6 Risk management methodologies

Risk management comprise all activities towards the assessing, mitigating (to acceptable levels) and monitoring of risks. According to the ISO/DIS 31000 standard, the process of risk management includes the establishment of the context or assessment base, the identification of potential risks and the assessment and treatment of identified risks. Furthermore, a risk-management plan should be developed and implemented, describing risk control implementation and responsibilities (ISO, 2009).

6.1 Risk assessment

The formal definition of risk assessment is: a systematic process for describing and quantifying the risks associated with hazardous substances, processes, action or events. (Covello and Merkhofer, 1993) Risk assessment is the process of identifying hazards and evaluating the risk of them. It includes estimates of uncertainty and is based on the scientific knowledge available. In a profound risk assessment framework, the following questions should be answered:

  • What can go wrong that may lead to any hazardous impact (scenario analysis - qualitative risk assessment)?
  • What would be the probability of this failure to happen (probability quantification - quantitative risk assessment)?
  • If it happens, which consequences are expected (impact assessment - quantitative risk assessment)?

6.2 Risk assessment methodologies

Risk management is an integral component of CO2 storage operations (including site selection, site characterization and storage system design), comprising risk assessment and the development of mitigation, monitoring and remediation strategies. The overall aim of risk assessment is to investigate the storage system’s behaviour over time on the basis of any potential scenario compromising the storage integrity. Because the geologic storage of CO2 is in the development phase, at this moment there is not one unique, standardized methodology for assessing associated risks as yet. For CCS projects, potential adverse effects on health, safety, and environment (HSE) could result from elevated CO2 levels. The exposure may stem from possible accidents/failures during and after CO2 injection. To identify the probability of occurrence and the impact of the undesirable events, and to deal with uncertainties in a comprehensive and transparent way, the application of a well-established methodology and framework is required for CCS projects.

A distinction needs to be made between qualitative and quantitative methods, as the evaluation of well integrity for CCS operations can be achieved by employing either assessment methodologies. In general a qualitative assessment will precede quantitative evaluations.

6.2.1 Qualitative assessment

Qualitative risk assessment constitutes of hazard identification, analysis and prevention. The hazards involved are site-specific events leading to undesirable consequences. A qualitative approach can be adopted for addressing well bore issues. Such methodology is aimed at the identification of hazards that could occur during injection operations as well as in the post-injection period. Hazards can be defined in terms of combinations of system properties and/or events that may lead to undesired consequences. In a CCS context, migration of CO2 out of the storage system, leading to elevated levels in the shallow subsurface or atmosphere, is considered to be the most significant general threat.

In order to define all conceivable events and processes that could be relevant, dedicated assessment techniques are being used that aid their identification. After the identification of scenarios numerical models can be created to quantitatively assess them, predicting the performance of critical parameters over a specified time frame. Critical parameters regarding well integrity are, for instance, cement thickness, corrosion rate, plug length etc.

Qualitative risk assessment methodologies aimed at the evaluation of CO2 storage operations (e.g. TNO FEP analysis and Quintessa FEP analysis; see Table 6.1) are described in more detail in Sections 6.3.1 and 6.3.2.

6.2.2 Quantitative assessment

Quantitative assessment is defined as the analysis to predict the evolution and performance of a given system with respect to relevant parameter selection. In the case of well integrity the assessment comprises the analysis of the behaviour of a single or multiple wells in a CO2 storage system over appropriate time scales and volumetric quantities. Quantitative risk assessment possesses deterministic and probabilistic elements. The choice between the two approaches depends among others on the number of the wells in the CO2 storage project. If a storage option exhibits few existing abandoned wells, each well can be addressed by an individual deterministic risk analysis. Considering a storage option revealing numerous wells, a deterministic approach on an individual well base is hardly applicable. Instead, wells properties (e.g. cement permeability) are generalized for all existing wells or groups of wells (‘well families’) and are expressed with uncertainty ranges.

6.3 Assessment of wellbore integrity

The integrity of the well bore in CO2-rich environments has been raised as the area of the highest concern with respect to the long-term effectiveness of CO2 storage in geological reservoirs.

Most of the well integrity issues stem from casing pressure and cement problems which eventually hinder the hydrocarbon production and significantly raise the operational expenses. Moreover, the presence of CO2 in any wellbore environment will introduce more issues and should be addressed for CCS applications. Previous meetings addressed the importance of using CO2 resistant materials and other equipment specifically designed for CCS applications to reduce the frequency of the failure of wells. It was clearly stated that promising a ‘leak-free well’ with conventional methods will not be possible. Therefore prospective drilling operations may focus on remedial and preventive tailor-made protocols.

However, there is a major issue playing a crucial role for the uncertainty of the wellbore integrity and the current status of wells. It has to be noted that CO2 injection wells, constructed according to recent standards and monitored during after injection operations, do not represent a major risk factor. Instead, existing older abandoned wells, highly abundant at many potential storage sites, e.g. depleted oil fields, pose a significant risk due to obsolete non-CO2 specific construction and abandonment practices. In the following, several state-of-art risk assessment approaches are described.

Table 6.1 Characteristics of Well Assessment methodologies

Well Assessment methodology Qualitative analysis Quantitative analysis HSE analysis Simulator Decision tool
TNO-CASSIF1 FEP-based Scenario Analysis N/A X
The Generic CO2 Geological Database2 FEP-based Scenario Analysis N/A X
Data mining3 Semi- quantitative X
Performance and Risk Management methodology (P&RTM)+ SimeoTM-Stor4 Probabilistic X X X
Semi-Analytical Model and Monte Carlo Simulation5 Probabilistic X
CO2-PENS6 Probabilistic X X X

1 TNO; Yavuz et al. (2008)

2 Quintessa; Preston et al. (2005)

3 Watson and Bachu (2007)

4 Le Guen et al. (2008); Meyer et al. (2008)

5 Kavetski et al. (2006)

6 Viswanathan et al. (2008); Stauffer et al. (2009)

6.3.1 TNO’s CASSIF and FEP database

The Carbon Storage Scenario Identification Framework (CASSIF; Yavuz et al., 2008 ) is based on three major CO2 release scenarios (well, fault or seal) from where the relevant risk factors are identified. The risk factors of the major scenarios are forming the assessment basis. A questionnaire, to be completed by the experts, has been designed to get an initial overview of potentially important risk factors. The interface facilitates the online use of the so-called FEP database and a risk terminology glossary. Together with real-time scenario formation, these improvements will greatly enhance the speed, transparency and comprehensiveness in the creation of subsurface CO2 release scenarios.

