Quantitative assessment of monitoring strategies for conformance verification of CO₂ storage projects

Highlights

• Quantitative model-based workflow for conformance verification of CO₂ storage projects.

• Description of key elements for quantifiable conformance notions and computational workflows.

• Demonstration of proposed workflow for a priori assessment of candidate monitoring strategies.

• Discussion of potential research directions for (quasi) real-time conformance monitoring.

Abstract

We propose a quantitative model-based workflow for conformance verification of CO₂ storage projects. Bayesian inference is applied to update an ensemble of simulation models that capture prior uncertainty based on mismatches with measured data. Conformance assessments are derived by comparison of updated model predictions with storage permit requirements and confidence criteria. Two examples, one conceptual and one based on a real candidate storage site, are provided in which the quantitative workflow is applied to the a priori assessment of candidate monitoring strategies. The examples illustrate the limitations of pressure monitoring in the presence of realistic subsurface uncertainties, and the potential for cost saving by informed design of geophysical monitoring surveys. Approximate methods are discussed that could make the workflow also applicable for (quasi) real-time conformance monitoring.

Introduction

The main storage-related challenges for deployment of large-scale Carbon Capture and Storage (CCS) are capacity, confidence and cost. There must be high certainty that the site has the capacity to permanently hold the volumes that are planned to be injected, and that it can be operated safely and economically. These elements are normally addressed in plans that operators submit to the regulating authority before approval to operate a storage site is granted. In Europe, the basis for assessment of these plans is provided by the European Commission (EC) CCS Directive (EC, 2009). It states that a monitoring plan is required to enable comparison between actual and modelled behavior, detection of irregularities, detection of the migration of CO₂, and the assessment of the effectiveness of corrective measures in case of leakages or significant irregularities. It furthermore states that reports need to be submitted, at a frequency determined by the competent authority, that contain all information relevant for assessment of compliance with storage permit conditions, and for ‘increasing the knowledge of CO₂ behavior in the storage site’.

The specific requirements for safe storage operations are identified in a storage permit. A general requirement is that injected CO₂ remains within the storage complex permanently. The storage permit will also specify maximum injection rates and pressures, as well as the maximum allowed reservoir pressure. Additional requirements may result from site-specific risks and could therefore differ from one project to another. Key concepts in the monitoring of these risks are captured by the notions of containment and conformance. Containment refers to the basic requirement that CO₂ must remain permanently within the storage complex or within clearly identified boundaries inside that storage complex. According to the EU CCS Directive, conformance (or conformity) refers to consistency of the actual behavior of the injected CO₂ with the modelled behavior. Conformance is a requirement for the transfer of responsibility after site closure, together with the absence of detectable leakage, and a demonstration that the site is evolving towards a situation of long-term stability. A slightly broader definition of conformance, that we will adopt here, also includes compliance with any additional requirements as specified in the storage permit. This definition is in agreement with that used by the government of Alberta, Canada, in its the regulatory framework for CCS, and combines the concepts of concordance (agreement between models and data) and performance (agreement with permit requirements) as discussed by Oldenburg (2018). We will define conformance verification to mean the activity aimed at establishing if a situation of conformance exists at any moment during operation of the site and will remain to exist in the future. Since a storage permit will not be granted if initial modelling suggests that conditions qualifying as non-conformance are likely to develop, it may be assumed that initial models prepared by the operator will suggest that the site can be operated in accordance with storage permit requirements.

The actual behavior of the CO₂ after injection in the reservoir will be strongly controlled by conditions (fluid pressure, temperature and composition, stress field) and rock properties (porosity, permeability, connectivity, fault stability) inside the storage reservoir, which are site specific. While these properties may be approximately known at the locations of wells that have been drilled into the reservoir before the start of injection, they will generally be poorly known everywhere else. As a consequence, also the dynamic behavior of the CO₂ will be uncertain. This will also be the case to some extent for storage in depleted gas reservoirs for which information obtained from monitoring data is generally limited to storage volume dimensions and pressure. Some of the uncertainties can be associated with risks of non-conformance situations, such as poor injectivity (preventing injection of the planned volumes), leakage through old wellbores or through the overburden, or migration of CO₂ outside of the licensed storage area (all leading to non-containment), and high seismicity (violating safe operation standards). Some of these risks may be more relevant in some sites than in others.

Monitoring programs should facilitate the identification of irregularities that could point at non-conformance situations and trigger mitigating actions. Information extracted from monitoring data will generally be uncertain due to measurement or interpretation errors and the sparsity or limited resolution of data, requiring the use of (also uncertain) models to fill the gaps (Harp et al., 2019). Conformance verification activities involve the comparison between modelled behavior of the CO₂ and its ‘actual’ behavior, which must be inferred from monitoring data, and they must therefore take this uncertainty into account.

