i
A Methodology to Assess the Reliability of Hydrogen- based
Transportation Energy Systems
By
RYAN WILLIAM McCARTHY
B. S. ( University of California, San Diego) 2002
THESIS
Submitted in partial satisfaction of the requirements for the degree of
MASTER OF SCIENCE
In
Civil and Environmental Engineering
In the
OFFICE OF GRADUATE STUDIES
of the
UNIVERSITY OF CALIFORNIA
DAVIS
UCD- ITS- RR- 04- 36
Committee in Charge:
Prof. Joan Ogden
Prof. Daniel Sperling
Prof. Patricia Mokhtarian
December 2004
ii
ACKNOWLEDGEMENTS
I am in the debt of my colleagues and friends who volunteered to participate in this study:
Matthew Caldwell, Anthony Eggert, David Grupp, Courtney Harter, Jonathan Hughes,
Nils Johnson, Michael Nicholas, Nathan Parker, Brett Williams, and Christopher Yang;
my mentors, whose wisdom has guided me throughout: Dr. Joan Ogden, Dr. Daniel
Sperling, and Dr. Patricia Mokhtarian; and my family and friends, without whose love
and support I would never have the opportunities I so much enjoy. My heartfelt thanks
goes out to you all.
iii
ABSTRACT
This paper introduces a method to assess the reliability of hydrogen supply systems for
transportation applications. It relies on a panel of experts to rate the reliability and
importance of various metrics as they pertain to selected hydrogen systems. These are
aggregated to develop broad reliability scores to be compared across systems. A trial
application of the methodology is presented, where a group of hydrogen researchers at
the Institute of Transportation Studies at the University of California, Davis comprise the
expert panel. Two hydrogen pathways supplying a hypothetical network of refueling
stations in Sacramento were compared. The first uses centralized steam reforming of
imported liquefied natural gas and pipeline distribution of hydrogen. The second
electrolyzes water onsite from electricity produced independent of the grid, and no
hydrogen transport is required. The panel determined the second pathway to be more
reliable, primarily due to the lack of imports, the distributed nature of the system, and the
lack of hydrogen transport. This preliminary application only intends to demonstrate
how the method is applied, however, and the results presented here should not be taken as
definite.
iv
TABLE OF CONTENTS
LIST OF TABLES........................................................................................................... vi
LIST OF FIGURES........................................................................................................ vii
INTRODUCTION ............................................................................................................ 1
Motivation and Background ........................................................................................ 1
BACKGROUND............................................................................................................... 4
Statistical Approaches to Reliability Assessments ..................................................... 4
Quantitative Reliability Assessments......................................................................... 5
Qualitative Reliability Assessments........................................................................... 6
Reliability in the Energy Sector................................................................................... 7
Electricity Sector ........................................................................................................ 9
Reliability – Adequacy ........................................................................................... 9
Applied Probabilistic Methods ......................................................................... 11
Reliability – Security ............................................................................................ 13
Security Planning.............................................................................................. 13
Governance and Oversight ............................................................................... 16
Managing Security............................................................................................ 17
Natural Gas Sector................................................................................................... 19
Natural Gas Supply.............................................................................................. 19
Recent Trends.................................................................................................... 20
Future Projections ............................................................................................ 21
Liquefied Natural Gas ( LNG)........................................................................... 22
Infrastructure Reliability ..................................................................................... 24
Pipelines............................................................................................................ 25
LNG................................................................................................................... 27
Interdependencies ............................................................................................. 27
Summary............................................................................................................... 28
Petroleum Sector...................................................................................................... 29
Reliability Perspectives from the Petroleum Industry ........................................ 30
The New Business Environment........................................................................ 30
Risk Management.............................................................................................. 31
Risks .................................................................................................................. 33
U. S. Petroleum Dependence and Its Economic Implications ............................ 33
Measures of Petroleum Dependence ................................................................ 34
Measures of Vulnerability to Supply Disruption .............................................. 38
Costs of Oil Dependence................................................................................... 43
Reliability of Global Supply Infrastructure ........................................................ 47
Supply Outlook.................................................................................................. 47
Geopolitics ........................................................................................................ 48
Threats .............................................................................................................. 50
Infrastructure Risks........................................................................................... 51
Summary............................................................................................................... 56
METHODOLOGY ......................................................................................................... 56
Methodology Overview............................................................................................... 56
1. Define Scope of Study and Select Participants.................................................... 57
v
2. Define Reliability in Hydrogen Energy Systems ................................................. 59
3. Select Metrics to Value Reliability in Hydrogen Energy Systems..................... 60
4. Specify Hydrogen Energy Systems to Evaluate .................................................. 61
5. Develop Evaluation Matrix ................................................................................... 63
6. Develop Rating Scales and Rating Criteria......................................................... 65
7. Collect Expert Reliability and Importance Ratings ........................................... 67
8. Aggregate Expert Ratings to Determine Reliability Scores............................... 68
9. Compare Reliability Scores across Pathways...................................................... 73
APPLYING THE METHODOLOGY.......................................................................... 73
1. Define Scope of Study and Select Participants.................................................... 75
2. Define Reliability in Hydrogen Energy Systems ................................................. 76
3. Select Metrics to Value Reliability in Hydrogen Energy Systems..................... 77
Adequacy .................................................................................................................. 79
Security..................................................................................................................... 79
4. Specify Hydrogen Energy Systems to Evaluate .................................................. 82
5. Develop Evaluation Matrix ................................................................................... 83
6. Develop Rating Scales and Rating Criteria......................................................... 85
7. Collect Expert Reliability and Importance Ratings ........................................... 87
8. Aggregate Expert Ratings to Determine Reliability Scores............................... 91
9. Compare Reliability Scores across Pathways...................................................... 96
CONCLUSIONS........................................................................................................... 101
Lessons Learned from Trial Application................................................................ 102
Opportunities for Future Research......................................................................... 106
BIBLIOGRAPHY......................................................................................................... 110
APPENDIX A: GEOPOLITICAL OVERVIEW OF OPEC MEMBER
STATES ......................................................................................................................... 115
APPENDIX B: DESCRIPTION OF INTERNATIONAL OIL TRANSPORT
CHOKEPOINTS........................................................................................................... 128
APPENDIX C: MATERIALS PROVIDED TO THE EXPERT PANEL.............. 133
APPENDIX D: AUTHOR’S RELIABILITY RATINGS......................................... 164
vi
LIST OF TABLES
Table 1. Natural gas supply projections through 2025 ( adapted from: EIA, 2001b,
pp. 22- 23)............................................................................................................ 22
Table 2. Natural gas reserves by selected country. Current LNG exporters are
darkly shaded, potential LNG exporters are lightly shaded ( adapted from:
EIA, 2003, p. 5)................................................................................................... 24
Table 3. Top five petroleum supplying nations into U. S. from 1973 to 2003 ................. 39
Table 4. Physical U. S. oil infrastructure components ( adapted from: NPC, 2001,
p. 32) .................................................................................................................. 51
Table 5. Reliability and importance ratings for two hypothetical pathways ................... 71
Table 6. Reliability scores for two hypothetical hydrogen pathways using two
aggregation methods .......................................................................................... 72
Table 7. Scale used to rate the reliability of each metric as it applies to each pathway
component.......................................................................................................... 86
Table 8. Scale used to rate the importance of the metrics to reliability of the pathway
component.......................................................................................................... 87
Table 9. Sample rating criteria for the metric intermittency............................................ 87
Table 10. Average and standard deviation of experts’ reliability ratings ........................ 94
Table 11. Average and standard deviation of experts’ aggregated reliability scores ...... 96
Table 12. Average and standard deviation of experts’ maximum possible
aggregated scores ............................................................................................. 97
Table 13. Aggregated reliability scores showing percentage of maximum score
possible ............................................................................................................ 98
vii
LIST OF FIGURES
Figure 1. Reliability networks: a) series network, b) parallel network............................. 6
Figure 2. Net U. S. imports of natural gas, 1990- 2025 ( EIA, 2003, from AEO 2004
reference case) .................................................................................................. 23
Figure 3. The National Petroleum Council’s assessment of physical vulnerabilities
facing natural gas infrastructure ( NPC, 2001, p. 34) ......................................... 25
Figure 4. Natural gas sector interdependencies ( NPC, 2001, p. 29)................................ 28
Figure 5. U. S. net petroleum imports since 1970 ............................................................ 34
Figure 6. U. S. petroleum stocks and their coverage against imports and consumption .. 35
Figure 7. U. S. petroleum stocks and their coverage against imports and consumption,
minus Lower Operational Inventory Levels ..................................................... 36
Figure 8. Percentage of total energy consumption met by petroleum in the U. S. ........... 37
Figure 9. U. S. oil expenditures as a percent of GDP ....................................................... 38
Figure 10. Concentration of U. S. petroleum imports from its top five supplying
countries.......................................................................................................... 40
Figure 11. OPEC share of global crude oil production.................................................... 41
Figure 12. Persian Gulf share of global crude oil production.......................................... 42
Figure 13. World excess petroleum production capacity vs. price .................................. 43
Figure 14. U. S. expenditures on imported oil and the trade deficit, in 2003 $................ 44
Figure 15. Distribution of global crude oil reserves ........................................................ 48
Figure 16. The National Petroleum Council’s assessment of physical vulnerabilities
facing oil infrastructure ( NPC, 2001, p. 33) .................................................... 52
Figure 17. Structure of hydrogen reliability evaluation matrix ....................................... 63
Figure 18. Sample importance ratings: a) different importance ratings for each
pathway component, b) same importance ratings for each pathway
component....................................................................................................... 65
Figure 19. Comparison of reliability scores for two hypothetical hydrogen pathways
using the two aggregation methods................................................................. 72
Figure 20. Hydrogen reliability metrics considered in this study.................................... 78
Figure 21. Evaluation Matrix for Pathway # 1 and Pathway # 2 used in this study .......... 84
Figure 22. Sample question excerpted from survey, ascertaining expert opinions on
the importance of two metrics to the subcategory capacity............................ 90
Figure 23. Sample question excerpted from survey, ascertaining expert opinions on
the reliability of three metrics corresponding to the subcategory flexibility
in Pathway # 1.................................................................................................. 91
Figure 24. Aggregation steps used to determine aggregated adequacy scores ................ 93
Figure 25. Comparison of adequacy and security scores for Pathways # 1 and # 2
( unscaled)........................................................................................................ 99
Figure 26. Comparison of adequacy and security scores for Pathways # 1 and # 2
( scaled according to maximum possible reliability scores) .......................... 100
Figure 27. Chokepoints for international petroleum transport ( International Institute
for Strategic Studies, 2001)........................................................................... 129
1
INTRODUCTION
A transition to hydrogen as a primary transportation fuel offers potential societal benefits
over the current paradigm. Some advocates claim that hydrogen would provide a more
reliable energy system. But reliability benefits associated with a switch to hydrogen have
not been studied. This research introduces a method to assess the reliability of hydrogen
supply systems for transportation applications. The discussion here is limited to
comparing reliability between hydrogen supply systems (“ hydrogen pathways”), but the
methodology itself is not so constrained. It could be applied to compare the reliability of
other energy systems to hydrogen as well.
Motivation and Background
Existing energy infrastructures tend toward massive, highly integrated systems which can
catastrophically fail with any link. The electric grid delivers energy from large, isolated
power plants via a limited number of high- voltage transmission lines connected at a few
critical nodes. Massive blackouts, such as the one that hit the East Coast on August 14,
2003, exemplify the fragility of the electric grid. During the outage, 61,800 MW of
power serving 50 million people were lost, resulting in costs estimated between $ 4 billion
and $ 10 billion ( ELCON, 2004).
Petroleum systems are similarly centralized, with pipelines reliant on a few pumping
stations delivering products from remote, aging refineries. The consequences of the
centralized delivery system were felt nationwide when gasoline prices soared to record
highs in the spring of 2004. Compounding reliability concerns is the concentration of
2
petroleum resources in the tumultuous Middle East, and several “ chokepoints” along
delivery routes from the region.
