). A report
commissioned by The Count Administration of Vastra Gotaland, Sweden. .pp 30.
Bennett, M. 1998. The Effect of Plume Intermittnecy Upon Differential Absorption LIDAR
Measurements. Atmospheric Environment. v 32, n 15. pp. 2423-2427.
Brake, D. 2005. Detection and Measurement of Fugitive Emissions Using Differential
Absorption Lidar (DIAL). A presentation made by Active Imaging Solutions of ITT Industries
Space Systems Division at the EPA Gas STAR Program – Annual Implementation Workshop, 25
October 2005.
Chambers, A.K. 2003. Well Test Flare Plume Monitoring Phase II: DIAL Testing in Alberta.
ARC Contract Report No. CEM 7454-2003, December, 2003.
(available at www.ptac.org/env/dl/envp0402fr.pdf ).
Chambers, A.K. 2004. Optical Measurement Technology for Fugitive Emissions from Upstream
Oil and Gas Facilities. ARC Contract Report No. CEM – P004.03, December, 2004.
(available at www.ptac.org/env/dl/envp0403.pdf ).
Chambers, A.K., and M. Strosher. 2006. Refinery Demonstration of Optical Technologies for
Measurement of Fugitive Emissions and for Leak Detection. A report prepared by Alberta
Research Council for Environment Canada. .pp 43.
CONCAWE. 1995. VOC Emissions from External Floating Roof Tanks: Comparison of Remote
Measurements by Laser with Calculation Methods. Prepared for the CONCAWE Air Quality
Management Group, based on work performed by the Special Task Force on DIAL
measurement of gasoline tanks (AQ/STF-44). Report No. 95/52. .pp 70.
(www.concawe.org/1/MAJDFIPABLJPHMMLHJHILPDIVEVC7191P3PDBK9DW3GK9DW3
571KM/CEnet/docs/DLS/Rpt_95-52-2004-01744-01-E.pdf)
Egeback, A., K.A. Fredriksson, and H.M. Hertz. 1984. DIAL Techniques for the Control of
Sulfur Dioxide Emissions. Applied Optics. v 23, n 5. March 1984. pp. 722-729(8).
European Commission. 2006. Reference Document on Best Available Techniques on Emissions
from Storage. A report on an information exchange carried out under Article 16(2) of Council
Directive 96/61/EC (IPPC Directive). .pp 432. (http://www.jrc.es/pub/english.cgi/d1254315/)
Fredriksson, K., B. Galle, K. Nystroem, and S. Svanberg. 1979. LIDAR System Applied in
Atmospheric Pollution Monitoring. Applied Optics. v 18, n 17. September 1979. pp. 29983003(6).
20
Final Report
IMPEL. 2000. Diffuse VOC Emissions: Emission Estimation Methods, Emission Reduction
Measures and Licensing and Enforcement Practice. A report prepared by Tebodin assisted by
Schelde Leak Repairs Specam and Cowi. Brussles. .pp 124.
Keder, J., M. Strizik, P. Berger, A. Cerny, P. Engst, and I. Nemcova. 2004. Remote Sensing
Detection of Atmospheric Pollutants by Differential Absorption LIDAR 510M/SODAR PA2
Mobile System. Meteorology and Atmospheric Physics. v 85, n 1-3. January 2004. pp. 155164(10).
Lamb, B., J.B. McManus, J.H. Shorter, C.E. Kolb, B. Mosher, R.C. Harriss, E. Allwine, D.
Blaha, T. Howard, A. Guenther, R.A. Lott, R. Siverson, H. Westberg, and P. Zimmerman. 1994.
Measurement of Methane Emissions from Natural Gas Systems Using Atmospheric Tracer
Methods. Presented at the 1994 International Workshop on Environmental and Economic
Impacts of Natural Gas Losses, March 22-24, 1996, Prague, Czech Republic. pp. 26.
Minnich, T.R., R.J. Krocks, P.J. Solinski, D.E. Pescatore, and M.R. Leo. 1991. Determination of
Site-Specific Vertical Dispersion Coefficients In Support of Air Monitoring at Lipari Landfill. A
paper presented at the 1991 AWMA/EPA International Symposium on the Measurement of
Toxic and Related Air Pollutants, Durham, NC, May 1991. .pp 8.
O’Connor, S., H. Walmsley, and H. Pasley. 1998. Differential absorption LIDAR (DIAL)
measurements of the mechanisms of volatile organic compound loss from external floating
roofed tanks. EUROPTO Conference on Spectroscopic Atmospheric Environmental Monitoring
Techniques, Barcelona, Spain, SPIE Vol. 3493. [abstract]
Piccot, S.D., S.S. Masemore, W. Lewis-Bevan, E.S. Ringier, and B.D. Harris. 1996. Field
Assessment of a New Method for Estimating Emission Rates from Volume Sources Using OpenPath FTIR Spectroscopy. J. Air & Waste Manage. Assoc. v46 .pp 159-171.
Richardson, S.A., and V.R. Phillips. 2001. A Comparison of Petrochemical and Agricultural
Approaches to Emission Inventorisation and Uncertainties. Report No. OG.01.47049R . A report
prepared by Shell Global Solutions. Chester, England. OG.01.47049R
Sira Ltd. 2004a. Recommendations for Best Practice in the Use of Open-path Instrumentation - A
Review of Best Practice Based on the Project: Remote Optical Sensing Evaluation (ROSE)
August 2001-July 2004. A report prepared for the European Commission by the ROSE
Consortium. Contract No. G6RD-CT2000-00434. .pp 131.
Sira Ltd. 2004b. Recommendations for Performance Standards for Open-path Instrumentation –
Recommendations Generated Based on the Project: Remote Optical Sensing Evaluation (ROSE)
August 2001-July 2004. A report prepared for the European Commission by the ROSE
Consortium. Contract No. G6RD-CT2000-00434. .pp 174.
US Environmental Protection Agency. 2006. Final ORS Protocol: Optical Remote Sensing for
Emission Characterization from Non-Point Sources. .pp 44.
(www.epa.gov/ttn/emc/prelim/otm10.pdf).
21
Final Report
Walmsley, H.L. and S.J. O’Connor. 1998. The Accuracy and Sensitivity of Infrared Differential
Absorption LIDAR Measurements of Hydrocarbon Emissions from Process Units. Pure Appl.
Opt. v 7. pp. 907-925(19).
Warren, R.E. 1989. Concentration Estimation From Differential Absorption LIDAR Using
Nonstationary Wiener Filtering. Applied Optics. v 28, n 23. December 1989. pp. 5047-5051(5).
Weibring, P., C. Abrahamsson, M. Sjoholm, J.N. Smith, H. Edner and S. Svanberg. 2004. Multicomponent Chemical Analysis of Gas Mixtures Using a Continuously Tuneable LIDAR System.
Applied Physics B. v 79, n 4. September 2004. pp. 525-530(6).
Weibring, P., M. Andersson, H. Edner, and S. Svanberg. 1998 Combination of lidar and Plume
Velocity Measurements for Remote Sensing of Industrial Emissions. Department of Physics,
Lund Institute of Technology, Sweden, SPIE vol. 3104, 0277-786X/97
22
This page intentionally left blank.
OCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCV
Fugitive VOC-emissions measured at Oil Refineries
in the Province of Västra Götaland in South West Sweden
- a success story
development and results 1986 – 2001
commissioned by The County Administration of Västra Götaland
County Administration
Report 2003:56
Fugitive VOC-emissions measured at Oil Refineries
in the Province of Västra Götaland in South West Sweden
- a success story
development and results 1986 – 2001
commissioned by The County Administration of Västra Götaland
County Administration
Report 2003:56
PRODUCTION | THE COUNTY ADMINISTRATION OF VÄSTRA GÖTALAND TEXT | LENNART FRISCH, AGENDA ENVIRO AB LAYOUT | CILLA ODENMAN PUBLICATION | 2003:56 ISSN | 1403-168X PRINT | GÖTEBORGS LÄNSTRYCKERI AB
PREFACE
This report describes the environmental trends that have been on the agenda of the
Swedish oil refineries in recent years, specifically focusing on emissions of Volatile Or­
ganic Compounds (VOC). In the case of oil refineries this is more or less also synony­
mous with hydrocarbons and in most cases VOC is synonymous with NMVOC (Non­
methane VOC). If methane is included this is clearly stated in the report.