The acronym ‘FEP&Rsquo; refers to ‘Features’, ‘Events’ and ‘Processes’ relevant to describing the state of a system of interest at any time, or the processes that take place. In the scope of this study the ‘system of interest’ consists of all elements characterizing the storage site. The application of Features, Events, and Processes (FEP) to describe the evolution of underground storage systems evolves from its first state. This evolution is called a scenario where case specific events with undesirable outcomes are addressed.

As was described above, the Carbon Storage Scenario Identification Framework is based on the three major CO2 release scenarios (well, fault or seal) from where the relevant risk factors are identified. The structure of CASSIF has been designed to get an initial overview of potentially important risk factors. CASSIF facilitates the online use of the FEP. For the well integrity case, 3 subjects are being addressed in qualitative terms (Figure 6.1). Subjects include various risk factors from cement, casing and operational themes which might change the overall integrity of well during injection and post injection phases.

Figure 6.1 Well integrity section from CASSIF.

6.3.2 The Generic CO2 Geological Database

The Generic CO2 Geological Database was originally developed by Quintessa Ltd., UK as a contribution to the IEA GHG Weyburn CO2 Monitoring & Storage Project (e.g. Preston et al., 2005). Continued development of the database was supported by IEAGHG and BERR. The database represents a qualitative tool to support risk assessment, based on the principle that CO2 storage systems can be described in terms of features, events and processes (FEPs) associated with geological storage of CO2. The database was inspired by the OECD / Nuclear Energy Agency FEP Database for Geologic Disposal of Radioactive Waste (http://webnet.oecd.org/wbos/).

Due to the generic character of the database, the FEPs relate neither to specific storage concepts nor to specific sites. Instead, the approach is to use the database as a cross referencing and auditing tool for site-specific FEP studies. The database currently contains 178 described FEPs, selected with respect to long-term safety and performance of CO2 storage, after injection has ceased and boreholes have been closed. The tool is accessible in the public domain (http://www.quintessa.org/co2fepdb/PHP/frames.php). The database exhibits a hierarchical structure consisting of 1) categories 2) classes 3) FEPs and (when necessary) 4) sub-FEPs. The eight categories are arranged as follows:

  1. Assessment Basis
  2. External Factors
  3. CO2 Storage
  4. CO2 Properties, Interactions & Transport
  5. Geosphere
  6. Boreholes
  7. Near-Surface Environment
  8. Impacts

The borehole category, which describes FEPs associated with wellbore leakage-related risks, is subdivided in two classes: 1) Drilling and completion, 2) Borehole seals and abandonment, each containing several FEPs (Figure 6.2).

Figure 6.2. Screenshot of the FEP database showing the structure of the borehole category (www.quintessa.org).

Figure 6.3. Screenshot of the FEP 5.1.1 ”Formation damage” (www.quintessa.org).

The boreholes category “is concerned with the way that activity by humans alters the natural system. Boreholes used in the storage operations and those drilled for other purposes are relevant to the long-term performance of the system” (www.quintessa.org). Each FEP entry contains 7 kinds of information: 1) name of the FEP; 2) FEP description, in many cases accompanied by an illustration; 3) relevance of the FEP to performance and safety; 4) references; 5) web links; 6) FEP number; and 7) date of last modification. As an example, Figure 6.3 shows the content of FEP 5.1.1 ‘Formation damage’.

A search engine has been implemented and the database can be browsed for FEPs. A list of references and links can be displayed. More detailed insights into the approach are provided on the Quintessa Ltd. website (www.quintessa.org). This approach covers the qualitative aspect of the risk assessment allowing selecting the major well bore integrity issues to be addressed in the quantiative assessment later on.

6.3.3 Leakage evaluation by data mining

Potential CO2 storage reservoirs are often intersected by a vast number of abandoned wells, each representing a potential leakage pathway. One strategy for evaluating and ranking abandoned wells relevant to the integrity of an envisaged storage option is data mining of well information collected by regulatory agencies, particularly with focus on surface casing vent flow (SCVF) and gas migration data (GM). A corresponding approach conducted by Watson and Bachu (2007) considered stored data on more than 315,000 oil, gas and injection wells in the province of Alberta, Canada recorded up to the end of 2004. The data have been collected by the Alberta Energy and Utilities Board (EUB), the regulatory agency in Alberta (now the Energy Resources Conservation Board). The agency records well and production data including construction details and SCVF/GM data. SCVF is caused by gas entering the exterior production casing annulus from a source formation below the surface casing shoe, resulting in a gas flow to or pressure built-up at the surface. SCVF can be measured at the surface casing vent of a well (Figure 6.4). GM is an effect of gas migrating outside of the cemented surface casing. It may be caused by deep gas originating from formations below the surface casing shoe migrating upwards past the surface casing shoe, potentially caused by poor surface casing cement or fractured cement or rock. GM may also originate from shallow gas accumulations located above the surface casing shoe and leaking through a poorly cemented surface casing. In Alberta, testing of SCVF and GM is conducted as per regulatory requirements. More details on the testing procedures are described in Watson and Bachu (2007).

Major focus of the investigation has been to find factors contributing to wellbore leakage in the context of CO2 storage. To reach this goal, correlations between SCVF/GM and economic activity, technological changes, geographic parameters, completion and abandonment practices as well regulatory changes have been considered. Additionally, casing inspection logs indicating internal as well as external corrosion for 500 wells have been considered and compared to cement bond logs. Groundwater level records have been used for correlation with surface casing, annular cement and casing failure depths.

Generally, three conditions have to be met for a leak to occur: 1) A leak source, 2) a driving force, such as buoyancy or a head differential, 3) a leakage pathway. CO2 storage operations fulfil two of these conditions: a potential leak source is represented by the injected and stored CO2 which, at the same time, delivers the driving force due to CO2 buoyancy and an increased reservoir pressure caused by injection operations. Thus, the additional presence of a leakage pathway will inevitably result in a leakage scenario. The main leakage pathways associated with wells include: 1) Poorly cemented casing/hole annulus 2) casing failure 3) abandonment failure.

Figure 6.4. Schematics and picture of a typical wellhead with surface casing vent installed (Watson and Bachu, 2007).