As energy systems apparently grow more vulnerable, the prevailing business climate is
such that reliable energy supply is valued more than ever. A new business environment
characterized by automated operations, just- in- time logistics, and rapid changes has
emerged with the coming of information technologies. Business today is utterly
dependent on the numerous systems that support it, and cannot function without their
reliable operation. Consequences stemming from infrastructure disruptions have grown
more severe, and often no feasible manual backup processes exist ( NPC, 2001).
Energy reliability has gained increased focus in political and social realms as well. Issues
dominating the news and political debate include volatile gasoline prices and
developments in the Middle East. The tragic events of September 11, 2001 prompted the
creation of a new Cabinet position, overseeing the Department of Homeland Security.
One of the Department’s five major directives is the protection of “ critical
infrastructure,” including energy systems ( NPC, 2001, p. 1). Since the attacks, the U. S.
has gone to war and has seen anti- American sentiment rise. More attacks have been
threatened, and energy systems are perceived as high- value targets. The result is
increased public awareness and demand for reliable energy systems.
Many suggest that a switch to hydrogen as an energy carrier can relieve the
environmental and reliability problems posed by current energy systems. Since hydrogen
3
can be produced from any number of resources – including renewable electricity – and
utilized essentially pollution- free in a fuel cell, it certainly presents the potential to serve
as an environmentally sustainable fuel. But, hydrogen can also be produced and used in
ways that would significantly increase emissions over their current levels. Several
studies have considered hydrogen supply scenarios from the environmental slant, and
confirmed these findings ( e. g., NRC [ 2004], Weiss et al. [ 2000], GM et al. [ 2002]). But
none have investigated in detail claims that hydrogen affords a more reliable system. A
systematic assessment of hydrogen reliability is needed to assess these claims and to
properly account for reliability in the potential development of a widespread hydrogen
infrastructure.
This study introduces a methodology to assess the reliability of hydrogen energy systems.
First, reliability is defined for hydrogen energy systems and metrics are selected to value
it. Next, hydrogen pathways are selected and described. Three constituent components
of the pathways are assessed by a panel of experts – the primary energy supply system,
the hydrogen production process, and the hydrogen transport process. They rate the
reliability and importance of each pathway component in terms of the metrics. Finally,
their ratings are aggregated to determine broad reliability scores that can be compared
across pathways.
The intent of this work is to provide a tool to guide decision makers to properly consider
and design reliability into hydrogen systems for the public good. Selecting and
promoting an individual pathway as the most reliable is not the goal. Indeed, results from
4
an application of the methodology to two unrelated pathways are given, but they should
not be considered definitive. The motivation of this preliminary application was to test
the methodology and demonstrate its use, not to reach definite conclusions about the most
reliable hydrogen pathways. Nevertheless, the results are interesting, and indeed telling
of hydrogen reliability.
To the best knowledge of this author, the work here represents the first effort to examine
hydrogen reliability in depth. It is that – a first attempt – and will undoubtedly benefit
from future revision and the insights of others. But the hope is that the methodology will
promote the fair consideration of reliability between hydrogen pathways, and potentially
between energy sectors. We are in the unique position of creating an entirely new energy
system where energy security, environmental awareness, safety, and infrastructure
reliability can be ingrained in the system from the onset. At a time when these concepts
have never been more highly valued in society, this opportunity should not be
overlooked.
BACKGROUND
Statistical Approaches to Reliability Assessments
Reliability assessments are well developed for systems applications in the field of
statistics. They generally define reliability in terms of the likelihood of a failure, and
determine the reliability of a system based on the known reliabilities of its elements.
Reliability assessments are usually quantitative, and results take the form of a probability,
but when data is lacking they can take on a qualitative form.
5
Quantitative Reliability Assessments
Traditional reliability assessments use probabilistic techniques to establish the likelihood
that a system will be found in some state of non- operation within a given time period. In
that context, reliability is defined as “ the probability that an item ( component, equipment,
or system) will operate without failure for a stated period of time under specified
conditions” ( Andrews and Moss, 2002, p. 3). Reliability is measured as a probability –
that is, a value between 0 and 1 – over a given time period. So output from a
probabilistic reliability assessment might read: “ the 5000- hour reliability of item x is
0.95,” meaning that item x has a 95% chance of operating without failure over the course
of 5000 hours.
From this definition, the reliability of a simple system can be determined quantitatively. 1
Reliability networks represent the dependencies between components in a system. The
simplest networks are series networks and parallel networks. A series network is a
system that cannot tolerate component failure. There is no redundancy in the system, and
if one component fails, the entire system fails. A parallel network includes redundancy,
and all parallel components must fail for the system to fail ( Andrews and Moss, 2002,
pp. 167- 169). The two configurations are depicted in Figure 1. If the reliability of the
two components is known, reliability of the system can be determined. Let r1 be the
reliability of component 1 ( i. e., probability that component 1 works over a given time
frame), and r2 be the reliability of component 2 over the same period. Then reliability
can be determined quantitatively for the series network as follows:
1 Leemis ( 1995) describes five ways to calculate reliability quantitatively, but that discussion is beyond the
scope here.
6
Reliabilityseries = Prob[ 1 works AND 2 works]
= r 1
r2 .
Similarly for the parallel network:
Reliabilityparallel = Prob[ 1 works OR 2 works]
= r 1
+ r2 – r1r2 .
Figure 1. Reliability networks: a) series network, b) parallel network.
Qualitative Reliability Assessments
When probabilities cannot be quantified due to a lack of data, reliability assessments can
take a qualitative approach, using expert opinion to establish elemental reliabilities.
Contadini ( 2002) suggests several ways to collect expert opinions, including traditional
surveys and the Delphi process. The Delphi process is used to build consensus among a
panel of experts while avoiding the drawbacks of face- to- face interaction. Contadini
reviews the literature, and summarizes four key features that characterize the process:
7
• Anonymity – allows more diverse responses
• Controlled feedback – multiple rounds of surveying are conducted, to build the
experts’ knowledge of the material and the process
• Interaction – meant to promote open discussion and aid in building consensus
• Statistical aggregation – group member responses are weighted, combined, and
analyzed
When relying on expert opinion, proper selection of the expert panel is crucial. Ideally,
the panel should include members from all slants on a particular topic. But in some
cases, a more accurate analysis may result if representatives of some schools are actually
excluded, if they are thought to be biased ( Bedford and Cooke, 2001, p. 192). The results
of any qualitative study will be sensitive to the selection of the panel, and the level of
expertise possessed by panel members. One method to minimize error is to include a
weighting factor to account for the confidence an expert has in his or her responses. A
more rigorous method is performance based weighting ( Cooke, 1991). Experts are asked
a series of questions whose responses are known to the analyst, but not the expert. Based
on their responses to these questions, a weighting factor is computed to calibrate their
responses to the survey questions.
Reliability in the Energy Sector
In Brittle Power, Amory and Hunter Lovins describe the “ brittleness” of existing energy
systems, and explain how to best design energy systems to be resilient against failures.
According to the Lovins, energy systems in the U. S. are made up of complex components
8
that are prone to failure, difficult to diagnose and fix, and interact with interdependent
components in complicated ways. They also tend to be inflexible, and are unable to
easily adapt to changes in demand or primary energy supply. These characteristics make
energy systems incredibly vulnerable to potentially catastrophic failures. The Lovins
argue that failures are inevitable, but resilient energy systems can minimize the damage
by rapidly isolating and repairing disruptions. They claim that resilience can best be
achieved in an energy system with numerous small modules which each have a low
individual cost of failure.
The National Research Council ( NRC) published a report following September 11th that
includes many of the same concepts as Brittle Power ( NRC, 2002). The report
recognizes vulnerabilities in energy systems and describes ways in which science and
engineering can work to protect against malicious attacks. It recommends actions that
can be undertaken to reduce vulnerability in energy systems, and identifies further
research areas to reduce risks. A key recommendation throughout is to increase
cooperation with the national security and defense communities, who have dealt with
such threats for many years.
These references apply broadly throughout the energy sector, but most of the literature
reviewed focused on specific sectors. Below, background and literature reviews specific
to the electricity, natural gas, and petroleum sectors are provided. Each considers the
existing state of the sector and looks at how reliability is defined, valued, and assessed.
9
Electricity Sector
Reliability in the electricity sector is defined in terms of two components – adequacy and
security. Adequacy considers average supply and demand over the long term, while
security is concerned with dynamic operating conditions in the immediate term. The
North American Electricity Reliability Council ( NERC) defines the terms as follows:
Reliability – The degree to which the performance of the elements of the
system results in power being delivered to consumers within
accepted standards and in the amount desired ( as cited in:
Kirby and Hirst, 2002, p. 9).
Adequacy – The ability of the electric system to supply the aggregate
electrical demand and energy requirements of customers at
all times, taking into account scheduled and reasonably
expected unscheduled outages and system elements ( NERC,
2002, p. 7).
Security – The ability of the electric system to withstand sudden
disturbances such as electric short circuits or unanticipated
loss of system elements ( NERC, 2002, p. 7).
Reliability – Adequacy
The NERC produces annual assessments of the adequacy of the North American
electricity system ( NERC, 2002). They reduce the electricity system into its resource,
transmission, and fuel supply components, and determine adequacy by comparing the
projected capacity of each component to projected average demands over ten years.
Resource ( i. e., generation) adequacy considers the ability of projected electricity
generation facilities to supply future demand. Growth of peak demand is projected over
the time frame of the study, primarily based on the expected future economic growth of
10
the region. 2 Generation supply additions are also predicted over the time period. From
these projections, the capacity margin ( the percentage by which resource capacity
exceeds peak demand) is predicted. If capacity margins are within acceptable levels,
resources are deemed adequate.
Transmission adequacy considers the ability of the transmission system to handle new
load patterns resulting from increased electricity transfers and demand. Similar to
resource adequacy, demand levels are projected over the time frame of the study and
compared to projected capacity expansions. 3 Another gauge of transmission adequacy is
the number and severity of transmission line relief ( TLR) procedures. They are classified
according to severity, on a scale of 0 to 6 ( 6 being the most severe), and indicate a degree
of instability in the electric grid. Although the procedures are used to maintain security
in the system, studying their trends can shed light on its adequacy as well.
Fuel supply adequacy depends on several factors for each resource. The availability of
fuel resources can be projected in a similar fashion as generation and transmission were
above, but it also depends on characteristics far more uncertain. For example, the
availability of fossil resources is influenced by geopolitics, environmental regulations,
extraction technologies, and weather. The availability of renewable resources similarly
depends on future policy measures, conversion technologies, and weather patterns. End
2 These forecasts are probabilistic in nature, and planners usually use a 50% projection, which indicates that
there is a 50% chance that demand will exceed the projection, and a 50% chance that demand will fall
below the projection.
3 New capacity includes line construction, voltage upgrades to existing lines, utilization of empty tower
positions, additional capacitor banks or transformers, and upgrading limiting circuitry at substations.
11
use technologies and consumer behavior affect all fuel resources, and are impossible to
predict.
Applied Probabilistic Methods
The percentage reserve method and others described above can be extended to include
the probability of future service interruptions. Probabilistic methods allow the stochastic
nature of system behavior, customer demands and component failures to be included in
analyses. Understanding the likelihood of service interruptions also allows a balance to
be reached between economics and reliability, according to a cost/ benefit framework.
Probabilistic assessments consider adequacy of the electricity system on three
“ hierarchical levels.” Debnath and Goel ( 1995) describe the assessments and outline
reliability indices at each level. Hierarchical Level I ( HLI) evaluates the adequacy of
generation facilities, ignoring that of the transmission and distribution systems. 4 Multiple
indices can be used to evaluate reliability at HLI. Loss of Load Expectation ( LOLE)
captures the average number of days in which the daily peak load is expected to exceed
available generating capacity. It is determined from the daily peak loads and the
probability that a generating unit will be found in some state of incapacity. A benchmark
adequacy index used by many utilities is LOLE = 0.1 days/ year. LOLE is the most
common index, but it does not translate to customer losses and cannot be used in a
cost/ benefit analysis. Loss of Energy Expectation ( LOEE), and Frequency and Duration
( F& D) extend LOLE and can be used in a cost/ benefit framework, but are less common.