The issue of VOC-emissions has been high on the agenda for the Swedish oil refi­
neries since the mid 1980’s, when the first major discussions started on how to carry
out measurements at the sites. Later the issue also has been raised for, among others,
oil harbours and other main tank storage areas. The total crude oil throughput of the
Swedish oil refining sites is about 20 million ton per year.
Today we have more than 15 years of measurement experience with the laser based
DIAL-system (Differential Absorption Lidar). The system has been shown to be a very
powerful tool in the measurement, as well in the combat, of the true VOC-emission.
Other systems have also been tested (DOAS, HAWK) but have been shown to be nonreliable in performance.
This report is written by Lennart Frisch, MD at the environmental consulting bureau
Agenda Enviro AB, and is commissioned by the County Administration of Västra
Götaland (former the Provincial Government of Göteborg and Bohus) and the Swedish
Environmental Protection Agency.
The Author is fully responsible for the content in the report.
Gunnar Barrefors,
Department of environmental protection
County administration of Västra Götalands län
About the author:
Lennart Frisch, MSc. and certified environmental lead auditor according to ISO
19 011, is the managing director of the environmental consultancy bureau
Agenda Enviro AB, www.agendaenviro.se
Between 1981 – 1986 he was process engineer and head of computer systems at
the Shell Refinery in Göteborg, later environmental officer at the regional autho­
rities of the Province of Göteborg and Bohus, and since 1996 an environmental
consultant for mainly industrial clients but also for the Swedish environmental
ministry, the Swedish EPA as well as regional and local environmental authori­
ties. He has amongst others been Swedish representative at the EU-commission
network IMPEL (Implementation and enforcement of environmental law) and
the Article 19 committee of EMAS at the EU-commission. He has been a multiannual member of the Swedish EPA advisory board on implementation and en­
forcement of environmental law and of the Swedish EPA scientific committee on
air quality and emissions to air.
CONTENTS
1. SHORT HISTORICAL BACKGROUND
7
2. SHORT-CUT INFORMATION ON THE SITES FOR OFFICIAL DIAL-MEASUREMENTS
9
2.1 Preem Raffinaderi AB, Göteborg
2.1 Skandinaviska Raffinaderi AB, Scanraff, Lysekil
2.3 Shell Raffinaderi AB, Göteborg
2.4 Nynäs AB, Göteborg
2.5 Nynäs AB, Nynäshamn
2.6 Gothenburg Port, Oil Harbour, Göteborg
9
9
9
10
10
10
3. INITIAL MEASUREMENTS
12
4. MEASURES UNDERTAKEN TO REDUCE EMISSIONS
14
4.1 Preem Raffinaderi AB, Göteborg
4.2 Skandinaviska Raffinaderi AB, Scanraff, Lysekil
4.3 Nynäs AB, Göteborg
5. DESIGNING A MEASUREMENT SURVEY
5.1 VOC’s to be included
5.2 Meteorological measurements
5.3 Measurement strategy
6. MEASUREMENT RESULTS
6.1 State of the art methodology
6.2 Presented data
6.3 Preem Raffinaderi AB, Göteborg
6.4 Skandinaviska Raffinaderi AB, Scanraff, Lysekil
14
15
17
18
18
20
20
25
25
26
28
28
SHORT HISTORICAL BACKGROUND
1. SHORT HISTORICAL BACKGROUND
In Sweden there are three fuel producing oil refineries. On top of that there are also oil
refining facilities for other products like bitumen and lube oil.
Out of the total of five oil refineries in Sweden four lie in the Province of Västra
Götaland and out of these, three are situated in the town of Göteborg (Gothenburg),
the Capital of the Province and the second biggest town of Sweden. The fourth refinery
in the province - with the highest capacity - is situated in the municipality of Lysekil
some 100 km north of Göteborg. The fifth oil refinery is mainly producing lube oil
and is situated in Nynäshamn, some 100 km south of Stockholm. Crude oil and the
products received when processing it are also handled at a number of Oil Harbours
along the Swedish coast. The largest facilities for this are the Gothenburg Port and the
oil harbour at the Scanraff oil refinery in Lysekil.
The first refinery in the area, the Koppartrans refinery, later bought by Shell, was on
stream in 1953. Originally this plant was planned and designed for China, but with
the changing political realities at that time, the facilities were redirected to Sweden and
Göteborg. Prior to that the Nynäs oil refinery in Nynäshamn had already opened in
1928, at that time also being a fuel producing refinery. The second Nynäs-refinery, was
opened in Göteborg in 1956 aiming at a production of mainly bitumen.
In the mid 1960’s the Shell refinery was revamped doubling its capacity and in
1967 BP got its own refinery on stream (later sold to OK Petroleum and later renamed
Preem). Until the beginning of the 1970’s there were no refineries in the province having
other than low skimming facilities. In 1972 Shell installed a thermal cracker unit and
1975 the Scanraff facilities in Lysekil came on stream with about the same production
outline as the Shell refinery, but with significant higher capacity (7 Mton/a). In 1984
the Scanraff refinery was extended with a catalytic cracker unit. Scanraff is also today
within the Preem Group.
At the beginning of the 1970’s there were plans for major extensions of the Shell
and BP refineries (up to 13 and 15 Mton/a respectively). These plans were however
subsequently turned down because of the energy crisis in 1973/74 as well as the startup
of Scanraff. There were also plans for a second refinery, “Statsraff” close to Scanraff.
These plans were also never fulfilled.
Environmental issues were not really on the refinery agenda in the beginning, alt­
hough equipment for the removal and recovery of sulphur in process streams - such
as Claus-units - were installed all over during the 1960’s and 1970’s. The function of
these units in the BP and Shell case though left some doubt, leaving BP to slaughter the
old Claus-units, installing a new (smaller) one in the early 1980’s and Shell revamping
its units also in the 1980’s.
The turning time in environmental thinking at the refineries came during the second
half of the 1980’s with some court cases on sulphur emissions. This lead to a subsequent
change of policy at the oil refineries towards an environmental image. After substantially
reducing overall emissions in the late 1980’s advanced facilities for sulphur removal tail gas treating units - were installed at both the Shell and Scanraff refineries in 1993/4
and soon after that also at Preem. Scanraff also reduced the use of oil as internal fuel
early on so that the entire refinery - with the exception of the FCC-unit using coke
– was normally fired on gas only.
7
SHORT HISTORICAL BACKGROUND
In the beginning of the 1990’s low NOx-burners were introduced at the refineries,
starting with some mixed experience. Clearly though that introduction, as well as the
increased knowledge in the control rooms of the impact of firing conditions to the
creation of NOx-emissions, also reduced NOx -emissions substantially although it has
been difficult to describe exactly how much as historically NOx never was measured.
Later also SCR-units, beginning with the FCC at Scanraff, were installed, today also
being used for boilers at the Shell and Preem refineries.
Emissions of volatile organic compounds (VOC) historically were only roughly calcu­
lated either as a figure based on throughput, or on the number of certain process-units
in the plant multiplied by certain theoretical emission data. Historically emissions from
storage facilities, such as tanks were only very rarely thought of being of any magnitude
to count with. Because of hard pressure from the Provincial Government in the second
half of the 1980’s sophisticated measurement devices were taken out of the laboratories
to be used for field measurements.
Measured – true – VOC-emissions showed to be substantially higher than what could
be thought of based on the old calculations, especially for the storage facilities. Based on
the first measurements with the laser technique in 1988 and 1989, later measurements
in 1992, 1995, 1996 and 1999 have shown tremendous reductions of VOC-emissions
during these years. The reduced emissions clearly follow fruit-bearing actions taken by
the companies to reduce emissions.