These potential leakage pathways can become effective without additional chemical interactions of CO2 such as cement degradation and casing corrosion. The study does not differentiate for seepage/leakage to different “compartments” (e.g. atmosphere, aquifers). Instead it is assumed that any leak of CO2 from the storage site or natural gas from the reservoir is undesirable.

Abandonment methods in Alberta include three scenarios: 1) Wells drilled and abandoned, 2) wells drilled, cased, completed and abandoned, 3) wells drilled, cased and abandoned. Descriptions of these scenarios and further information on regulatory requirements and testing methodologies with respect to SCVF/GM can be found in Watson and Bachu (2007).

The data mining approach relates a number of well features to the magnitude of impact on the leakage risk. The well features were grouped into three categories: 1) Factors showing no apparent impact 2) factors showing minor impact 3) factors showing major impact.

6.3.3.1 Factors showing no impact

  • Well age: The expectation of higher leakage rates of older wells due to less elaborate construction and materials was not supported by the data. While the oldest recorded abandoned well in Alberta dates back to 1893 and the first commercial gas field was developed in 1901, the compulsory SCVF/GM testing did not come into effect until 1995. Thus, corresponding reports for many wells abandoned prior to 1995 were not available, resulting into a serious distortion of the analysis of the age factor. However, the latter is indirectly covered by other factors relating to construction and abandonment practices.
  • Well operational mode: Well operational mode (oil and gas production, water and solvent injection, disposal of liquid waste or acid gas) did not reveal any effect with respect to wellbore leakage based on SCVF/GM data. Thermal operation modes such as steam assisted gravity drainage (SAGD) and steam injection were expected to exhibit an enhanced impact due to thermal stress on the well materials. However, this could not be verified, probably because these wells are more recent and largely still operational. This results in a lack of data due to few abandoned wells of this operation mode and too short observation periods. The impact on thermal operation modes will not be ascertainable until a significant portion of such wells will be subject to abandonment.
  • Completion interval: Depth of the SCVF/GM source and depth of completion intervals revealed no correlation. This was supported by casing and cement logs, exhibiting that the majority of wells are characterized by a good cement quality and zonal isolation deep in the wellbore.
  • H2S or CO2 presence: The presence of hydrogen sulphide and CO2 in produced hydrocarbons was investigated for a possible impact on internal and external casing corrosion, which was not supported by the available data. Sour gas well operations in Alberta are characterized by certain features enhancing casing protection. First, such operations require the well equipped by packers, to protect the inner casing surface. Second, in Alberta, H2S usually occurs in deep formations, where the majority of wells exhibit a good cement bond.

6.3.3.2 Factors showing minor impact

  • Licensee: Various operators using different abandonment practices may result in different sealing efficiencies of wells subject to abandonment. Table 6.2 reveals the licensee-dependent variation of the leakage rates. However, an unequivocal correlation was not evident.

Table 6.2. Licensee comparison in terms of well leakage occurrence (Watson and Bachu, 2007).

Licensee % Toal Well % Reported SCVF % Reported GM Ratio SCVF Well Total Ratio GM WEll Total
Licensee A 11.3 7.5 36.2 0.66 3.2
Licensee B 35.4 43.2 52.6 1.2 1.5
  • Surface casing depth: Surface casing depth was not found to influence the sum of leakage by SCVF/GM. However, there is an influence on whether leakage occurs as SCVF or GM. Generally, greater surface casing depth reduces the occurrence of SCVF while GM increases, supporting the indication that GM sources are typically situated above the surface casing shoe depths and that GM is influenced by surface casing cementing practices.
  • Total depth: SCVF/GM increases slightly with total well depth, attributed to the generally increasing uncemented upper interval of deeper wells, providing enhanced hydraulic communication with source formations.
  • Well density: In areas exhibiting high well densities, occurrence of well-to well crossflow can potentially result in enhanced leakage rates of wells subject to inflow due to such migration phenomena. This hypothesis was not supported by the available data but was indicated from other work. This might refer to the existence of more recent wells not yet sufficiently tested and characterized by better cementation.
  • Topography: River valleys may represent zones of higher leakage risks due to removal of overburden and corresponding decline of hydrostatic pressure, potentially resulting in shallow overpressured zones. However, the data set investigated revealed no significant influence of topography on SCVF/GM.

6.3.3.3 Factors showing major impact

  • Geographic area: In certain areas within the province of Alberta testing of all wells is required, whereas in other areas the requirements are less strict. One area within the province subject to testing requirements for all wells exhibited a more frequent occurrence of leakage compared to the entire province. However, it is not clear whether this finding refers to the different testing conditions.
  • Wellbore deviation: Figure 6.5 shows that leakage by SCVF/GM occurred significantly more often related to deviated wells, while the impact of well deviation on the ratio of SCVF to GM was minor. This may be caused by mechanical aspects, such as casing centralization and cement slumping. Any well revealing a total depth exceeding true vertical depth was considered a deviated well. It has to be noted that in case of vertical drilling, non intended well deviation may lead to an increased penetration surface within the caprock. This may impact on the cap rock integrity.

Figure 6.5. Comparison of leakage rates between deviated wells and the average of all wells in the test area (Watson and Bachu, 2007).

  • Well Type: Cased abandoned wells account for 98% of all leakage cases reported. The rest refers to wells drilled and abandoned. This significant difference may rely on more stringent abandonment requirements for drilled and abandoned wells. Completed wells exhibit an additional leakage potential due to perforated intervals.
  • Abandonment Method: In Alberta, cased and completed wells are predominantly abandoned by bridge plugs capped with cement. Based on the data set and experience 10% of these bridge plugs will fail over a period of centuries allowing formation fluids to enter the well bore. Alternative methods, such as placing cement plugs across completed intervals using a balanced plug method, or setting a cement retainer and squeezing cement through perforations are expected to reveal lower failure rates. The final barrier in a well is the welded casing cap, known to be highly unreliable. However, leaking caps may contribute to reduce well overpressure and can act as an early warning system for compromised well integrity.
  • Oil price and regulatory changes: The data set reveals a significant positive correlation between SCVF/GM occurrence and oil price between 1973 an 1999. This can be explained by the relation between exploitation activity and equipment availability. Satisfaction of a high demand with limited equipment resources impacts on primary cement placement practices. The implementation of heavy oil production by thermal recovery, high well densities and application of diverse well technologies which accompanied the oil price rise additionally increased the likelihood of well leakage. The correlation of the oil price and SCVF/GM starts to diverge in 2000 (oil price raises while SCVF/GM occurrence ceases). This may be due to SCVF/GM being detected primarily at the time of abandonment. Wells drilled since 2000 are generally not yet abandoned these well leakages may not yet be reported or detected.
  • Uncemented casing/hole annulus: A low cement top was found to be the most important indicator for SCVF/GM. Low cement top is also the main cause for external casing corrosion. Watson and Bachu (2007) stated the following conclusions based on the analysis of well logs for casing inspection and cement bond quality:
    • The majority of significant corrosion occurs on the external wall of the casing
    • A significant portion of the wellbore is uncemented
    • External corrosion is most likely to occur in areas of no or poor cement

Furthermore, it was determined that the top 200 m of the cement annulus is generally of poor quality and that the vast majority of SCVF/GM originates from formations not isolated by cement.