4 Akin to resource adequacy as defined by the NERC ( 2002).
12
LOEE, defined as the ratio of energy supplied to energy demanded, includes the severity
of an interruption. F& D identifies the expected frequency and duration of deficiencies.
Hierarchical Level II ( HLII) considers the ability of generation and transmission together
to supply electricity at bulk supply points ( Billinton, 1969). HLII assessments are usually
performed using analytical techniques or Monte Carlo simulation. Reliability indices can
be considered either at load points or on the system level. Load point indices are used to
identify weak points in the system, and include the probability, frequency and duration of
outages, unsupplied energy, and curtailed loads. System indices are used to describe the
adequacy of the complete system, without regard to specific load points. Some system
indices are system unsupplied energy, bulk power supply disturbances
( occurrences/ year), bulk power interruption index ( MW/ MW yr), and system- minutes
( annual unavailability if all interruptions occurred at peak loads).
Hierarchical Level III ( HLIII) considers the adequacy of electricity generation,
transmission, and distribution facilities altogether. This presents an enormous task, and is
rarely conducted. As in HLII, indices are determined at load points and on the system
level. Load point indices include: expected rate of failure, the average duration of
failure, and the average annual outage time. System performance indices are: system
average interruption frequency index, customer average interruption frequency index,
system average interruption duration index, customer average interruption duration index,
energy not supplied index, average service availability index, and average service
unavailability index ( Billinton and Allan, 1984).
13
Reliability – Security
Security assessments look at the ability of the system to prevent disruptions of service to
end users in real time. Important to assessing security is defining normal ( i. e., non-disrupted)
operating conditions. Normal operation of the electricity grid can be described
as the condition when frequency and voltage are within acceptable bounds, no component
is overloaded, and no load is involuntarily disconnected ( Alvarado and Oren, 2002, p. 3).
Conditions that deviate from these suggest a security failure.
Providing security in the electricity sector is complicated by the passive nature of the
transmission network and the need to continuously balance generation and load in real
time ( Kirby and Hirst, 2002). These force readiness for the next contingency, rather than
current operating conditions, to dominate the design and operation of the grid. They also
require instantaneous actions, which imposes a dependency on automatic computing,
communication, and control actions.
Security Planning
Securing the bulk electric supply system requires preparing for contingencies. A single
contingency is almost always planned for, regardless of cost. To protect against a single
contingency, the “ N- 1 criterion” must be satisfied. It requires systems to have sufficient
reserve capacity to withstand the loss of any ( i. e., the largest) generator or transmission
line in the system. Maintaining N- 1 security requires having sufficient spinning reserves
to meet demand following the loss of generation, and sufficient supplemental reserves to
14
then restore spinning reserve margins. 5 These reserves must be located so that power
may be delivered under any possible outage condition. Systems may design for N- 2 or
N- 3 security ( i. e., multiple contingencies), but only when it is determined cost effective
to do so ( Alvarado and Oren, 2002, pp. 6- 7).
Increasingly, security planning is also taking on the role of protecting the system against
deliberate attacks. Leading this effort are federal agencies with the intent of establishing
guidelines for industry participants to follow. The Office of Energy Assurance within the
U. S. Department of Energy ( U. S. DOE) has spearheaded this effort with the development
of the Vulnerability and Risk Analysis Program. This program aims to develop and
validate vulnerability assessment methodologies in response to increased concern about
the security of the nation’s critical infrastructure. Upon its completion, the Program will
outline assessment methodologies for the electric, natural gas, and petroleum sectors.
Methods for the electricity sector exist, but are still under development for the natural gas
and petroleum sectors.
The Program uses a three- phase approach to assess the vulnerability of industry assets in
the electricity sector ( U. S. DOE, 2002). First is the pre- assessment, where the scope and
objective of the assessment are defined. It involves the collaboration of individuals from
all sectors of the company to define the concept of criticality, rank assets according the
criticality definition, and determine the consequence of disruption or loss of each asset.
Next is the assessment, which addresses ten items:
5 “ Spinning reserves are generators that can instantaneously increase their output when a decrease in
frequency signals that load is exceeding generation” ( Alvarado and Oren, 2002, p. 7).
15
1. Network architecture. Evaluate existing security plans and identify concerns
with the system architecture or operating procedures.
2. Threat environment. Characterize threats, trends in threats, and mechanisms
by which threats can exploit vulnerabilities.
3. Penetration testing. Identify vulnerabilities in information systems, and test
to determine whether access can be gained.
4. Physical security. Evaluate existing or planned physical security systems.
5. Physical asset analysis. Examine physical assets for vulnerabilities.
6. Operations security. Identify and protect information pertaining to sensitive
activities.
7. Policies and procedures. Review policies and procedures, and identify areas
for improvement.
8. Impact analysis. Determine the consequences of exploitation of critical
facilities or information systems on markets and/ or physical operations.
9. Infrastructure interdependencies. Examine the interdependencies and
vulnerabilities of infrastructures supporting critical facility functions.
10. Risk characterization. Provide a framework to prioritize investment and
implementation recommendations.
The final phase is the post- assessment, where recommendations from the assessment are
prioritized based on an evaluation of the costs and benefits of each, and an action plan is
developed. Lessons learned and best practices are captured here, as well.
16
Similarly, the NERC has proposed a four- tiered model to guard against physical and
cyber threats ( NERC, 2001). The four tiers are avoidance, assurance, detection, and
recovery. Avoidance is the most cost effective means of action. It aims to prevent the
exploitation of threats by promoting awareness and sharing information and data through
an Electricity Sector Information Sharing and Analysis Center ( ES- ISAC). Assurance
promotes reliability through the regular evaluation of physical and cyber security
measures. Detection focuses on monitoring, identifying, reporting, and analyzing
operational, physical, and cyber threats or incidents. Recovery encourages timely
investigation of incidents and rapid recovery and restoration of services.
Governance and Oversight
Governance and oversight are fundamental to the notion of security in a deregulated
electricity market, where reliability decisions have shifted from vertically- integrated
utilities to a system operator. In the past, large utilities controlled generation,
transmission, and distribution operations, and could make reliability- based decisions
relatively easily. But in the deregulated environment, assets are distributed among
several more industry players, and reliability is now under the control of an independent
system operator ( ISO). Kirby and Hirst ( 2002, p. 10) offer six questions to guide
reliability decisions in a deregulated environment:
• What risks to take?
• When to take those risks?
• How much money to spend on risk mitigation?
17
• Who pays for reliability?
• Who is exposed to any remaining risks?
• Who decides on these matters?
Managing Security
Managing security in the electricity system is mainly a real- time effort by operators to
manage transience in the system. Transmission operators have two basic ways to ensure
reliability – by deploying reserves ( Kirby and Hirst, 2002), or controlling commerce
( Alvarado and Oren, 2002). Security in the electricity sector is currently managed
primarily through the deployment of reserves. Reserves insure against the sudden loss of
a generator or transmission line, and include additional generation and transmission, or
load that is willing to curtail. Most regional reliability councils set contingency reserve
requirements equal to the largest single contingency within the region ( N- 1 criterion), and
require at least half to be spinning ( Kirby and Hirst, 2002).
Transmission operators can also ensure reliability through the control of commerce, by
redistributing generation away from the typical pattern of the free market. Generators
can indicate a price at which they are willing to increase or decrease production, creating
a market for contingency reserves. This is attractive in a deregulated environment, and
might push reliability to be increasingly managed through the control of commerce.
18
Summary
Reliability in the electricity sector encompasses two concepts – adequacy and security.
Adequacy refers to the sufficiency of system throughput to supply long- term, average
demands. Security refers to the ability of the system to withstand disruption under
dynamic conditions. Factors influencing the adequacy of the system are primary energy
resource availability, and generation and transmission capacities. Sufficiency of capacity
can be measured deterministically in terms of reserve margins, or probabilistically in
terms of expected outages.
Although security predominately involves real- time management of system operations, it
has recently taken on a long- term planning approach as well, to secure assets against
vulnerabilities. Vulnerability assessments and mitigation plans can identify threats and
vulnerable assets early, and prevent future disruptions. Another concept important to
security in the electricity sector is that of governance and oversight. Increased
competition from industry deregulation has reduced the incentive for independent
reliability assurance measures in the industry. Thus, the role of an independent authority
to assure reliability has grown significantly. This body must be independent and fair in
its directives. Two mechanisms exist to manage security in the electric grid. Most
common is the deployment of reserves. Mandatory reserve margins are set so that the
loss of any generation or transmission facility ( or sometimes set of facilities) will not
cause a disruption of service. The other mechanism is to ensure reliability through
market- based principles. One example would be the creation of a reserve market, where
reserves could be brought online or taken off, according to real- time demands.
19
Natural Gas Sector
Unlike in the literature pertaining to the electricity sector, no recurring definition of
reliability was found in the natural gas sector. Perhaps the most concise definition was
found in the Infrastructure Reliability Program of the DOE. It suggests that reliability
efforts in the natural gas sector focus on securing the physical infrastructure, and are less
concerned with the concept of adequacy ( U. S. DOE and NETL, 2002, pp. 3- 4):
Ensure Reliability – Allowing operators to prevent damage or disruption,
to detect and diagnose leaks and failures more
quickly, and to enhance the flexibility and
responsiveness of the system in response to losses in
capacity
Another important factor weighing on reliability in the natural gas sector is cost. Price
fluctuations strongly influence natural gas reliability considerations. Indeed, the Energy
Information Administration ( EIA) has said that a key challenge facing the natural gas
industry over time is “ moderating the recurrence and severity of ‘ boom and bust’ cycles
while meeting increasing demand at reasonable prices” ( EIA, 2001a, p. 20).
Natural Gas Supply
Recent trends in the natural gas industry have seen significant demand increases and
price volatility, resulting in projections of future shortages. Exacerbating bleak
projections is a cyclic behavior commonly visible with commodities, and beginning to
manifest itself with natural gas. The trend sees a cycle of surpluses and shortages, and
low and high prices. These considerations have prompted calls for reviving and
20
expanding the liquefied natural gas ( LNG) infrastructure in the U. S., which has been
essentially dead since the early 1980s.
Recent Trends
The recent price spikes can be partially attributed to the increase in the construction of
natural- gas- fired power plants and cogeneration that has significantly increased natural
gas demand. The expansion was initially obscured by abnormally warm winters in 1997-
1998 and 1998- 1999, but in the two very cold winters that followed, demand
skyrocketed. Prices spiked in the winter of 1999- 2000, and remained high through the
beginning of April 2000, the beginning of storage refill season. High prices encouraged
operators to delay injecting gas into storage, and by November, storage was at a 20- year
low. When the cold winter hit, demand soared and prices spiked. On the coldest days in
December of 2000, utilization reached 90– 100% in some areas, and prices exceeded $ 10
per million Btu at the Henry Hub ( compared to the average price for the entire year,
which was $ 2.40 per million Btu) ( EIA, 2001b).
These price fluctuations might indicate that natural gas is entering a trend of cyclic
pricing behavior. Such trends are typical in commodity markets, but until recently, have
not affected the natural gas sector. The cycles follow periods of overinvestment or
underinvestment in production, and might develop as follows. A surge in demand during
a cold spell results in a price spike due to the inelasticity of supply. Sustained high prices
encourage producers to invest in new production. Peak demands fall during subsequent
warm winters, causing a surplus of supply and prices to fall. Sustained low prices
21
discourage investments in new production. When a cold season hits, production lags
demand causing a price spike, and the process repeats ( EIA, 2001b).
