Review of flow rate estimates of the Deepwater Horizon oil spill.

The unprecedented nature of the Deepwater Horizon oil spill required the application of research methods to estimate the rate at which oil was escaping from the well in the deep sea, its disposition after it entered the ocean, and total reservoir depletion. Here, we review what advances were made in scientific understanding of quantification of flow rates during deep sea oil well blowouts. We assess the degree to which a consensus was reached on the flow rate of the well by comparing in situ observations of the leaking well with a time-dependent flow rate model derived from pressure readings taken after the Macondo well was shut in for the well integrity test. Model simulations also proved valuable for predicting the effect of partial deployment of the blowout preventer rams on flow rate. Taken together, the scientific analyses support flow rates in the range of ∼50,000-70,000 barrels/d, perhaps modestly decreasing over the duration of the oil spill, for a total release of ∼5.0 million barrels of oil, not accounting for BP's collection effort. By quantifying the amount of oil at different locations (wellhead, ocean surface, and atmosphere), we conclude that just over 2 million barrels of oil (after accounting for containment) and all of the released methane remained in the deep sea. By better understanding the fate of the hydrocarbons, the total discharge can be partitioned into separate components that pose threats to deep sea vs. coastal ecosystems, allowing responders in future events to scale their actions accordingly.

Strontium isotopes test long-term zonal isolation of injected and Marcellus formation water after hydraulic fracturing.

One concern regarding unconventional hydrocarbon production from organic-rich shale is that hydraulic fracture stimulation could create pathways that allow injected fluids and deep brines from the target formation or adjacent units to migrate upward into shallow drinking water aquifers. This study presents Sr isotope and geochemical data from a well-constrained site in Greene County, Pennsylvania, in which samples were collected before and after hydraulic fracturing of the Middle Devonian Marcellus Shale. Results spanning a 15-month period indicated no significant migration of Marcellus-derived fluids into Upper Devonian/Lower Mississippian units located 900-1200 m above the lateral Marcellus boreholes or into groundwater sampled at a spring near the site. Monitoring the Sr isotope ratio of water from legacy oil and gas wells or drinking water wells can provide a sensitive early warning of upward brine migration for many years after well stimulation.

Molecular dynamics simulations of turbostratic dry and hydrated montmorillonite with intercalated carbon dioxide.

Molecular dynamics simulations using classical force fields were carried out to study energetic and structural properties of rotationally disordered clay mineral-water-CO2 systems at pressure and temperature relevant to geological carbon storage. The simulations show that turbostratic stacking of hydrated Na- and Ca-montmorillonite and hydrated montmorillonite with intercalated carbon dioxide is an energetically demanding process accompanied by an increase in the interlayer spacing. On the other hand, rotational disordering of dry or nearly dry smectite systems can be energetically favorable. The distributions of interlayer species are calculated as a function of the rotational angle between adjacent clay layers.

A machine learning framework for rapid forecasting and history matching in unconventional reservoirs.

We present a novel workflow for forecasting production in unconventional reservoirs using reduced-order models and machine-learning. Our physics-informed machine-learning workflow addresses the challenges to real-time reservoir management in unconventionals, namely the lack of data (i.e., the time-frame for which the wells have been producing), and the significant computational expense of high-fidelity modeling. We do this by applying the machine-learning paradigm of transfer learning, where we combine fast, but less accurate reduced-order models with slow, but accurate high-fidelity models. We use the Patzek model (Proc Natl Acad Sci 11:19731-19736, https://doi.org/10.1073/pnas.1313380110 , 2013) as the reduced-order model to generate synthetic production data and supplement this data with synthetic production data obtained from high-fidelity discrete fracture network simulations of the site of interest. Our results demonstrate that training with low-fidelity models is not sufficient for accurate forecasting, but transfer learning is able to augment the knowledge and perform well once trained with the small set of results from the high-fidelity model. Such a physics-informed machine-learning (PIML) workflow, grounded in physics, is a viable candidate for real-time history matching and production forecasting in a fractured shale gas reservoir.

A system model for geologic sequestration of carbon dioxide.

In this paperwe describe CO2-PENS, a comprehensive system-level computational model for performance assessment of geologic sequestration of CO2. CO2-PENS is designed to perform probabilistic simulations of CO2 capture, transport, and injection in different geologic reservoirs. Additionally, the long-term fate of CO2 injected in geologic formations, including possible migration out of the target reservoir, is simulated. The simulations sample from probability distributions for each uncertain parameter, leading to estimates of global uncertainty that accumulate through coupling of processes as the simulation time advances. Each underlying process in the system-level model is built as a module that can be modified as the simulation tool evolves toward more complex problems. This approach is essential in coupling processes that are governed by different sets of equations operating at different time-scales. We first explain the basic formulation of the system level model, briefly discuss the suite of process-level modules that are linked to the system level, and finally give an in-depth example that describes the system level coupling between an injection module and an economic module. The example shows how physics-based calculations of the number of wells required to inject a given amount of CO2 and estimates of plume size can impact long-term sequestration costs.

Development of a hybrid process and system model for the assessment of wellbore leakage at a geologic CO2 sequestration site.

Sequestration of CO2 in geologic reservoirs is one of the promising technologies currently being explored to mitigate anthropogenic CO2 emissions. Large-scale deployment of geologic sequestration will require seals with a cumulative area amounting to hundreds of square kilometers per year and will require a large number of sequestration sites. We are developing a system-level model, CO2-PENS, that will predict the overall performance of sequestration systems while taking into account various processes associated with different parts of a sequestration operation, from the power plant to sequestration reservoirs to the accessible environment. The adaptability of CO2-PENS promotes application to a wide variety of sites, and its level of complexity can be increased as detailed site information becomes available. The model CO2-PENS utilizes a science-based-prediction approach by integrating information from process-level laboratory experiments, field experiments/observations, and process-level numerical modeling. The use of coupled process models in the system model of CO2-PENS provides insights into the emergent behavior of aggregate processes that could not be obtained by using individual process models. We illustrate the utility of the concept by incorporating geologic and wellbore data into a synthetic, depleted oil reservoir. In this sequestration scenario, we assess the fate of CO2 via wellbore release and resulting impacts of CO2 to a shallow aquifer and release to the atmosphere.