The majority of casing failures have been attributed to regions of poor and no cement in the annulus. Evaluation of well logs revealed that cement quality typically improves deeper in the well and particularly across completed intervals.

6.3.3.4 Prediction of wellbore potential for leakage based on well attributes

Based on the results discussed, a decision tree was developed in order to estimate and rank leakage probability with respect to SCVF/GM (Figure 6.6) for abandoned wells. This scheme employs the following aspects:

  • Well type: Drilled and abandoned wells are far less prone to leakage than cased wells.
  • Regulatory changes: As of 1995 regulations became more stringent; wells abandoned after 1995 should exhibit less probability of leakage, as any detected leakage would have provoked counter measures prior to abandonment.
  • Oil price: Wells drilled before 1995 exhibit positive correlation between leakage and oil price due to reasons discussed above.
  • Geographic position: Is the well situated in a region where wells statistically reveal enhanced leakage or where the testing conditions are applied in a stricter manner?
  • Cement top requirement: Data mining revealed that absence of cement at the upper wellbore is probably the most reliable predictor for SCVF/GM and casing failure.

In general, this method is suitable for measuring uphole leakage. It is a decision tool for distinguishing between zones in intended storage area exhibiting different risk levels.

Figure 6.6. Decision tree for assessing the potential for well leakage inside and outside surface casing (Watson and Bachu, 2007).

6.3.4 Performance and Risk Management methodology (P&Amp;RTM)

The Performance and Risk methodology (P&Amp;RTM), developed by Oxand S.A., represents a quantitative risk-based approach for well integrity management, allowing identification and quantification of risks within CO2 injection and storage operations over various time scales (co-injection, post-injection, abandonment). A detailed description reaching beyond the scope of this overview can be found elsewhere (Le Guen et al., 2008; Meyer et al., 2008). The methodology is intended as a decision support tool for stakeholder parties involved in planning or operating a CO2 storage project as it can be applied for setting up risk mitigation strategies and emergency plans with respect to CO2 leakage or seepage through well completions. Additionally, this risk approach provides elements to demonstrate performance and safety of CO2 storage to authorities, which is necessary to get a permit for operations. Major focus lies on evaluating risks of CO2 leakage into subsurface compartments (e.g. aquifers) or seepage to the atmosphere caused by ageing processes (e.g. cement degradation, casing corrosion), potentially leading to migration channels.

The major working steps of the method are:

  • Identifying system and sources of degradation by system characterization and functional analysis
  • Quantifying the criticity (defined below) of scenarios by modelling approaches in terms of probability and severity
  • Establishing a risk mitigation plan

Performance assessment refers for example to the ‘containment performance’ of a well, defined as the “capability to ensure a good zonal isolation in order to contain the injected CO2 in the geological reservoir over the intended lifespan of a storage reservoir” (Le Guen et al., 2008). Risk is perceived as the probability of a loss in containment performance resulting in an impact on specific stakes. Criticity is defined in this context as the multiplication of the impact on health, safety and environment of a hazardous situation times the probability of the situation to occur. The essential component of the approach is a well completion and leakage simulator (SimeoTM-Stor) allowing the prediction of the quantitative impact of leakage paths along the wellbore. This implies the opportunity to establish a prognosis with respect to well integrity as a function of time covering periods from decades to millennia and thus, the feasibility for assessing the containment performance of a potential storage site. One of the main features is the expression of risks in terms of criticity (defined above). Additionally, the approach can indicate the requirement of preventive or corrective measures or monitoring for mitigating unacceptable risk levels. To summarize, main objectives are the identification and quantification of the risk-associated criticities and risk-treatment by selection and implementation of risk mitigation actions.

The process chain of the approach, illustrated in Figure 6.7, consists of a data collection survey followed by a functional analysis of the system, serving as input for a static well model. The latter, in turn, acts as an input for a dynamic model, able to predict degradation of well components as a function of time and to quantify CO2 leakage along the wellbore.

Figure 6.7 Workflow of the P&Amp;RTM methodology (Le Guen et al., 2008).

Based on the parameters of the static and dynamic models, and associated uncertainties, scenarios are defined and then evaluated numerically. Simulation results enable to identify leakage pathways along the wellbore and the amount of leaking gas towards different targets (fresh water aquifers, surface, etc) over time. Severity levels for each identified scenario are assessed from CO2 leakage simulation and a consequence grid relating a certain performance loss to the severity of the resulting consequence for different stakes (project performance, safety, environment, public opinion, etc.). The probability level of a risk is given by the probability of the scenario and a frequency grid. Finally, risk mapping allows identification and ranking of risks for all wells relevant to CO2 injection and storage operations at the site to be assessed. Practically, risk mapping is performed by filling a colour coded grid with each couple (probability, severity level) corresponding to all scenarios which lead to CO2 leakage. These results then lead to recommendations and conclusions to support decision–making and to establish site-specific risk mitigation measures. The workflow will be described in the following in more detail.

6.3.4.1 Data collection

The data collection survey, representing the initial step in the process chain, targets on data retrieval from all available documents referring to wells and their surroundings, including well descriptions (trajectory, completion details) and characteristics of the wells’ components (e.g. features of tubulars, packers and cement). Such information can be found in well completion design documents, drilling and cementing reports, cement and corrosion logs, production history and workover reports. Furthermore, all formations intersected by the well are to be characterized (by log and core data) in order to define boundary conditions for the modelling approach later on.

Table 6.3. Examples of a components and functions resulting from a functional analysis with corresponding failure modes, causes and effects (Le Guen et al., 2008).