Future Projections
The EIA developed a model projecting natural gas supplies in the U. S. through 2025
( EIA, 2001b). The model considers six scenarios, including cases where restrictions to
natural gas exploration in the Rocky Mountains and the Outer Continental Shelf ( OCS)
are eased, and where carbon dioxide ( CO2) emissions are limited. The reference case for
the model uses projections from the Annual Energy Outlook 2002, and assumes no policy
changes. Table 1 shows the results for the reference case and the limited CO2 emissions
cases. All models predict an increasing reliance on imports over levels today ( about 16%
in 2003), especially the limited CO2 emissions cases. 6 The model also predicts higher
prices and greater price volatility in the CO2 emissions limit cases. Similar effects as
seen in the CO2 emissions limit models might be expected with a burgeoning hydrogen
economy, as both add marginal natural gas demand. 7
The reference case is based on models the EIA uses in their Annual Energy Outlook to
generate future projections of energy markets. Their most recent projections, in the
Annual Energy Outlook 2004 ( AEO2004), extend from 2002 to 2025 ( EIA, 2004f). They
project an increase in U. S. natural gas demand from 22.8 trillion cubic feet ( tcf) in 2002
to 31.4 tcf in 2025. But domestic production is only expected to grow from 19.1 tcf in
6 Although not shown here, supply and demand both increased in the Rocky Mountain and OCS access
cases, but absolute imports were about the same as the reference case
7 Policies limiting CO2 emissions increase natural gas demands because some coal- fired power plants that
emit large amounts of CO2 would likely be replaced with natural gas- fired electricity generation.
22
2002 to 24.1 tcf in 2025. They conclude that “ growth in U. S. natural gas supplies will be
dependent on unconventional domestic production, natural gas from Alaska, and LNG”
( EIA, 2004f, p. 8).
Table 1. Natural gas supply projections through 2025 ( adapted from EIA, 2001b, pp. 22- 23).
Liquefied Natural Gas ( LNG)
LNG is projected to become a larger source of natural gas supply in the U. S. as domestic
supplies are expected to lag and the availability of Canadian imports is projected to
decline ( see Figure 2). Increasing LNG import levels carries interesting implications for
reliability in the natural gas sector. They could have a positive effect by leveling costs
and supplying demands that would otherwise be met with production from higher cost
sources ( EIA, 2001b, p. 37). With sufficient infrastructure, seasonal price spikes could be
moderated by increasing LNG imports. Similarly, during periods of low demand, LNG
imports could be curtailed to push prices up. But reliance on imported energy supplies
creates a dependence on foreign suppliers, thus detracting from reliability. Natural gas
reserves are concentrated in a few regions of the world. Ten countries control 77% of
global natural gas reserves, and the top three over 55% ( see Table 2). Conceivably, as
world natural gas demand grows and countries rely more on LNG imports, a natural gas
23
cartel could form that could control global trade with monopolistic power, similar to the
Organization of Petroleum Exporting Countries ( OPEC) ( EIA, 2001b, p. 29).
Figure 2. Net U. S. imports of natural gas, 1990- 2025 ( EIA, 2003, from AEO2004 reference case).
Table 2 lists global reserves by country and current ( darkly shaded) and potential ( lightly
shaded) exporters ( EIA, 2003, p. 5). It can be seen that current and potential export
capacity resides predominantly in countries with somewhat unstable political and/ or
social situations. This is similar to current conditions in the petroleum sector, and
introduces geopolitical threats into the reliability of natural gas supply. 8
8 Geopolitics is discussed in greater depth in the petroleum section of the literature review.
24
Table 2. Natural gas reserves by selected country. Current LNG exporters are darkly shaded,
potential LNG exporters are lightly shaded ( adapted from: EIA, 2003, p. 5).
Infrastructure Reliability
The National Petroleum Council ( NPC) addresses issues of natural gas infrastructure
security in their report, Securing Oil and Natural Gas Infrastructures in the New
Economy ( NPC, 2001). Part of the report investigates physical vulnerabilities facing the
natural gas infrastructure. Figure 3 outlines the natural gas infrastructure generally, and
25
presents the Council’s vulnerabilities ratings for some physical assets. The ratings are
based on the following scale ( NPC, 2001, p. 33):
Low – Key assets that if damaged could cause disruptions with local
impacts of short duration.
Medium – Key assets that if damaged could cause disruptions that would
have regional impacts. These disruptions would last long
enough to cause end users hardship, economic loss, and
possible loss of human life.
High – Key assets that if damaged could cause major disruptions that
would have regional and possibly national or international
impacts, and of sufficient duration to cause death and end users
major hardship and economic loss.
Figure 3. The National Petroleum Council’s assessment of physical vulnerabilities facing natural gas
infrastructure ( NPC, 2001, p. 34).
Pipelines
The DOE and the National Energy Technology Laboratory ( NETL) sponsored two
industry- based workshops focused on security concerns facing natural gas pipeline
networks. The first workshop identified security concerns and technological solutions
( SCNG, 2000). Predominant concerns included reducing the cost and incidence of
26
damage to underground pipelines, 9 and expanding and improving the flexibility of
pipeline networks. Technological solutions were posed to address these concerns, such
as developing better monitoring capabilities and integrity assessments, improving
pipeline and storage systems, developing cost- effective construction techniques, and
developing the ability to detect underground facilities and provide real- time proximity
warnings. The other workshop focused on securing the natural gas infrastructure against
malicious attacks ( U. S. DOE and NETL, 2002). The large, diffuse, and remote nature of
the infrastructure makes it quite vulnerable to attack. While much of the network is
somewhat protected underground, several portions are not. Those that are underground
can be easily located from warning markers. Few technologies exist to detect intrusions
or evaluate, inspect, and respond to pipeline problems. Automated control systems are
also vulnerable, lacking secure technologies or industry standards to direct information
and communication protocols. The group concluded that few options exist to prevent
physical attacks in the near term, but with increased coordination, effective steps can be
taken to better secure the infrastructure.
The level of utilization in the pipeline network conveys the degree to which end user
demands can be met, and the extent of consequences that might stem from a disruption
( EIA, 1998, p. 9). Utilization can be determined in a number of ways. One common
measure is average- day utilization, which is determined by dividing the average daily
throughput ( annual flow between states divided by the number of days in the year) by the
estimated capacity in the system. An obvious shortcoming in this measure is that it tells
9 More than half of all subsurface pipeline damage is caused accidentally by third parties, usually
construction crews ( SCNG, 2000, p. 5).
27
nothing of availability during peak demand periods. The use of monthly, weekly, or
daily throughput data helps circumvent this limitation. If several measures are developed
– for example, peak- day, high month, low month, average month, and average summer
( i. e., off- peak) – one can gauge variability throughout the system.
LNG
The implications of widespread LNG infrastructure are not well known. But it is thought
that the high capital costs and fuel concentrations associated with LNG infrastructure
make it an attractive target to attack. Natural disasters, especially earthquakes, are
significant threats as well. In the case of an LNG spill, a potentially very serious
situation could ensue. If LNG pools on water and is ignited, the resulting fire would burn
uncontained until all of the gas was consumed. Experimental spills of 10,000 gallons
resulted in cylindrical fires 50 feet wide and 250 feet high. This is quite intimidating
considering that an LNG tanker may carry up to 33 million gallons ( Havens, 2003).
Interdependencies
The natural gas sector is interdependent with several other infrastructures, and vulnerable
to disruptions in them. Five types of failure can occur between interdependent systems
( NPC, 2001, p. 30):
• Cascading failures – failure in one infrastructure leads to failure in another
• Escalating failures – duration of outage in one infrastructure increases due to a
failure in another
28
• Common mode failures – one incident impacts multiple infrastructures
• Marketplace failures – e- commerce links multiple infrastructures in the same
market
• Compounding failures – multiple independent incidents lead to additional failures
Figure 4 illustrates some of the many infrastructure interdependencies with natural gas.
A disruption in any of the eight other infrastructure systems shown in the ovals could
have consequences for the natural gas system described in the boxes. For example, if a
disruption occurred in the water supply system, the natural gas system would lose its
ability to control emissions, and production and cooling processes would be inhibited.
Figure 4. Natural gas sector interdependencies ( NPC, 2001, p. 29).
Summary
Unlike the electricity sector, no set definition of reliability was found in literature specific
to the natural gas sector. Nevertheless, reliability efforts throughout the sector revolve
29
around common concerns: securing sufficient supplies, securing the infrastructure
( especially pipelines), and moderating prices. The U. S. and much of the developed world
will likely grow increasingly dependent on imported LNG in the mid- term. This prospect
exposes natural gas supplies to threats and vulnerabilities on the global scale, 10 but may
also enhance reliability by mitigating prices. Another major concern for reliability in the
natural gas sector is securing widespread pipeline networks from accidental and
malicious attacks. Such a task is daunting, and its success may require technological
solutions which do not yet exist.
Petroleum Sector
Reliability concerns in the petroleum sector center around broad issues such as national
and international security and economic prosperity. The differences from the other
sectors reviewed stem from the global nature of petroleum supply. Petroleum importers
depend on global suppliers to feed their demand and maintain their economy. An
interruption in production from any major suppler has consequences that can ripple
through the global market, and have damaging effects on national and global economies.
Growing dependence in developed nations on petroleum links national security with
petroleum supply security. Dwindling petroleum reserves and lagging extraction rates in
those same countries exacerbate the problems, and lead to conflicts which can threaten
international security.
10 A more detailed discussion involving reliability concerns associated with global trade follows in the
section covering the petroleum sector.
30
In recent years, risks facing the sector have changed substantially. The transformation is
due in large part to changing business practices, brought by increasing globalization and
the influx of information technology. Traditionally, reliability efforts focused on
protecting assets from human error and natural disasters. But in this new business
environment, the focus has shifted to securing foreign supply sources and guarding
against cyber attacks. The post- September 11th atmosphere has invigorated efforts to
secure the physical infrastructure as well, but now with a focus on malicious attacks,
rather than accidents and natural disasters.
Reliability Perspectives from the Petroleum Industry
The NPC report Securing the Oil and Natural Gas Infrastructures in the New Economy
details the petroleum industry’s perspective on reliability in the petroleum sector. Its
recommendations intend to protect companies from financial loss, which somewhat
conflicts with our efforts to develop a hydrogen reliability assessment which places
society as a whole as the stakeholder. Nevertheless, the issues addressed carry over to
the end user and provide insight for our study.
The New Business Environment
The assimilation of information technologies and telecommunications in the petroleum
sector has dramatically altered the way the industry conducts business. The business
environment today is characterized by automation, rapid changes, new business models,
new business organizations, and globalization. These trends create new markets and
make business more efficient, but also compound reliability concerns. In the new
31
environment, reliability cannot be examined or planned for from a domestic slant alone.
Increasingly, reliability in the petroleum sector depends on that of the weakest link in the
global supply system. Interdependencies between the petroleum sector and other critical
infrastructures have grown more intricate as information technologies and
telecommunications take on dominant roles. The new environment has also expanded
potential consequences of incidents. Disruptions historically resulted in primarily local
consequences. But today the potential for regional, national or even global ones exists.
Compounding matters is the fact that increased automation and retirement of individuals
with the necessary skills makes a return to manual methods of business almost impossible
( NPC, 2001).
Risk Management
The NPC recommends that companies address risk proactively through routine risk
management. Typically, risks are measured in terms of likelihood of occurrence and
expected level of financial loss. The Council offers a six- step risk management process
to mitigate risks in the new business environment ( NPC, 2001, pp. 40- 47):
1. Identify and characterize key assets. Key assets include facilities, information,
people, processes, programs, and services. Each is assigned a value reflecting the
consequence of losing that asset.
2. Identify and characterize vulnerabilities and threats. Identify targets and
weaknesses, and review the ability of security measures to guard against them.
32
Usually covered are cyber systems, supervisory control and data acquisition
( SCADA) systems, physical assets, security measures, and interdependencies.
Threat assessments should consider ability to access an asset, ability to harm an
asset, intent to harm an asset, history ( including the past targeting of an asset), and
the effectiveness of existing security measures against the threat.
3. Perform risk assessments. Risk is the product of the probability of an incident
and the consequence of the incident, and can be determined by multiplying the
value of the asset ( i. e., the consequence) as determined in Step 1, with the
likelihood of an incident ( i. e., the vulnerability) as determined in Step 2. Risk can
be measured qualitatively, quantitatively, or using a mixture of both methods.