Starting in 1996, VOC-measurements with the DIAL-technique were also carried
out in the oil harbour of the Göteborg Port, giving the same principal results as at the
oil refineries. In 1999 also the first DIAL-measurement was carried out at the Nynäs­
hamn oil refining site of Nynäs, also here giving the same principal results as the early
measurements at the other oil refineries some 5-15 years earlier.
The only emissions not in accordance with the diminishing trend are the emissions of
carbon dioxide. As a basic rule, further refining of the crude oil needs more energy than
if no such refinement took place. This was already the case for the deeper conversion
introduced by the Shell refinery in the early 1970’s and with the introduction of the
FCC at Scanraff in 1984 and of course with an increased production in itself. In recent
years CO2-emissions have increased further based on the demand from society on the oil
refineries to produce new, less environmentally disturbing products. This of course is a
contradictory situation, which politically has been shown not to be all too easy to handle.
8
SHORT-CUT INFORMATION ON THE SITES FOR OFFICIAL DIAL-MEASUREMENTS
2. SHORT-CUT INFORMATION ON THE SITES FOR
OFFICIAL DIAL-MEASUREMENTS
Below the main features are described for the different refineries as well as for the Oil
harbour at the Göteborg Port.
2.1
Preem Raffinaderi AB, Göteborg
Being a BP Refinery until 1991, since its start on stream in 1966, the refinery has had
a low-skimming profile until the mid 1990’s. In 1994 an isomerization unit was set on
stream as the first new major process change since the startup. In 1996 facilities for
the desulphurization of gasoil as well as for the production of “Environmental diesel“
were installed. At the same time new big tail-gas treating units for process-sulphur
came on stream.
The licensed throughput is 6 Mton/a although a practical limit could be assumed
at somewhat more than 5 Mton/a. The normal annual throughput has been around 4
Mton/a, with the exception of some years in the beginning of the 1980’s, when then
throughput dropped below 3 Mton/a as a result of a major fire.
The number of people employed is about 250. The refinery is situated in the muni­
cipality of Göteborg on the Hising Island.
Measurements with DIAL (Spectrasyne) have been executed in 1988, 1989, 1992,
1995/96 and 1999.
2.2
Skandinaviska Raffinaderi AB, Scanraff, Lysekil
This refinery was planned in the 1960’s and got its licensing in the early 1970’s. In the
early years the licensed throughput was 7 Mton/a. After a period having the limit on
8.3 Mton/a it is now set at 10 Mton/a.
In 1984 the refinery was extended with a FCC unit, originally with a licensed capacity
of 1.3 Mton/a. In 1992 this was raised to 1.5 Mton/a, with 1.75 Mton/a from 1995
and onwards.
The owners have differed throughout the years, now being owned by Preem, the
same owner as for the Preem refinery.
The refinery is situated in the municipality of Lysekil, without any other industry of
its size in the neighborhood and being the industrial facility of highest importance in
the area. The refinery employs some 550 people.
Measurements with DIAL (Spectrasyne) have been executed in 1992, 1995 and
1999.
2.3
Shell Raffinaderi AB, Göteborg
The equipment for the refinery was originally built in the USA with destination for
mainland China just after the 2nd World War. Due to the political changes in China at
that time an alternative destination was thought of. Starting under the name of Kop­
partrans with a shared ownership by two Swedish companies, Kopparberg and Trans­
atlantic a small fuel producing refinery with two minor crude oil units was set up in
the 1950’s. The maximum capacity at this time was about 2 Mton/a.
9
SHORT-CUT INFORMATION ON THE SITES FOR OFFICIAL DIAL-MEASUREMENTS
After being bought by Shell, new facilities were installed in the mid 1960’s more
than doubling the throughput. Licensed throughput is 5 Mton/a, but the practical
maximum could be set at around 4 Mton/a. The refinery is situated in the municipality
of Göteborg on the Hising Island.
The number of people employed is a little less than 200.
Measurements with DIAL (Shell Research) have been executed in 1996 and 1999.
2.4
Nynäs AB, Göteborg
Nynäs AB is a refinery in Göteborg (Gothenburg) situated at the Hising Island produ­
cing mainly bitumen and related products. The licensed throughput is 450 000 ton/a.
The facilities were built in 1956 and subsequently put on stream in 1957 slowly in­
creasing the throughput from some 100 - 200 kton/a in the early years to around 400
kton in recent years. On a monthly basis the throughput is about 50 kton, but as the
plant normally has a winter shut down the possible level of some 600 000 ton/a is ne­
ver reached at the present situation. As the winter shut down is based on the needs of
the domestic market, changes could though be brought about in the future if the mar­
ket picture is being altered.
The Nynäs refinery is normally referred to as the “small bitumen plant“ in the Pro­
vince as the facilities for the fuel producing refineries are much bigger.
The number of people employed is about 50 at the refinery.
Measurements with DIAL (Spectrasyne) have been executed in 1995 and 1999.
2.5
Nynäs AB, Nynäshamn
The refinery is situated in Nynäshamn some 100 km south of Stockholm and has the
longest history of the Swedish refineries. The refinery was started in 1928 and was a
fuel producing refinery until 1983. At that time Nynäs, as a company, left the Swe­
dish fuel market and the refinery was revamped in order to produce bitumen products
and naphtenic special oils including lube oils using a very heavy crude oil. The license
for the refinery is limited to 1,8 Mton of crude oil intake, and the refinery is equipped
with, amongst others, one vacuum distillation and three hydrations units, the latter be­
ing one hydrofinisher and two hydrotreaters.
The desulphurization capacity has been increased during the last years and new equip­
ment has been installed for the removal of sulphur. The refinery uses external fuel.
Measurements with DIAL (Spectrasyne) have been executed in 1999.
2.6
Gothenburg Port, Oil Harbour, Göteborg
The Gothenburg Port has a long history dating back to the time when Göteborg was
founded in the 17th century. Since about 1850 the Gothenburg Port has held the posi­
tion of being the largest Swedish – as well as Nordic – port, now being about number
10 in Europe. Annually some 34 Mton of goods is handled in the port of which the Oil
harbour is handling close to 20 Mton of crude oil and oil products.
10
SHORT-CUT INFORMATION ON THE SITES FOR OFFICIAL DIAL-MEASUREMENTS
The Oil harbour is situated on the northern shore of the Göta Älv river and thus
nowadays lies more or less as a part of the Göteborg city, although a bit west of the
centre. Within the oil harbour site also a number of handling and distribution compa­
nies have their facilities including tank storage and off-loading of products to trucks
and railway.
Measurements with DIAL (Spectrasyne and Shell Research) have been executed in
1996 and 1999 respectively.
11
INITIAL MEASUREMENTS
3. INITIAL MEASUREMENTS
In the early and mid 1980’s the problem with the – at that time – unknown real emis­
sions of Volatile Organic Compounds (VOC) from the oil refineries, lead to a number
of discussions between representatives from the environmental enforcing authority,
the Provincial Government of Göteborg and Bohus, and representatives from the re­
fineries.
The discussions finally lead to a decision on January 19th 1988 by the Provincial
government, that one of the oil refineries had to start doing measurements. The oil
refinery chosen, at that time the BP refinery in Göteborg (nowadays Preem Raffinaderi
AB) , was considered to have the best location for a first trial of measurements. The
reason for this was that the refinery at the time was a simple low-skimming facility,
with the geographical positioning of the process area and the tank farms well separated.
Also the infringement of emissions from other sources in the area could easily be taken
care of as the distance to other emitting sources – also taking in account the prevailing
wind direction – was considered to be more than sufficient. The decision was coupled
with a fine of SEK 2 million – at the time some USD 300 000 - in case measurements
and reporting were not carried out as decided by the authorities.