6.3.4.2 System description and functional analysis

Before being able to conduct a risk assessment, the physical environment of the well has to be taken into account, including sub-systems potentially able to interact with the well (formations above the storage reservoir and the cap rock, subsurface fluids, shallow subsurface/soil, surface or sea floor, atmosphere). The time of the P&Amp;RTM approach initiates when CO2 enters the well and covers decades to millennia, depending on the purpose of the study. The functional analysis requires defining components, functions and sub functions, and associated failure modes. Well components are for example tubulars, packers, cement sheaths, and cement plugs. Such well components are related to specific functions, e.g. a packer refers to hydraulic separation. This function can be decomposed in several subfunctions such as to resist subsurface pressure and temperature as well as to resist chemical degradation by pore fluids. Specific failure modes can be allocated to each function or subfunction of a component, leading to deterioration or complete failure of the function. Furthermore, the specific failure modes can be referred to causes end effects (Table 7).

6.3.4.3 Consequence grid

As outlined above, criticity is defined as the interaction of probability of occurrence of a scenario and the severity of its impact when occurring (related to the amount of CO2 leaking from a reservoir to a specific target (e.g. environment, aquifer pollution, humans, economy, etc). A consequence grid relates stakes involved into CO2 storage operations with specific severity magnitudes and their qualitative or quantitative impact (Table 6.4).

The representation of a consequence grid refers to a matrix: each column represents an individual stake involved in CCS operations (stakes to be defined by project management). The rows of the matrix refer to different impact levels on each stake (impact levels to be defined by each stakeholder). The matrix cells reveal a stake-specific qualitative or quantitative degree of impact. Consequence grid is project and site specific.

Table 6.4. Example of a consequence grid including different stakes and severity levels (Le Guen et al., 2008).

Severity levels stakes
Personal injury Public opinion Additional OPEX-Financial CO2 storage performance goals Corporate Perception of know how Environment
1: Minor / < 0.1 M$ Loss < 0.01 % of injected CO2 no impact
2:Low no impact [0.1 - 0.5[M$ Loss =[0.01-0.05[ % of injected CO2 Technical skill non affected (project is considered as a test)
3: Serious First aid [0.5 - 1[M$ Loss =[0.05-0.1[ % of injected CO2 Top Management becomes suspicious about technical skill
4: Major Medical treatment [1 - 5[M$ Loss =[0.1-0.5[ % of injected CO2 Lack of confidence from the Top Management - Request for a demonstration of technical feasibility
5: Critical Serious personal injury [5 - 10[M$ Loss =[0.5-1[ % of injected CO2 Questioning from the Top Management about the technical capability to assume CO2 storage projects
6: Extreme Serious pers. injury, possible permanent injury >= 10 M$ Loss >= 1 % of injected CO2 Termination of the project - Field is not considered as a CO2 storage field

6.3.4.4 Static model building

A preliminary assessment for well integrity is based on using all available characterization measurements and their associated uncertainties. For example, the component casing is associated with the characteristics diameter, thickness, overlaps, shoe depths, etc., whose quantities may reveal uncertainties. Cement evaluation can be achieved by interpretation of logging data. To be able to put the well data into the geological context and to keep the model simple, a segmented well model is generated, representing a 2D axi-symmetric description of the well, where each layer is characterized by constant properties. Segmentation processing is applied for well components, formation layers, heterogeneity of cement sheaths and the quality of cement bonding. Wells revealing similar designs, completions, cement sheaths and intersect similar geological settings can be grouped to ‘well families’ described by a ‘typical well’, representative for the well family.

6.3.4.5 Dynamic modelling

The static model is used as an input for a dynamic simulator, targeting on the prediction of the degradation of the well components exposed to CO2 and/or other interactors (e.g. formation fluids) as a function of time, achieved by numerical simulations. Furthermore, CO2 leakage rates to any point of interest (e.g. aquifer, atmosphere) can be calculated. Further input for the simulator are the various component-specific degradation mechanisms, the associated chemical kinetics and boundary conditions. The model can be calibrated by correlation with laboratory experimental results and/or in-situ data, e.g. on steel corrosion. Initial and boundary conditions like a hydrostatic pressure profile and fluid saturations of intersected formations have to be specified. Cement sheaths are regarded as porous media, saturated with water and CO2. Darcy’s law is applied to describe the fluid flow within well’s system. The model refers to the Van Genuchten law to relate cement pore water saturation to flow properties.

6.3.4.6 Risk assessment

The initial state of risk assessment implies the definition of leakage scenarios, resulting from state of well components. Quantitative descriptions of well components, degradation mechanisms, kinetics, and initial as well as boundary conditions are subjects to uncertainties which are considered during static and dynamic modelling. Consequently, CO2 leakage scenarios are calculated relying on probability distribution functions to account for the uncertainty range in each the input parameters. This will result in a specific probability for each scenario.

The outcome of a leakage scenario modelling is a CO2 leakage rate as a function of time referring to a selected spatial point, e.g. a point in an aquifer or at the surface. The leakage rate calculated for each relevant leakage scenario is converted into a severity level according to the consequence grid and, after involving the scenario probability, transferred into a criticity value. This allows the quantification of all pre-defined risks. Compilation of all risk scenarios for a certain well in a colour coded graph and subsequently relating probability and severity of all identified risks, leads to risk map, which is a powerful tool to visualize the envelop of risks associated to the well integrity performance.

Such an approach for several wells leads to a high resolution risk assessment for all wells relevant for a CO2 storage site and enables their ranking in terms of criticity. When using ‘typical well’ approach for a well family (group of wells presenting similar characteristics), a reverse analysis of the risk map of a ‘typical well’ allows the generation of a risk map for each individual well of the family. This process is referred to as risk distribution.

6.3.4.7 Risk treatment

As full integrity of a storage option is neither realistic nor relevant, there has to be a definition of an acceptable risk threshold. A definition can be achieved relying on the stakeholders’ perception of non-tolerable risks levels, by consideration of the objectives developed within Corporate Integrated Management Systems (CIMS) and through legal and regulatory requirements.