4. Identify and characterize potential risk abatement options. Risk abatement
generally focuses on deterring threats, reducing vulnerabilities, reducing
consequences, reducing severity during an incident, and ensuring rapid recovery
after the incident.
5. Select cost- effective risk abatement options. The options identified in Step 4 are
analyzed and prioritized on a cost/ benefit basis.
6. Implement risk management decisions. Attractive abatement options identified in
Step 5 are implemented. Implementation involves preparing plans and
procedures, training staff, and continuing to monitor the risk environment.
33
Risks
The new business environment has transformed the risks facing the petroleum industry.
Traditionally, primary risks in the petroleum sector were incidents resulting from human
error or natural disaster, and were mitigated by hardening assets ( NPC, 2001, pp. 2- 4).
But industry operations in the new business environment face an entirely new set of risks,
against which the industry remains unprepared. The NPC ranks seven risks facing the
industry today, in decreasing order of preparedness against them ( NPC, 2001, pp. 17- 37):
1. Information technology and telecommunications
2. Globalization
3. Business restructuring
4. Interdependencies
5. Legal and regulatory issues
6. Physical and human factors
7. Natural disasters
U. S. Petroleum Dependence and Its Economic Implications
Dependence on foreign energy sources has imposed tremendous costs on the U. S.
economy over the past 30 years. Metrics exist to gauge the level of petroleum
dependence in an economy, and its vulnerability to a supply disruption. These measures
indicate that the U. S. is more dependent on petroleum and more vulnerable to an
interruption in its supply than ever before.
34
Measures of Petroleum Dependence
Greene and Tishchishyna define U. S. petroleum dependence as “ the product of ( 1) a non-competitive
world oil market strongly influenced by the OPEC cartel, ( 2) high levels of
U. S. oil imports, ( 3) the importance of oil to the U. S. economy ( especially the
transportation sector), and ( 4) the absence of economical and readily available
substitutes” ( Greene, 2000, p. 2). It can be measured several ways. Alhajji and Williams
( 2003) gauge dependence according to four metrics, which consider imports, reserve
levels, and the percentage of total energy consumption met by petroleum.
Imports
One measure of petroleum dependence is the percentage of petroleum consumption met
by imports. Figure 5 shows the average annual U. S. petroleum consumption met by
imports. According to this metric, U. S. petroleum dependence hit a record high in 2001
when net imports averaged 57% of petroleum supplied.
U. S. Net Petroleum Imports vs. Consumption
0
5,000
10,000
15,000
20,000
25,000
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Thousand bbl/ d
0%
10%
20%
30%
40%
50%
60%
Percentage Imports
Petroleum Consumption Net Petroleum Imports Percentage Imports
Figure 5. U. S. net petroleum imports since 1970 ( EIA).
35
Number of Days Stocks Cover Imports and Total Consumption
Two additional measures suggested by Alhajji and Williams are the amount of total
petroleum reserves compared to net imports and total consumption. Figure 6 shows
average annual U. S. petroleum stock levels since 1970, and their average coverage
against imports and consumption. Stocks here include both commercial stocks and
reserves such as the Strategic Petroleum Reserve ( SPR), which was created in 1977.
Total petroleum stock coverage against imports has constantly decreased since the mid-
1980s, from a peak of 300 days in 1985 to 116 days in January of 2004. Against total
consumption, total petroleum stock coverage has also decreased, from a peak of 102 days
in 1984 to 77 days in January 2004.
U. S. Total Petroleum Stocks
0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
1,400,000
1,600,000
1,800,000
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Total Stocks
( Thousand barrels)
0
50
100
150
200
250
300
350
Stock Coverage ( Days)
Petroleum Stocks Stock Coverage vs Imports Stock Coverage vs Consumption
Figure 6. U. S. petroleum stocks and their coverage against imports and consumption ( EIA).
A minimum stock level, known as the Lower Operational Inventory Level ( LOIL), is
required to operate and maintain the system. 11 If it is included ( see Figure 7), coverage
11 The LOIL in the U. S. is currently 862 million barrels of crude oil and petroleum products.
36
levels drop compared to Figure 6. As of January 2004, coverage against imports was 52
days and coverage against consumption was 34 days when the LOIL was included.
U. S. Total Petroleum Stocks Above LOIL
0
200,000
400,000
600,000
800,000
1,000,000
1,200,000
1,400,000
1,600,000
1,800,000
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Total Stocks
( Thousand barrels)
0
50
100
150
200
250
300
350
Stock Coverage ( Days)
Petroleum Stocks Stock Coverage vs Imports Stock Coverage vs Consumption
Figure 7. U. S. petroleum stocks and their coverage against imports and consumption, minus Lower
Operational Inventory Levels ( EIA).
Percentage of Petroleum in Total Energy Consumption
The final measure of petroleum dependence according to Alhajji and Williams is the
percentage of total energy consumption met by petroleum. It indicates the importance of
petroleum to an economy. Total energy and petroleum consumption are shown in Figure
8. The percentage of total energy consumption met by petroleum is also shown. It
peaked in the late 1970s at 48% before falling to 38% in 1995. Since then, it has slowly
increased to its current level of approximately 40%.
37
U. S. Petroleum Share in Total Energy Consumption
0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
80,000
90,000
100,000
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
Year
Trillion Btu
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Petroleum Percentage
Total Energy Consumption Petroleum Consumption
Petroleum/ Total Energy Consumption
Figure 8. Percentage of total energy consumption met by petroleum in the U. S. ( EIA).
Oil as a percent of GDP
A similar measure of the importance of petroleum to an economy is the percentage of
gross domestic product ( GDP) of petroleum expenditures ( Green, 2000, p. 3). Higher
expenditures ( as a percentage of GDP) indicate a greater dependence of an economy on
petroleum. Figure 9 shows annual U. S. petroleum expenditures in nominal dollars from
1970 to 2000, and their percentage of GDP. Expenditures as a percentage of GDP
peaked in 1982 at about 5.3%, and most recently were about 4% in 2000.
38
U. S. Oil Expenditures
0
50,000
100,000
150,000
200,000
250,000
300,000
350,000
400,000
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
Year
Million 1996 Dollars
0%
1%
2%
3%
4%
5%
6%
Petroleum Percentage
Petroleum Expenditures Petroleum/ GDP
Figure 9. U. S. oil expenditures as a percent of GDP ( EIA).
Measures of Vulnerability to Supply Disruption
Similar to petroleum dependence, Alhajji and Williams define measures of vulnerability
to a supply disruption. While the previous measures related to the importance of
petroleum to an economy, the measures here reflect the likelihood that imports might be
disrupted. They are based on the global distribution of supply sources, and essentially
gauge the influence of large suppliers on the global market.
Degree of Import Concentration
Alhajji and Williams define import concentration as the percentage of imports coming
from the top five suppliers. The consequences of a disruption from a supplying country
increases with import concentration. The top five exporters of petroleum to the U. S. over
the past thirty years are shown in Table 3. Canada, Saudi Arabia, Mexico, Venezuela,
and Nigeria have generally dominated U. S. petroleum imports.
39
Table 3. Top five petroleum supplying nations into U. S. from 1973 to 2003 ( EIA).
The average annual concentration of U. S. imports from its top five supplying countries
over the last thirty years is illustrated in Figure 10. After a decline in import
concentration following the energy crisis in 1973, import concentration has been steadily
increasing since the late 1970s. Import concentration in the U. S. from its top five
suppliers peaked near 71% in 1991, and averaged about 63% in 2003.
40
U. S. Import Concentration
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
Year
Average Imports
( Thousand bbl/ d)
0%
10%
20%
30%
40%
50%
60%
70%
80%
Percentage of Total
Imports
Total Petroleum Imports Total Petroleum Imports from Top Five
Percentage from Top Five
Figure 10. Concentration of U. S. petroleum imports from its top five supplying countries ( EIA).
OPEC Share of World Petroleum Supply
The Organization of Petroleum Exporting Countries ( OPEC) is a collection of several oil
rich countries that together exert tremendous influence on global supply. As their control
of global production increases, so does the vulnerability facing each importing nation.
Figure 11 shows OPEC’s average daily crude oil production from 1970 to 2004, and its
share of global production. Its percentage of global production declined dramatically in
the late 1970s and early 1980s, from a peak of 55% in 1973 to a low of 30% in 1985.
Since then, their share has been increasing, and as of January 2004, constitutes about
41% of global production.
41
OPEC Share of World Production
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Average Production
( Thousand bbl/ d)
0%
10%
20%
30%
40%
50%
60%
Percentage of World
Production
OPEC Crude Oil Production OPEC Share of Global Production
Figure 11. OPEC share of global crude oil production ( EIA).
Persian Gulf Share of World Petroleum Supply
Social and political turmoil have afflicted several Persian Gulf nations for years, and
incidents in the region have been responsible for each energy crisis over the last 30
years. 12 Growing animosity in the region against western states compounds matters and
increases the vulnerability of a supply disruption in the region. Figure 12 shows the
average daily crude oil production in the Persian Gulf from 1970 to 2004, and its share of
global production. The trends essentially mirror those from OPEC over the same period,
but with a peak of about 38.2% in 1974 and a low of 17.8% in 1985. In 2003, Persian
Gulf supplies averaged 27.7% of global production.
12 Energy crises followed the Arab oil embargo in 1973, the Iran- Iraq war in 1979, and the Iraqi invasion of
Kuwait and subsequent war with the U. S. in 1990- 1991.
42
Persian Gulf Share of World Production
0
5,000
10,000
15,000
20,000
25,000
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
Year
Average Production
( Thousand bbl/ d)
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Percentage of World
Production
Persian Gulf Crude Oil Production Persian Gulf Share of Global Production
Figure 12. Persian Gulf share of global crude oil production ( EIA).
World Excess Production Capacity
Excess production capacity provides an element of flexibility in the global market to
withstand disruptions from individual suppliers. Essentially all spare production capacity
in the world is controlled by OPEC and Persian Gulf countries ( Kreil, 2004). Figure 13
shows the annual average world excess production capacity versus price since 1970. It
can be seen that current excess capacity is lower than any other time during that period
except the Gulf War in 1991.
43
World Excess Production Capacity
0
2
4
6
8
10
12
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Spare Production
Capacity ( Million bbl/ d)
$ 0
$ 10
$ 20
$ 30
$ 40
$ 50
$ 60
$ 70
$ 80
Average Cost per Barrel
( 2003 $)
Spare Production Capacity Crude Oil Price
Figure 13. World excess petroleum production capacity vs. price ( EIA).
Costs of Oil Dependence
Dependence on oil supplies from other countries has profound consequences on the U. S.
economy. It increases the trade deficit, the costs of securing resource supply, and slows
GDP growth. Figure 14 shows annual U. S. expenditures on imported petroleum and the
U. S. trade deficit since 1970, based on real prices in 2003 dollars. Expenditures on
imported petroleum are approaching record values not seen since the second energy
crisis, when the U. S. spent approximately $ 145 billion on net imports in 1980. In 2004,
if the price of oil averages $ 40 per barrel and net imports remain close to 11 million
barrels per day, the U. S. will spend $ 160 billion on imported oil. Since 1975, the last
year the U. S. had a trade surplus, expenditures on net imports of petroleum have
consistently accounted for over 20% of the total trade deficit. Over the last decade,
increases in spending on imported oil have corresponded well with increases in the trade
deficit. The connection is especially apparent since 1997. In 2003, with spending on
44
imported oil supplies amounting to $ 128 billion and the trade deficit at $ 490 billion,
dependence on imported oil accounted for over 25% of the total trade deficit.
U. S. Expenditures on Imported Oil vs. Trade Deficit
$ 0
$ 20
$ 40
$ 60
$ 80
$ 100
$ 120
$ 140
$ 160
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Year
Import Expenditures
( 2003 $, Billions)
($ 100)
$ 0
$ 100
$ 200
$ 300
$ 400
$ 500
$ 600
Trade Deficit
( 2003 $, Billions)
Expenditures on Net Petroleum Imports U. S. Trade Deficit
Figure 14. U. S. expenditures on imported oil and the trade deficit, in 2003 $ ( EIA and the Bureau of
Economic Analysis).