By coincidence BP at the time had already developed an in-house laser based DIALsystem (Differential Absorption Lidar) which had already been used inside of the BP
group under the flag of BP Research.
The first measurements were carried out at the BP refinery (later Preem) in May 1988,
in June 1989 and also in February 1992, before it was considered that it was without
any doubt possible and feasible to use the system in an appropriate way to determine
the true VOC-emissions also for the other Swedish oil refineries.
At the time of the initial measurements at the BP refinery, theoretical (API- and
Radian-based) calculations had been used to get some rough idea of the VOC emission
level. The emissions calculated showed that some 700 ton VOC/a could be estimated
to be emitted. This was virtually turned upside down when the figures of the real emis­
sions – based on the DIAL-measurements – were released during the autumn of 1988.
The emission level at the refinery turned out to be about 10 000 tons/a instead, and
in this figure the product tank farm was not included. With that included (it was first
measured in the measurements in 1989), the real emission level in 1988 for the BP
refinery could be estimated at some 14 000 tons/a, ie. 20 times higher than what the
calculations showed.
The presentation of the measured figures to the public – in Sweden all these data
are open to the public domain – resulted in a heated discussion in the papers and in
subsequent meetings between the representatives of the environmental authorities of
the Province and the management of the then BP refinery. These discussions resulted
in a number of decisions, which showed to be of great value in the coming combat of
the VOC-emissions, namely:
• the management of the BP Refinery confirmed that the measured values, although
high, were reliable
• the management of the BP Refinery confirmed that they felt obliged to undertake
actions in order to reduce emissions. As a matter of fact the measurements showed
amongst others one single leak corresponding to some 4 000 ton VOC/a in itself.
This leak was subsequently tightened up by the end of the measurements
12
INITIAL MEASUREMENTS
• the management of the BP Refinery declared that they were willing to undertake a
new measurement in about one years time in order to both confirm the results of
the undertaken measurements and to receive a proof of the impact due to measures
planned to be undertaken in the meantime before that measurement.
With this declaration by the BP Refinery a good basis, between the environmental
authorities at the Provincial Government and the refinery itself, was laid for a mutual
cooperation climate on these issues.
It was agreed that the coming measurements should also consider the possible impact
of such ambient factors as wind speed and outdoor temperature as well as the impact
of a shining sun. The results that were achieved showed that the impact of wind speed
for some installations could be other than negligible, namely the tank storage area, but
that a knowledge of normal average wind speed could be of good value in assuming
normal average emisisons. Outdoor temperature as well as the impact of the sun rays
on the other hand was shown to be of a negligible impact, this being specifically – and
amongst others – proved by the measurements at the now Preem refinery during one
of the coldest February-periods in 1992.
13
MEASURES UNDERTAKEN TO REDUCE EMISSIONS
4. MEASURES UNDERTAKEN TO REDUCE EMISSIONS
Information on the current situation on implemented measures has been gathered for
the two main fuel producing oil refineries and for the small bitumen-producing plant.
Generally it can be noted that the different actions started with a major implementa­
tion phase as a result of the 1988 measurement results, which – as noted above - were
staggering high.
On top of what is presented here it is also obvious that the refineries now pay much
more attention to problems of VOC’s and to the emissions of these pollutants compa­
red to the situation only a decade ago. Today it is a normal part of life in the crude oil
processing to think of solutions to keep down the VOC emissions, especially in case
new facilities are designed, constructed and taken on stream.
The following measures to reduce emissions could specifically be noted:
4.1
Preem Raffinaderi AB, Göteborg
4.1.1 Tanks and other storage
• For oil pumps the sumps are covered and the trays tilted. At the place of the crude
oil tanks the clear water pumps are vented directly to air.
• Inner floating roofs are equipped with primary seals (4 tanks)
• Blanket gas is used for three different tanks containing naphta and equipped with
fixed copula roofs. In recent years the blanket gas has been changed from hydro­
gen-rich reformer off gas to nitrogen.
• Secondary seals on the outer floating roofs of the crude oil tanks.
• All product tanks with outer floating roofs have been equipped with a secondary
sealing (excl. tanks with kerosene) as well as with equipment to reduce evaporation
around the piping for level control.
• External fixed cupola roofs with internal floating roof equipped with primary and
secondary seal on gasoline components, gasoline and slops tanks. The cupola roofs
also play a role to avoid rain water entering the tanks.
•Drainage of tanks being better surveyed during operations. The drainage is led from the
crude oil tanks to other tanks, ie. led back to a sludge tank and not directly to the WWT.
• A new type of roof drainage system is installed on the crude oil tanks. The size
of the drainage devices have been decreased, allowing also a decrease of the area
where VOC is exposed to the atmosphere.
• Changed roof drainage systems on all other tanks to abolish old piping which was
due to leak VOC.
• Caverns are kept with a low filling degree and designed with a common gas phase
to keep the pressure low and thereby diminishing the risk of evaporating or ven­
ting VOC’s at all filling levels.
4.1.2
Process area
• Piston rod seals have been exchanged for products of the latest technique, mainly
related to the material of the seal on the piston compressors.
14
MEASURES UNDERTAKEN TO REDUCE EMISSIONS
• All control valves on the refinery re equipped with live-loading packing. All valves
for manual operation on the new parts of the process area are from 1996/97 also
equipped with live loading packing as well as also some other valves which – due
to other reasons - have been up for exchange in recent years. All valves in service
with light hydrocarbons are equipped with live-loading packing.
• Safety relief valves are led to the flare due to basic design by the plant in 1967.
• Pumps: In 1994 the first LPG-pump was equipped with magnetic drive. Now all
LPG-pumps have been equipped with tandem seals (not pressurized). Pumps wor­
king with a magnetic drive amongst others are being used in service where H2S is
present in more than negligible concentration levels.
• All flanges serving light hydrocarbon streams are equipped with expanding grap­
hite seals.
• For new process equipment the number of flanges are reduced by design.
• Flanges to purge or drain ends are either equipped with caps, blinded or plugged.
• Streams of product samples sent to on-line instruments to control specifications are
returned to the processes are led to the flare.
• Most of the sampling stream to places for manual caught analyses are returned to
the process or to the flare.
• A flue gas compressor installed in 2002.
• In line mixing of products is the general means of establishing final products.
• A leak detection and repair programme has been in full implementation for about
10 years.
4.1.3 Waste water treatment
• A settling tank of 10 000 m3 has been installed before the WWT to reduce the hy­
drocarbon content to the API also enabling an uncovering of the API. Measure­
ment tests will be undertaken to see if the uncovering is a possible option or not.
• The well to gather incoming water to the WWT is covered.
• The PPI-separator is kept covered by water, by which no further coverage is neces­
sary.
4.2
Skandinaviska Raffinaderi AB, Scanraff, Lysekil
4.2.1 Tanks and other storage
• A balancing line in between tanks in light hydrocarbon service to improve pressure
balancing and to reduce the risk of venting through safety relief valves.
• Secondary seals are being used on all tanks with floating roofs which are in service
for products with a higher evaporating pressure than kerosene, in total 14 storage
tanks.
• A new liquefied secondary seal installed on one of the crude oil tanks, following
very high measured emissions by the DIAL trial in 1999.
• Vent gas from caverns is led to the flare instead of to the atmosphere.
15
MEASURES UNDERTAKEN TO REDUCE EMISSIONS
4.2.2 Process area
• For four centrifugal compressors the vent gas is led to the fuel gas/flue gas system.
• For 13 piston compressors the leakage from the piston rod is led to the fuel gas/
flue gas system.
• All pumps used for hydrocarbons with a density below 0,65 (at 200°C) are revam­
ped and have new axis seals of tandem type.
• About 250 control valves in service with naphta and lighter hydrocarbons are
equipped with improved packing material (graphite) which in some cases also is
combined with a system based on springs.
• Valves run manually are all equipped with a new type of glandered packing (grap­
hite rings in combination with a plait of carbon fibre)
• For flanges a spiraled graphite packing is used.
