3

Conclusions

The goal of this project to predict the combustion efficiency of wellhead flames was ambitious, and the Naval Research Laboratory (NRL) report shows that both modeling and experimental frameworks have been successfully initiated and are ongoing. Without a well-defined problem, however, it is difficult to define all of the submodeling components and their required complexity so as to meet the goals expressed by the sponsors of the work. Furthermore, it will be difficult to improve the modeling and experimental programs without more concise information and consideration of the variability of the physical and chemical properties of crude oil that appear to be of interest in terms of future predictive uncertainties and the foreseen utility of the NRL model.

The committee identified key concerns regarding the NRL modeling approach and experimental methods in three categories:

1. Gaps in the study approach and the assumptions chosen to represent the physical system of wellhead combustion limit the utility and accuracy of the approach and the model.
2. Several modeling approaches employed are not the state of the art.
3. Other modeling methods employed are the state of the art, but their related uncertainties and known weaknesses are not considered.

The objective of predicting the combustion efficiency of wellhead fires was not well scoped in the NRL study. The envelope of multiphase flow conditions and the physical and chemical properties of crude oils were not adequately accounted for in terms of submodel property considerations either in developing the model or in choosing fuels to be used in bench-scale experiments. The feedback of results of experimental efforts into the development of submodel component needs, as well as the validation of model predictions, was very limited. Hence, the relevance of the selected laboratory- and bench-scale experiments and computational fluid dynamics (CFD) simulations to actual field conditions remains highly uncertain. While the authors of the NRL report made reasonable a posteriori comparisons between the Reynolds-averaged Navier-Stokes (RANS) CFD and the bench-scale experiment, many of the conventional submodels used and their assumptions were not validated for the conditions at hand, and sensitivities and uncertainties of key quantities of interest to the tunable constants of the submodels and boundary conditions were absent. A more systematic verification and validation approach would have been beneficial and would have instilled trust in the predictive nature and level of uncertainty of the CFD approach to conditions outside of the bench-scale experiment. Future studies would benefit from a community survey (involving, e.g., experts and stakeholders) bounding the relevant conditions of wellhead fires, and from the selection of a set of hierarchical unit problems addressing specific aspects of this complex problem.

Regarding the utility of the approach and assumptions applied in developing the model and its components, the wellhead system is not well defined in the NRL interim report, and the level of accuracy required or desired for the model is never identified. For example, what is considered sufficient for the prediction of combustion efficiency—would an order of magnitude suffice? These are fundamental concerns that dictate which approaches for the modeling and experimental work are acceptable—e.g., whether RANS or Large Eddy Simulation (LES) modeling methods are appropriate or how crude properties are to be considered. Another key concern is the lack of well-defined initial and boundary conditions in the context of wellhead combustion.

Designing appropriate experiments with which to validate the model or help scale the results to actual wellhead combustion conditions is difficult without a well-defined problem. Without a well-defined problem, moreover, it is not possible to evaluate the adequacy of the model.

Other high-level technical findings as to the completeness of the modeling results for predicting wellhead oil-burning efficiency identified by the committee are as follows:

1. It is unclear whether the authors considered the correct flow system.
   1. Are the authors considering the correct configuration for the multiphase flow? Specifically, is co-annular two-phase flow appropriate for representing wellhead oil flow? Can that configuration be justified, e.g., for the range of mass flow, pipe dimensions, fluid properties, and so on? The boundary and initial conditions require justification, and dimensionless parameters (e.g., Re, We) need to be applied to define the system and to leverage work in the literature where appropriate.
      1. The two-phase wellbore flow modeling focused on film thickness, entrainment, and droplet diameter, but the role of these parameters in combustion efficiency was not established, and it is not clear that these parameters are sufficient to describe combustion efficiency (quantitatively or qualitatively).
      2. The authors provide a reasonable review of the literature on correlations for film thickness, liquid entrainment, and droplet diameters. However, it is unclear whether the correlations are valid for the regimes under consideration. Additionally, while the correlations for Weber numbers may be valid for the bench-scale simulations, it is unclear whether they are applicable for the actual wellhead.
   2. What are the thermophysical and chemical properties of the crude oil? How are those properties captured (or not captured) by the simpler fluids used in the study?
      1. Surrogates may be a good approach for considering certain aspects of the system; however, the current work emphasized lighter hydrocarbon components that will likely significantly affect the combustion efficiency. Specifically, the soot characteristics for crude oil are not encompassed by n-heptane, and sooting propensity will affect several critical physical and chemical transport mechanisms in the model, including the radiative energy balance. In addition, preferential evaporation of lighter components in crude oil may induce composition and thermal stratification in the mixture, affecting combustion rates, which are not captured by heptane, and CFD mixing and combustion models need to account for stratification.
      2. Few assumptions regarding the selection of properties are specified in the interim report, and the modeling and experimental work did not validate the properties used.
   3. Wellhead conditions were applied based on results from a worst-case discharge (WCD) model by Hilcorp, details of which are either not adequate or not provided. Annular-mist flow regime was considered based on the WCD results and downscaled to apply to the problem at hand. Wellbore modeling methods and input details could be provided to verify output, capture uncertainties, and cross-validate WCD output among different wellbore flow modeling methods, and to improve the assumptions and approach of the modeling and experimental efforts. An independent WCD model could be developed using data from the Liberty (Hilcorp report) and analog reservoirs (for which the Hilcorp report is insufficient). Doing so would help characterize the wellhead conditions and flow parameters and provide confidence in the input/boundary conditions used for the study.
      1. The impact of nonannular flow regimes needs to be considered or justified for exclusion in modeling and experimental efforts.
      2. Importantly, the authors assume annular-mist flow behavior for the sake of brevity and applicability, as these sprays may atomize well. However, the pools or fountains emerging from lower-speed flows may not burn well, as evidenced by the experimental results. Thus, the modeling may not consider the “worst-case” conditions

for combustion efficiency (i.e., conditions in which significant oil droplets drop out of the flow).

2. Naturally imposed external flows and induced flows were not considered. Specifically, cross-flow was not considered in the modeling or the experiments, and Arctic wind speeds are extremely high (average of 5.5 m/s between September and May on the North Slope, up to 30 m/s during polar lows over the Arctic Ocean, according to the National Oceanic and Atmospheric Administration [NOAA]¹). Buoyancy effects also were not considered, and may be comparable to or exceed the effects of wind cross-flow. These effects may aid or impair burning rates. Additionally, the imposed external flows will lead to significant multidimensional behavior, and the flame/plume evolutions are not well-represented by axisymmetric assumptions. In terms of induced flows, the forced flow of the wellhead fluid at the exit of the pipe will induce external flows that the authors did not consider.
3. The verification and validation processes were not rigorous. For example, the experiments were not designed to validate any of the subprocess models (i.e., the turbulence, turbulence/combustion interactions, combustion chemistry, droplets, radiation, and soot submodels). The sensitivity of the submodels and experiments to boundary and initial conditions was not considered and could provide extremely valuable information to guide future work. There were some opportunities to validate portions of the model with some of the experimental data (e.g., droplet behavior) that were not explored in depth.

The consensus conclusion of the committee is that the model is not adequate for predicting the combustion efficiency of wellhead flames. A broad-based research program may be appropriate to address the complex challenges of wellhead combustion. To this end, identifying better unit problems to frame such a research program will require more substantive understanding of the underlying conditions of wellhead combustion, as well as the goals for the stakeholders of such work.

___________________

¹ National Snow & Ice Data Center. 2020. Patterns in Arctic Weather. https://nsidc.org/cryosphere/arctic-meteorology/weather_climate_patterns.html#:~:text=Wind%20speeds%20average%20around%2050,over%20relatively%20warm%20open%20water.