No clear guidelines or frameworks currently exist for identifying monitoring tasks and setting monitoring performance requirements in a conformance verification context (Bourne et al., 2014). However, experience with monitoring of industrial-scale CO₂ storage operations has been gathered in a number of projects that include Sleipner (e.g. Arts et al., 2008; Furre et al., 2017), Snøvhit (Maldal and Tappel, 2004), In-Salah (Mathieson et al., 2011), Otway (Jenkins et al., 2015) and Quest (Bourne et al., 2014). Decisions about which monitoring technologies to adopt to address identified risks in these projects have generally been based on site-specific monitoring technology feasibility studies that employ qualitative or scenario-based approaches for risk assessment involving collective expert judgement (Bourne et al., 2014). A method often used in support of such approaches is the bowtie method which aims to map and rank threats, consequences, and safeguards. Safeguards include monitoring that can detect irregularities, and a decision logic to interpret the monitoring data and suggest control measures. The possible monitoring technologies are ranked by experts, and an ultimate selection is made by balancing benefits against costs. Applications of such a risk-based framework for developing site-specific measurement, monitoring and verification plans were presented by Bourne et al. (2014) for Quest, by Dean and Tucker (2017) for the Peterhead CCS project (Spence et al., 2014) and by Metcalfe et al. (2017) for the White Rose project. This framework aims to produce a systematic risk assessment and to establish monitoring performance targets that will reduce storage risks to a desired level. As part of this approach extensive modelling based on variation of identified uncertain model aspects was performed to assess the risk of CO₂ migration beyond spill points or storage site boundaries. Active seismic surveys were planned to establish a baseline for repeat surveys and detect lateral movement of the CO₂, and to identify changes in the formations above the storage formation. Revision of the monitoring plans would be necessary in the case of unexpected plume shape or migration velocity, injection pressures, or other deviations from the modelled behavior. Mismatches between dynamic models and monitoring data could then lead to model updates, additional monitoring or corrective measures. It may not always be clear a priori what the character of mismatches would need to be in order to trigger one or more of these actions. The current practice appears to be to conduct a qualitative assessment of such mismatches, leading to an expert-based judgement of required actions.

Quantitative assessment of risks, and the design of monitoring strategies that address them, is often considered complex and computationally demanding since it requires extensive evaluation of mathematical models that capture the full range of possible behavior of the system, identification of the ranges of possible values of all parameters that influence system behavior, and a derivation of risk under uncertainty. However, there are important benefits to such an approach. Perhaps the most important are the possibility to properly account for all uncertainties and their interdependencies, both in models and data (without the need to rank them based on extensive sensitivity experiments), and the possibility to assess the contribution of different monitoring technologies in quantitative terms.

Several examples of quantitative workflows related to risk management in CO₂ storage operations can be found in the literature (see Section 2). This includes applications of optimization approaches to above-zone monitoring well placement aimed at minimizing the time to leak detection (Sun et al., 2013; Cameron, 2013; Yonkofski et al., 2016, 2017; Jeong et al., 2018) and workflows for leakage location estimation and uncertainty reduction based on monitoring data (Sun and Nicot, 2012; Jung et al., 2013; Jung et al., 2015; Hu et al., 2015; Chen et al., 2018; Chen et al., 2020). Some of these methods were also applied to the FutureGen 2.0 candidate storage site (Vermeul et al., 2016; Bacon et al., 2019) illustrating the feasibility of the application of quantitative life-cycle management workflows to realistic settings.

Chadwick and Noy (2015) considered modeling-monitoring convergence as an indicator of the possibility to demonstrate conformance. This is based on their observation that model predictions tend to become more reliable when models are in sufficient agreement with historic data. Following this idea, Harp et al. (2017) proposed an approach for identifying robust pressure management strategies based on decision gap theory. Rather than representing the uncertainty probabilistically, the methodology aims to quantify the degree that one can be incorrect in the characterization of the system and still ensure that performance (conformance) criteria are met. This could be viewed as a metric for comparing alternative monitoring strategies. This approach was adopted by Harp et al. (2019) in a follow-up study that demonstrated an application based on pressure monitoring at a single well where a conformance criterion was defined in terms of the pressure at the well. The workflow uses derivatives of observables with respect to uncertain model parameters which will not be generally available for more complex simulators and measurements or for conformance criteria that are not directly observable. The approach appears to require that a critical value for the conformance criterion is not exceeded in any of the model realizations, which seems quite restrictive. Also, it is not entirely clear how the approach should be extended to applications with multiple conformance criteria and large numbers of uncertain parameters, such as heterogeneous permeability fields.

As an alternative to the workflow of Harp et al. (2019), we develop and demonstrate a generic conformance analysis approach that can be applied to cases with arbitrary numbers and types of uncertainties and arbitrary combinations of monitoring techniques. The approach is inspired by ensemble-based workflows for quantitative evaluation of monitoring strategies for oil field management (e.g. Le and Reynolds, 2013; 2014; He et al., 2020, Barros et al., 2016; Barros, 2018). These studies have suggested that such approaches are indeed capable of providing useful comparisons and rankings of alternative strategies under realistically complex uncertainty scenarios. Here we will therefore propose a quantitative ensemble-based workflow that can be applied to objectively assess the current and future conformance state of the storage system. The workflow can also be used on top of practical risk identification and mitigation frameworks as currently used, to produce objective assessments of the validity of simulation models, the effectiveness of measurement acquisition strategies, and ultimately support decisions about contingency monitoring or corrective measures. We demonstrate an application of the workflow prior to the start of injection with the aim of evaluating possible monitoring strategies in terms of their capability to correctly assess the conformance of the site at any time in the future.

The remainder of this paper is organized as follows. In the next section we review the literature on quantitative workflows for evaluation of monitoring designs in subsurface applications. In Section 3 we define terminology related to conformance that will be used in this document. In Section 4 we introduce our proposed quantitative workflow for conformance assessment. Section 5 provides a relatively simple example application of the workflow to illustrate the main steps as well as an application to a complex case based on a real potential storage site. In Section 6 we discuss possible extensions and alternative applications of the workflow that could be useful in support of CO₂ storage operations. Finally, the main conclusions are provided in Section 7.