• Streams for on-line samples to GC’s are led to the flue gas system.
• Streams for samples of LPG are equipped in such way that purge gas is led either
back to the flare or returned back to the product.
• A leak detection and repair programme has been in full implementation for about
10 years.
4.2.3 Waste water treatment
A number of changes have in recent years been undertaken on the WWT in order to
both reduce the amount of oil led to the plant and to reduce the amount of open space
where oil can evaporate. This has been done by the following actions:
• Settling tanks with inner floating roofs prior to the waste water treatment to reduce
the oil led to the WWT.
• Installing skimmers in a pre treatment basin to the API-separators system.
• Removal of oil at different underground culvert systems leading to the WWT.
• The waste water stemming from the product quay is led to the settling tanks ins­
tead of directly to the WWT.
• Total coverage of the API-separators.
4.2.3.1 Actions in 2002
During 2002 the WWT was rebuilt to enable the refinery to fulfil new emission limits
set out in the license for the plant, specifically concerning the amount of suspended
material and nitrogen in the effluent water. These changes were also used in order to
improve the balance of the emissions to air at the WWT. Existing API-separators and
flotation units were exchanged for new flotation units. The basin for pumping of was­
te water to the settling tanks, as well as to the flotation units, was completely covered
and the gas recovered sucked off and led to the biological cleaning stage of the WWT.
Also the biological cleaning stage was renewed. Existing equipment for supplying air
were replaced with systems entering the air to the bottom of the basin. The new system
for the cleaning of nitrogen in the water also should lead to a situation where the air
supply is turned off from time to time in both of the basins where the air is supplied.
This is presumed to also reduce the VOC-emission to air.
16
MEASURES UNDERTAKEN TO REDUCE EMISSIONS
4.3
Nynäs AB, Göteborg
The refinery has in the late 1990’s, after some staggering measurement results on the
VOC-emissions in 1995, been introducing a complete system for vapour recovery for
nearly all tanks on the refinery. The system is also continuously extended.
As the initial measurements were carried out in 1995 there were not any major emis­
sions expected from the site, as nearly only heavy products were being produced and
fed through the system. On the contrary very high emission levels were encountered
o
due to the raised temperature in the bitumen tanks, held at around some 200 C. This
was contrary to all the old techniques for calculating emissions, where emissions from
storing such heavy products by these calculation methods as a definition were set to
zero. The measurements proved this completely wrong.
In the mid 1990’s the refinery subsequently decided to introduce a complete system
for vapour recovery at the tanks of the refinery. The system is divided in two parts, one
where all tanks with non-oxidized products are put together, and one where the oxidized
products are taken care of. There is also a connection in between the both systems to
level the pressure out. The system uses nitrogen as blanket gas.
By later measurements it has been shown that the carried out actions substantially
have decreased the emissions and on top of that an improved reliability in the proces­
sing has been achieved as the tanks, mainly those with oxidized products, now do not
get choked at all, allowing for far fewer shut-downs of tanks and for far fewer cleaning
operations than before.
Roughly the reduction in emissions from the tanks being put together in the vapour
recovery system was reduced by half from 1995 until 1999, due to the system described
above.
17
DESIGNING A MEASUREMENT SURVEY
5. DESIGNING A MEASUREMENT SURVEY
Measurements of fugitive VOC emissions need both sufficient time to be carried out,
and to be sufficient in area coverage. They also need to take into account variations in
the meteorological circumstances during the measurement survey as well as its relation
to the meteorological normal conditions.
It will not be possible to defend continuous measurements on the site by DIAL or
any – if so – equivalent measurement technique at today’s cost. The costs for such an
exercise will be too high. On the other hand too short measurement periods will not
give sufficient data, and will make the data received doubtable in both accuracy and
relevance.
The methods for a good survey, in that the aim is to really sort out and define the
real emission levels, vary from site to site depending on differences in both localisation
and possible interference from other sources, topography and meteorology as well as
fluctuations in the normal running of the facilities at the site. Never the less, below
are proposed some basic rules to run a successful measurement exercise, based on the
Swedish experience.
VOC’s to be included
5.1
Define at an early stage which VOC’s are to be included! For a petrochemical plant it
might sometimes be possible to distinguish this to a few and well defined number of
specified VOC’s due to the production of well distinguished hydrocarbons. This does
on the other hand not mean that in case of an ethylene-cracker you only can go for et­
hylene. You need also to measure ethane, propane, propylene, butane and aromatics
and maybe also some other well defined VOC’s to get the major part of the emission.
For oil refineries on a general basis there is a vast spread on which VOC’s really are
emitted. This means that a measurement should be covering the widest scope possible.
With current existing equipment it is possible to measure alkanes and alkenes in the
span C2 – C22. In case a too narrow span is used the figures measured will be too low
compared to the real situation. As normally the share above C15 is low, it is sufficient
to measure C2 – C15.
Aromatics should also be included and it is possible to measure at least up to some
C10 – C11 with today’s techniques. On a GC that would correspond to about C15 when
talking about the retention times of the straight hydrocarbons. The normal way is to use
the DIAL-equipment for measuring one typical aromatic substance, normally toluene
or benzene, and the other aromatics present are measured by sorption tube equipment
in order to get a sufficiently proper value on their presence related to that aromatic
substance measured directly by the DIAL.
Other VOC’s to be taken care of are the cyclic ones with a cycle less than C6, which
could be included in the alkanes/alkenes-measurement set, although at a maximum
they look to account for some 5-7% of the total, which on the other hand cannot be
said to be negligible.
In case certain interest is lying within the field of methane as a green-house gas, this could
of course also be measured by the DIAL, and should be done so in case it cannot be defined
as a less important parameter for the plant. When describing emission figures it should
however preferably be done separately for methane as the environmental impact of dif­
18
DESIGNING A MEASUREMENT SURVEY
ferent types of VOC vary quite substantially. A proposed division would be the following:
SUBSTANCES
MEASURED (KG/H)
ANNUALIZED (TON/A)
REMARKS
Alkanes C2 – C8
Alkanes C9 – C15
Σ Alkanes
Alkenes C2 – C8
Alkenes C9 – C15
Σ Alkenes
Aromatics Benzene
Aromatics Toluene
Aromatics C8 – C11
Σ Aromatics
Cyclic hydrocarbons
Σ NMVOC
Methane
Σ VOC
This division should preferably be done for – primarily – the site as a whole but also
for each of the main subsections defined for the site, normally at least the crude oil
tank storage, the process area, the waste water treatment plant and the product tank
storage.
The DIAL-system gives by its measurements in the normal operation mode, at a
refinery or the like, levels of a sum of those VOC’s that are detected for a certain
wavelength. To get a picture of which VOC’s are present it means that the DIAL has to
be added to tube sorption measurements or other methods to get the full picture and
the distribution. For this it is very important that the equipment used is able to detect
all VOC’s fully in the whole of the above described span, i.e. not only up to a level of
C8 – C10 when talking about straight hydrocarbons, but up to about C15 instead, and
subsequently also such a range for aromatics.
Below the measurement strategy is described mainly in terms of the use of the DIAL
– or another equivalent system – as this gives the basic variation in emissions on a
mass flux basis. To get the real figures in mass flux it is nearly equally important that
the sorption tubes – or equivalent equipment – are used more or less in parallel to get
the full picture. Without this, or in case a too narrow range is used for the VOC’s, the
received data will not give the full picture.
19
DESIGNING A MEASUREMENT SURVEY
5.2
Meteorological measurements
The meteorology, as wind speed and direction, should be continuously measured at at
least three heights during the whole time of each measurement activity. Normally the
levels should be something like 5-8 m, around 10 m and 15 – 25 m above ground le­
vel, to get an accurate picture of the wind profile.
Continuous reliable information about the wind profile is necessary for getting an
accurate measurement of the emissions from the facilities and continuous data on the
wind direction is also basic information for defining the plume during the measurement
as a whole, as it is the flux perpendicular to the plane that counts.