In addition to compounding the trade deficit, oil dependence increases the burden of
securing supply. The average annual peacetime cost to the U. S. of maintaining a military
presence in the Middle East is about $ 50 billion ( e. g., IAGS [ 2003a], Delucchi and
Murphy [ 1996]). Military conflicts add additional costs. The cost of the 1990- 1991 Gulf
War to the international community totaled about $ 80 billion ( IAGS, 2003b). Final cost
figures for current operations in Iraq will be in the hundreds of billions. 13 Another cost
associated with international suppliers is insurance. Increased fear of attack on
supertankers has caused insurance rates to skyrocket. Insurance rates recently tripled for
13 The author does not intend to suggest motives for the current operations in Iraq, nor necessarily attribute
their financial costs to securing oil supplies. But they certainly carry implications for the global oil market.
45
tankers passing through Yemen, adding about $ 0.15/ barrel ( bbl) to the price of petroleum
traveling through the region ( IAGS, 2003c).
The EIA has established “ rules of thumb” to assess the impacts of oil supply disruptions
on economic growth, specifically GDP. First, every 1 MMbbl/ day of lost oil causes
world oil prices to increase by $ 3-$ 5 per barrel. Second, each 10% increase in the price
of oil lowers the real U. S. GDP growth rate by 0.05 percentage points in the first year and
0.10 percentage points in the second year. So, if 1 MMbbl/ day were disrupted and
prevailing oil prices were $ 30 per barrel, oil prices could increase to $ 33-$ 35 per barrel.
This is equivalent to a price increase of 10%- 17%, which equates to possible reduction in
the U. S. GDP growth rate of 0.05- 0.08 percentage points in the first year, and 0.10- 0.17
percentage points in the second year ( EIA, 2004g).
Multiple studies have aggregated these and other costs to estimate the true cost of U. S. oil
dependence. Greene and Tishchishyna present a model developed by Oak Ridge
National Laboratories to estimate the costs of oil dependence to the U. S. from 1970 to
1999 ( Greene, 2000). They consider three categories of cost in their study: ( 1) loss of
potential GDP, ( 2) macroeconomic adjustment losses, and ( 3) wealth transfer. The loss
of potential GDP results from monopolistic pricing practices by global oil suppliers, who
keep oil prices above the level which would exist in a competitive market. Higher oil
prices constrain the economy, allowing less production with the same amount of capital,
labor, and materials than if oil was less expensive. Macroeconomic adjustment costs
account for delays in adjusting prices, wages, and interest rates following oil price spikes,
46
during which there is a less than optimal use of available resources. They depend on
policy responses to price shocks, and are sensitive to the elasticity of GDP with respect to
the price of oil. Wealth transfer is equal to the quantity of imported oil times the
difference in the actual and competitive prices. Combining these costs, Greene and
Tishchishyna conclude that oil dependence cost the U. S. $ 3.4 trillion from 1970 to 1999.
The National Defense Council Foundation ( NDCF) also studied the economic impacts of
oil dependence, and presents the costs on a per- gallon of gasoline basis to determine the
“ real price” of gasoline ( Copulos, 2003). The study includes three hidden imported oil
costs: ( 1) military expenditures in the Persian Gulf, ( 2) a diversion of financial resources,
and ( 3) periodic oil price shocks. Military expenditures are defined in terms of the
portion of the budget of U. S. Central Command ( whose area of responsibility is the
Middle East and the Horn of Africa) that goes towards defending Persian Gulf oil. It
does not include the cost of the current engagement in Middle East. The diversion of
financial resources includes direct costs from the transfer of wealth, and indirect costs
from lost employment and investment. The costs stemming from the oil price shocks of
1973- 74, 1979- 80, and 1991 were estimated to be $ 2.3 trillion – $ 2.5 trillion, and
amortized over three decades to determine an annual cost. They conclude that the real
price of gasoline paid by the U. S. consumer, when taking oil dependence into account, is
between $ 5.01/ gallon and $ 5.19/ gallon.
47
Reliability of Global Supply Infrastructure
The oil supply chain is composed of a vast infrastructure of interdependent physical
assets that stretches worldwide. Supply resources tend to be centralized in tumultuous
regions far from the final demand, creating a long and complicated transportation
network of ships, trains, trucks, and pipelines. Geopolitics influence oil extraction rates,
transportation routes traverse dangerous terrain and hostile territory, refineries are aging
and are not being replaced, and global oil consumption is expected to increase by 50%
over the next twenty years ( EIA, 2004f, p. 2). Every asset throughout the infrastructure
faces unpredictable threats presented by the new business environment, natural disasters,
human error, and hostile attacks. This section investigates the reliability of the physical
petroleum supply infrastructure, and discusses its vulnerabilities and threats.
Supply Outlook
As world consumption continues to grow and reserves deplete, global distribution of
petroleum resources should grow more concentrated. Members of OPEC stand to gain an
even greater share of the world market, and nations dependent on imported oil will grow
increasingly vulnerable to a disruption in supply. Figure 15 shows the estimated
distribution of oil reserves as of January 1, 2001. Over half of the remaining oil in the
world is located in the Middle East.
48
World Crude Oil Reserves
0
200
400
600
800
1000
1200
1400
World
OPEC
Persian Gulf
No rth America
Ce ntral & South America
We stern Euro pe
E. Europ e & Former U. S. S. R.
Middle East
Afr ica
Asia & Ocean ia
Reserves ( Billion barrels)
Figure 15. Distribution of global crude oil reserves ( EIA, from Oil & Gas Journal).
Geopolitics
The Oxford American Dictionary defines geopolitics as “ the politics of a country as
determined by its geographical features.” Here, the geographical feature of concern is the
abundance – or lack thereof – of oil. Geopolitics weighs heavily on international energy
markets, and will impose increasing threats on global oil supply as reserves grow more
concentrated and demand continues to increase.
The Center for Strategic and International Studies ( CSIS) investigated the “ symbiotic
relationship” between oil and politics from 2000 to 2020 ( CSIS, 2000). Four geopolitical
trends could have significant impacts on global energy demand and supply reliability
before 2020 ( CSIS, 2000, pp. 7- 13):
• World powers and conflict. The wake of the Cold War has left the role of the world’s
major powers still somewhat undefined, and as they each pursue their national
49
interests, conflicts could disrupt world energy supplies. The politics of global and
regional powers will shape oil production from the Caspian Sea and Central Asia.
• Political instability among key energy suppliers. Several key oil producing states
face internal conflict, which could disrupt global oil supplies.
• Economic globalization. The globalization of all forms of trade is increasingly
making producers and consumers interdependent.
• The growing impact of non- state actors. Information technology has allowed non-governmental
organizations to gain greater control in the political process.
Similarly, trends in energy usage effects geopolitics ( CSIS, 2000, pp. 13- 18):
• Swings in energy demand. The economies of oil producing states are heavily
dependent on oil revenue. A drop in revenues could cripple these countries and make
them more vulnerable to internal crises.
• Competition for energy supplies in Asia. Competition for oil imports and territorial
disputes over regions rich in oil could ignite tensions between Asian countries that
have deep, historical roots. China’s increased oil dependence could lead to strategic
relationships with Middle Eastern countries and Russia, which could be damaging to
relations with the U. S., Europe, and other Asian countries.
• Energy and regional integration. Energy can also serve to strengthen ties between
rival countries. Infrastructure projects and trade liberalization can cut through
boundaries and bring economies together, serving to ease conflicts in many regions.
50
• Energy and the environment. Debates regarding the role of the environment in
energy supply and consumption could create conflicts between nations, especially
between developed and developing countries.
A brief evaluation of the geopolitical situations in each OPEC member state is given in
Appendix A. Similar looks into the socio- political situations in other significant oil-producing
and - consuming states could provide further insights into the future reliability
of global petroleum supply.
Threats
Changes in the global business and political climates intensify threats facing oil supply
infrastructure. The new business environment has exposed the industry to great threats,
as discussed earlier. Natural disasters and human error also continue to threaten
operations. An increasing source of threats is from malicious attacks, whether from
disgruntled employees, thieves, or ideologues. Oil infrastructure provides an attractive
target because it is so vital to global economies, and the infrastructure is dispersed and
generally unprotected. One source of increasing attacks is “ oil terrorism.” Most are
kidnappings, but attacks on personnel, pipelines, rigs, and wells are also included
( Adams, 2003, pp. 5- 12). Acts of piracy are also increasing, and have tripled in the last
decade ( Luft and Korin, 2003). According to the International Maritime Bureau ( IMB),
445 attacks were reported in 2003. Pirates have become better organized, and
coordinated attacks involving several boats are on the rise ( ICC, 2004). Strategic
51
shipping passages, especially the Strait of Malacca, 14 experience frequent piracy which
threatens oil tankers traversing their waters.
Infrastructure Risks
Oil infrastructure is vast and difficult to harden, creating vulnerabilities throughout the
supply chain. The extent of the U. S. infrastructure is described in Table 4, and its
vulnerabilities are classified in Figure 16 ( NPC, 2001, pp. 32- 33). Compounding supply
vulnerability are global interdependencies and trans- oceanic supply lines.
Table 4. Physical U. S. oil infrastructure components ( NPC, 2001, p. 32).
Production 602,200 wells
Gathering
74,000 miles of crude pipeline
30,000 miles of gathering pipeline
74,000 miles of product pipeline
Processing 161 petroleum refineries
Transmission 74,000 miles of crude pipelines
74,000 miles of product pipelines
Storage 2,000 petroleum terminals
Distribution
Modes
616.5 billion ton miles via pipeline
295.6 billion ton miles via water
27.2 billion ton miles via road
16.7 billion ton miles via railroads
14 See the discussion regarding international chokepoints below and in the Appendix.
52
Figure 16. The National Petroleum Council’s assessment of physical vulnerabilities facing oil
infrastructure ( NPC, 2001, p. 33).
Reservoirs
A direct attack on a reservoir would be highly unlikely and difficult to carry out, but a
successful attack on a reservoir could devastate the producer state, and severely reduce
global production ( Adams, 2003, p. 102).
Wells
Adams ( 2003) estimates that onshore wells are the most vulnerable component of the
supply system. Wells can be highly pressurized, posing a continuous fire risk. If ignited,
well fires create pollution and toxicity problems. Most wells are remotely located,
minimizing the consequence of an incident beyond lost production. But this also makes
them difficult and impractical to secure.
Offshore wells often provide attractive targets for attack, as they tend to be expensive and
have high output flow rates. They have been attacked on numerous occasions, especially
53
in Africa. Higher- producing wells far offshore are more hardened and less attractive for
attack than the softer targets offered by the often unstaffed wells closer to shore. Besides
lost production, the primary consequence of an offshore attack is pollution. Some wells
are equipped to continuously ignite any released product to avoid water pollution. But
burning oil presents toxicity and air pollution problems ( Adams, 2003, pp. 125- 127).
Transport
According to the IAGS, the “ transportation system has always been the Achilles heel of
the oil industry,” and it has become even more so in recent years ( IAGS, 2003c). Long
haul distances typical of the petroleum supply system increase vulnerabilities to every
hazard. Three- fifths of internationally- traded oil is transported by sea, and the rest
primarily via pipeline ( EIA, 2002). Both methods face considerable vulnerabilities and
threats, and pose serious consequences. But, unlike other components of the supply
system, the transport system is somewhat flexible. Trucking capacity can easily be
expanded, and provides the most flexibility, followed by rail and waterway, and finally
pipelines ( Lovins, 1982, p. 40).
• Pipelines. Pipelines tend to be unsecured in remote areas and are incredibly
vulnerable. They are often buried, but are exposed at junctures and where terrain
dictates. Signage calls out the location of buried lines to warn against inadvertent
third- party damage, but similarly alerts wrongdoers. Oil pipelines often follow
the same paths as natural gas pipelines, so an incident on one line could damage
the other as well ( Adams, 2003, pp. 106- 114). One especially vulnerable pipeline
54
in the U. S. is the Trans- Alaska Pipeline System ( TAPS), which is currently the
only route to deliver Alaskan oil to the contiguous U. S. TAPS has been bombed
twice and shot more than 50 times in recent years, and cannot be repaired in the
winter ( Luft and Korin, 2003).
