Hydraulic fracturing is a typical and vital technique applied in shale gas reservoir development. Numerical simulation used to be a common tool to optimize the parameters in hydraulic fracturing design determining the stage numbers, injection pressure, proppant amount, etc. However, the current understanding of shale gas storage and transport mechanism (e.g. adsorption/desorption, diffusion) is basically adopted from the lessons learned from coal seams through past experience, which might not help an efficient numerical simulation development.
In this study, how artificial intelligence assisted data driven models assist the hydraulic fracturing design in shale gas reservoir is discussed. It starts by collecting field data and generate a spatial-temporal database including reservoir characteristics, operational/production information, completion/stimulation data and other variables, Neural Network models are then developed to study the impacts of all parameters on gas production as well as perform history matching of the field history. The AI assisted model with acceptable matching of field data can be used to model different hydraulic fracturing design scenarios and provide predictions on well production.
Application of horizontal drilling and hydraulic fracturing technique has made development of shale gas reservoir successful in the United States during the past decade. Chasing its operational success, researchers have been studying to understand the fundamentals of shale gas production, which will provide valuable information to assist in optimization of shale reservoir development. Unfortunately, the mechanism of shale gas production has not been fully revealed so far, and most reservoir simulation models are adopting the mechanism of coalbed methane production to forecast shale gas development process, which might not be the real case.
In this paper, instead of using numerical simulation model, artificial intelligence and data mining techniques are implemented to study the controlling factors of shale gas production and understand the impacts of reservoir, completion and stimulation parameters in a dynamic manner only according to the field data. A database of Marcellus shale reservoir is generated by integrating information such as well locations, well trajectories, reservoir characteristics, completion, hydraulic fracturing, and production parameters, etc. Neural network models are trained to learn the key performance impacting factors on shale gas production in a dynamic manner, which could assist reservoir management decisions.
The solid skeleton of the mudcake consists of fine-grain particles; therefore, a mudcake plug is expected to have a very low permeability and a very good ability to isolate the fracture from wellbore pressure. This requires a relatively permeable formation for two reasons: Mudcake buildup requires fluid loss into the formation, and fracture pressure needs to dissipate after being isolated from the wellbore (Kumar et al. 2010).
Application of horizontal wells and multi-stage fracturing has enabled oil recovery from extremely low permeability shale oil reservoirs, but the decline in production rate is more than two thirds in the first two years. We are trying to develop chemicals that can be injected into old wells to stimulate oil production before putting the well back in production. The goal of this work is to evaluate chemical blends for such a process at the laboratory scale. The chemical blend contains surfactants, a weak acid, a potential determining ion, and a solvent. Six different solvents were screened: Cyclohexane, D-Limonene, Dodecane, Kerosene, Turpentine, and Green Solvent®. Most of the chemical blends with the solvents extracted about 60% of the oil from shale chips, but the Green Solvent® extracted about 84%. Spontaneous imbibition tests were performed with outcrop Mancos shale cores. Oil was injected into these outcrop cores at a high pressure. NMR T2 distributions were measured for the cores in the original dry state, after oil injection and after imbibition. The aqueous phase from the chemical blend imbibed into the cores and pushed out a part of the oil and gas present in the cores. The surfactant in these blends can change wettability and interfacial tension. The solvent can mix with the oil and solubilize organic solid residues such as asphaltenes. The weak acid can dissolve a part of the carbonate minerals and improve permeability. The synergy can make these chemical blends strong candidates to stimulate oil recovery in shale formations.
Shahri, Mojtaba (Apache Corp.) | James, Moisan (Apache Corp.) | Vasicek, Alan (Apache Corp.) | De Napoli, Roy (Apache Corp.) | White, Matthew (Apache Corp.) | Behounek, Michael (Apache Corp.) | D'Angelo, John (University of Texas at Austin) | Ashok, Pradeep (University of Texas at Austin) | van Oort, Eric (University of Texas at Austin)
Given the intensity of drilling operations in the North American unconventional reservoirs and the quality and amount of data gathered during a drilling operation, leveraging those data along with advanced modeling techniques for optimization purposes is becoming more feasible. In this study, historical data and advanced physical modeling are utilized to better understand and optimize the bottom-hole assembly (BHA) performance in drilling operations. A comprehensive data set is gathered for more than 300 BHA runs in the span of three years. This extensive data set enables thorough examination of the variation in the operational parameters and its effect on the drilling performance.
Different indices are used to determine and evaluate drilling performance, such as rate of penetration (ROP). Excessive tortuosity in a well can have many detrimental effects while drilling such as excessive and erratic torque and drag, poor hole cleaning (cuttings removal), low ROP, along with problematic casing and/or liner runs and associated cementing procedures. In this paper, a tortuosity index (TI) is used to quantify the drilled well quality and correlate it to ultimate drilling performance. In the next step, patterns are extracted and used along with physical modeling for optimizing drilling performance before the well is drilled.