The component characteristics contributing to unacceptable criticity scenarios, termed risk sources, represent the base for setting up a mitigation plan. This results in recommendations for measures assuring that all risks are decreased and remain below the acceptable risk levels of wells relevant for a CO2 storage option. Such measures are targeting on lowering the probability of a risk, its severity or both. They can be allocated to four types:

  • Conduct additional characterization/inspection measures to reduce uncertainties
  • Lowering Migration by workovers, treating risk sources detected after conducting additional characterizations/inspections
  • Conduct operational best practices , e.g. casing pressure tests during injection operations
  • Conduct monitoring approaches to observe the evolvement of the system during operations and to detect potential hazards prior to their occurrence

Such recommendations have to be established for each individual well within a storage site. The recommendations represent the final step to approve operations from the perspective of risk management.

6.3.5 Semi-Analytical Model and Monte Carlo Simulations

This summary is obtained and compiled from Kavetski et al. (2006) and details beyond the scope of this summary can be found in this article.

6.3.5.1 Introduction

Because of geological conditions and the existence of appropriate infrastructure, mature sedimentary basins are likely candidates for injection and storage of CO2. These basins, for example in North America, more than a century of oil and gas exploration and production. This has resulted in many wells drilled; 400,000 in the Alberta Basin and more than one million in Texas. Thus, estimating the probability of CO2 leakage along existing (abandoned) wells is an important part of any risk assessment study.

6.3.5.2 Risk assessment

In order to perform a reliable risk assessment, it is necessary to estimate the likelihood and magnitude of potential leakage out of the storage formations. This, in turn, requires the ability to model the migration of CO2 plumes during and after injection. In addition, it is necessary to include all the wells in the mathematical description because they may be encountered by the CO2 plumes. Furthermore, multiple geological formations need to be included in the description of the surface, because vertical migration of CO2 is considered to be a central feature of the leakage problem.

In general, performing calculations based on this list of requirements is extremely time-consuming when using traditional numerical simulators. On top of that, the high degree of uncertainty associated with the hydraulic characteristics (for instance, poorly understood leakage pathways and effective permeabilities) and often even the location of the abandoned wells, make traditional simulations even more difficult, because many models have to be investigated.

Because of the high degree of uncertainty, a probabilistic analysis is necessary to estimate the likelihood of leakage and the confidence limits on these predictions. The system is strongly non-linear with respect to its (highly uncertain) hydraulic properties. In particular, it is believed that the effective permeability in abandoned wells is the dominant source of uncertainty in leakage predictions. Considering these uncertainties, the use of Monte Carlo methods becomes necessary to obtain leakage estimates under different permeability scenarios. Monte Carlo analysis requires multiple runs (hundreds or thousands) with different system properties sampled from a priori distributions. Existing, time consuming, numerical algorithms for multiphase flow are, therefore, not the most suitable way in Monte Carlo based risk assessments, because of time and hardware constraints. Therefore, a semi-analytical model for computationally fast estimation of CO2 migration has been developed to replace the existing numerical algorithms.

6.3.5.3 Semi-analytical model

The semi-analytical method specifically focuses on the wells as the dominant transport mechanism and derives an approximation of the general multiphase equations. A detailed numerical description can be found in Nordbotten et al. (2005; 2009). A central component of the methodology is a model of radial CO2 plumes developing around injection and leaky wells. The semi-analytical model uses plume masses around the injector and leaky wells as the primary variables. This model is based on the following assumptions:

  1. Only two fluids are present in the system; saline water (brine) and CO2.
  2. The CO2 migration is solely advection driven.
  3. The geological formations are homogeneous and horizontal, with alternating permeable (aquifer) and impermeable (aquitard) layers.

In addition, it is assumed that:

  1. CO2 plumes are radial symmetric in all formations,
  2. All layers are spatial homogeneous,
  3. Caprock formations are impervious,
  4. The formation is horizontally layered and the wells are vertical,
  5. Capillary pressure and thermal effects are ignored,
  6. Leaky wells fully perforate each aquifer,
  7. The flow rate is constrained by the available mass and the well segment permeability across the aquitard

Figure 6.8 shows the schematic of the system modelled by the semi-analytical approach. In general, any well may leak CO2 to the overlying formations. In addition, it uses another semi-analytical model to determine the shape of the plumes and the pressures at all well locations in all layers. The latter model is used to identify the wells contacted by CO2 plumes and to estimate the leakage flow rates. The magnitude of leaking is computed using a Darcy-type permeability function.

Figure 6.8. Schematic of injection and leakage, including leakage plumes and multiple layers of alternating permeable aquifers and impervious aquitards. The flow rates are denoted by Q(t) and the amount of mass by M(t).

The semi-analytical model consists of the following (coupled) primary equations, adapted to the above listed assumptions:

  1. Mass balance equations,
  2. Well flow rate equations,
  3. Pressure equations,
  4. Plume shape equations,
  5. Saturation equations.

This set of equations is solved numerically using time stepping algorithms and iterative solvers, because these equations are coupled nonlinear differential-algebraic equations.

6.3.5.4 Case study: Alberta Basin

Kavetski et al (2006) implemented the semi-analytical model into a Monte Carlo analysis. The Wabamun Lake formation in Alberta (30x30 km2 and 500 wells) has been modeled to demonstrate the capabilities of the semi-analytical model. The distribution of the effective permeability is assumed to be bi-Gaussian. This means that two peaks are present; one corresponds to well-formed cement and the other to degraded cement. For this case study, 600 independent Monte Carlo realizations were performed based on this bi-Gaussian distribution to determine the leakage profile over 32 years. From these simulations output statistics, including statistics related to well leakage, can be generated. Because of the nature of the demonstration, no conclusions are drawn on the leakage profile. This demonstration, however, shows that the simulations based on the semi-analytical method can be performed within a reasonable amount of time. This is a major improvement compared to the traditional numerical methods.

6.3.5.5 Future work

So far, the capabilities of the semi-analytical model in a Monte Carlo framework are considered in case the the distribution of well permeabilities is well-known a priori. In practice, this is not the case. It is, therefore, intended to formulate the Monte Carlo problem in inverse form. This would imply that the well permeability distribution would be derived from given pre-defined maximum leakage rates to a known degree of confidence. Secondly, the computational runtime has significantly improved with respect to the runtime of traditional methods. Still, the runtime is strongly dependent on the well permeabilities and on the ratio of well permeabilities in adjacent well segments. This phenomenon is being studied in more detail. Further improvements of the method consist of:

  1. More physically-based estimation of saturations in the wells,
  2. Improved numerical algorithms for the nonlinear equations that arise in the semi-analytical formulation,
  3. A new hybrid approach that allows traditional numerical solutions in the injection layer to be coupled with semi-analytical leakage solutions for the overlying formations.