Another basic requirement for the measurement of the meteorological conditions is
that the free air wind is given and that the met stations are placed in the scan plane so
that the effect of possible partial wind shadows are accounted for.
5.3
Measurement strategy
5.3.1 The whole site
The running of an oil refinery, or the like, in itself contains a lot of parameters which
in different ways can be varied and thus differently affect the operations and thereby
also the emission levels. This is true for both the storage areas and the process area,
although the general influence of day-to-day variations for the storage area, on a gene­
ral level, is definitely greater than for the process area in the case where we do not talk
of sudden leaks in the process or shut-down operations.
To receive reliable data, measurements therefore have to be undertaken in such a
way that variations in normal operations are taken care of and, as much as possible,
also are included in and analyzed during the measurement campaign. This means that
surveys need to be undertaken over such a length of time that variations can be taken
care of, and under such operative conditions within the site so that it during the mea­
surements is possible to gather all necessary and relevant data for the later analyses
and determination of emission mass flux levels. This has to be emphasized even more
when measuring the tank storage area and the different off-loading operations. It is
always recommended that, initially, an overall measurement of the whole site is carried
out. This is also possible to undertake if the site is not all to big in size (up to some 1,2
* 1,2 km). In case the site is very large it is still of importance to get an initial overall
picture of the emissions situation. This then has to be done by splitting the area up in
sections with sizes which are possible to cover by the measurement device.
Each initial overall measurement should not be less than half a day, preferably one
whole day. The time to be spent is also depending on the number of repeat visits aimed
for and is of course also wind direction dependant. When the wind curtails the initial
measurements, this could also be made up for later by the following measurements
during the total measurement survey.
5.3.2
Division into Sub-sections
Having a broad picture of the overall emissions situation the next step is to focus on
the defined sub-sections of the main area. For an oil refinery this would generally mean
20
DESIGNING A MEASUREMENT SURVEY
a division in at least the following areas:
• the storage facilities for raw material (crude oil etc.)
• the process area
• the waste water treatment area
• the product storage and
• other specific areas which might be of certain interest or by other means large emit­
ters like ship, truck or railway loading
For other types of sites non-relevant parts of these could of course be omitted, i.e.
for an oil harbour the “process area” generally is a non existent part.
The geographical and topographical parameters of course also have a general impact
on this choice and generally there is no problem in a further division into sub-sections
other than what is indicated above. Typically it could on top of this also be of interest
to study just one or a few storage tanks or parts of the process area, mainly perhaps
where there are new installations or parts of the plant which are suspected to have
higher emissions than others. See also below chapter 5.3.3. In case the geographical
area for the site is small (in relative terms) the number of specific sub-sections could be
reduced, but then only when taking into account that the possibilities for interpreting
the results are not hampered.
Measurement quantity is a tricky issue. As a general rule you can never have too
much data but this has to be balanced against economics. The idea is to get sufficient
data to cover the day to day or hour to hour normal operations and peaks. Overlaid on
this will be the more abnormal peaks due to a whole range of accidental or maintenance
activities etc. It is rare that you don’t detect one or two of these ‘abnormals’ during the
course of a survey, which however is positive as they anyway have to be taken care of
in a correct manner to address a real annual emission level. An exclusion is debatable
because, although the specific incident may be very unusual, there is infinite potential
for other unusual incidents. Something unusual will be happening with a relative high
frequency at complex sites, and it needs to be recognized that unusual incidents will
add to the emission total of “normal operations”.
As a guideline for a survey, a minimum of two to three days should be devoted to
each sub-section. This time should be split up into at least four separate visits of 3 - 4
hours each at random choice of time during the total survey, but in conjunction with
situations when the right conditions for measuring are met. Where a sub-section is very
large it may be necessary to sub-divide it into even smaller parts with consequently less
time spent at each spot. If several sub-divisions are necessary then the total time devoted
to the whole main area preferably should be proportionately increased.
Specifying the number of scans to be carried out, i.e. single shot measurement of
about 10 – 15 minutes, is often counter productive because scans can be shortened and
coarsened, so a measurement time utilization is better to specify. This should consist
of the system utilization time per day or for the whole survey. What is required is the
actual measurement time of the system excluding between-scan setting and relocation
time, although provision needs to be made for these. A good system should give over 4
hours a day of integrated measurement time, which is then also the basis for the timings
described in this report.
21
DESIGNING A MEASUREMENT SURVEY
As the emissions performance of the different sub-sections at an oil refinery are dif­
fering quite substantially the following advice would be given based on which type of
sub-section we are focusing at:
6.3.2.1 Tank storage
Especially in case of outer floating roofs, emissions are expected to vary with wind
speed and liquid level in the tanks. This proposes specific measurement activities to
cover the impact of these parameters. Measurements should be carried out in such a
way that they in the analyzing can be split up in different single and/or groups of tanks
in order to enable a reception of data which can be used to implement measures to re­
duce emissions.
Typically one division is normally for crude oil tanks and product tanks respectively.
Here it is very important to point out that the old traditional calculation methods more
or less say that emissions are virtually zero for products heavier than kerosene. This
is a huge mistake in these calculation methods, as the real measurements will show
substantial emissions and this especially if such products are heated up above normal
ambient temperature. Another mistake by the old calculations is the misinterpreting of
the huge influence individual variations in between tanks, due to construction, history
and maintenance, although they at a first glance look very similar.
The individual conditions of a tank, especially when looking at the larger tanks with
outer floating roofs, has a sometimes tremendous impact on the real emissions. Some­
times emissions can be about fifty times higher or more compared to what is predicted
by the old calculation methods, even if the liquid used is of kerosene type and lighter.
Each measurement activity should be divided up into a sufficient number of scans so
that enough information is gathered to enable an annualization as well as to have a good
picture of the individual tanks with the highest emissions as well as a general picture of
the variability of emissions due to wind speed and the filling height of the tanks. The
latter especially is important for tanks with outer floating roofs.
5.3.2.2 Process area
Emissions do nearly not at all vary with meteorological conditions, but could be va­
rying due to sudden leaks, changes in leak pattern and – in some cases – throughput
as well as due to major changes in operational conditions. The relatively constant ex­
pected processing conditions could indicate that in some cases – when equal emission
levels are measured from one time to another – the number of measurements to cover
this subsection during the survey could be reduced to as low as two measurements of
the above indicated length if the sub-divisions is not too large.
Measurements need on the other hand normally to be divided up into different parts
of a site as the processing at many sites geographically is split up into different and
well divided sub-divisions. If so, for each of these further sub-sections measurements
have to be carried out as specified above. In a normal situation we talk of some two
to three sub-sections.
5.3.2.3 Waste water treatment plant
Measurements of VOC-emissions from the WWT should be carried out in an analogy
with those done for the tank storage. The variation of emissions with wind speed nor­
22
DESIGNING A MEASUREMENT SURVEY
mally is far less compared to that of storage tanks. In the case where the WWT con­
sists of a large open surface, emissions to air will normally both be high (“cleaning the
water by letting the pollutants evaporate to air”) and to some extent also affected by
meteorological conditions. If settling tanks and other – intermediate storage facilities
– are used in combination with the traditional WWT, their emissions should be mea­
sured separately.
The content of hydrocarbons in the effluent water and the mix of different types of
hydrocarbons – and thereby also the corresponding mix in the emissions to air - will
vary more than what is the case for the other areas. This indicates that the use of mea­
surement device facilities to speciate the hydrocarbons need to be frequently used for
the WWT plant.
5.3.2.4 Other facilities
Measurements around loading facilities should be carried out in such a way as to
making it possible to arrive at some statistical sound level when looking at the nor­
mal operation, the working hours and other general performance parameters for the
trucks as well as the railcar or ships being used. It is expected that there will be good
possibilities to arrive at such measurement data when talking of trucks, as such opera­
tions are quite frequent, and railcars where they are frequently used. The aim has to be
to arrive at typical emission levels for the specific operations and then sum that up to
annual values depending on the number of such operations which are carried out as a
whole, also taking into account typical daily start-up and shut-down situations.