The corresponding tortuosity index can be used as a proxy for the well path smoothness and may be used for quantifying parameters affecting drilling performance. According to historical drilling performance data, there appears to be a strong relationship between wellbore tortuosity and ROP. If drilling operating parameters (e.g., BHA configuration, directional company's performance, target formations, bit specification, mud types, etc.) can be related to the TI based on historical data, such parameters can be modified for optimizing the performance before the well is drilled.
By investigating the historical data, different trends have been extracted. In addition, different models can be built to predict drilling performance (e.g., TI) prior to drilling and according to new well design specifications. Based on data from more than 300 BHA runs and using advanced physical modeling, the most strongly correlated parameters to drilling performance have been determined and shown using different case studies. Such a historical database along with modeling techniques are used to predict well quality and drilling performance during the design phase. Using this method, well design specifications can then be optimized to enhance drilling performance and reduce the cost.
The oil & gas industry uses production forecasts to make a number of decisions as mundane as whether to change the choke setting on a well, or as significant as whether to develop a field. As these forecasts are being used to develop cashflow predictions and value and decision metrics such as Net Present Value and Internal Rate of Return, their quality is essential for making good decision. Thus, forecasting skills are important for value creation and we should keep track of whether production forecasts are accurate and free from bias.
In this paper we compare probabilistic production forecasts at the time of the development FID with the actual annual production to assess whether the forecasts are biased; i.e., either optimistic, overconfident, or both.
While biases in time and cost estimates in the exploration & production industry are well documented, probabilistic production forecasts have yet to be the focus of a major study. The main reason for this is that production forecasts for exploration & production development projects are not publicly available. Without access to such estimates, the quality of the forecasts cannot be evaluated.
Drawing on the Norwegian Petroleum Directorates (NPD) extensive database, annual production forecasts, given at time of project sanction (FID), for 56 fields in the 1995 – 2017 period, have been compared with actual annual production from the same fields. The NPD guidelines specify that the operators should report the annual mean and P10/90-percentiles for the projected life of the field at the time of the FID; that is, the forecasts should be probabilistic. The actual annual production from the fields was statistically compared with the forecast to investigate if the forecasts were biased and to assess the financial impact of such biases.
This paper presents the results from the first public study of the quality of probabilistic production forecasts. The main conclusions are that production forecasts that are being used at the FID for E&P development projects are both optimistic and overconfident. As production forecasts form the basis for the main investment decision in the life of a field, biased forecasts will lead to poor decisions and to loss of value.
Objectives and Scope: Natural fractures were observed in core and image logs from the Hydraulic Fracture Test Site (HFTS) in Reagan Co., Texas. This paper provides an analysis of these fractures, including their orientation, size, spatial distribution, and openness.
Methods: We measured kinematic aperture sizes of two sets of sealed, opening-mode natural fractures in a slant core taken through a stimulated volume, and we analyzed the population distribution using cumulative frequency plots. For the spatial organization study, in addition to fractures identified in the slant core, we used data from image logs from three nearby horizontal producing wells. The spatial organization of fractures was investigated using our statistical method, Normalized Correlation Count (NCC), and by calculating the Coefficient of Variation, Cv, which is a measure of clustering.
Results: In the slant core 197 Set 1 (NE-SW) fractures are present (154 kinematic apertures measured), and there are 112 Set 2 (WNW-ESE) fractures (62 measured). The aperture-size distribution for Set 1 fractures follows a negative-exponential function, whereas Set 2 fractures follow a weak power-law. Only two fractures, both in Set 1 and ~ 1 mm wide, were open in the subsurface, although many more are now parted, mostly in Set 2. Linear intensity, P10, for measured fractures ≥1 mm wide is 0.01 frac/ft (Set 1) and 0.006 frac/ft (Set 2). Both natural fracture sets in an FMI image log from a nearby well have spatial arrangement patterns of regularly-spaced fractal clusters and Cv greater than 3 (3.22 to 4.05). Fracture cluster widths are 100–200 m, and cluster spacings range from 350–600 m. Fractures in COI image logs in two other wells have lower Cv (1.59 to 2.32). Both sets in the 6U well and Set 1 in the 6M well show elevated intensity along the middle section of the well and NCC indicates broad, but weak non-fractal clustering, likely related to lithological control of fracture growth. In the slant core Upper Wolfcamp Set 1 fractures are indistinguishable from random; Set 2 show a log-periodic clustering but with Cv less than 2.
Significance: Incorporation of Discrete Fracture Networks (DFN's) into hydraulic fracture modeling and reservoir simulation requires high-quality natural fracture data from image logs and core. This paper provides such data and provides information on natural-hydraulic fracture interaction at the HFTS site.
