Download or read book Multi-Sensor Data Assimilation for Geological Carbon Storage Monitoring Design written by Shams Joon and published by . This book was released on 2022 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: Geological carbon storage (GCS) is a climate change mitigation strategy that provides an innovative solution to offset the rising atmospheric CO2 concentrations. This process involves the injection of CO2 into underground geological formations where it is permanently trapped, thereby avoiding CO2 to be emitted into the atmosphere. The tax credit for CO2 sequestration (IRC Code: 45Q) has incentivized the feasibility of such operations and GCS is gaining substantial investment interest. The potential for CO2 to leak out and negatively impact the overlying environment is a primary concern for such operations and has motivated the development of risk-based monitoring, verification, and accounting (MVA) protocols around the world for Class II and Class VI wells. Fluid flow models are effective tools to simulate complex physical processes such as CO2 sequestration at a storage site. The accuracy of these models relies on multiple model parameters and state variables that are calibrated to reproduce the changing reservoir state. Geophysical monitoring data from multiple sources are used to further calibrate reservoir simulations and improve model accuracy. However, both the reservoir model and geophysical measurements produce uncertain predictions due to the underlying process and measurement errors. Monitoring tools can be evaluated based on their sensitivity, spatiotemporal coverage, cost, and regulatory requirements. Wellbore sensors, such as pressure gauges, provide high temporal sampling of the subsurface but are spatially limited to around the wellbore. In contrast, surface seismics can survey large volumes of the reservoir with a coarse spatial resolution and are costly which limits how frequently they can be conducted. Furthermore, using these types of geophysical monitoring tools to estimate changes in petrophysical properties is always subject to uncertainty due to inevitable ambiguities incurred during data acquisition, processing, and interpretation. Combining multiple sources of measurements can help reduce prediction uncertainty; however, quantifying the improvement afforded by such composite systems can be a challenging task when the true reservoir characteristics are unknown. Quantifying the reduction in prediction error from different monitoring tools and combinations of monitoring tools can also be useful to evaluate the efficacy of a proposed monitoring design. From a monitoring design perspective, this research validates the applicability of combining seismic attributes derived from full-waveform inversion of continuous active-source seismic monitoring (CASSM) data with pressure-based monitoring measurements to improve model state predictions. The improvement afforded by combining these two different types of measurements is quantified by computing the reduction in prediction error in an ensemble-based data assimilation environment. The first goal of this research is to develop and test out an ensemble-based data assimilation framework that takes advantage of rock physics models and combines numerical simulations with geophysical observations to predict subsurface changes at GCS sites. This proposed joint seismic-pressure-petrophysical data assimilation framework uses continuous geophysical measurements, in the form of seismic velocity (Vp) and seismic attenuation quality factor (Qp) along with wellbore pressure monitoring data (Pwf), to predict changes in the reservoir model state which is represented by CO2 saturation and reservoir pressure distributions. One of the challenges of using seismic data is the non-unique relationship between CO2 fluid properties and seismic attributes which introduces ambiguity (multiple solutions) during inversion. Rock physics models can be used to forward model seismic attributes but due to the highly non-linear nature of these models and the multidimensionality of reservoir rock and fluid properties, standard linear models are rendered unusable for inversion purposes. Combining different types of measurements (seismic with pressure) helps further constrain this non-uniqueness and improves the forward-modeled estimates. These multi-sensor measurements are assimilated using an ensemble Kalman filter (EnKF) which propagates the model state and uncertainty forward using an ensemble of reservoir realization and relies on ensemble-based sample statistics of the model state and measurement error to calibrate estimates when new measurements are made available. One of the novelties of this workflow is that the forward operator of the EnKF is replaced with rock physics models (RPMs). The choice of rock physics model depends on the geological context, the rock and fluid properties, operational parameters of the seismic survey, and available seismic attributes. I use one particular RPM i.e., White's patchy gas saturation model that we use for demonstration purposes, but one could use this general framework to employ any one of a variety of RPMs. I conduct a series of observation system simulation experiments (OSSEs) to demonstrate the effectiveness of this joint data assimilation framework by evaluating different monitoring tools and combination of monitoring tools on three different models. The OSSEs are first conducted on a lab-scale "sandbox" model before being tested on field-scale reservoir models like the Frio II brine pilot, near Houston, Texas and the Cranfield Site in Mississippi. In general, including seismic attributes improves the prediction estimate of CO2 saturation while Pwf measurements improve pressure prediction results by calibrating the well constraints and improving model state forecasts. Jointly assimilating both seismic and pressure data produces the greatest reduction in prediction error and the high temporal resolution afforded by continuous seismic measurements allows for shorter assimilation windows. Reducing the assimilation frequency increases the prediction error which is observed when CO2 injection is halted and the post-injection assimilation time window is increased. This improvement afforded by jointly assimilating multi-sensor observations is consistently observed in all three synthetic case studies even when different data assimilation parameters are varied such as type, ensemble size, assimilation frequency etc. After successfully implementing the multi-sensor, rock physics-based data assimilation framework in an OSSE environment, I integrate the framework with full-waveform inversion (FWI) results from the CASSM dataset at Frio II. In this work, the CASSM-derived FWI seismic attributes and wellbore pressure monitoring data are jointly assimilated to predict CO2 plume movement and reservoir pressure changes over a 5-day injection period. A comprehensive comparison of using a multi-sensor approach as compared to just wellbore pressure sensors is carried out to conclude that the error reduction afforded by using multiple sensors is valuable both from a perspective of risk as well as cost. Lastly, the multi-sensor, rock physics-based data assimilation framework is reconfigured for additional operational applications at GCS sites like observation targeting. In particular, this modified workflow takes advantage of ensemble-based sensitivity analysis to evaluate how changing the placement location of monitoring wells influences the prediction uncertainty of model state variables. Furthermore, by evaluating the efficacy of pre-existing and/or limited monitoring tools and designs, one can identify regions of the reservoir with highest uncertainty and subsequently find optimal locations for drilling new monitoring wells. A series of OSSEs of the Frio II reservoir model are used to demonstrate the applicability of this observation targeting approach.