Pump stations along pipelines are similarly vulnerable. They are located
approximately every 50 miles, and are often remote and unsecured. The loss of a
pump station would have the same effect as losing the pipeline it serves, but pump
stations take longer to repair ( Adams, 2003, pp. 15- 16).
• Tankers and ports. Tankers are vulnerable to attack and are facing greater and
more frequent threats. They serve as large, expensive and symbolic targets, and
often travel through dangerous waters. Loading terminals are critical to supply,
and vulnerable to interruption. They are difficult to secure, and if damaged,
would disrupt infrastructure facilities served by the port. Loading terminals may
pose a greater risk than refineries or storage sites ( Adams, 2003, p. 124).
• International chokepoints. Chokepoints are vulnerable transportation routes
through which the flow of oil could be easily disrupted. Most only have long,
inaccessible alternate routes, if any at all. If flow through any chokepoint were
disrupted, it could carry significant consequences for the global market. About
40% of total world petroleum consumption and more than 55% of all exports flow
55
through these chokepoints daily. Descriptions of each chokepoint, and threats and
consequences facing each, are given in Appendix B.
Storage
Storage facilities can include tank farms or underground storage. Tank farms are more
vulnerable and tend to be located in oil fields, refineries, loading terminals, or even
residential areas. They are visible, and their contents highly flammable. If ignited, toxic
fumes pose health risks to proximate populations. Underground storage sites have larger
capacities, but better security ( Adams, 2003).
Refineries
Refineries are probably the most vulnerable component of the supply system aside from
wells. Major damage can be done without many explosives, as refineries contain hot,
pressurized, and explosive gases and liquids. They also depend on one type of crude, and
are vulnerable to impurities ( Lovins, 1982). Refineries in the U. S are aging, and are no
longer being built due to environmental constraints and financial risks ( NPC, 2001, p. 32).
Refineries employ a large number of workers ( usually 1000- 2000 people on average) and
tend to be less remote than wells. Consequences stemming from an incident may be
more likely to reach populated areas, and include significant direct financial costs
associated with rebuilding, a high loss of life potential, and possible costs associated with
lawsuits if incident damages reach surrounding communities ( Adams, 2003, p. 27).
56
Summary
Reliability in the petroleum sector is considered in terms of broad concerns such as
national and economic security. This designation emanates from the dependence of
developed economies on imported petroleum supplies, which often originate in volatile
regions. Reliability in the sector is measured in terms of imports, origin of imports,
storage levels, and reserve levels. Economic indicators exist as well, such as petroleum
expenditures as a fraction of GDP, wealth transfer, military expenditures, and the effects
of oil price spikes. The sector faces quickly- evolving risks as a result of automation and
globalization, and the supply infrastructure is incredibly vulnerable – due to age, location,
size, and long haul distances typical of global trade.
METHODOLOGY
Methodology Overview
This study aims to develop a methodology to assess the reliability of hydrogen energy
systems. The intention is to promote fair consideration of reliability in hydrogen
discourse by introducing methods allowing complete, ordered assessments. To the best
knowledge of the author, it represents the first systematic effort in this regard.
This study uses qualitative methods to assess the perceived reliability of hydrogen energy
systems. First, reliability is defined and metrics are selected to value it. Next, hydrogen
pathways are selected and described. Three constituent components of the pathways are
assessed by a panel of experts – the primary energy supply system, the hydrogen
57
production process, and the hydrogen transport process. They rate the reliability and
importance of each pathway component in terms of the metrics. Finally, their ratings are
aggregated to determine broad reliability scores that can be compared across pathways.
The methodology is summarized by the following steps, each detailed separately below:
1. Define scope of study, and select participants
2. Define reliability in hydrogen energy systems
3. Select metrics to value reliability in hydrogen energy systems
4. Specify hydrogen energy systems to evaluate
5. Develop evaluation matrix
6. Develop rating scales and rating criteria
7. Collect expert reliability and importance ratings
8. Aggregate expert ratings to determine reliability scores
9. Compare reliability scores across pathways
The discussion in this section introduces the method and generally describes its
application. The next section details the methodology for a specific application.
1. Define Scope of Study and Select Participants
The first step of an evaluation of a system is to define the scope of study. The scope will
depend on details of the system being considered, the objectives of the organization
conducting the study, and the motivation for the research. Some parameters of the energy
systems being evaluated will be known or postulated. These include geographical extent,
58
volume of hydrogen demand, geographical- or time- distribution of demand, and others.
The composition and reach of the systems as described by these parameters shape the
boundaries and processes of the assessment. The objectives of the organization and its
motivation for conducting the study will also influence the scope. The organization could
be a company, a governmental organization, an industry group, a non- governmental
organization ( NGO), a research institution, or a university. Each holds a different slant
and motivation, and would define the scope uniquely.
The organization conducting the study also selects experts to evaluate reliability, and
determines their involvement in the assessment process. The organization may select to
use in- house experts, involve a wide group of experts comprising all stakeholders and
schools of thought, or a combination of the two. If a panel of experts representing
multiple parties is used, there are three roles it could take ( Contadini, 2002, p. 62). First, a
single modeler could decide on the inputs for the analysis, and involve other parties later
in the process. The modeler could define reliability and select the metrics and pathways
to consider, and the expert panel could rate reliability. This method allows the
organization to shape the study to its liking. But Contadini warns that this practice can
lead to missed information, and to large modifications late in the process.
The other two roles Contadini describes involve the experts in the entire process. In
addition to rating the reliability of the metrics, the expert panel also defines reliability and
selects the metrics and pathways to be evaluated. These options add a greater level of
consensus, but also introduce complications and could allow an overrepresented group to
59
bias the results. They could also reduce the ability of the organization conducting the
study to define reliability in line with its objectives. The two vary by the method in
which consensus is reached. In one, selections are made by majority vote. In the other,
final decisions are established via technical discussion based on information provided by
the organizations with which the experts are affiliated.
2. Define Reliability in Hydrogen Energy Systems
The participants selected to develop the inputs for the analysis begin by defining
reliability in hydrogen energy systems. A thorough definition is essential to set a
foundation for the assessment. It establishes boundaries and outlines key parameters to
include in the study. The definition could vary among organizations. Each is likely to
perceive reliability differently, to encapsulate concepts it feels are important.
Important issues of semantics emerge when defining reliability. Leemis discusses these
as they apply to defining reliability of any system, not specific to hydrogen ( Leemis,
1995, pp. 2- 4). He emphasizes the importance of clearly specifying within the definition
the item of interest, what constitutes adequate performance ( or non- failure), a time
duration, and the environmental conditions in which the item operates. The item can be a
component or an entire system. It should be clearly specified exactly what the item is,
and the boundaries that delineate components comprising the item. Adequate
performance must be clearly defined for the item as well. The simplest way is to
establish a binary criterion, that the item is either operational or has failed. An example
of a binary criterion in a hydrogen transport subsystem might be that a pipeline is either
60
able or unable to deliver hydrogen. But this model can be difficult to apply, because
performance of an item often degrades over time. In these cases, Leemis suggests setting
a threshold below which the item is considered to have failed. Here, the example above
might be modified to include a level of throughput under which the subsystem is
considered “ failed”. A time period should also be clearly specified in the definition. Any
item has a finite lifespan after which it will invariably fail, so adequate performance
cannot be defined without providing a context of time. Finally, the environmental
conditions under which the item is expected to operate profoundly affect the reliability of
an item, and must be specified. Two identical items operating under different
surrounding conditions will undoubtedly fail at different times. For example, a garaged
pickup truck used as a commuter vehicle will probably demonstrate greater reliability
than the same truck kept outside and used on a farm or construction site.
3. Select Metrics to Value Reliability in Hydrogen Energy Systems
Once hydrogen reliability has been thoroughly defined, metrics to value it are selected.
They are what the experts ultimately rate for each system. The idea is to decompose the
broad reliability concepts captured in the definition into tangible elements that can be
easily evaluated. Upon measuring and rating these basic elements, they are recombined
to develop overall reliability scores. The number of metrics selected and their precision
depends on the level of specificity included in the definition, the objectives of the study,
and the resources and time available. Limiting the number of metrics reduces the burden
on the experts significantly, but can also limit the scope of the assessment. Conversely,
including superfluous elements could skew the results. Conflicting issues should be
61
balanced to develop measures which fully encompass the concepts in the reliability
definition, while accounting for real- world constraints such as time, resources, and
human cognitive ability.
Several methods can be used to select the metrics. A somewhat systematic one is
outlined in the field of hazard analysis. Hazard analysis is a qualitative method used in
risk analyses to identify components deserving detailed review. It often takes the form of
a checklist evaluation completed by industry experts. Andrews and Moss define hazard
analysis as a process used for “ identifying events which lead to materialization of a
hazard, analysis of mechanisms by which these events occur, and estimation of the
likelihood and extent of harmful effects” ( Andrews and Moss, 2002, pp. 59- 60). It
provides a formulaic method to prioritize metrics to include in the assessment given
limited time. Metrics can be selected that best capture events and mechanisms deemed
most likely to produce harmful effects. Less formal methods can be used as well. These
include literature reviews, interviews with experts, and group discussions.
4. Specify Hydrogen Energy Systems to Evaluate
The metrics developed in the previous step are used to assess the reliability of hydrogen
pathways. The pathways should be detailed to the extent possible to allow accurate and
consistent reliability ratings. Descriptions should include demand scenarios, primary
energy supply systems, hydrogen production processes, and hydrogen transport
processes. End use – including energy use associated with compression or liquefaction,
required purity and pressure, and risks at the refueling station – also affects reliability, but
62
is beyond the scope of this study. This analysis only considers hydrogen reliability
upstream from the consumer.
An important aspect of reliability is the demand scenario under which the hydrogen
systems operate. It should be defined over the entire time frame established in the
reliability definition. If the pathways are expected to operate under different demand
scenarios, each needs to be clearly specified. Items to consider when defining the
demand scenario include:
• Total volume demanded
• Demand profiles ( variation of demand with time and season)
• Geographical distribution of demand
• Geographical distribution of supply sources and systems
• End use ( not considered here)
The primary energy supply system must also be clearly defined. Hydrogen is similar to
electricity and gasoline in that it does not exist by itself, and must be created from
another energy resource. The primary energy supply system encompasses the entire
system used to deliver an energy product to the point of hydrogen production. It includes
the primary energy feedstocks, their extraction and transport processes, and the
production, transportation, and/ or refining of the final energy product. Primary energy
feedstocks include any naturally occurring fossil or renewable energy resource. If
electricity is used as the primary energy supply system, it also has a primary energy
63
supply system which must be defined in this step. That is, the feedstocks used to create
the electricity ( and the systems used to extract, transport, and produce those feedstocks)
should be specified along with the systems used to generate and transport it to the
hydrogen production facility.
Similar considerations apply for defining the hydrogen production and transport
processes. The technologies used, the size and geographical extent of the processes, and
other details should be specified. Greater detail allows more accuracy and consistency in
the ratings.
5. Develop Evaluation Matrix
The metrics selected in step 3 can be related to the pathways defined in step 4 in a matrix.
The matrix displays the ratings for each metric for each component of each pathway. The
structure of the matrix is depicted in Figure 17.
Figure 17. Structure of hydrogen reliability evaluation matrix.
Associated with each metric is an importance rating. It allows the expert to evaluate the
degree to which he or she perceives the metric to contribute to the reliable operation of
the system. These ratings are used to weight the reliability ratings during aggregation.
The idea is similar to the use of saliency weights in consumer behavior research ( Day,
64
1973, p. 310). They weight consumer beliefs about a product and represent the degree to
which the item being rated relates to another item or concept, such as preference for the
product ( Fishbein, 1967, p. 489). The importance ratings should be independent of the
reliability rating for each element of the matrix. One way to think of the difference
between the two ratings is to consider the reliability rating as the likelihood that the
element will perform with a certain level of reliability, and the importance rating as the
consequence that unreliable performance of that element would have on the system.