US unconventional resource production has developed tremendously in the past decade. Currently, the unconventional operators are trying many strategies such as refracturing, infill drillings and well spacing optimization to improve recovery factor of primary production. They are also employing big data and machine learning to explore the existed production data and geology information to screen the sweet spot from geology point of view. However, current recovery factor of most unconventional reservoirs is still very low (4~10%). A quick production rate decline pushes US operator to pursue gas EOR for unconventional reservoirs, lifting the ultimate recovery factor to another higher level. The goal of this work is to improve oil recovery by implementing gas Huff and Puff process and optimizing injection pattern for one of the US major tight oil reservoirs - Eagle Ford basin. Gas diffusion is regarded as critical for gas Huff and Puff process of tight oil reservoirs. Utilizing the dual permeability model, gas diffusion effect is systematically analyzed and compared with the widely used single porosity model to justify its importance. Transport in natural fractures is proved to be dominated recovery mechanism using dual permeability model. Uncertainty studies about reservoir heterogeneity and nature fracture permeability are performed to understand their influences on well productivity and gas EOR effectiveness. Moreover, three alternative gas injectant compositions including rich gas, lean gas and nitrogen are investigated in gas Huff and Puff processes for Eagle Ford tight oil fractured reservoir. The brief economic evaluation of Huff and Puff project is conducted for black oil region of the Eagle Ford basin.
Hong, Aojie (National IOR Centre of Norway and University of Stavanger) | Bratvold, Reidar B. (National IOR Centre of Norway and University of Stavanger) | Lake, Larry W. (University of Texas at Austin) | Ruiz Maraggi, Leopoldo M. (University of Texas at Austin)
Aojie Hong and Reidar B. Bratvold, National IOR Centre of Norway and University of Stavanger, and Larry W. Lake and Leopoldo M. Ruiz Maraggi, University of Texas at Austin Summary Decline-curve analysis (DCA) for unconventional plays requires a model that can capture the characteristics of different flow regimes. Thus, various models have been proposed. Traditionally, in probabilistic DCA, an analyst chooses a single model that is believed to best fit the data. However, several models might fit the data almost equally well, and the one that best fits the data might not best represent the flow characteristics. Therefore, uncertainty remains regarding which is the "best" model. This work aims to integrate model uncertainty in probabilistic DCA for unconventional plays. Instead of identifying a single "best" model, we propose to regard any model as potentially good, with goodness characterized by a probability. The probability of a model being good is interpreted as a measure of the relative truthfulness of this model compared with the other models. This probability is subsequently used to weight the model forecast. Bayes' law is used to assess the model probabilities for given data. Multiple samples of the model-parameter values are obtained using maximum likelihood estimation (MLE) with Monte Carlo simulation. Thus, the unique probabilistic forecasts of each individual model are aggregated into a single probabilistic forecast, which incorporates model uncertainty along with the intrinsic uncertainty (i.e., the measurement errors) in the given data. We demonstrate and conclude that using the proposed approach can mitigate over/underestimates resulting from using a single decline-curve model for forecasting. The proposed approach performs well in propagating model uncertainty to uncertainty in production forecasting; that is, we determine a forecast that represents uncertainty given multiple possible models conditioned to the data. The field data show that no one model is the most probable to be good for all wells. The novelties of this work are that probability is used to describe the goodness of a model; a Bayesian approach is used to integrate the model uncertainty in probabilistic DCA; the approach is applied to actual field data to identify the most-probable model given the data; and we demonstrate the value of using this approach to consider multiple models in probabilistic DCA for unconventional plays. Introduction Although numerical techniques for forecasting hydrocarbon production have developed rapidly over the past decades, DCA remains an industry-accepted method and is used extensively in the oil and gas industry. Decline-curve models are very computationally attractive because only production data, which can be easily acquired, are required for determining a few parameter values through history matching.
Field data have shown the decline of fracture conductivity during reservoir depletion. In addition, refracturing and infill drilling have recently gained much attention as efficient methods to enhance recovery in shale reservoirs. However, current approaches present difficulties in efficiently and accurately simulating such processes, especially for large-scale cases with complex hydraulic and natural fractures.
In this study, a general numerical method compatible with existing simulators is developed to model dynamic behaviors of complex fractures. The method is an extension of an embedded discrete-fracture model (EDFM). With a new set of EDFM formulations, the nonneighboring connections (NNCs) in the EDFM are treated as regular connections in traditional simulators, and the NNC transmissibility factors are linked with gridblock permeabilities. Hence, manipulating block permeabilities in simulators can conveniently control the fluid flow through fractures. Complex dynamic behaviors of hydraulic fractures and natural fractures can be investigated using this method.
The proposed methodology is implemented in a commercial reservoir simulator in a nonintrusive manner. We first present one synthetic case study in a shale-oil reservoir to verify the model accuracy and then combine the new model with field data to demonstrate its field applicability. Subsequently, four field-scale case studies with complex fractures in two and three dimensions are presented to illustrate the applicability of the method. These studies involve vertical- and horizontal-well refracturing in tight reservoirs, infill drilling, and fracture activation in a naturally fractured reservoir. The proposed approach is combined with empirical correlations and geomechanical criteria to model stress-dependent fracture conductivity and natural-fracture activation. It also shows convenience in dynamically adding new fractures or extending existing fractures during simulation. Results of these studies further confirm the significance of dynamic fracture behaviors and fracture complexity in the analysis and optimization of well performance.