Typically loading operations to truck often to a high degree take place outside of
normal operation hours – quite frequently during early morning hours – which means
that measurements need to cover this period also in an appropriate way.
5.3.3 Dividing sub-sections
When having a good picture of a certain sub-section, or even prior to that, it is re­
commended to also focus measurements on different already detected or expected hot
spots. These could be defined due to many reasons of which some could be:
• Newly constructed plant at the site
• Plants with old equipment
• Plants where certain measures have been carried out to reduce emissions
• Plants where the strategy to reduce emissions would differ from other parts at a
site
• Specific tank operations (such as major tanks with outer floating roofs)
• Parts of a sub-section or sub-division where specifically high emissions are expected
or already have been initially measured
The latter is quite often due to “surprising” bad operations, typically in facilities like
splitters, distillation towers or due to poor maintenance of storage tanks.
For each of these single spots, irrespective of its size, there should be counted up to
2-3 hours of effective measurements in order to get accurate data. In case spots with
very high emissions are detected - and which are possible to tighten within short notice
23
DESIGNING AV MEASUREMENT SURVEY
- this should of course be done. The measurements however still have to be included
in the reporting to show the actual measures that have been undertaken on the plant
in order to display the real situation.
5.3.4
Conclusions
As described above the amount of time and the necessity to, for a single site, split me­
asurements up is different from site to site. Still it is possible, based on the experience
at the Swedish refineries, to foresee what an average measurement survey would look
like in time and methodology when following the guidelines described in the chapters
above.
It has also to be noted that the risk of arriving at disputable data due to not working
according to the points outlined above cannot be ruled out as there always tend to be
a discussion about the final contents, possibly leading to the need for new repeated
measurements, thus making the whole story more expensive than it had been if it had
been done in the right way from the beginning. Good planning and contact with the
measuring team by personnel at the site therefore is an essential part of any survey.
As noted above a measurement day at the site should normally mean at least 4 hours
of real collecting of data, which means concentration and meteorological data, the rest
of the time allowing for accurate placing of the measurement devices (normally in a
truck), tuning of instruments and adhering to the right wind directions. Data should
also be analyzed daily, to in the best way configure the measurements for the coming
days. Summing up the time needed for a measurement survey would thus – as a rule of
thumb – look like the following number of days at the site for a measurement team:
• Measurement of the whole site: 2 days
• Subsections:
- Crude Oil Tank storage: 2-3 day
- Process Area: 2-6 days (for 1-3 sub-sections)
- Waste water treatment: 2-3 days
- Product Tank Storage: 2-3 days
- Loading operations etc.: 2-3 days
• Other certain hot spots: 2-4 days
This makes out a total of 14 – 24 days which for small sites could be reduced to
about one third, but for large sites even more time could be needed.
Preliminary reporting should already be made by the measurement team to the site at the
end of the survey, but there should of course also be daily discussions with the responsible
personnel at the site on the ongoing findings and the proposed coming measurements.
A final written report should normally be presented within one month from the last
day of measurements.
24
MEASUREMENT RESULTS
6. MEASUREMENT RESULTS
As mentioned above DIAL measurements have been undertaken at 6 different Swedish
sites since 1988. There have been 15 measurement surveys using two different systems,
the Spectrasyne system with 12 surveys and Shell Research with the remaining three.
Out of the 15 measurements, 10 are on fuel producing oil refineries, three on bitumen/
lube oil producing oil refineries and two on an oil harbour.
The easiest way to present the measurement results of all these surveys in a report
like this would be to – for each single measurement survey – just present the data that
were reported at the time of each measurement. As the systems and the methodology
have been continuously improved throughout the years, this would however not give a
really fair description of the results when comparing them with each other.
The presentation of, and the abilities to assess, the achieved measured data has
been continuously improved. Initially only hydrocarbons in the range of C3 – C8 were
measured as well as one of the aromatics, normally toluene or benzene. The aromaticscontent has been shown to vary quite substantially with different areas of the refineries
and there has also turned out to be a non-negligible content of hydrocarbons in the
volatile part being heavier than C8, which was the initial upper limit of chain length to
be measured. Now up to C22 is measured. On the light side in recent years C2 is now
also included, initially only reaching down to C3.
To make a true presentation of the real emission values the information of the old
measurement surveys has to be processed together with the once recorded and presen­
ted figures to arrive at comparable and even – with today’s knowledge - more exact
data and to assess trends. It would be of very limited value, with current knowledge
of what is being emitted and with the improved techniques of recording met data and
tube sorption analyses, to go back to old methods for measurement and reporting. The
methods for displaying results in this report are thus discussed below.
6.1
State of the art methodology
Today techniques exist to measure NMVOC’s from C2 up to about C22 in the non-aro­
matic range. Hydrocarbons in the heavy range above C15 only make out a small por­
tion of the emission and could thus be exempted as they can be difficult to analyze.
Aromatics are possible to measure up to the same retention times as straight C15. It
is therefore no reason for not measuring these, as aromatics of this size seem far from
negligible for the total mass flux.
For storage facilities it is also important that wind measurements are correct, and
that it is possible to normalize emissions to air from the tank storage facilities to what
is defined as the normal meteorological situation, mainly talking in terms of wind speed.
There is an impact of wind-speed on the emission levels at the storage facilities. Spe­
cifically this impact is high when the wind-speed is very high and the impact has been
shown to also have a slight exponential profile. On the other hand, very high wind speed
is normally not the predominant situation but the impact of such situations should at
least be addressed, in case there temporarily is a high windspeed when measuring. At
the initial Swedish measurements, data was not specifically related to any normalized
wind speed, but emphasis has been put on the issue in recent years establishing windnormalized emission data. Normalization should then be done to a situation typical to
25
MEASUREMENT RESULTS
the specific spot within a site where measurements are carried out, which means that
for one refinery site there could be different average wind speed levels depending on
the place of measurement.
The means of reducing this possible impact is of course by, during one measurement
period, doing a number of repeated measurements at each of the different measurement
spots. By experience the wind normalization at tanks could as a rule of thumb mean up
to about +- 10-20% of the measured emission level, but of course less in comparison
with the total measured emissions for the site as a whole.
As the DIAL-measuring system works in a real life situation, i.e. measuring the VOC
which are passing through the measurement-plane, there are of course also emitted
VOC’s that do not get across that plane (the lower the wind, the higher the degree),
which means that the DIAL-measured emissions always can be expected to be on the
low side due to this. Trials carried out to get a better understanding of this phenomenon
show that the maximum “lost” emission due to this normally would be about 10% of
the real emission. In the figures below this “loss” is not taken into account, but is here
anyway mentioned as it indicates that emission-levels in fact could be even higher than
what is being measured and presented below.
6.2
Presented data
The data presented below consists of both comparable data for a number of years at
those sites where the highest number of measurements have been carried out, the Preem and Scanraff refineries, as well as data showing typical changes in emission levels
due to variations in the conditions of some of the equipment at the Scanraff refinery.
All presented figures are exclusive of methane, as methane has a completely different
environmental impact than does the other VOC’s – although they also within themselves
show big differences in environmental impact – as methane is more of a green house gas
than anything else. A rough guess is however that methane could add some 10-20% on
top of the total emission as a rough approximation, and then of course varying with site,
equipment, service and with time. Some specific measurements have also been carried
out on the Swedish refineries to indicate typical levels of methane emissions for crude
oil tanks and the waste water treatment. Methane in these exercises have amounted to
12-33% of the total NMVOC-emissions for the crude oil storage and being as high as
50-80% of the NMVOC-emissions from the WWT.
Measurements on the Swedish sites have been carried out with two different systems,
the Spectrasyne (former BP Research) system which can measure both in the infrared
(alkanes etc.) and the ultraviolet (aromatics) and the Shell Research system which only
can measure in the infrared (alkanes etc.).