The importance metrics should be the same across pathways, but can vary between
components. That is, Metric 1 can be given an importance rating of a for the primary
energy system, an importance rating of b for the hydrogen production process, and an
importance rating of c for the hydrogen transport process. But across pathways, the same
a, b, c ratings apply ( see Figure 18a). Varying the importance ratings across pathway
components adds detail to the assessment and conveys the notion that the importance of a
metric depends on the component of the system being considered. But it also increases
the burden on the experts, and is sometimes difficult to distinguish the importance of a
metric among pathway components. These drawbacks were made apparent in the trial
application of the methodology, discussed in later sections. The alternative is to rate the
importance of the metric only once, to the entire pathway ( see Figure 18b). The selection
of the technique depends on the level of information desired from the experts and the
time available for the study.
65
Figure 18. Sample importance ratings: a) different importance ratings for each pathway
component, b) same importance ratings for each pathway component.
6. Develop Rating Scales and Rating Criteria
After forming the evaluation matrix, rating scales and criteria to evaluate its elements are
developed. Rating scales for both the reliability ratings and importance ratings should be
specified, though they can be the same. If more are desired, such as different scales for
different metrics, then more can be incorporated into the evaluation. While it adds
complexity and may make the evaluation more confusing for the experts, various scales
could be beneficial in some cases, such as when some metrics can be evaluated
quantitatively, and others qualitatively.
The scale used should accurately capture the degree to which the system operates reliably
according to the definition established in step 1. Several scales exist to capture different
types of measurements. The primary difference between scales is the level of
information that can be inferred from the rating. Behavioral researchers identify four
scales conveying increasing levels of information ( e. g., Summers, 1970, p. 11). Nominal
66
measurements are the simplest. They are categorical and simply distinguish between
responses. They are not appropriate for this study, and are not considered here. Ordinal
measures are the next most powerful and simply convey a ranking of elements. That is, a
1 comes before a 2, comes before a 3, and so on. Interval measures include an extra
degree of information – the interval between numerical ratings is meaningful. That is, the
difference between a 2 and a 3 is the same as the difference between a 3 and a 4. The
last, and most powerful, is the ratio measure. This scale includes an absolute origin, so
all mathematical operations, including multiplication and division, can be performed on
the ratings. That is, a rating of 2 implies twice as much as a rating of 1. The literature
covers the advantages, disadvantages, and semantics of each scale in depth. Here, it
suffices to say that care should be taken when developing a rating scale, to properly
capture the desired information contained in the expert opinions.
Criteria for rating the elements must also be clearly specified. This allows for consistent
ratings and reduces the subjectivity of expert opinion. The criteria may be qualitative,
quantitative, or a mixture of both. The selection of the criteria depends on the level of
knowledge among the experts and the quantity and quality of data available regarding the
metric. Quantitative criteria are often desirable to remove ambiguities that may emerge
in subjective ratings. But for somewhat abstract metrics or for those on which little data
exists, qualitative criteria may be needed. The type of criterion selected does not
necessarily depend on the type of rating scale selected. For example, although a
qualitative rating scale of good, fair, and poor might be applied to a metric weather,
supporting criteria could be quantitative. Good might correspond to a mid- day
67
temperature above 85° F, fair to temperatures between 60° F and 85° F, and poor to those
below 60° F.
7. Collect Expert Reliability and Importance Ratings
With all inputs and procedures defined and selected, the method proceeds to the experts.
They rate the reliability of each metric as it pertains to the components of each pathway,
and the importance of each. Their ratings are based on the scales previously established.
If the experts have not been involved in the process until this point, the method and their
task should be clearly described to them. This includes clearly defining the metrics,
pathways, scales, and criteria involved in the assessment. If multiple experts are
involved, the methodology should be similarly described to each.
The shape of future hydrogen energy systems remains unknown and little data exists
publicly on their reliability. Thus, expert opinions rely heavily on subjective assumptions
about future systems, taking the form of cognitive beliefs. Specific definitions of
cognitive belief vary in the literature, 15 but here it is defined to encompass what an expert
thinks, knows, or believes about each metric.
Cognitive beliefs can be ascertained through the use of attitudinal surveys. Attitudinal
surveys gauge feelings, intentions, and opinions towards concepts, objects, or persons
( Mokhtarian, 2003). The process by which the survey is administered is up to the
organization, and depends on the scope of the study, the desired results, and the time and
resources available. The organization may want to bring the experts together to
15 Some examples can be found in Sudman and Bradburn ( 1982, p. 123) and Dillman ( 1978, pp. 80- 86).
68
encourage discussion and consensus, or have the experts conduct the evaluations
separately if anonymity is desired. Formal surveys, informal surveys, group discussion,
facilitated exercises, or personal interviews can all be used, each suited for different
situations.
8. Aggregate Expert Ratings to Determine Reliability Scores
After expert ratings are collected, they are statistically aggregated to develop broad
scores for the reliability of each pathway. Specific ratings – of which there could be
hundreds or thousands from each expert – are combined to generate general scores
applicable to the original definition that can be easily compared across pathways.
The method used to aggregate the scores depends on the scope and intention of the study
and the definition of reliability. Two possible techniques are described here, though any
number of others could be substantiated as well. One is to take a weighted average of
each expert’s responses. The idea is to capture the importance- weighted average
perception of each respondent, using the following formula:
Importance- weighted average perception
( )
Σ
Σ ×
=
=
=
n
i
i
n
i
i i
I
R I
1
1 ,
where: Ri= Reliability rating of metric i,
I i
= Importance rating of metric i,
n = Number of metrics included in the aggregation.
69
The other method is to establish a “ utility” function to capture each expert’s overall
evaluation of reliability. Day discusses this method in terms of consumer attitudes and
purchasing behavior ( Day, 1973, p. 312). He defines consumer attitudes toward an object
as the product of a belief score multiplied by an importance rating. The belief score
represents the degree to which the consumer feels that the object possesses a specific
quality. The importance rating is the degree to which the consumer feels that the specific
quality is important to an overall purchasing decision. These products are summed across
the several attributes important to the object. The nomenclature of his model can be
adapted to apply to expert opinions on reliability:
= Σ( × )
=
n
i
i i Utility R I
1
.
The additive model proposed by Day is conceptually elegant, but poses problems when
comparing pathways in which not all metrics apply. If some metrics apply to one
pathway but not another, then the first pathway is bound to receive a greater score than
the next pathway. If a high score corresponds to poor reliability, the argument could be
made that this does not pose a significant problem. One could contend that because not
all of the metrics apply, there are fewer opportunities for a loss of reliability and such a
pathway deserves a lower score. This claim could be true in many cases. But to argue
that the utility model properly captures the degree to which reliability improves relies on
the dangerous assumption that the metrics encompass reliability perfectly. In cases
where a low score corresponds to poor reliability, then the additive model makes little
70
sense. The pathway with fewer applicable metrics would likely appear less reliable than
a pathway where more metrics apply.
This problem arose between the pathways assessed in the next section. Many of the
metrics were thought to apply to one pathway but not the other. To alleviate this
problem, and put the utility model on a similar scale as the importance- weighted average
perception model for comparison purposes, the utility model can be scaled by the number
of metrics and the maximum reliability rating:
Scaled utility
( )
m n
R I n
i
i i
×
Σ ×
= = 1 ,
where: m = Maximum reliability rating.
The difference between the models is subtle, but noteworthy. Let us assume that a scale
of 1- 5 is used for both the reliability and importance ratings, where 5 corresponds to high
importance and low reliability, and 1 corresponds to low importance and high reliability.
Comparatively, both models show identical differences among pathway options. The
percentage difference between reliability scores for different pathways is the same under
both models. Also, the percentage of the maximum possible reliability score allowed by
each model is the same. But the maximum possible aggregated score differs between the
two models. Under the importance- weighted average perception model, the maximum
score is 5, but maximum score for the scaled utility model depends on the importance
ratings. It is equal to the score obtained for a given set of importance ratings if all of the
reliability ratings are 5. That is:
71
Maximum possible aggregated score ( scaled utility)
( )
m n
I n
i
i
×
Σ ×
= = 1
5
.
The difference appears on an absolute scale, where the scores using the scaled utility
method will always be lower ( unless every metric received an importance rating of 5).
The similarities and differences between the two scales are depicted in Table 6. Using
the reliability and importance ratings listed in Table 5, reliability scores are aggregated in
Table 6 using both techniques. It can be seen that the maximum score possible using the
scaled utility model is only 2.8, but in both methods Pathway # 2 scores 1.79 times higher
than Pathway # 1. The scores obtained using the scaled utility model are lower than those
using the importance- weighted average perception model, but both aggregation
techniques yield scores that are 47% of the maximum possible score for Pathway # 1, and
76% of the maximum possible in Pathway # 2. Figure 19 illustrates the similarities
between the methods if both are plotted in terms of their maximum possible score.
Table 5. Reliability and importance ratings for two hypothetical pathways.
72
Table 6. Reliability scores for two hypothetical hydrogen pathways using two aggregation methods.
Figure 19. Comparison of reliability scores for two hypothetical hydrogen pathways using the two
aggregation methods.
The difference between the techniques stems from the fact that metrics of low importance
serve to improve the reliability score under the scaled utility model, but in the
importance- weighted average perception model, they are scaled down and influence
reliability to a lesser extent. In the scaled utility model, the reliability of a component is
determined equally by its reliability rating and its importance to the overall system. That
is, a component with an importance rating of 1 and a reliability rating of 5 contributes the
same to reliability as a component with an importance rating of 5 and a reliability rating
of 1. The importance- weighted average perception model determines component
reliability only by its reliability ratings. Under this model, importance ratings serve to
73
weight the reliability ratings in terms of their effect on reliability of the system. The
reliability score for the pathway can only be improved by improving the reliability rating
of the component.
The differences in the models may be negligible if the assessment looks only to compare
pathway options, since both produce the same percentage difference between pathways.
But if the reliability scores are to be put on an absolute scale, the differences are no
longer negligible. Careful consideration should be taken when selecting the aggregation
method, to assure the results are portrayed accurately.
9. Compare Reliability Scores across Pathways
Finally, the aggregated reliability scores are compared across pathways to determine
reliable or unreliable aspects. This can be done graphically, numerically, or statistically.
APPLYING THE METHODOLOGY
The methodology was tested using a group of hydrogen researchers from the Institute of
Transportation Studies at the University of California, Davis ( ITS- Davis) as the expert
panel. The primary objective was to refine the methodology and identify opportunities
for improvement.
The scope of the assessment and the participation of the panel were limited by time and
logistical constraints. First, only three hours were allotted for the study. In practice,
74
vulnerability or risk assessments involving an expert panel often last multiple days at
workshops. 16 Due to time limitations, the definition of hydrogen reliability, the metrics
to value it, and the specification of pathways were established prior to meeting with the
panel. The role of the expert panel was to rate the reliability metrics and provide
feedback on the method. Second, although ITS- Davis arguably boasts one of the largest
and most diverse groups of hydrogen infrastructure researchers in the world, many are
not completely familiar with reliability. An ideal panel would include reliability experts
from all relevant sectors, not just hydrogen. Despite these limitations, the test application
did serve its purpose. It further developed the methodology and brought to light
particular strengths and weaknesses.
Inputs provided to the panel in this assessment were purposefully vague. Certainly, when
considering real systems, the panel should be provided with as much information as
possible to allow an accurate assessment. But due to the limited time during which the
panel was available, descriptions and definitions of reliability, the scope of study, and the
supply and demand scenarios were not specified to the degree desired for an assessment
of real systems. 17 For the developmental purposes of this application and the
hypothetical scenarios considered, specific details were not required. In fact, they would
likely not have supplied the experts with extra useful information, and could have biased
the results. Many of the researchers comprising the panel do not have a background in
reliability studies, and may have not been able to translate specific details about a system
16 For example, the U. S. DOE routinely hosts workshops of natural gas industry experts to identify issues
with infrastructure reliability and R& D opportunities to address those issues ( e. g., U. S. DOE and NETL
[ 2002] and SCNG [ 2