6.3.6 Probabilistic assessment of wellbore leakage using CO2-PENS

CO2-PENS (predicting engineered natural systems) is a probabilistic simulation tool designed to incorporate CO2 injection and sequestration knowledge from the petroleum industry to perform risk assessment of sites. CO2-PENS includes economic tools, as well as models for the physical and chemical interactions of CO2 in a geologic reservoir (Viswanathan et al., 2008; Stauffer et al., 2009). The model links high level system models (i.e. a reservoir model) to the process level (wellbore leakage, chemical interaction of CO2) and thus represents a hybrid coupled process and system model designed to simulate the following CO2 pathways:

  • Capture from the power plant
  • Transport to the injection site
  • Injection into geologic storage reservoirs
  • Potential leakage from the reservoir
  • Migration of escaped CO2 either to compartments in the vicinity of the storage reservoir or into the atmosphere

Due to its modular architecture, the tool allows incorporation of additional process models by linking to dynamic linked libraries (DLL). Viswanathan et al. (2008) demonstrated the applicability of CO2-PENS with respect to wellbore leakage simulation from a synthetic depleted oil reservoir by using a wellbore release model DLL, developed by the Princeton–Carbon-Mitigation initiative (CMI) group (Nordbotten et al., 2009). The simulated scenario assumes leakage of CO2 through a plugged well and subsequent migration into an overlying aquifer and to the atmosphere. For the latter, the atmospheric model developed by Los Alamos National Laboratories has been used. Details of the applied atmospheric dispersion model are described in Viswanathan et al. (2008). Within the scope of the case study, leakage of 0.01% of the initially injected CO2 has been assumed as the acceptable upper limit.

Simulation of wellbore leakage is complicated since the associated interactions and processes are not yet entirely understood. Wellbore cement permeability is identified as a key parameter in a wellbore leakage scenario and is difficult to estimate. Additionally, as the seals of storage sites are usually intersected by numerous wells, simulation approaches require probability distribution functions (PDF) with respect to potential failure mechanisms as input parameters to take account of uncertainties. A conceptual model of CO2 leakage may be developed for any given well relying on PDFs of the quantities of the following processes:

  1. Flow at cement-casing interface
  2. Flow through the cement matrix
  3. Flow through pathways created by bulk chemical dissolution of the cement
  4. Flow through fractures in the cement
  5. Flow through an open annular region due to inadequate cement placement
  6. Flow at the cement-cap rock interface

For the case study, observations and experiences from an investigation on wellbore cements at Scurry Area Capital Reef Operations Committee (SACROC) have been adopted for a first attempt to establish PDFs enabling modelling of CO2 leakage through wellbore cement. In order to obtain reliable PDFs, key processes have to be identified through a combination of experimental and theoretical information, which in turn, has to be validated with field analogues. Carey et al. (2007) reported on a cement core retrieved from a 55-year old well which was exposed to a CO2 environment for 30 years due to EOR operations. The investigators concluded on the one hand that cement can maintain an adequate hydrological barrier after decades of CO2 exposure. On the other hand, cement samples revealed unequivocal evidence for CO2 interactions at the cement-casing and cement-cap rock interfaces. For the single well investigated, Carey et al. (2007) found that processes (2) and (3) were not important for the observed interaction phenomena, while processes (1), (4) and (6) turned out to be significant. When dealing with old wells, information on construction details is often patchy. Consequently, the well age has been used as a proxy for its integrity, i.e. the probability of well failure. The investigation continued with the PDF constructions focused on the six leakage pathways discussed above and by considering the findings obtained from the SACROC samples, discussed in the following.

6.3.6.1 Pathway 1: cement-casing interface

Occurrence of a microannulus at the cement-casing interface is probable due to pressure and thermal stress during operation, raising the possibility that microannulus generation increases with operation time. The leakage effect may be mitigated by carbonate precipitation with respect to a small annulus aperture. Larger apertures may cause significant flow. Due to a lack of corresponding data, an aperture of 5 mm has been approximated as a maximum. For the case study, a cubic relationship between effective permeability and annulus aperture (k = (aperture)3/12) has been adopted (Snow, 1968). In CO2-PENS a PDF referring to the effective permeabilities along the cement-casing interface is used.

6.3.6.2 Pathway 2: Flow through the cement matrix

Matrix flow is not considered a significant leakage pathway, due to the very low effective permeability (magnitude of 10-9 Darcy) of appropriately placed cement. However, cement permeability can be modified by CO2 diffusion and interaction. Within the SACROC project, cement permeabilities of 10-4 Darcy where measured in old wells, which provides sufficient fluid retardation, considering the relatively large thickness of the well bore cement.

6.3.6.3 Pathway 3: Flow through pathways created by bulk chemical dissolution of the cement

Although carbonate brine causes dissolution of cement, typical thicknesses of wellbore cements are likely to prevent significant leakage. This consideration is in accordance with the observations in SACROC. Due to the minor importance of this process, no corresponding PDF is required.

6.3.6.4 Pathway 4: Flow through fractures in the cement

Fracturing of wellbore cement is likely due to thermal and mechanical stress associated with well operations. However, CO2 flow through fractures is considered self limiting due to three phase flow of CO2 resulting from pressure and temperature decrease during upward migration. Within SACROC, fractures have been observed to be filled with calcium carbonate or hydroxide, indicating an initial fluid flow ultimately terminated by mineral precipitation, leading to the conclusion that no PDF is required to quantify the contribution of this migration process.

6.3.6.5 Pathway 5: Flow through an open annular region due to inadequate cement placement

Inadequate cement placement depends on quality assessment and is more likely to be present in older wells. It may be a result of incomplete casing coverage. The SACROC project does not provide further insights in this matter. Viswanathan et al. (2008) propose to apply PDFs exhibiting a bimodal distribution, covering older and newer wells. For simplicity, the wells in the study have been assumed to be properly cemented. Thus, no PDFs have been created.

6.3.6.6 Pathway 6: Flow at the cement-cap rock interface

Formation of a porous interface between wellbore cement and cap rock is probable. A tight bond between these two compartments can be impeded by mechanical tension, for instance created by shale swelling. A further circumstance provoking a porous interface is ‘wall cake’, i.e. non-removed drilling mud and cap rock debris. Fluid migration through the interface is governed by the effective aperture and is related to effective permeability by the cubic equation mentioned above. Leakage through the cement-cap rock interface is considered more probable for older wells. Migrating CO2 may cause self sealing to a certain extent at moderate flows. In CO2-PENS, a PDF referring to typical cement permeability has been applied.