Combined with these systems meteorological measurements have been carried out
as well as tube sorption measurements. The range of hydrocarbons covered vary with
the system used. Generally it could be noted that emissions of VOC’s could be expected
up to at least the C15-level (pentadecane). The Spectrasyne measurements have been
carried out to meet this requirement, whereas Shell Research only reaches the level of
some C8– C10. Spectrasyne, in contrary to Shell Research, also includes C2 in the total
and has also made some spot measurements on methane, although this is not included in
26
MEASUREMENT RESULTS
normal VOC-figure, i.e. VOC-figures should in this report be looked at as NMVOC
if nothing else is stated.
The conclusion is that it already from the beginning are different results to be expec­
ted from the two systems as the measurements are not done in an equivalent way. It is
quite obvious that the Shell Research measurement device will record too low values
compared to the real life situation as hydrocarbons heavier than C8-C10 are not detec­
ted, and also not C2. As the Shell Research DIAL does not include the ultraviolet it
does not measure aromatics either which also makes the presented figures a bit more
doubtable. Being aware of this, it does seem to be possible to in some way interpret the
Shell Research measurements, although it will not be possible to in an accurate way
say how much too low the presented figures are compared to the true total emissions
as C2 and hydrocarbons above C8/C10 are left out and no aromatics are measured with
the Shell Research DIAL. As noted above however in contrary though the Spectrasyne
measurements fulfil the needed requirements.
Presenting single measurement surveys at a site only gives a rough indication of
the emissions level at the site – still however much more reliable than any theoretical
calculation – and is therefore not the best way of describing the emissions situation as
the level of emissions may vary from time to time due to the condition of the site, both
when talking about the tanks (i.e. conditions of seals and tanks as a whole) and the
process area (sudden leaks etc.).
A better way of describing the emissions situation is to describe it for those sites where
at least three measurement surveys have been carried out and where the measurement
results can be compared with each other. The measurements should then preferably
also be compared bearing in mind the current state of the art of measurements, which
means that DIAL-measurements for aromatics, and hydrocarbons in the range of C2
– C15 also should be taken into account.
To make this possible the choice in this report has been to present the data of
measurement surveys for the Preem and Scanraff refineries during the period of 1992
– 1999, bearing in mind the current state of the art of the measurements. This means
that the older and historically reported measured data is, where it has been shown to
be needed, recalculated to the current standards to be comparable and to show trends.
Data for the tank storage area has also been normalized to the wind speed on average
being accurate for the area of the specific refinery. For Scanraff data is also presented
for a few practical situations, showing the impact and value of real life emissions and
the uselessness of old calculated data.
In the tables data are transferred from measured kg/h to tons/a by presuming an
average of 8 600 hours/a of emissions a year. By this periods of maintenance and shut
downs are taken into account. Of course minor individual differences do exist, but
their impact can anyway be assumed to be below the total level of accuracy of the
presented figures.
27
MEASUREMENT RESULTS
6.3
Preem Raffinaderi AB, Göteborg
AREA
1988
kg/h
Crude Oil Tanks
(Wind Normalized)
1989
tons/a
kg/h
1992
1995
tons/a
kg/h
tons/a
kg/h
1999
tons/a
kg/h
tons/a
410
3500
350
3000
180
1590
90
790
80
700
Process Area
1640
14100
530
4600
115
1000
130
1150
170
1490
Waste Water
Treatment
56
480
55
470
9
80
17
140
37
320
Product Tanks
(Wind Normalized)
Total
750
6400
750
6400
310
2700
270
2330
170
1470
2860
24500
1680
14500
620
5360
510
4410
460
3980
Note: In 1988 emissions from Product Tanks were not measured, as the emissions were presumed to be low. As to make figures
comparable, figures for these emissions have 1989 have been put in the table to also represent 1988.
6.4
Skandinaviska Raffinaderi AB, Scanraff, Lysekil
AREA
1995
kg/h
South Tanks incl Crude
oil (Wind normalised)
310
2700
90
770
350
3020
Process Area
380
3270
260
2270
230
2000
Waste Water Treatment
160
1350
80
690
55
480
Main Tanks (Wind
Normalized)
660
5660
320
2760
430
3670
1500
12900
760
6500
1060
9160
Example of tank storage conditions – tanks
with outer floating roofs
A set of comparative measurements and anal­
yses have been undertaken at a sertain set of
tanks at the Scanraff refinery. The results are
shown below in a table. As a background to
the table the following should be noted: At the
Scanraff refinery in 1992 high levels of emis­
sions were recorded from tanks with Vacuum
gasoil, because light hydrocarbons had slipped
to the tank with the product, i.e. due to poor
upstream operation. High emissions from cru­
de oil tank Tk-1401 were recorded due to high
liquid level being kept in the tank. In 1992 only
two gasoline component tanks were equipped
with secondary seals.
28
1999
tons/a
Total
6.4.1
1992
kg/h
tons/a
kg/h
tons/a
At the 1995 measurement survey all storage
tanks with outer floating roofs had recently been
equipped with secondary seals and thus main­
tenance work had been done quite close before
the measurements, so emissions were measured
to be very low.
In 1999 the emissions from the gasoline as well
as the gasoline component tanks (with outer floa­
ting roofs) had increased due to presumed poor
performance of the seals introduced. High emis­
sions were experienced from the crude oil tanks.
The inspection that followed due to the high
recorded measurement results proved a number
of leakages along the roof sealing of the crude oil
tanks. Major maintenance works were carried out
on two of the crude oil tanks and on the third the
MEASUREMENT RESULTS
seals were changed to new ones. Unfortunately however no follow-up measurements
were carried out to see the results of the latter installation.
Below the measured emissions at the tanks are presented as comparison-figures with
the results related as factors compared to the old rigid calculation methods, ie. the true
emission value is presented as a factor compared to what emission level is accounted
for when only relying on calculations. In the table measured data from 1992 and 1995
is in this case not compensated to the 1999 state of the art knowledge, but it anyway
quite clearly shows the dependence of good maintenance work and the need for fre­
quent control on emission levels. Recalculating all data to 1999 standards would even
clearer have shown the trends and the value of measured emission data compared to
only calculated data, the latter not varying at all from time to time.
TANKS WITH OUTER FLOATING ROOFS
DIAL MEASUREMENT (YEAR)
1992
(factor)
1995
(factor)
Gasoline tanks
2,7
2,0
2,4
Gasoline component tanks
1,9
1,7
2,2
Crude oil tanks (Tk-1401 and Tk-1402)
26
2,6
13
-----
1,1
52
Crude oil tank (Tk-1406)
6.4.2
1999
(factor)
Rearranging the waste water treatment unit
Also for the waste water treatment unit a follow-up study has been carried out for
the Scanraff the existing refinery. In 1992 at Scanraff settling tanks were used for
ballast water only and they were also equipped with fixed roofs only. The APIseparator was only partially covered.
At the 1995 measurement survey one of the settling tanks was equipped with an
inner floating roof and was also put into a somewhat different service, being used as a
settlingtank for both ballast and all other waste water produced at the site.
In 1999 both settling tanks were used for all waste water produced at the site and
had also become equipped with inner floating roofs. The API-separators were now
completely covered.
29
kg/h
45
1992
1995
1999
40
35
30
25
20
15
10
5
0
Wast water treatment
Settling tanks
In the table data from 1992 and 1995 is not compensated to 1999 state of the art conditions, but anyway quite clearly show the dependence of good maintenance work and need for frequent control on emission levels. Recalculating all data to 1999 standards would have shown the decreased emissions even more clearly. The County Administration of Västra Götaland | The Environmental Protection Section | 403 40 Göteborg | tel +46-(0)31-60 50 00 | fax +46-(0)31-60 58 97 | www.o.lst.se
OCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCVOCV
County Administration
This page intentionally left blank.