The limited data available complicates PDF-creation for interface apertures. For CO2-PENS a standard aperture of 3 mm accompanied by a standard deviation of 2 mm has been applied. The CO2-PENS wellbore release module is capable of predicting CO2 release based on the given wellbore cement effective permeability. Simulations can be conducted in three ways:

  1. User specified distribution of leakage ratesIf the user has obtained information on leakage by other means than CO2-PENS they can be used as input for further simulations.
  2. Finite Element Heat and Mass Transfer (FEHM)The FEHM-model embedded in CO2-PENS represents a multidimensional multiphase reservoir simulator. This tool supports the user in estimating CO2 leakage rates relying on detailed numerical simulation of leakage. Application consumes considerable computation time. However, if cement-associated leakage is accompanied by other leakage pathways, application of the FEHM model is required.
  3. Calculation of leakage rates by a semi analytical modelWith respect to semi analytical modelling, a model developed by Nordbotten et al. (2009), referred to as the ‘Princeton Model’, has been embedded in CO2-PENS. The advantage lies in the moderate computation time required for calculating reliable leakage rate estimations.

Table 6.5. Key Parameters of the synthetic reservoir used for injection and leakage simulation (according to Viswanathan et al., 2008).

Depth to bottom of sequestration reservoir 3 km
Pressure in sequestration reservoir 30 MPa
Temperature in sequestration reservoir 155 °C
Max injection pressure 45 MPa
Injection duration 50 years
Injection rate 50 kg/s
Simulation duration 50 years
Number of Monte Carlo realizations 1000
Mean of permeable layers porosity (normal distribution) 0.14
Standard deviation of permeable layers porosity (normal distribution) 0.03
Mean of effective aperture in cement (normal distribution) 3 mm
Standard deviation of effective aperture in cement (normal distribution) 2 mm

The case study, based on the synthetic reservoir, encompasses a sequestration target reservoir, impermeable and permeable layers in the saturated zone, the vadose zone and the land surface (Figure 6.9). The scenario further consists of one injection well. Leakage out of the storage reservoir is assumed to occur through eight plugged and abandoned wells. Additionally, 10 shallow wells have been added to the scenario not intersecting the storage reservoir. CO2-PENS relies on numerous data inputs, which may be available from various data base systems. For this purpose tools have been implemented allowing import, selection, pre-processing, and usage of imported data at the system level.

Reservoir permeability and porosity have been assumed similar to the SACROC site. In absence of direct permeability measurements, permeability has been calculated from porosity. The implemented GIS tools have been used to extract necessary information from the SACROC site in order to build a spatial data base.

Figure 6.9. (a) Hypothetical reservoir cross section and (b) location of existing deep wellbores and shallow groundwater wells (top view), adopted from Viswanathan (2008).

The key input parameters for the Princeton model are revealed in Table 6.5 Within the area of interest, the individual porosity values were generalized and characterized by standard deviations. For the porosity-PDF, a normal distribution was assumed while permeability was assumed to have a log-normal distribution. The cement fracturing has been conservatively assumed to be considerable. After setting up the permeabilityPDFs, 1000 Monte Carlo Simulation runs have been performed to simulate wellbore leakage over a period of 50 years.

The results in terms of leakage rate as a function of aquifer layer and time are depicted in Figure 6.10. Total leakage was quantified as high as 6 × 104 kg CO2 over 50 years. Compared to the simulated injection of 8 × 1010 kg of CO2 the leakage rate is way below the committed benchmark of 0.01% a-1. As expected, leakage to the topmost layer reveals the lowest quantities due to retention by the intermediate layers. Seepage to the surface is additionally moderated by the top permeable layer and the vadose zone. The development of a module enabling comprehensive simulation of the vadose zone is under way (Viswanathan et al., 2008).

Figure 6.10. Average (dotted line), standard deviation (green) of the accumulation rates of CO2 in the storage reservoir (a) and in the two shallower reservoirs (b, c) as functions of time. These results are based on 1000 Monte Carlo simulations and cover a simulated period of 50 years (adopted from Viswanathan et al., 2008).

One significant feature of the leakage model is the capability to calculate the spatial extension of a plume, generated in any CO2 containing aquifer. This information is subsequently processed by CO2-PENS to evaluate intersection of CO2 plumes by groundwater wells. The simulation study on the synthetic reservoir revealed that up to 3 out of the 10 shallow wells are impacted by CO2 after 50 years. The radii of the modelled plumes ranged between 1 and 20 m. Such information can be subsequently processed to evaluate the hydrochemical impact on aquifers. Recently, the geochemical model PHREEQC has been linked to CO2-PENS for such purposes. However, this was not considered in the investigation.

Figure 6.11. The average (dashed line) and standard deviation (green) increase in predicted atmospheric CO2 concentration due to wellbore leakage obtained from 100 Monte Carlo simulations. CO2 concentrations include contributions from both, the biosphere and a diffuse leak (adopted from Viswanathan et al. (2008).

Due to the modular architecture of CO2-PENS, coupling of the well leakage module with the atmospheric model was feasible. To achieve this, a fraction of the leakage rate calculated by the well leakage module was passed to the atmospheric mixing module for every time step. This was constrained by the Princeton model which is not capable of calculating CO2 migration to the vadose zone, where the fluid is subject to phase transition. Therefore it was assumed that 10% of the leakage rate into the top permeable layer would be subject to seepage to the atmosphere, taking the considerable retention potential of the vadose zone into account. At each time step in the system model, the wellbore module is queried to predict the leakage rate into the top aquifer. This leakage rate is passed back to the system model, which subsequently passes a fraction of 10% of this leakage rate to the atmospheric mixing module. The atmospheric model then processes the seeped CO2 masses with highly resolved time steps, taking the diurnal near-surface CO2 concentration variations into account. Once the envisaged vadose zone module is established, estimations can be substituted by model simulations. The simulated atmospheric CO2 concentrations including the seasonal variations are depicted in Figure 6.11.

The modular structure of CO2-PENS holds the potential for a comprehensive risk assessment tool with respect to CO2 storage. It can be used as a screening tool as well as for performance and risk assessment of individual sites, when site-specific information becomes available.